a generic simulation-based perioperative decision support tool for tactical decisions by daphne
TRANSCRIPT
A Generic Simulation-based Perioperative Decision SupportTool for Tactical Decisions
by
Daphne Sniekers
A thesis submitted in conformity with the requirementsfor the degree of Doctor of Philosophy
Graduate Department of Mechanical and Industrial EngineeringUniversity of Toronto
Copyright c© 2013 by Daphne Sniekers
Abstract
A Generic Simulation-based Perioperative Decision Support Tool for Tactical Decisions
Daphne Sniekers
Doctor of Philosophy
Graduate Department of Mechanical and Industrial Engineering
University of Toronto
2013
In Canada and around the world, there has been an increased focus on the efficiency,
cost and access to health care services. One area of particular focus is surgical procedures,
often with government funding and policies focused on reducing wait times through pay
for performance and volume target initiatives. In Ontario, an expert panel was assem-
bled to evaluate the current state of surgical processes and provide recommendations to
improve access, efficiency and quality. This thesis addresses the panel’s recommendation
for a simulation-based decision tool to help hospitals inform decisions that can lead to
improved access and efficiency.
A generalised, simulation based perioperative decision tool is presented that can be
used to test a variety of tactical decisions. The generic model has been applied to six
hospitals of varying sizes, ranging from large academic centres to small rural community
hospitals. The model remains in use at some of the hospitals to regularly inform decisions.
The model is also being applied to additional hospital sites.
During application of the generic model, challenges in design decisions and validation
were encountered. As a result, a series of principles are proposed to guide future generic
modelling design and achieving user acceptance. These principles add to the generic
simulation modelling and healthcare modelling research fields by laying some groundwork
for a formalised approach to designing effective generic simulation models and achieving
confidence in results.
ii
Finally, the research demonstrates two uses of the generic model: as decision tool
and as a demonstrative tool. As a decision tool the model is used to compare numerous
potential tactical decision options under consideration. As a demonstrative tool, the
model is used to quantify the effect of poor practices on hospital performance. The design
of the generic model only considers efficient processes and best practices. When model
results are compared to historical performance, decision makers are able to quantify the
effect of existing poor practices on their performance and decision making. The tool
enables users to base their tactical level decisions on the assumption that good practices
and procedures are followed.
iii
Dedication
To my husband Sal Stranges, my parents Ghislaine Cleiren and Jan Sniekers and my
sister Karen Sniekers for supporting me through this journey.
iv
Acknowledgements
First, I would like to thank my supervisor Professor Michael Carter for his encour-
agement and support throughout this research endeavor. He has been a great mentor in
my academic and industry pursuits. I would also like to thank him for his patience when
I chose to pursue an industry career while completing my PhD. Prof. Carter’s dedication
and enthusiasm for improving Canada’s healthcare system using operation research is
inspiring.
I would also like to thank my doctoral committee members, Professors Chris Beck
and Daniel Frances for their insights into improving the academic rigor of this research.
Further, I would like to acknowledge Erwin W. Hans who acted as the external reviewer;
his comments and appraisal lead to important improvements in the final thesis prod-
uct. Finally, I would like to extend my gratitude to the final member of the doctoral
examination committee Professor Dionne Aleman for her participation and review.
To the granting agencies (HTX and CIHR), hospitals (St. Michael’s, Hamilton Health
Sciences, Mount Sinai Hospital, and William Osler Health System), and industry partner
(Visual8) who contributed both financially and by providing valuable information, data
and feedback for this research program and resulting generic simulation model. I am
indebted to your generosity, without which this research would never have been possible.
Finally, I would like to thank my family and friends who have supported and guided
me to pursue my doctorate. Specifically, I would like to thank my parents, Jan Sniekers
and Ghislaine Cleiren, for their support throughout my academic and personal pursuits,
including many much needed pushes along the way to stretch myself to reach higher
goals.
Finally, a big thank you to my husband, Sal Stranges, for his uplifting and continued
counsel, words of encouragement and love.
v
Contents
1 Introduction 1
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Research Contributions and Objectives . . . . . . . . . . . . . . . . . . . 5
1.4 Layout . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2 Project Overview and Organisation 7
2.1 Project Organisation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.1.1 Initial Pilot Model . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.1.2 Development of the Proposed Generic Model . . . . . . . . . . . . 10
2.1.3 Application to Multiple Sites . . . . . . . . . . . . . . . . . . . . 11
2.1.4 Current and Future Application . . . . . . . . . . . . . . . . . . . 12
2.2 Descriptions of the Hospitals Studied . . . . . . . . . . . . . . . . . . . . 12
2.2.1 Hamilton HS Juravinski Hospital . . . . . . . . . . . . . . . . . . 14
2.2.2 St. Michael’s Hospital . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2.3 Mount Sinai Hospital . . . . . . . . . . . . . . . . . . . . . . . . 16
2.2.4 William Osler HS Brampton Civic and Etobicoke General Hospitals 16
3 Background and Literature Review 18
3.1 The Perioperative Process . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.1.1 Patient Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
vi
3.1.2 Pre-Operative Activities . . . . . . . . . . . . . . . . . . . . . . . 20
3.1.3 Operative Activities . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.1.4 Post-Operative Activities . . . . . . . . . . . . . . . . . . . . . . . 21
3.2 Perioperative Decision Making . . . . . . . . . . . . . . . . . . . . . . . . 21
3.2.1 Strategic Perioperative Decisions . . . . . . . . . . . . . . . . . . 22
3.2.2 Tactical Perioperative Decisions . . . . . . . . . . . . . . . . . . . 22
3.2.3 Operational Perioperative Decisions . . . . . . . . . . . . . . . . . 23
3.3 Review of Models on Perioperative Planning and Scheduling . . . . . . . 23
3.3.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.4 Review of Existing Scheduling Tools and Software . . . . . . . . . . . . . 35
3.5 Review of Generalised Models and Frameworks . . . . . . . . . . . . . . 36
3.5.1 Defining a Generalised Model . . . . . . . . . . . . . . . . . . . . 36
3.5.2 Generalised Models and Frameworks within Healthcare . . . . . . 38
3.5.3 Generalised Models and Frameworks outside of Healthcare . . . . 43
3.5.4 Design Concepts for Generic Models . . . . . . . . . . . . . . . . 44
3.5.5 Validation and Verification of Generic Models . . . . . . . . . . . 47
3.5.6 Cost-Benefit Analysis of Generic Models . . . . . . . . . . . . . . 48
4 The Proposed General Model 51
4.1 The Generic Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.1.1 Model Objective . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.1.2 Model Scope, Level of Detail, Assumptions and Simplifications . . 52
4.1.3 Model Outputs . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
4.1.4 Model Inputs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.2 Application of the Generic Model . . . . . . . . . . . . . . . . . . . . . . 66
4.2.1 Use of the Model at Multiple Sites . . . . . . . . . . . . . . . . . 66
4.2.2 Issues Experienced and Changes Required while Implementing . . 70
4.2.3 Limitations Noted . . . . . . . . . . . . . . . . . . . . . . . . . . 73
vii
4.3 Cost Effectiveness of the Generic Model . . . . . . . . . . . . . . . . . . . 75
4.3.1 Cost analysis at six hospitals . . . . . . . . . . . . . . . . . . . . . 79
4.4 Proposed Guidelines for Designing Generic Models . . . . . . . . . . . . . 82
4.4.1 Clearly Define the Problem Description, Scope, Objectives, etc. . 82
4.4.2 Study Multiple Sites Prior to Design . . . . . . . . . . . . . . . . 83
4.4.3 Keep it Simple . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
4.4.4 Collaborate with End Users . . . . . . . . . . . . . . . . . . . . . 85
5 Generic Model Validation 86
5.1 Validation at the Test Sites . . . . . . . . . . . . . . . . . . . . . . . . . 86
5.1.1 Achieving Validity and User Acceptance at Brampton Civic Hospital100
5.2 Challenges of Achieving User Acceptance . . . . . . . . . . . . . . . . . . 110
5.3 Compared to a Specific Model . . . . . . . . . . . . . . . . . . . . . . . . 113
5.3.1 Validity of the Specific Model . . . . . . . . . . . . . . . . . . . . 114
5.3.2 Comparing the Specific and Generic Models . . . . . . . . . . . . 115
5.3.3 Differences in Model Design . . . . . . . . . . . . . . . . . . . . . 117
5.3.4 Challenges Compared to a Specific Model . . . . . . . . . . . . . 117
5.4 Guidelines for Achieving User Acceptance . . . . . . . . . . . . . . . . . 120
5.4.1 Step 1: Assemble an Engaged Working Group . . . . . . . . . . . 121
5.4.2 Step 2: Review of the Generic Conceptual Model . . . . . . . . . 121
5.4.3 Step 3: Calibrate Model . . . . . . . . . . . . . . . . . . . . . . . 122
5.4.4 Step 4: Reconcile Model Results . . . . . . . . . . . . . . . . . . . 123
5.4.5 Step 5: Repeat and Re-evaluate . . . . . . . . . . . . . . . . . . . 124
5.4.6 Step 6: Accept the Model as Correct . . . . . . . . . . . . . . . . 124
5.4.7 Final Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
6 A Decision and Demonstrative Tool: Case Studies 126
6.1 As a Decision Tool at the Juravinski Hospital . . . . . . . . . . . . . . . 127
viii
6.1.1 Problem Description . . . . . . . . . . . . . . . . . . . . . . . . . 127
6.1.2 Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
6.1.3 Application of the Generic Model . . . . . . . . . . . . . . . . . . 131
6.1.4 Results from the Generic Model . . . . . . . . . . . . . . . . . . . 132
6.1.5 Implementation of the Decision . . . . . . . . . . . . . . . . . . . 138
6.1.6 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
6.2 As a Demonstrative and Decision Tool at William Osler Health Sciences 144
6.2.1 Problem Description . . . . . . . . . . . . . . . . . . . . . . . . . 144
6.2.2 Application of the Generic Model . . . . . . . . . . . . . . . . . . 146
6.2.3 The Generic Model - Results as a Demonstrative Tool . . . . . . . 147
6.2.4 The Generic Model - Results as a Decision Tool . . . . . . . . . . 154
6.2.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
7 Conclusions and Future Work 165
7.1 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
7.1.1 Additional Functionality and Addressing Limitations . . . . . . . 169
7.1.2 Development of a Commercial Product . . . . . . . . . . . . . . . 169
7.1.3 Additional Uses of the Model . . . . . . . . . . . . . . . . . . . . 171
Bibliography 174
Appendices 186
A Conceptual Model Tables 186
B Change Database 208
C Generic Model Detailed Description 214
C.1 Introduction to the Proposed Framework . . . . . . . . . . . . . . . . . . 214
C.2 Overall Framework Structure . . . . . . . . . . . . . . . . . . . . . . . . 214
ix
C.2.1 Flow of Patients and Information . . . . . . . . . . . . . . . . . . 216
C.2.2 Modelled Patients . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
C.3 General Model Set Up Information . . . . . . . . . . . . . . . . . . . . . 221
C.4 Pre-surgical Processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
C.4.1 Pre-Surgery Module . . . . . . . . . . . . . . . . . . . . . . . . . 223
C.5 The Surgical Day Processes . . . . . . . . . . . . . . . . . . . . . . . . . 233
C.5.1 Surgical Day Module . . . . . . . . . . . . . . . . . . . . . . . . . 233
C.6 Post Surgical Processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
C.6.1 Post-surgery Module . . . . . . . . . . . . . . . . . . . . . . . . . 236
C.7 Hospital Resources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242
C.7.1 The Operating Room Resource . . . . . . . . . . . . . . . . . . . 243
C.7.2 Bed Resources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
C.8 Start of the Day Module . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
C.8.1 Step 1: Generate the OR schedule . . . . . . . . . . . . . . . . . . 250
C.8.2 Step 2: Schedule patients . . . . . . . . . . . . . . . . . . . . . . . 250
C.8.3 Step 3: Bed Management . . . . . . . . . . . . . . . . . . . . . . 250
C.9 Scheduling Details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
C.9.1 Schedule Inputs . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254
C.9.2 Scheduling Rules . . . . . . . . . . . . . . . . . . . . . . . . . . . 263
C.9.3 Daily Schedule Creation . . . . . . . . . . . . . . . . . . . . . . . 268
C.9.4 Elective Patient Scheduling Algorithm . . . . . . . . . . . . . . . 268
C.10 Additional Framework Details . . . . . . . . . . . . . . . . . . . . . . . . 270
C.10.1 Determine next OR Activity Decision Tree . . . . . . . . . . . . . 270
C.10.2 Further Information about Cancellations . . . . . . . . . . . . . . 273
C.11 Framework Output Files . . . . . . . . . . . . . . . . . . . . . . . . . . . 275
C.11.1 Wait List Report . . . . . . . . . . . . . . . . . . . . . . . . . . . 276
C.11.2 Cancellation Report . . . . . . . . . . . . . . . . . . . . . . . . . 277
x
C.11.3 OR Reports . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 278
C.11.4 PACU Report . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279
C.11.5 Ward, ICU and SDU Reports . . . . . . . . . . . . . . . . . . . . 281
C.11.6 Throughput Report . . . . . . . . . . . . . . . . . . . . . . . . . . 282
C.11.7 Census Report . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282
D Generic Model Implemented in Simul8 284
D.1 Screen Shots of the Generic Model Implemented in Simul8 . . . . . . . . 284
E Juravinski Process Map 288
F St. Mike’s Process Map 292
G Mt. Sinai Process Map 297
H William Osler Process Map 302
xi
List of Tables
2.1 Key characteristics of the hospitals studied. . . . . . . . . . . . . . . . . 13
4.1 Costs (in terms of time spent) of the specific model applied to the Juravin-
ski orthopaedic service and the generic model applied to seven instances
at six hospital sites. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
5.1 Validation results from the Juravinski Hospital’s all services generic model
application. Bold type indicated when the confidence interval from the
model results does not include the historical value. . . . . . . . . . . . . . 90
5.2 Validation results from St. Mike’s Hospital all services generic model ap-
plication - Throughput. Bold type indicated when the confidence interval
from the model results does not include the historical value. . . . . . . . 91
5.3 Validation results from St. Mike’s Hospital all services generic model ap-
plication - Cancellations. Bold type indicated when the confidence interval
from the model results does not include the historical value. . . . . . . . 92
5.4 Validation results from Mt. Sinai Hospital general surgical service generic
model application. Bold type indicated when the confidence interval from
the model results does not include the historical value. . . . . . . . . . . 93
5.5 Validation results from Prince Albert Hospital generic model application.
Bold type indicated when the confidence interval from the model results
does not include the historical value. . . . . . . . . . . . . . . . . . . . . 94
xii
5.6 Validation results from Brampton Hospital generic model application -
Throughput. Bold type indicated when the confidence interval from the
model results does not include the historical value. . . . . . . . . . . . . . 96
5.7 Validation results from Brampton Hospital generic model application -
Cancellations. Bold type indicated when the confidence interval from the
model results does not include the historical value. . . . . . . . . . . . . . 97
5.8 Validation results from Etobicoke Hospital generic model application -
Throughput. Bold type indicated when the confidence interval from the
model results does not include the historical value. . . . . . . . . . . . . . 98
5.9 Validation results from Etobicoke Hospital generic model application -
Cancellations. Bold type indicated when the confidence interval from the
model results does not include the historical value. . . . . . . . . . . . . . 99
5.10 First attempt at validating at Brampton Hospital using the Fall 2009 MSS
schedule as planned. Bold type indicated when the confidence interval
from the model results does not include the historical value. . . . . . . . 101
5.11 First attempt at validating at Brampton Hospital using the Fall 2009 MSS
schedule as planned - Cancellations. Bold type indicated when the confi-
dence interval from the model results does not include the historical value. 102
5.12 Second round at validating at Brampton Hospital using the November
2009 MSS schedule as a representative schedule for Fall 2009. Bold type
indicated when the confidence interval from the model results does not
include the historical value. . . . . . . . . . . . . . . . . . . . . . . . . . 107
5.13 Second round at validating at Brampton Hospital using the November 2009
MSS schedule as a representative schedule for Fall 2009 - Cancellations.
Bold type indicated when the confidence interval from the model results
does not include the historical value. . . . . . . . . . . . . . . . . . . . . 108
xiii
5.14 Comparing validation results of the specific model to the generic model for
the Juravinski Hospital’s orthopaedic service application. Bold faced types
is used to indicate when the historical value is outside of the confidence
interval produced by the model. . . . . . . . . . . . . . . . . . . . . . . . 116
5.15 Comparing the challenges faced when modelling perioperative services,
using the proposed generic model and to using a specific, custom made
model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
6.1 The percent change compared to the base case for three of the scenarios
considered by the Juravinski - Throughput . . . . . . . . . . . . . . . . . 135
6.2 The percent change compared to the base case for three of the scenarios
considered by the Juravinski - Cancellations . . . . . . . . . . . . . . . . 136
6.3 Demonstrative Tool Report - Illustrating the effect of poor practices on
throughput volumes at Brampton. . . . . . . . . . . . . . . . . . . . . . . 149
6.4 Differences in elective OR time by service from the MSS to actual in April-
May 2010. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
6.5 The proposed base MSS for Brampton based on the demonstrative results
from the generic model. . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
6.6 The predicted throughput and cancellation rates achievable from the pro-
posed base MSS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
6.7 Model throughput results from short term MSS changes. . . . . . . . . . 156
6.8 Model cancellation rate results from short term MSS changes. . . . . . . 157
6.9 Results from comparing orthopaedic scenarios varying the number of TJR
reserved blocks. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
A.1 Model scope: components identified within the boundary of the study. . . 186
A.2 Level of detail: detail included for each component included in the model. 190
A.3 Model Assumptions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
xiv
A.4 Model Simplifications. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
B.1 Change data base. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
C.1 I: Wait List Expiry Information - Elective/ Urgent/ Inpatient . . . . . . 218
C.2 I: Patient Input File - Elective/Inpatient/Urgent . . . . . . . . . . . . . . 220
C.3 I: Wait List Rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
C.4 I: General Simulation Information . . . . . . . . . . . . . . . . . . . . . . 222
C.5 I: Model Initialisation - Wait List Sizes . . . . . . . . . . . . . . . . . . . 222
C.6 I: Arrival Patterns - Elective/Inpatient . . . . . . . . . . . . . . . . . . . 224
C.7 M: Current Patient File . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
C.8 I: Urgent Patient Scheduling Rules . . . . . . . . . . . . . . . . . . . . . 230
C.9 I: Inpatient Scheduling Rules . . . . . . . . . . . . . . . . . . . . . . . . . 232
C.10 I: Off-Servicing Nurse Ratios . . . . . . . . . . . . . . . . . . . . . . . . . 235
C.11 I: Turnover Times . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
C.12 I: Post OR Routes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240
C.13 I: Allowable Wait for PACU . . . . . . . . . . . . . . . . . . . . . . . . . 240
C.14 I: Off-servicing Rule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
C.15 I: Resources to Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
C.16 I: Resource PACU - weekday/weekend . . . . . . . . . . . . . . . . . . . 246
C.17 I: PACU on call flag - weekday/weekend . . . . . . . . . . . . . . . . . . 246
C.18 I: Bed Resources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
C.19 I: Non-surgical Patient Bed Occupancy - Ward/SDU/ICU . . . . . . . . 249
C.20 I: Urgent Patient Arrival Information . . . . . . . . . . . . . . . . . . . . 252
C.21 M: Beds Available Today - Ward/ICU/SDU . . . . . . . . . . . . . . . . 253
C.22 I: Schedule - Details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
C.23 I: Schedule - Function and Service . . . . . . . . . . . . . . . . . . . . . . 256
C.24 I: Schedule - Service . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
xv
C.25 I: Schedule - Surgeons . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259
C.26 I: Schedule - Surgeons # ORs . . . . . . . . . . . . . . . . . . . . . . . . 259
C.27 M: Schedule - Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
C.28 M: Schedule - Service . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
C.29 M: Schedule - Surgeon . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262
C.30 M: Schedule - Time Remaining . . . . . . . . . . . . . . . . . . . . . . . 262
C.31 M: Schedule - Patients . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263
C.32 I: Scheduling Rules List . . . . . . . . . . . . . . . . . . . . . . . . . . . 264
C.33 I: Scheduling Rules Schedule . . . . . . . . . . . . . . . . . . . . . . . . . 265
C.34 M: Schedule Rules Tracking . . . . . . . . . . . . . . . . . . . . . . . . . 265
C.35 I: Scheduling Ordering . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
C.36 I: Overtime Allowance - Elective/Urgent/Inpatient . . . . . . . . . . . . 272
C.37 I: Other Cancellations - Elective/Inpatient/Urgent . . . . . . . . . . . . . 274
C.38 M: Wait List Tracking . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276
C.39 O: Wait List Report . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277
C.40 O: Cancellation Report . . . . . . . . . . . . . . . . . . . . . . . . . . . . 278
C.41 O: OR Report - Utilisation . . . . . . . . . . . . . . . . . . . . . . . . . . 279
C.42 O: OR Report -Over and Under Time . . . . . . . . . . . . . . . . . . . . 280
C.43 O: OR Delay Report . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 280
C.44 O: PACU Report . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281
C.45 O: PACU Report - Diverts . . . . . . . . . . . . . . . . . . . . . . . . . . 281
C.46 O: Ward/ICU/SDU Report . . . . . . . . . . . . . . . . . . . . . . . . . 282
C.47 O: Throughput Report . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282
C.48 O: Census Report . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283
xvi
List of Figures
1.1 Top nine reasons for cancellations in the NHS. . . . . . . . . . . . . . . 4
1.2 Reasons for delay throughout perioperative patient flow according findings
from the SPAI report. (Zellermeyer, 2005) . . . . . . . . . . . . . . . . . 5
3.1 The three stages of perioperative activity. . . . . . . . . . . . . . . . . . . 19
4.1 High level surgical patient flow and generic model boundary shown. . . . 53
4.2 Pictorial summary of perioperative processes included and excluded in the
proposed generic model. . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.3 A more detailed pictorial summary of components included and excluded
in the proposed generic model. . . . . . . . . . . . . . . . . . . . . . . . . 56
4.4 Summary of model inputs of the proposed generic perioperative simulation
model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.5 Definitions of variables for cost analysis. . . . . . . . . . . . . . . . . . . 76
5.1 Example of a report provided for validation analysis comparing the amount
of time booked by surgeon versus the amount of time actually used during
elective block time. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
5.2 Example of a report provided for validation analysis comparing the per-
centage of patients by surgeon historically and within the input patient
file. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
xvii
5.3 Comparing the specific and generic model in terms of areas included and
key assumptions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
6.1 Daily midnight census results from selected scenarios presented to Juravin-
ski Hospital. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
6.2 Comparing average daily unit census of the simulation model and the
seven-week trial. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
6.3 Emergency Department length of stay during the trial compared to the
previous year. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
6.4 Model average census results from short term MSS changes. . . . . . . . 157
6.5 Total average census by service as predicted from the Monte Carlo bed
model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
6.6 Difference in average census results from long term orthopaedic MSS changes.162
C.1 The overall view of the framework, showing the three modules’ patient flows.215
C.2 Legend of shapes used in the pictorial framework. . . . . . . . . . . . . . 216
C.3 The pre-surgery module. . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
C.4 The elective patient management sub-module. . . . . . . . . . . . . . . . 227
C.5 The urgent patient management sub-module. . . . . . . . . . . . . . . . . 229
C.6 The inpatient management sub-module. . . . . . . . . . . . . . . . . . . . 232
C.7 The surgical day module. . . . . . . . . . . . . . . . . . . . . . . . . . . . 233
C.8 The OR Activities sub-module. . . . . . . . . . . . . . . . . . . . . . . . 234
C.9 The post-surgery module. . . . . . . . . . . . . . . . . . . . . . . . . . . 237
C.10 The determine next location decision tree. . . . . . . . . . . . . . . . . . 239
C.11 The Determine Next Activity Process. . . . . . . . . . . . . . . . . . . . 271
D.1 Screen shot of generic model in Simul8. . . . . . . . . . . . . . . . . . . . 285
D.2 Screen shot of generic model in Simul8 - focused in on arrival processes
and waiting lists. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 286
xviii
D.3 Screen shot of generic model in Simul8 - focused in on post-surgical units
(ICU, SDU, ward and flex beds) . . . . . . . . . . . . . . . . . . . . . . . 286
D.4 Screen shot of generic model in Simul8 - focused in on operative day pro-
cesses. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287
E.1 The pre surgical patient flow at Juravinski. . . . . . . . . . . . . . . . . . 288
E.2 The scheduling details at Juravinski. . . . . . . . . . . . . . . . . . . . . 289
E.3 The pre-op clinic details at Juravinski. . . . . . . . . . . . . . . . . . . . 289
E.4 The surgical day patient flow at Juravinski. . . . . . . . . . . . . . . . . 289
E.5 The decision tree on whether to proceed with a case, at Juravinski. . . . 290
E.6 The patient flow and decisions made during the procedure at Juravinski. 290
E.7 The post surgical patient flow at Juravinski. . . . . . . . . . . . . . . . . 291
F.1 The pre surgical patient flow at St. Mike’s. . . . . . . . . . . . . . . . . . 292
F.2 The elective patient scheduling process map at St. Mike’s. . . . . . . . . 293
F.3 The inpatient patient scheduling process map at St. Mike’s. . . . . . . . 293
F.4 The urgent patient scheduling process map at St. Mike’s. . . . . . . . . . 294
F.5 The surgical day process map at St. Mike’s. . . . . . . . . . . . . . . . . 294
F.6 The decisions made the morning of surgery at St. Mike’s. . . . . . . . . . 295
F.7 The process map within the OR at St. Mike’s. . . . . . . . . . . . . . . . 295
F.8 The decision tree to determine the next OR activity at St. Mike’s. . . . . 296
F.9 The post-surgical day patient flow at St. Mike’s. . . . . . . . . . . . . . . 296
G.1 The process flow map for DS patients at Mt. Sinai. . . . . . . . . . . . . 298
G.2 The process flow map for SDA patients at Mt. Sinai. . . . . . . . . . . . 299
G.3 The process flow map for inpatients at Mt. Sinai. . . . . . . . . . . . . . 300
G.4 The process flow map for urgent patients at Mt. Sinai. . . . . . . . . . . 301
H.1 The pre-operative process flow map for elective patients at William Osler. 303
xix
H.2 The pre-operative process flow map for urgent and in- patients at William
Osler. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 304
H.3 The operative (day of surgery) process flow map for patients at William
Osler. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 304
H.4 The post-operative process flow map for patients at William Osler. . . . 305
xx
Chapter 1
Introduction
To aid in perioperative tactical decision making, a generic simulation model of peri-
operative patient flow is proposed. The value of the model is demonstrated through
application to multiple sites to inform decisions and demonstrate possible best practice
performance. Being a generic model does not significantly impact the value, validity or
cost effectiveness of the model compared to a specific model.
1.1 Motivation
Over the last decade many countries, including Canada, have been concerned with the
amount of time their citizens wait for medical procedures and diagnostics such as surg-
eries, emergency department visits, and MRI and CT scans. Countless improvement
initiatives, money and time have been dedicated to reducing the amount of time patients
must wait.
Beginning in 2004 the Ontario provincial government chose to focus on five key wait-
ing lists: joint replacements, cardiac procedures, cancer procedures, cataract procedures
and MRI diagnostics. The government set up the Wait Time Initiative, which involved
contracting hospitals to meet specified volumes often with financial incentives attached
to meeting the target. If the hospital failed to meet the target, they could face potential
1
Chapter 1. Introduction 2
claw-back of the money received, and a reduced funded target the following year. If the
hospital exceeded their target, they could receive additional money and a higher tar-
get the following year. The initiative also included the development of a province-wide
Wait Time Information System (WTIS) that was later expanded to include all surgical
procedures (MacLeod et al., 2009).
Additionally, Ontario formed expert panels on a number of key areas within the health
care system to provide recommendations and to identify further areas of improvement.
One such panel was the Surgical Process Analysis and Improvement Expert Panel (SPAI).
In 2005, SPAI reported on 22 recommendations; recommendation 12, which is part of
the motivation of this thesis, reads as follows:
The Ministry of Health and Long-Term Care develop a request for pro-
posals to support the development of a Peri-Operative Simulation System
that would be accessible to all Ontario hospitals, and could be used by Local
Health Integration Networks for planning purposes. Hospitals should use this
system to simulate the impact peri-operative decisions after they have mapped
and clearly understand their peri-operative processes (see Recommendation
3). (Zellermeyer, 2005, p. 35)
This recommendation proposes that a patient-level model be created that will allow
hospitals to review and test changes to their processes and procedures in order to improve
the flow of surgical patients. The panel identified that the advantage of simulation is that
it allows one to run a number of different improvement ideas quickly and analyse outputs
without affecting patient care until a good choice is made. For example, the hospital
“can identify the impact of ten additional total joint cases per week on a hospital’s peri-
operative services (e.g. on pre-assessment services, operations, recovery room resources,
post-surgical bed utilisation, and wait times for other types of patients). (Zellermeyer,
2005, p. 34)” The model would serve as a decision making tool for managers to become
proactive instead of reactive about resource planning and reaching volume and wait time
Chapter 1. Introduction 3
targets.
SPAI further identified in recommendation three that in order to properly understand
their processes and issues, hospitals should map out their current processes in order to
analyse the results and “systematically identify areas for improvement” (Zellermeyer,
2005, p. 22). This ties in with the simulation model, as a detailed understanding of the
processes and issues is needed in order to accurately simulate patient flow.
SPAI identified that many hospitals do not have the capability to design and imple-
ment their own simulation model. Hence, the panel recommended that the Ministry of
Health and Long Term Care (MOHLTC) fund the development of a generic simulation
model that can be easily implemented at any Ontario hospital.
It is within this spirit that this research endeavour began; with the overreaching idea
to develop a generic model that could be commercialised to support hospitals in their
surgical patient flow planning. Moreover, partnering with hospitals, grant agencies and
a local simulation software consultancy group to encourage potential future commercial-
isation of the generic model.
1.2 Background
The peri-operative process refers to all the activities that occur from the time the decision
to perform surgery has been made up to and including any immediate post-operative
recovery. It can be sub-divided into three processes (Zellermeyer, 2005):
• Pre-Operative: diagnostics, routine testing, patient education, preparation for surgery
and preparation for discharge from the OR and hospital;
• Operative: the surgical day;
• Immediate Post-Operative: recovery room/post-aesthetic care unit (PACU).
Chapter 1. Introduction 4
Formally, the perioperative process does not include critical care, general ward units,
and discharge to home care, long-term care and rehabilitation. Many surgical models and
discussions find access to these beds out of scope when considering solutions to improving
perioperative patient flow. However, there is a significant amount of interrelatedness and
influence that these resources impart on perioperative flow. For example, many hospitals
experience a high surgical cancellation rate due to the lack of availability of ward or
critical care beds.
Due to the complexity of surgical care and the numerous players and resources re-
quired to complete a case, many bottlenecks exist; impeding flow through delays, cancel-
lations and inefficiencies. An NHS study found that more than 50% of surgical cancel-
lations were due to non-clinical reasons (Modernisation, 2001). Figure 1.1 lists the top
nine reasons for same-day cancellations found by the NHS. Five of the reasons are within
control of the hospital and physicians, accounting for 44% of cancellations. These five
can be reduced through improved co-ordination and planning.
0.00% 5.00% 10.00% 15.00% 20.00% 25.00%
inconvenient appointment
emergency/trauma case
equipment failure/unavailable
surgeon unavailable
no longer required
list overruns
unfit
no shows
no ward bed
Top Reasons for Surgical Case Cancella3ons in the UK
Figure 1.1: Top nine reasons for cancellations in the NHS.
Chapter 1. Introduction 5
The SPAI report identified many reasons that can cause procedures to be delayed or
cancelled throughout the perioperative flow, as demonstrated in figure 1.2 (Zellermeyer,
2005). These delays can all be considered under the control of the hospital and are
symptoms of underlying processes and procedures. Effort to reduce the occurrence and
severity of these delays is key to improving surgical patient flow. However, caution is
required to avoid reducing one delay at the expense of increasing the occurrence of another
type of delay somewhere else along the process. Further background on the perioperative
process is provided in chapter 3.Section C: The Context For Surgical Process Analysis and Improvements
!""#$%&'#(
)*+',-.#/&'%0#(
((-.#/&'%0#(()/#,-.#/&'%0#( )*+',-.(
1#2*0#/3(
! Patient does not show up
! Patient not screened
appropriately to ensure
readiness for surgery
! Patient and family not
educated to understand
procedure and participate in
care
! Incomplete diagnostic tests
! Paperwork incomplete
! Chart incomplete/not
reviewed
! Discharge process not begun
for in-patients
! Home care not arranged for
out-patients
)*'#4'%&5(65*27&8#+(95*48(':#()#/%,-.#/&'%0#(;'&8#((
! Surgeon not available
! Anesthesia not available
! Other members of surgical
team not available
! OR not prepared (supplies,
instruments, case carts)
! Insufficient time scheduled for
the surgeries
! Inaccurate scheduling or
booking
! Cases not sequenced into
blocks
! Number of cases capped
! Equipment failure
! Insufficient capacity (staff,
supplies, instruments, blood)
! Instrument tray inaccurate
! No flex for emergency cases
! Poor communications
! Post-anesthetic care unit or
critical care bed unavailable
! Insufficient nursing and post-
op staff
! Patient not discharged in a
timely fashion
! Transport delays
! Inappropriate utilisation of
post-op resources
! No ward bed
! Non-surgical
patients in
surgical beds
! No home
care
! No rehab
bed/service
! No long-
term care
bed
<=! 9>(->?-!>?(@-AB;(->()9C!D>C(;9@DCE((
International research indicates that the rate of potentially avoidable adverse events in
hospitals ranges from 7.5% to 17% of all hospital admissions.8 The recent Canadian
study of adverse events reported that 43% of adverse events among hospital patients were
related to surgery.9
Preventing and controlling infections is critical for patient safety. According to the
Institute for Healthcare Improvement (IHI), 40 to 60 percent of clean case infections are
preventable.10
Surgical site infections are the second most common adverse event in
hospitalised patients,11
and are known to increase mortality, readmissions, length of stay
8 Baker GR, Norton PG et al. “The Canadian Adverse Events Study: the incidence of adverse events among
hospital patients in Canada” CMAJ 2004; 25 May 170 (11): 1678-1686. Carthey J “Institutional Resilience
in Healthcare Systems” Quality in Health Care 2001; 10: 29-32. Leape LL, Berwick DM “Safe Health
Care: Are We Up to It?” British Medical Journal 2000; 320: 725-726. Vincent C et al. “How to Investigate
and Analyse Clinical Incidents: Clinical Risk Unit and Association of Litigation and Risk Management
Protocol” British Medical Journal 2000; 320: 777-781. 9 Baker GR, Norton PG et al., Ibid. 10 Institute for Healthcare Improvement. Getting Started Kit: Prevent Surgical Site Infections How-to
Guide. 100,000 Lives Campaign. www.ihi.org/IHI/Programs/Campaign/ 11 Brennan, New England Journal of Medicine 1991; 324: 370-376.
17
Figure 1.2: Reasons for delay throughout perioperative patient flow according findings
from the SPAI report. (Zellermeyer, 2005)
1.3 Research Contributions and Objectives
The work proposed herein contributes to healthcare modelling research in the following
three ways:
1. The model itself: The proposed generic model contributes to the healthcare oper-
Chapter 1. Introduction 6
ational research field by demonstrating that a generic model of a complex healthcare
system process is feasible. Moreover, a generic model can be meaningfully applied
to multiple sites to influence and inform decisions.
2. A validation process for generic modelling of large, complex real world
systems: A validation and reconciliation schema is proposed to provide guidance
on how to convince end-users (decision makers) that results are reasonable and
trustworthy, even if statistical prediction error remains.
3. An effective decision and demonstrative tool applied successfully to mul-
tiple sites: The generic model has been proven to support decision makers as a
decision and negotiation tool by allowing comparisons of decision options against
multiple selected key performance indicators. In addition, the generic model has
proven useful as a tool to quantify inefficiencies in current practices by demonstrat-
ing possible performance if those practices were improved.
1.4 Layout
The purpose of this document is to provide details into the contributions of this research.
Chapters 4, 5 and 6 will provide these details and relevant discussion. Before delving
into these discussions, an overview of the research project and participating hospitals is
provided in chapter 2. Next, a background on perioperative process and a review of the
relevant literature is provided in chapter 3. Limitations and future work are discussed in
the final chapter (chapter 7).
Chapter 2
Project Overview and Organisation
The purpose of this chapter is twofold. First it offers a brief history of the research
project in terms of what tasks were performed, their sequence and which research team
member(s) completed the tasks. Afterwards, a short description of the key characteristics
of each hospital involved in the study is provided.
2.1 Project Organisation
The research project began in the fall of 2005 when the idea to design a generic model
was formed based on one of the recommendations from the Surgical Process Analysis
and Improvement (SPAI) Expert Panel Report. The research began with an extensive
literature search for existing generic models of surgical processes and other healthcare
systems. A search of existing software was also conducted. The purpose of the literature
and software investigation was threefold. First, to determine if such a model already
existed or could be created based on an already existing tool. Second, to gather insight
on the complexities, lessons learned, etc. of modelling perioperative systems. Finally, to
learn about typical perioperative processes, procedures and best practices.
In 2006, three initial hospitals were recruited as pilot sites for the generic model: the
Juravinski Hospital of Hamilton Health Sciences, St. Michael’s Hospital in Toronto, and
7
Chapter 2. Project Overview and Organisation 8
Mount Sinai Hospital in Toronto. Together with the simulation consulting firm Visual8,
an industry partner research grant was awarded for the development of a prototype
model. Furthermore, the intent of this development is that the industry partner, Visual8,
would work towards commercialisation of the resulting prototype model. This allowed
for considerable resources to aid the study in terms of funding to recruit research team
members, including a simulation model programmer, research students to conduct data
collection and analysis, and technical and programming support for the chosen simulation
model platform, Simul8.
2.1.1 Initial Pilot Model
At Juravinski, this research project was part of a larger hospital-wide process improve-
ment initiative called the Innovation and Learning initiative (I&L). Accordingly, a large
working group was assembled to support the project and make changes to processes and
policies to improve surgical patient flow. The scope of the working group also included
other operational process improvement initiatives within surgical care. The working team
included administrators and nursing leaders from the surgical and inpatient areas; physi-
cians, including surgeons and anaesthesiologists; decision support (data) department
staff; and quality and patient safety department staff. The advantage of this working
group was that much of the information and data was available through the members of
the group, plus the team had the authority to make decisions. Unfortunately a similar
working group was not available at the other two sites and therefore a full impact analysis
could not be performed at St. Michael’s and Mount Sinai Hospitals.
For each hospital, the first step was to create a process map of their surgical patient
flow. This was done through an intensive information collection endeavour including
interviews with hospital staff (front line staff, managers, surgeons, etc.); observation of
processes and patients through their stay; and analysis of patient level data. Process maps
were reviewed with and approved by key stakeholders for accuracy and completeness.
Chapter 2. Project Overview and Organisation 9
In addition, a needs/wants assessment was conducted with the key stakeholders at
each hospital. The purpose of this assessment was to determine what type of questions
they would like to “ask” the model, what measures are important in their decision making
and what type of decision options they would like to input into the model.
At this time, inquiry into electronic data sets at each hospital was made to determine
what data was stored. This included not only what fields but also the format of each
field and the accuracy of the data. This information was used to guide the input data
requirements of the simulation model.
The next step was to build a model of the patient flow at Juravinski as an initial
pilot to gain understanding of what is entailed in modelling perioperative patient flow
for tactical decisions. In order to contain the scope and complexity of the pilot model,
only orthopaedic surgical patient flow was considered. The pilot model was created based
on the process map and the results of the needs/wants assessment. The pilot model was
largely coded by the simulation modeller, Carolyn Busby, with guidance based on a draft
conceptual model, the process map and notes, and model requirements.
Based on the objectives and goals of the Juravinski working group, three sets of out-
come measures were chosen: throughput, cancellation rates and ward census. Due to the
MOHLTC wait times initiative that provided additional funding for specific procedures,
Juravinski was particularly interested in the throughput (volume) of total joint replace-
ments (TJR) and the number of urgent fractured hip repair cases completed. Therefore,
these two numbers were provided along with the overall throughput and by patient type
(elective, urgent, inpatient). A second chief concern at Juravinski was the number of
cancellations as a result of no available ward or ICU bed for the patient following their
surgery. To monitor this concern, cancellation rates were provided for the following five
reasons: no ward bed, no ICU bed, bumped due to more urgent case, not enough elective
time, and other reasons. Finally, Juravinski was concerned with the number of inpa-
tient beds required to achieve the throughput and cancellation rates desired. The model
Chapter 2. Project Overview and Organisation 10
provides the average daily midnight census in each ward and ICU to evaluate decision
options.
The orthopaedic pilot model at Juravinski was validated against historical perfor-
mance of throughput and cancellation rates. The validation work with support from
the modeller, Carolyn Busby. Moreover, the validation effort was done with significant
participation from Juravinski stakeholders and the working group in order to understand
and adjust for discrepancies between the model and historical results.
In collaboration with the working group, the resulting validated model was manipu-
lated to test and compare numerous decision options including various schedule changes
and changes to ward capacity. The resulting simulation model continued to be adjusted,
updated and used to influence decisions from when it was first validated in 2007 until
2010.
2.1.2 Development of the Proposed Generic Model
Based on the experience from building the pilot model at Juravinski (see section 5.3),
along with the results of the process mapping, needs/wants assessments, data inquiries
at all three sites, literature review of existing models, and perioperative patient flow
alternatives, an initial conceptual model of the proposed generic model was designed.
This conceptual model was reviewed with the doctoral supervisor, as well as with some
stakeholders at the three hospitals to ensure that it appeared to be sufficiently inclusive
of differing hospital and patient characteristics and that simplifications and assumptions
were reasonable. Details on the proposed generic model are provided in Chapter 4 and
Appendix C.
In 2009, a management consulting company working with William Osler Health Sys-
tem approached the research team for help as they faced many inefficient practices and
poor performance. It was proposed that their two hospitals with surgical sites join the
research study. Funding was provided to re-hire Carolyn Busby to code the generic model
Chapter 2. Project Overview and Organisation 11
in Simul8. While Carolyn coded the generic model, work with William Osler HS to un-
derstand and map out their processes, analyse the data and create the input files, and
build credibility and trust in the model that was being performed.
In 2010, the generic model was successfully applied to both William Osler HS hospi-
tals: the Brampton Civic Hospital and Etobicoke General Hospital. During the validation
stage, it became clear that the inefficient practices at William Osler HS were making val-
idation challenging. This led to the discovery of a secondary use for the generic model:
to demonstrate the level of performance that can be achieved if practices were improved.
Validation challenges and the use of the model as both a decision and a demonstrative
tool are outlined in Chapters 5 and 6.
2.1.3 Application to Multiple Sites
After application at the two William Osler HS hospitals, the generic model was applied
to the three initial hospitals.
At Juravinski, the working group continued to be involved as they were interested in
understanding interactions between surgical services. They also wanted to update the
model to incorporate changes to their processes and patient mix. Application of the
generic model was performed while working closely with the working group. In addition,
support was provided for testing, creating reports and providing other support. In early
2012, the working group at Juravinski successfully proposed to test one of the decision
options that according to results from the generic model, produced favourable outcomes.
As described in chapter 6, the pilot study proved successful and the changes are still
in effect. At the time of writing, an industrial engineer at HSS was in the process of
updating the data in the model to test new decision options for Juravinski, and also
working to apply the model to the General Hospital of Hamilton HS. They are being
supported by the Centre for Research in Healthcare Engineering (CRHE) and Visual8.
The generic model was also applied to St. Mike’s and Mt. Sinai. However, at these
Chapter 2. Project Overview and Organisation 12
sites, interest and continued support in the model had dwindled. As a result, the data
and information provided by the hospitals during the initial phase of the research project
was used as model inputs and validation parameters. Testing of decision options or
implementation of any proposed solutions did not occur at these sites.
2.1.4 Current and Future Application
In addition to the five hospitals that were included in the doctoral portion of this research,
the generic model has been applied by CRHE research members to an additional three
sites including Prince Albert Hospital, a small rural hospital in Saskatchewan consisting
of only four ORs.
Finally, Visual8 continues to be interested in creating a commercial version of the
generic model to allow for more widespread application. They are currently working
with CRHE (including the modeller and others who have used the model) and Hamilton
HS to further test and refine the model. The vision is to produce a model with a user
interface that allows for end users to interact more directly with the model to test decision
options.
2.2 Descriptions of the Hospitals Studied
At the time of publication, the generic model proposed herein has been applied to data
at eight hospitals in Canada. The hospitals studied vary widely in size, type, services
provided and population served. Of the eight hospitals studied, this thesis considers the
first five implementations. The other hospitals studied were completed by a researcher
within the CRHE team, with guidance. Table 2.1 provides the different hospital char-
acteristics of the first six hospitals studied. Further detail on each of the hospitals is
provided in the subsections below.
Chapter 2. Project Overview and Organisation 13
HospitalHospital
Type
Number of
Surgical
Services
Number
of ORs
Number of
Surgical
Inpatient
Beds
Hamilton HS Juravinski
HospitalAcademic 5 8 98
St. Michael’s Hospital Academic 11 22 223
Mount Sinai Hospital Academic 1* 10 112
William Osler HS
Brampton Civic HospitalCommunity 9 16 80
William Osler HS
Etobicoke General HospitalCommunity 7 7 46
Prince Albert Hospital Rural 7 4 30
*Only the General Surgery service at Mount Sinai was studied.
Table 2.1: Key characteristics of the hospitals studied.
Chapter 2. Project Overview and Organisation 14
2.2.1 Hamilton HS Juravinski Hospital
The Juravinski Hospital is one of three acute hospitals belonging to Hamilton Health
Sciences. Hamilton Health Sciences is an academic research centre affiliated with Mc-
Master University and serves more than 2.2 million people in the Hamilton and Central
South and Central West Ontario area. The Juravinski has a strong orthopaedic program
and is linked with the Juravinski Cancer Centre. The Juravinski is composed of 228
beds with about 9000 admissions and nearly 30,000 emergency department visits per
year (Hamilton Health Sciences, 2011).
The perioperative service is composed of eight operating rooms (ORs), 23 post-
anaesthesia care unit (PACU) beds, as well as 169 budgeted surgical postoperative re-
covery beds in various wards. The Juravinski has two critical care units for a total of 18
critical care beds: an intensive care unit (ICU) with 12 beds and cardiovascular care unit
(CCU) with 6 beds. In addition, Juravinski provides 23 mobile telemetry units available
for patients staying in ward beds who require additional monitoring. These monitoring
units are meant for patients who require some additional care, but at a lower observation
level than intensive and cardiovascular care patients require. This helps reduce the de-
mand on ICU and CCU beds within Juravinski while still allowing access to monitoring
units when required. The perioperative service is composed of seven surgical specialities
who share the resources.
2.2.2 St. Michael’s Hospital
St. Michael’s hospital is located in the downtown core of Toronto and is one of University
of Toronto’s teaching hospitals. St. Mike’s has over 480 adult inpatient beds, includ-
ing around 60 intensive care beds, which together discharge almost 25,000 patients a
year. Additionally, St. Mike’s emergency department sees over 700,000 ambulatory and
emergency cases a year, including more than 500 trauma cases (St. Michael’s Hospital,
Chapter 2. Project Overview and Organisation 15
2011).
In order to facilitate the over 30,000 surgical procedures each year, St. Mike’s runs
two sets of operating rooms. The core OR consists of sixteen operating suites and is
linked directly to 20 PACU beds and 13 same day surgery beds. The ambulatory OR
consists of six suites, nine PACU beds and 20 same day surgery beds. The ambulatory
OR is normally meant for patients who are to be discharged on the same day as their
procedure (i.e. Same Day Home patients), patients with shorter, less invasive procedures
and patients not requiring anaesthetics. Generally, a patient who is scheduled in one of
the ambulatory ORs will visit the ambulatory PACU. However, in high occupancy times,
patients may be placed in the first bed available, regardless of the OR unit. St. Mike’s
wards are not specifically for surgical patients, but rather programs, such as cardiology,
gynaecology, etc. In general, surgical patients will go to one of eight wards that total
223 beds. Surgical patients who require critical care are sent to one of four critical care
areas, depending on their surgical service and their need for critical care. There are 64
budgeted critical care beds across the four areas.
St. Mike’s uses a regional block room and allows for planned extended stay in the
PACU. The five regional block rooms are used for patients who have their anaesthesia
induced prior to entering the OR in a separate room. This decreases the time between
surgical procedures as the patient enters the room already induced, only needing to be
positioned before the procedure can begin. This affects the flow of these patients as they
are called directly to the regional block room from the waiting area, and proceed to the
OR when it is ready.
Some of St. Mike’s orthopaedic surgeons are scheduled into parallel OR rooms, where
the surgeon is provided two ORs and OR teams. This allows running two case lists, where
the surgeon is able to move between the two OR rooms to perform surgery and supervise
the residents and interns. Since the surgeon is provided two staffed ORs and the residents
can perform cases under his guidance, it allows a surgeon to effectively double his daily
Chapter 2. Project Overview and Organisation 16
patient throughput.
2.2.3 Mount Sinai Hospital
Mount Sinai Hospital is a University of Toronto affiliated hospitals, also located in down-
town Toronto. Mt. Sinai has 472 beds that serve over 25,000 admissions a year. Mt.
Sinai also runs an ambulatory and emergency service, that sees more than 700,000 visits
per year (Mount Sinai Hospital, 2011). Within the surgical service, Mt. Sinai performs
more than 19,000 operations per year in ten operating rooms. However, for the purpose
of this study, Mt. Sinai asked to only concentrate on their general surgery patient popu-
lation, thus, the focus is on the five operating rooms assigned to general surgery and the
downstream inpatient bed resources. General surgery patients have access to a pool of
ICU beds, SDU beds, PACU and PACU holding beds and day surgery unit beds that are
shared with the other surgical and medical services. There is also an inpatient general
surgical unit of 15 beds.
2.2.4 William Osler HS Brampton Civic and Etobicoke General
Hospitals
William Osler Health System services part of the Greater Toronto Area (GTA) and
surrounding areas and is one of the largest hospital organisations in Ontario. William
Osler HS is composed of three community hospitals: Brampton Civic Hospital, Etobicoke
General Hospital and Peel Memorial Hospital. Surgical services are available at both the
Brampton and Etobicoke sites.
Brampton opened in 2007 to meet the increasing demand in the fast growing city of
Brampton (William Osler Health System, 2012). The perioperative service at Brampton
is composed of 22 ORs, 13 of which were staffed and running during the study. The
perioperative service includes four inpatient areas: the general surgery ward composed
Chapter 2. Project Overview and Organisation 17
of 28 beds, a surgical observation unit of four beds, 16 short stay unit beds and a 32 bed
orthopaedic surgical ward. For patients requiring critical care, the service has access to
the ICU that consists of 24 beds.
Etobicoke is an older hospital that is currently undergoing extensive renovations
(William Osler Health System, 2012). At Etobicoke, there are seven ORs currently
in operation, one of which is a highly specialised cystology room. The perioperative ser-
vice at Etobicoke operates two inpatient units, a general surgery ward of 26 beds and an
orthopaedic surgical ward of 20 beds. Patients also have access to an ICU and a Cardiac
Care Unit (CCU) of 12 and 6 beds, respectively.
Chapter 3
Background and Literature Review
Before going into detail into the contributions of this dissertation, this chapter serves to
provide a brief description of perioperative processes and decision making, and a review
of the pertinent research related to this work.
The purpose of this literature review is two-fold. First, it aims to give an overview
of recent work in modelling patient flow through perioperative processes for the purpose
of aiding tactical and strategic level decisions. Second, a review of generalised models
and frameworks is provided, focusing on the methodology of designing, validating, and
implementing such models both within and outside of healthcare applications.
3.1 The Perioperative Process
As mentioned previously, the perioperative process can be divided into three different
sets of activity, as described in figure 3.1. Each of these will be discussed in the sections
that follow after an introduction to the typical patient types within perioperative flow.
18
Chapter 3. Background and Literature Review 19
!"#$%&'"()&
*$+#&,-+$%&
.#$+)$(/0&
1$2-+3#$(3%&
4-+5%&
,+67/-8&,-+$%&
$3/9&
1$/6:6"(&3"&
"2$+-3$&
;+$<
=2$+-7>$&
;":3<
=2$+-7>$&
?$/">$+0&
@##$56-3$&
;":3<
=2$+-7>$&
=2$+-7>$&
16:/A-+)$&3"&
A"#$%&8"()&
3$+#&/-+$%&
+$A-B6863-7"(%&
$3/9&
;$+6"2$+-7>$&C3-)$&
Figure 3.1: The three stages of perioperative activity.
3.1.1 Patient Types
Typically, there are three different types of patients within perioperative flow: elective,
urgent/emergent and inpatient.
Elective patients are scheduled for surgery during regular operating room hours.
These patients generally first visit a surgeon at a clinic, where the surgeon determines
that they require surgery. Elective patients remain at home prior to their procedure and
arrive at the hospital a few hours prior to their scheduled procedure start time. They
are also often referred to as same day patients, as they arrive at the hospital the same
day as their procedure.
The second group of patients is emergent and urgent patients. These patients often
present at the emergency department and require surgery within a few hours to a few
days. These patients may require a procedure on a life-or-death basis, while others are
stable enough to wait a day or two. These patients are often admitted to the hospital
prior to their procedure. Their level of urgency determines how long they can wait
before their procedure must begin. Urgent/emergent patients procedures are typically
performed by a surgeon who is working that day or who is on call. For the remainder
of this document, urgent and emergent patients will be collectively referred to as urgent
patients.
Finally, inpatients are those patients who have already been admitted to the hospital
Chapter 3. Background and Literature Review 20
and require surgery, but are not considered urgent. These patients often have been
admitted for some other reason and now require surgery to progress their care. Some
hospitals treat inpatients as part of the urgent population, in their least urgent level, as
they can wait up to a week for their procedure. Other hospitals treat them more like an
elective patient who must be done within a week or so, but should be scheduled within
elective time whenever possible. Regardless of how the patients are scheduled, inpatients
often have a surgeon assigned prior to scheduling, thus cannot simply be grouped with
the urgent patients to be performed by the next available surgeon.
3.1.2 Pre-Operative Activities
Pre-operative patient flow begins when the decision to treat surgically has been made by
the surgeon and patient. This phase of the perioperative service involves all activities up
until the surgical day, including wait list management, scheduling the patient for surgery,
scheduling of and attendance at any pre-operative clinic visits, diagnostic tests, patient
education, and any other preparation required for the surgical procedure or post-surgical
recovery and eventual discharge.
For elective patients, this stage may last several months waiting for their scheduled
date. For urgent patients and inpatients, this stage will last between a few hours to a
week or so, depending on urgency, and the scheduling policies of the hospital.
3.1.3 Operative Activities
The day of surgery includes a number of activities, including registration, patient prepa-
ration, patient positioning, anaesthesia, the procedure itself, and post-surgical recovery.
Co-ordination of the activities on the day of surgery is important to patient flow.
Elective patients arrive on the day of their surgery, generally two or more hours ahead
of their scheduled procedure time to ensure that all pre-surgical activities are complete.
Urgent and inpatients are already registered and in the hospital. They most often stay in
Chapter 3. Background and Literature Review 21
their current hospital location until near the time of their procedure. Delays in patient
flow due to activities done before the patient is in the OR are generally due to poor
co-ordination and/or communication. These problems are generally operational, and can
be reduced through improved co-ordination and communication of staff, equipment and
patients.
Conversely, processes that occur post-procedure can often impede patient flow. For
example, OR delays can occur when the PACU or ICU is full, blocking the OR with a
patient waiting to get into the PACU, ICU or another bed. As mentioned in Chapter 1,
one of the most common reasons for cancellations is due to no bed. Resource capacity
decisions and co-ordination of resources through scheduling policies are made through
tactical level planning.
3.1.4 Post-Operative Activities
In the true sense of perioperative service, post-operative activities only include recovery
immediately following surgery. Typically, this includes the post-anaesthesia care unit
(PACU), the same day surgical area, and perhaps the intensive or cardiac care units
(ICU and CCU).
One of the main bottlenecks to perioperative patient flow however, includes non-
immediate recovery activity, such as step-down units, wards and critical care. To address
this, some hospitals have identified the importance of surgical inpatient units to peri-
operative flow, and have reflected this through their organisational chart and program
structures.
3.2 Perioperative Decision Making
There are three types of decisions that organisations make: strategic, tactical and op-
erational. These three types vary in terms of the types of decisions that are made, and
Chapter 3. Background and Literature Review 22
the frequency of which they are made, and the time horizon of the decision. Naturally,
within perioperative decision making, these same three types apply. This section will
review the types of decisions that are made at each type in terms of the perioperative
processes of a hospital.
3.2.1 Strategic Perioperative Decisions
Strategic decisions are long term decisions, typically made once every few years but
no more than once a year. Strategic decisions have great effect on the perioperative
service’s direction and function as they determine the capacity of resources and the
types of care provided (Wachtel and Dexter, 2008). Perioperative strategic decisions are
“mainly a resource allocation problem” (Guerriero and Guido, 2011, pg. 94) as they
involve decisions regarding the number and types of surgery provided (case mix) or the
amount of resources available. Resource decisions can include procurement or building
of new capacity or equipment, such as MRI, new operating rooms or inpatient beds, or
the availability of existing resources, such as the number of staffed elective OR hours.
3.2.2 Tactical Perioperative Decisions
Tactical decisions are made a few times per year to address “the organisation of the
operations/execution of the health care delivery process (i.e. the ‘what, where, how,
when and who’)” (Hans et al., 2012, pg. 310)’. Tactical decisions include calibrating
OR capacity through the Master Surgical Schedule (MSS) by adjusting the hours/blocks
available for elective cases and the assignment of OR capacity to patient types, services,
and surgeons. Tactical decisions also consider resource capacity adjustments of inpatient
units and ancillary services such as medical imaging and lab (Blake and Carter, 1997;
Hans et al., 2012; Wachtel and Dexter, 2008; VanBerkel et al., 2011b). Tactical capacity
decisions typically involve determining if the resource will be staffed, whereas strategic
capacity decisions determine how many beds/machines/ORs are physically available. In
Chapter 3. Background and Literature Review 23
other words, deciding to build a new OR suite or ward is a strategic decision, whereas
determining when and how much that resource will be staffed, and thus available to be
used, is a tactical decision.
3.2.3 Operational Perioperative Decisions
Finally, operational decisions are made frequently and directly affect the day-to-day
operation of the perioperative service. Operational decisions can be further broken down
into online and offline decisions (Hans et al., 2012, pg. 310).Offline decisions are made in
advance and involve the co-ordination of activities. Scheduling and sequencing of elective
and other planned cases are the main activities of offline decision making. On the other
hand, online decisions are reactive in nature, and involve “monitoring the process and
reacting to unforeseen or unanticipated events”. Examples of online decisions include
scheduling emergency surgeries, cancellation decisions, and use of overtime to complete
cases.
3.3 Review of Models on Perioperative Planning and
Scheduling
The focus of this section of the literature review is on surgical planning and schedul-
ing models within the operational research field. Specifically, surgical planning refers
to matching supply of resources such as operating room time, inpatient beds, equip-
ment, diagnostic imaging tests, etc., with demand. Whereas surgical scheduling seeks
to determine the time and sequence of cases in the operating room (Cardoen et al.,
2010; Guerriero and Guido, 2011). As a result, this review will exclude work on process
improvement, the impact of new medical technologies, estimation of surgical procedure
duration, staff scheduling and facility design. Furthermore, based on the scope of the
research, the review will focus on addressing tactical decisions. If the reader is interested
Chapter 3. Background and Literature Review 24
in strategic and operational decision models, a number of literature reviews have been
published recently (e.g. Guerriero and Guido, 2011; Cardoen et al., 2010)
A significant amount of interest, especially in the past ten years, has been focused on
addressing perioperative tactical planning and scheduling decisions (Guerriero and Guido,
2011). Within the operational research literature, many papers were found that look
solely at the ORs and the effect of changes to the MSS on OR efficiency measures such as
utilisation, over time, or the number of ORs required. Many of these use mathematical
programming to try to find an optimal MSS based on the objective and constraints
considered (e.g. Blake and Donald, 2002; Kharraja et al., 2006; Kuo et al., 2003; Testi
and Tanfani, 2008; Belien and Demeulemeester, 2008; Blake et al., 1995). The effect of
case sequencing guidelines on OR performance has also been considered by some authors
(e.g. Dexter and Traub, 2002; Lebowitz, 2003). These models are interesting in terms of
their choice of variables, objectives, constraints, assumptions, etc., however, they are not
explored further within this literature review as they do not directly consider resources
beyond the operating rooms themselves, which a key goal of this research.
While it may be easier and less complex to limit one’s study to the surgical depart-
ment, or specifically the operating rooms, conclusions may be suboptimal since effects
on other areas are not accounted for (VanBerkel et al., 2010). Studying up and down-
stream resources, in other words a holistic approach, is essential to understanding and
improving surgical patient flow (VanBerkel et al., 2011b; Sobolev et al., 2008). Opera-
tional research papers that consider a holistic view of surgical planning vary in breadth of
inclusion. Some consider a limited breadth by only including immediate downstream re-
sources such as the post-anaesthesia care unit (PACU). Meanwhile, there are some papers
that consider a broader set of resources by not only including immediate post-surgical
recovery but also inpatient resources such as ward and ICU beds.
As noted in a recent review by Guerriero and Guido (2011), the aim of most tactical
decision models is to produce a MSS. Further, various studies note that “an efficient MSS
Chapter 3. Background and Literature Review 25
is obtained when not only demand of [surgical specialities], equipment restrictions, staff
(surgeon-nurse) availability, surgeons preferences but also other resources, such as ICU
and ward capacity are considered.” (Guerriero and Guido, 2011, p.109)’
Many of the perioperative tactical decision models proposed in the literature use
combinatorial optimisation models (i.e. Integer Programming (IP), Mixed Integer Pro-
gramming (MIP), Integer Quadratic Programming (IQP)) to find an optimal or near-
optimal solution (Guerriero and Guido, 2011). The activities, constraints, objectives and
resources included in these models vary widely.
Some models found do not directly include resources outside of the OR, for example
(Zhang et al., 2009) presented a MIP model to set the MSS, taking into consideration the
cost of an inpatient’s length of stay (LOS) by minimising the inpatient’s LOS prior to
surgery, i.e. the patient’s waiting time. Belien and Demeulemeester (2008) proposes an
IP to set the MSS and the inpatient nurse schedule by taking into account the workload
of various surgery types and the requirements from the nurses’ collective agreement.
As noted by Cardoen et al. (2010) and Guerriero and Guido (2011), most models
found in the literature are deterministic, and do not take into account the uncertainty and
variability in demand, LOS, arrival time, no show rate, etc.. For example, three different
models are proposed in Gupta (2007) for perioperative decision making based on three
common problems: allocating time to surgical specialities, managing elective bookings
and sequencing cases. Gupta (2007) proposes to manage elective bookings utilising a
model to determine the amount of downstream resources to reserve for future high-
priority cases, taking into account urgent patient demand and maximum wait allowable
for elective patients before requiring use of overtime in the ward. None of the three
models were applied using real or generated data to test for solvability and usability.
Due to the high cost of hospital resources, such as inpatient beds and the ORs,
utilisation is often considered an important measure to hospital decision makers. Vissers
et al. (2005) formulated a Mixed Integer Linear Programming (MILP) model to determine
Chapter 3. Background and Literature Review 26
the MSS for the cardiothorasic speciality in terms of the number of patients and case
mix to allocate during each OR block. The model takes into consideration the amount
of OR time available, the capacity of the intensive and intermediate care units and the
staffing levels in the ICU. The objective of the model is to minimise the maximum over
and under utilisation of the four resources considered.
Levelling the demand for beds over the days of the week is often considered as an
objective when adjusting the MSS. Belien et al. (2009) proposed a multi-objective MIP
model that assigns the blocks to surgical specialities while minimising the weighted sum
of peaks and variance of bed occupancy of the wards. The goal of this study was to
build an MSS that assigns blocks in a simple and repetitive fashion, assigning OR blocks
to a single surgical service (no sharing a block) and to level the demand for ward beds.
The MIP was applied using real data of a Belgian hospital. The real power of this work
however is not the MIP model proposed, but rather a GUI created to visualise various
proposed MSS and their effect on bed occupancy levels. Belien et al. (2009) propose
that there is no optimal solution as no mathematical program can account for all the
complexities and restrictions of reality, and that the value lies in a decision support tool
that helps compare various MSS options.
The last interesting deterministic model found in the literature is a MIP developed
to consider the system wide trade-offs between OR availability, bed capacity, surgical
booking privileges and wait lists (Santibanez et al., 2007). This model purports to set
the MSS for twelve hospitals within a health authority. The model considers which
surgical speciality can be accommodated in the ORs available, the number of blocks
to assign to each speciality, throughput targets and the capacity of the wards and step
down units. The objective of the model can be to either maximise throughput or level
bed occupancy. This model is also considers the demand for resources of emergency cases
when setting the MSS.
The stochastic models found vary not only in terms of the objective, constraints and
Chapter 3. Background and Literature Review 27
activities considered, but also in what stochastic elements are included. Most models
considered the stochasticity of the inpatient length of stay (Belien and Demeulemeester,
2007; Adan et al., 2009; Calichman, 2005), while one model found considered the stochas-
ticity of the procedure length (van Oostrum et al., 2008b).
Clearly, incorporating stochasticity in the model should improve performance and
provide robust solutions, especially considering the uncertainty in healthcare. When
compared to a deterministic model proposed by Vissers et al. (2005), Adan et al. (2009)
found that incorporating variation into the decision model resulted in better resource
utilisation.
In his paper, Calichman (2005) presents a set of guidelines and sample spreadsheets
for a simple MSS optimisation model that can be adjusted based on the needs and
goals of the hospital . The basic model structure presented includes inpatient LOS
probabilities. Little detail is provided on the formulation behind the model’s spreadsheets
as the purpose is to simply present general guidelines for hospital consultation.
A MIP model and metaheuristic is presented by Belien and Demeulemeester (2007)
that includes stochastic inpatient length of stay with the aim of levelling inpatient oc-
cupancy. The objective of the proposed model is to produce a MSS that minimises the
total expected bed shortage. This model consists of only two sets of constraints: the
surgical service or surgeon block requirements and the total number of blocks available.
Alternatively, accounting for the variability of the procedure length allows for planned
slack in the OR schedule, reducing the need for costly overtime. A two phase mathemat-
ical model using a IP and a MIP to produce a cyclical MSS that levels ICU and ward
workload and minimises the OR capacity required is proposed by van Oostrum et al.
(2008b). This model allocated OR time to the identified elective patient types such that
there is a planned amount of slack in each OR day to account for the uncertainty in case
duration. The objective is to minimise both the amount of OR time required and the
maximum demand of bed resources. Moreover, the model is presented as a case study
Chapter 3. Background and Literature Review 28
at a Dutch hospital that considered levelling the demand for ICU beds (Houdenhoven
et al., 2008).
Another consideration is the welfare or the societal implications of setting a MSS. Testi
and Tanfani (2008), Tanfani and Testi (2010) and Testi et al. (2007) present mathemati-
cal programming models that aim to improve the welfare of the patient by considering the
effect on the patient’s waiting time and access to their surgical procedure. The models
presented propose both the assignment of surgical specialities to blocks of OR time, and
scheduled specific patients to the time, resulting in a tactical and operational decision
making model. The assumption of these models is that an open scheduling method is
employed that allows for flexibility in the speciality assignment each week as based on
realised elective case demand.
Simulation is another popular method and is typically proposed as a decision aid to
compare multiple decision options on a number of different performance indicators. This
method allows for the decision making to remain in the hand of the decision maker who
must weigh the results from the different options and make the final decision. As with
the mathematical programming models proposed, the objectives and considerations in
perioperative simulation models vary widely.
In both simulation and mathematical programming models found, most consider re-
sources downstream from the OR and use statistical distributions to represent upstream
effects (VanBerkel et al., 2010). Only one model was found that included activities that
occur prior to the day of surgery. Vasilakis et al. (2007) studied the effect of scheduling
methods at the surgical clinic on the number of patients waiting for an appointment and
the patients’ waiting time to their first appointment and to their scheduled procedure.
The simulation model included elective, urgent and emergent patient flow for the cardiac
surgical speciality. The model considers a high level view of the operations as it does not
consider detailed processes at the clinic or OR, but rather considers only the schedule
and capacity of the two resources.
Chapter 3. Background and Literature Review 29
A number of the simulation models found focused on tactical resource capacity deci-
sions, such as the number of staffed inpatient beds. For example, several Monte Carlo
simulation models developed studied the relationship between the operating rooms and
recovery rooms. The models were used to determine the number of ORs and recovery
rooms were required to meet increased volumes (Kuzdrall et al., 1981), and how admis-
sion policies and case sequencing could affect volumes and utilisation measures (Kuzdrall
et al., 1981; Kwak et al., 1976).
Another resource decision option considered is the use of a quota system to control
the occupancy of resources by surgical patients. In Kim and Horowitz (2002), a Discrete
Event Simulation (DES) model was built to examine the use of a quota system on the
performance of the ICU. The model also considers the effect of employing a scheduling
window where patients must receive surgery within a specified time from when initially
referred. To evaluate ICU performance the simulation model considers a number of
different outputs including the utilisation, time patients spend in the system, cancellation
rate and volumes. This model was applied and discussed in more detail in a case study
(Kim et al., 2000).
Surgical case sequencing and configuration options for the pre- and post-recovery
areas of an ambulatory surgical centre were evaluated using simulation (Ramis et al.,
2001). The goal of the study was to determine the best process and layout to maximise
throughput of the centre with a fixed number of total beds. Similarly, Marcon and Dexter
(2006) used a DES model to evaluate the effect of sequencing cases on OR utilisation
and overtime and PACU staffing (represented by the number of patients in the PACU
per hour).
In addition to including the PACU, Marcon et al. (2003) and McAleer et al. (1995)
include the porters as a resource in their simulation models. The question posed to
the model by McAleer et al. (1995) was whether surgical throughput could be increased
without additional recovery room (PACU) beds. The model found portering was not
Chapter 3. Background and Literature Review 30
in fact a bottleneck of the system, and instead improvements in throughput could be
achieved through better scheduling practices and process improvement initiatives. On
the other hand, the purpose of the simulation model by Marcon et al. (2003) was to
determine the number of PACU beds and porters required to achieve desired OR and
PACU performance.
A final model found that studied resource capacity and sequencing decisions was
the application of a general hospital simulation framework, PROMPT, to the general
surgery service in the UK (Harper, 2002). The simulation model includes the ORs,
recovery area and inpatient units. The author was interested in the OR and bed capacities
required, cancellation rates and the forecasted patient mix (i.e. type and number of cases
and inpatient versus day patient). Decision options studied included case sequencing
and scheduling rules such as performing admitted cases early in the week. The general
framework applied allows for a high level simulation model where patients are classified
into a number of groups based on characteristics including procedure length, inpatient
length of stay, route and resource requirements.
In addition to resource and sequencing decisions, some simulation models study the
effect of the MSS on performance. One work found compared the OR utilisation, through-
put and number of overruns of the current MSS and a MSS suggested from a previous
study using an IP model (Sciomachen et al., 2005). The simulation model was also used
to study the effect of introducing pre- and post-recovery room areas and case sequencing
in the OR. The decision options were evaluated based on volume, wait list size, number
of delayed cases, the number of OR overruns and OR utilisation.
Alternatively, Lowery (1992) was concerned with both the ORs and the critical care
units of a hospital. The simulation model considered the ORs, PACU, and seven different
types of critical care units. The primary question of this study was to determine the
number of critical care beds required. Since the surgical service has a high demand for
these beds, patient flow from the service was a key input. The model included both
Chapter 3. Background and Literature Review 31
elective and urgent/emergent patient flows from the surgical service as well as other
sources. The surgical elective patient flow was modelled as a function of the MSS and
historical scheduling patterns. The model logic was coded such that when no beds were
available, off servicing patients to another critical care area or discharging patients from
their current level of care early were considered in order to place the patient. Scenarios
of changes to the MSS, other patient flow patterns and number of beds in the units
were evaluated based on unit utilisation, the number of emergent patient turn aways,
the number of patients bumped from the units and the number of patients off serviced.
Unfortunately though the model was useful to hospital decision makers in terms of gaining
understanding of the trade-off between unit occupancy and turn away rates, a decision
on the number of critical care beds based on simulation results had not been made at
the time of publication.
As part of a health district’s cost reduction strategy, the number of beds available
for general and urology surgical services was to be reduced (Wright, 1987). A simulation
model was employed to determine how the reduction in beds and other changes would
effect performance at the hospitals and the district overall. The chief concerns were the
average, minimum and maximum bed occupancy by day of week, the number of admitted
patients per bed, the number of patients off serviced to other units and the number of
rejected cases due to no beds. A number of scenarios were considered in the study
including various reductions in the number of beds within the district, improvement in
LOS, smoothing admissions by performing elective cases on the weekend, changes to the
patient mix and changes to the emergency arrival rate. The model was used by the health
district to help inform decisions on how to implement their bed reduction initiative. The
district plans to continue using the model for future decision making.
Four models were found that included multiple downstream resources and the MSS
within their scope, each using a different solution methodology. First, a discrete event
simulation model was built to evaluate changes to the MSS, surgeon case mix, number
Chapter 3. Background and Literature Review 32
of inpatient beds, and number of nurses scheduled (Blake et al., 1995; Carter and Blake,
2005). The simulation model tracked patients from when they entered the waiting list,
were scheduled for surgery, the procedure itself and recovery afterwards. It also included
a nursing workload model in order to estimate the nurse staffing requirements. The model
included logic to cancel patients if not enough beds or nursing staff were available to care
for them post operatively. In order to make portability and validation easier, the authors
chose to generate patient characteristics such as LOS and route based on a sample of real
patients instead of attempting to fit statistical distributions. During validation of the
model at the sites, it was found that the model helped identify some inefficient practices.
For instance, at one site it was found that the ENT service regularly did not use all the
time assigned to them in the MSS; time which general surgery would often book. This
resulted in the model producing higher than historical throughput values for ENT, and
lower volume of general surgery cases. The model helped not only uncover this practice,
but also quantify the effect on system performance. Finally, the intention of the authors
was to design a generic model that could be applied easily to all four sites under study.
However, they found that the differences in process and model needs between the sites
required significant changes and adjustments to the model. As a result, building a generic
model was not possible in this case.
A software program was developed in Belien et al. (2006) to visually demonstrate
the impact of changes to the MSS on resource utilisation. The deterministic tool was
intended for hospital decision makers to understand the effect of their tactical decisions
on the utilisation patterns of various resources. A user friendly GUI was built that allows
easy changes to the MSS by dragging and dropping OR blocks into the schedule. Users
can view the resulting resource consumption patterns by surgeon, surgical group, day of
week, and OR block. The software was designed with feedback from one of the hospital’s
decision makers, however, has not yet been used on site to help inform decisions.
Another article presented an analytical model of the ORs and inpatient units based
Chapter 3. Background and Literature Review 33
on queueing theory concepts (VanBerkel et al., 2011b). The authors chose an analytical
model as they felt that it could be more precise and be developed faster than a simulation
or other type of model. The arrival rate is based on the inputted MSS and the scheduling
trends of the surgical specialities. The inpatient wards are modelled as infinite servers,
i.e. they can accommodate any number of patients. One of the key objectives of this
model was to understand the effect of changes to the MSS on nursing workload. As
a result, the outputs of the model were chosen in order to predict workload. These
outputs included the number of patients in the system per day in the MSS schedule,
daily occupancy levels in the wards, the daily number of admissions and discharges and
the number of patients in each post-operative recovery day. Working with the decision
makers at the hospital, a new MSS was derived based on the outputs of the model and
the resulting MSS was implemented. In VanBerkel et al. (2011a), the authors compare
the results of the analytical model to 33 weeks of data collected after the new schedule
was enacted. This is one of the only models found that not only reported that a decision
was implemented based on results from the model, but also evaluated the model results
compared to results from implementing the solution.
The final model found considering multiple inpatient units was a Monte Carlo simu-
lation and MIP model to examine the relationship between the MSS and inpatient bed
requirements (Saremi et al., 2013). The purpose of the simulation model is to predict
the bed requirements in the various inpatient units, including ward, step down units and
critical care units. Since the purpose is to determine the demand for beds the simulation
model does not restrain the capacity of the units. Similar to the model engineered by
(Blake et al., 1995), the simulation model randomly selects patients from a patient file of
historical patient records, instead of using a set of statistical distributions to determine
the patient’s characteristics. On the other hand, the purpose of the MIP is to build
a MSS in terms of assigning blocks to surgeons and patient types with the objective
to reduce the peak occupancy levels of the inpatient units. The intention is that the
Chapter 3. Background and Literature Review 34
MIP model is used initially to suggest an optimal schedule, then the simulation model
is used iteratively to test both the MIP-suggested schedule as well as other proposed
schedules to compare performance and ultimately select a schedule to implement. The
two models have been applied to a number of hospitals within a health region. As part of
the application, it was determined that since hospital decision makers are not experts in
mathematical programming, a set of guidelines be presented in place of running the MIP.
These guidelines help identify possible MSS to test in the simulation model. The paper
reports that the two models were successful in demonstrating to decision makers the
effect of MSS decisions on inpatient unit utilisation. A final note, the aim of the model
design was to be transparent and portable which lead to a simple design of both models.
However, during application to the site described in the case study, a few changes to the
models were required.
3.3.1 Summary
In conclusion, a search of the operational research literature found that though there
exists significant demand and focus on improving surgical planning and scheduling deci-
sions, much of the focus has been limited in the breadth of perioperative flow considered.
In other words, very few works consider a holistic approach to surgical decision making
by including perioperative and other hospital resources beyond the OR. In addition, little
attention has been given to creating generalised models that can be applied easily and
cost-effectively to many sites. Finally, only a handful of solutions proposed have been
successfully implemented at the sites under study, indicating the need to ensure solutions
produced are considered credible and useful by hospital decision makers.
One of the challenges faced in healthcare operational research is that many of the
recommendations are not implemented into the decision making and processes of the
hospital. For example, VanBerkel et al. (2011a) found that many studies on improved
MSS were never actually implemented. A key of strategy that has shown to encourage
Chapter 3. Background and Literature Review 35
more implementation in future studies is frequent and meaningful involvement of the
stakeholders and decision makers. Unfortunately, many of the research studies do not
closely involve them. In a review of simulation models of surgical patient flow, Sobolev
et al. (2011) noted that out of the 34 papers reviewed, only 19 were made to meet the
needs of decision makers, and only 9 of those 19 reported that the decision makers were
actually involved in the study. (van Oostrum et al., 2008a, p. 2) also made note of this
phenomenon:
“the influence of various stakeholders, with varying degrees of autonomy,
is often substantial. In the case of OR scheduling, surgical services (with
surgeons, surgery and anaesthesia assistants, and anaesthesiologists), all have
a considerable influence on OR management. Intelligent OR planning and
scheduling approaches proposed in the literature often fail to account for
this, which explains their relatively marginal impact in practice and the small
number of successful implementations in the literature.”
Implementation of a model is futile if the people who they are intended to help are not
considered or involved in the creation of the tool.
3.4 Review of Existing Scheduling Tools and Soft-
ware
A survey of existing commercial products and literature did not produce any commer-
cially available tools that aid in tactical or strategic operating room scheduling decisions.
The search for tactical planning and scheduling decision tools, as described above, has
demonstrated the need for further study into these decisions. There is a noted need
for the development of commercially available products to help inform hospital admin-
istration of policies and methods to determine strategic and tactical policy decisions of
Chapter 3. Background and Literature Review 36
OR scheduling that work towards achieving the overall strategic goals and objectives of
the organisation. In the era of an ageing population and constrained healthcare bud-
gets, improved planning and coordination of operating room scheduling and resources
in conjunction with alignment to the strategic goals of the organisation is imperative to
the financial and operational health of an organisation and of the healthcare system in
general.
3.5 Review of Generalised Models and Frameworks
The theme of this section is two-fold; firstly to briefly describe the work to date in the
field of general simulation models. Secondly, to demonstrate that though many models
exist that claim genericity, there is little done in the way of proposing methods around the
process itself, abstraction decisions and model verification and validation. The discussion
will start with defining generic modelling.
3.5.1 Defining a Generalised Model
A generic simulation model can be broadly defined as a model that is “designed to apply
to a range of systems which have structural similarities (Pidd, 1992, p. 238)”. Lowery
(1998) states that generic models should be “general, flexible, intuitive and simple, and
include default values for system parameters (qtd. in Fletcher and Worthington, 2009,
p. 376)”. Related to generic models is model or code reuse. Reuse can vary from small
portions of code, to entire models (Robinson et al., 2004).
Sinreich and Marmor (2004) use three levels to classify models. The most generic
type, generic activities, is flexible enough for any system and scenario, and requires
the highest level of abstraction. At the other end, models requiring the lowest level of
abstraction are referred to as fixed process. These models can only be used to analyse
the system it was designed for. In the middle are generic process models, which can be
Chapter 3. Background and Literature Review 37
used to model and analyse any system that uses similar processes.
More recently, Fletcher and Worthington (2009) propose a spectrum of genericity
based on previous works in classification of generic models, survey results from experts
and a literature review of generic models. While their research focused largely on simula-
tion within healthcare, the classification system of generic modelling can easily be applied
to simulation modelling as a whole. The authors propose four levels of genericity:
1. A broad generic principle model - This level describes models that are not setting or
industry specific, and are often used to demonstrate general principles. An example
of a level one model is a generalised theoretical queuing model that demonstrates
that as server utilisation increases above 80%, wait times will greatly increase as
well.
2. A generic framework - Level two describes models that have grouped together com-
mon issues within a toolkit. These types of models can be used in combination with
local data and knowledge to build understanding of system issues. The model can
also be used as a base to create a locally specific model, or to demonstrate general
principles similar to level one models. For example, the tool kit can contain a model
of a generic OR that the user can customise with hospital data and information.
The resulting model can be used to demonstrate the effect of OR utilisation on the
number of inpatient beds required.
3. A setting-specific generic model - These are intended to provide general insights
into the issues faced in delivering specific services, such as surgical care. Input data
can be adjusted such that these models can be used multiple times, by multiple
providers of the same service, without changing the model structure. An example
would be a model of a typical emergency department. Users would input data
specific to their hospital, such as the number of physicians, beds, etc. The model
can then be used to gain insights into the processes of the site under study.
Chapter 3. Background and Literature Review 38
4. A setting-specific specific model - Level four models largely differ from level three
models by the objective of transportability. Level four models are created to repre-
sent the particular system under study, with no intent to use it for another system
or provider.
The intention of this research is to design and test a level three generic model of
perioperative patient flow, processes and decisions. For the remainder of this review,
attention will be given to generic models within level three, specifically those intended
for use at a local site level (as opposed for central use by health planning authorities for
example).
3.5.2 Generalised Models and Frameworks within Healthcare
The purpose of this review is first to understand how this research adds to current
literature on generic modelling in healthcare, specifically in perioperative tactical decision
making. Secondly, to gather techniques on generic model design, implementation and
validation.
Generalised Models Focusing on Perioperative Flow
During the review, only one paper was found that set out purposely to create a general
surgical model in collaboration with multiple hospitals. Blake et al. (1995) set out to
create a simulation based decision support tool that modelled the effects on inpatient
bed usage and nursing resources of changes in the operating room block schedule. Based
on their needs, they chose to use the model from Wright (1987) as a base, and add func-
tionality as required. The resulting model followed surgical patient flow from admission
to discharge. The model made use of a GUI that linked a database and a scheduling
model into an animated simulation model. The database populates the model with ac-
tual patient discharge records, the OR schedule and hospital resources levels, including
Chapter 3. Background and Literature Review 39
beds and available nursing staff on each of the surgical wards modelled. The data can
be altered to represent alternative designs of hospitals studied and of possible solutions.
The scheduling model randomly schedules patients into the block schedule and pa-
tients flow through the various departments and areas based on the flow rules and patient
characteristics.
Outputs of this model include admission rates, bed utilisation and staff workload. The
model was validated at the pilot hospital based on the historical admission rate and case
load, by service. The authors note that validation efforts of the model at each site faced
challenges typical to validation of complex, real world systems. Through involvement
with hospital decision makers and thorough analysis of the model, data and processes,
it was determined that the outputted results were credible and realistic. The model was
used at all four sites to propose improved surgical block schedules. Unfortunately, there
was significant resistance from some of the users. As a result, none of the results were
implemented.
The focus of this model was for strategic decisions around surgical scheduling and
resource levels. The paper does not provide any insight into the level of detail modelled
in terms of scheduling and processes. However, based on the information given and its
strategic intention, the model appears to use a simplistic scheduling method and high
level process mapping. Also of note, the model’s structure was used as a base that was
customised to each site, thus was not entirely generic.
Another model found that purported to be general was developed by Saremi et al.
(2013) who presented three simulation-optimisation formulations. While the author’s
intention was to develop a generic model, the paper did not mention any significant
collaboration with hospital decision makers during design. Furthermore, the paper only
mentioned application to a single hospital site.
Chapter 3. Background and Literature Review 40
Generalised Models Focusing on Perioperative Flow or Hospital-wide
Two generic models were found that encompass multiple areas of a hospital. Harper
(2002) created a generic framework to model hospital resources including beds, operating
rooms and workforce needs. This level two model used patient groupings to describe the
different patient groups with similar characteristics and care paths; each with its own
distribution parameters to describe patient characteristics such as length of stay.
The paper claims that any healthcare system, or sub-system can be modelled, from
ORs and inpatient beds, to the complex care pathways of critical care patients, by in-
putting the flow diagram. The resulting model collects resource usage statistics at the
resource and patient level for output statistics and graphs. The paper gives the example
of modelling a high level strategic model of the operating rooms to evaluate the oper-
ating hours, patient mix and scheduling rules of the OR. The model was used to test
various schedule ordering rules (i.e. FCFS, LTF, etc.) in order to smooth daily demand
of inpatient beds. The model is also able to adjust the case mix each day based on the
three patient groups of major, intermediate and minor.
Similar to Blake et al. (1995), this general framework is unable to capture some of
the details required for tactical decisions including cancellations, emergent patients and
detailed scheduling rules. The model is mostly focused on resource usage. Furthermore,
the use of patient groupings result in a loss of some detail as they have been grouped in
high level categories of minor, intermediate and major, loosing service and surgeon level
detail. This model is appropriate for broad, strategic decision making.
The second generic model (Moreno et al., 1999) is of a whole-hospital view and is
intended to be used for high level management decisions such as the number of beds, lab
capacity, etc. The generic simulation model is based on patient flow and data inputs. It
does not include any details such as scheduling rules. The model entities include patients,
resources, workforce and processes (related to their pathology and tests). This generic
model was tested on a single hospital to identify bottlenecks and under-utilised resources
Chapter 3. Background and Literature Review 41
and to test scenarios altering the availability of those resources identified in an effort to
keep waiting times for services within an acceptable range.
Generic Models of Other Healthcare Processes
A number of generic models were found that represented other healthcare processes. For
example:
• Bed capacity decisions: Costa et al. (2003); Pitt (1997); Nguyen et al. (2005)
• Laboratory and outpatient clinics: Berchtold et al. (1994); Ramis et al. (2002);
Paul and Kuljis (1995)
• Emergency department: Fletcher et al. (2007); Miller et al. (2004); Sinreich and
Marmor (2004); Centeno et al. (2003)
The design and implementation approach taken by each varied widely. A common
design feature of these generic models is that they are data-driven; where parameters
are determined based on inputs as opposed to hard coded into the model (e.g. Costa
et al., 2003; Pitt, 1997). Users are able to specify such things as the number of resources
available or whether to use a specific rule or not. This feature not only allows for easy and
wide application across numerous sites, but also for flexibility in the possible scenarios
to be tested.
A subset of generic models found included the objective to design models that were
easy to use and understand by the end user, empowering the decision makers to gener-
ate scenarios and evaluate model outputs to inform decisions (Pitt, 1997; Sinreich and
Marmor, 2004).
Validation of simulation models of real world processes is often difficult and generic
modelling is not immune. The validation approaches documented for generic models vary
widely. Some were only tested on a single site (e.g. Costa et al., 2003), while others were
validated at multiple sites (e.g. Pitt, 1997; Paul and Kuljis, 1995).
Chapter 3. Background and Literature Review 42
Implementation of simulation models to aid decisions and the continued use of models
is often faced with many challenges, resulting in many solutions never being used. There
are many possible reasons for this including user buy-in and perceived model credibility,
the feasibility of the solution, and the ease of understanding and use of the model. This
problem is not unique to generic models. Of the generic models found, few document
successful implementation and ongoing use. One notable generic model is the generic
emergency department model created by Fletcher et al. (2007). The model was initially
intended to be used by the UK department of health to inform, at a national level,
barriers to reaching ED targets and to evaluate polices and solutions that address these
barriers (level three generic model). The success of the model at the national level lead
to a secondary study to apply the model locally at ten hospitals. With some changes
to the model structure, data from the local hospitals was inputted and the model was
validated at each site. The models were then used to test a number of possible solutions.
Implementation at each site varied, however. Only five of the sites tested scenarios,
and of those sites, only three implemented any of the changes suggested. The authors
noted a number of barriers to implementation of the generic model; the key barrier being
when hospital processes differed significantly from model assumptions. Though this issue
was often resolved through flexible use of data and data interpretation, small changes
to model structure and additional off-line analysis of model outputs. Other obstacles
faced were not related to the generic model itself. These obstacles included data quality,
‘organisational disfunction’, and motivation of the site.
The authors conclude that generic modelling of a ‘typical’ department and of a spe-
cific department are both valuable undertakings. At the local level, generic models are
especially valuable to “generate interest and facilitate debate of alternative methods.
(Fletcher et al., 2007, p. 1562)” However, for the purpose of studying the impact of
local decisions more work is required to calibrate, validate, and modify the model to the
specific site.
Chapter 3. Background and Literature Review 43
3.5.3 Generalised Models and Frameworks outside of Health-
care
Outside of healthcare there exists a large body of research in generic modelling across
all levels of genericity. However, within the area of domain-specific generic models (level
three) as described by Fletcher and Worthington (2009), there is a small body of research.
Similar techniques for design, validation and implementation were found.
Many of the generic models found purported to create a generic model of a specific
domain, often with the objective to produce an easy to use simulation model that can
either be used many times by the same company for different products, or to be used
by different companies for the same type of process. Though it is likely that specific
models have been created prior for similar processes at the same company or another,
most works found did not use these as a basis for their generic model, nor did they
compare results from the specific and generic models. A few papers were found that
specifically mention a previously created specific model, whether as a base for a generic
model, or as a comparator (Brown and Powers, 2000; Drake and Smith, 1996; Steele
et al., 2002). For instance, Drake and Smith (1996) found that compared to a previously
created specific model, the generic model proposed separated the model from the end-
user’s needs. As a result, end users who are not necessarily familiar with simulation
are able to run and compare alternate scenarios. Further, the authors found that their
generic model provided more flexibility and ease of use for testing a wider variety of ‘what
if’ scenarios.
No work was found that directly compared the results of a generic model to a specific
model.
For some works the objective of creating a generic model was a secondary goal, thus
a specific model was first built, then adjusted to be more generic. For example, Brown
and Powers (2000) proposed generic reusable maintenance model was first created as a
Chapter 3. Background and Literature Review 44
specific model in order to better understand the processes and appropriate scope of the
model. As understanding of the problem situation increased, the model was adjusted to
increase functionality and to be more generic in nature. In addition, Schroer et al. (1996)
set out to create a generic model for the apparel manufacturing industry, however, noted
and subsequently tested that the same model could be applied to other manufacturing
industries including electronics and electromechanical. Centeno et al. (2003) is an exam-
ple of a model within healthcare that was initially designed for a specific site, that can
easily be adopted to other sites because of its data-driven nature.
3.5.4 Design Concepts for Generic Models
The literature survey found two reoccurring themes for generic model design: data-
driven and component-based. Data-driven implies that the specifics of the model, in
terms of the layout, quantity, rule selection, etc., should be done using input variables
allowing for quick and easy changes based on the particular process under study, and the
‘what-if’ tests under consideration. Jain (2008) defined data-driven simulators as models
that are completely parameterised through data. This can be done through the use
of forms, tables, spreadsheets or templates. Jain further notes that the key advantage
to data-driven models is that they are quick and easy to develop and use. However,
disadvantages include some inflexibility in the model if an option was not modelled, and
that the user interface cannot always be easily customised to meet the specific needs of
the system under study.
Component-based design, also referred to as modular design or object-orientated de-
sign refers to the structures of the model. Using components allows for flexibility in the
design of the model to suit the system under study. It allows for changes in process
order, number of components used, etc. Modular design can also allow for incremental
enhancements to the model (Pidd, 1992).
Pidd (1992) further describes a number of other features that are important to con-
Chapter 3. Background and Literature Review 45
sider when designing generic simulation models. He defines the key features to any
data-driven model including:
• The model should be suited for a range of applications.
• There should be no programming required by the user.
• The user provides data to the model that can be numerical, logical or text-based.
• The model contains logic for all anticipated instances of the domain.
A number of other papers found commented briefly on design steps to create generic
models. Kaylani et al. (2008) laid out four steps to generic model building. First, key
performance measures are identified. Then system boundaries and level of detail are
defined. Third, the typical instance that is required to be captured for the initial design
is identified. Finally, typical scenarios are iteratively generalised to accommodate all
instances in the domain.
Steele et al. (2002) present a higher level set of steps: First, the domain of interest
is selected. In other words, the problem and domain that address the objectives of the
study. This step helps to make the appropriate abstraction decisions. Here, the authors
caution to keep a narrow domain, as broad domains can lead to overly large models that
are difficult to understand and maintain. The second step proposed is to map out the
conceptual level diagram of the system processes and the interrelations between those
processes. Finally, Steele et al suggest to garner information from many different experts
in the system to ensure that different options are considered. To promote usability, Steele
et al also advocate to allow users to enter input in their own system-specific terminology,
and that outputs reflect their terminology as opposed to the terminology chosen by the
designer of the generic modeller.
Additionally, Brown and Powers (2000) offer a series of questions to help define the
scope and structure of generic models: What’s important to the model? What has crit-
ical impact on operations and what is touched by key functions? In what way should
Chapter 3. Background and Literature Review 46
the model be flexible and how much flexibility should the model allow? What is more
important, fidelity or speed? What inputs and outputs are required? In addition to an-
swering these questions when designing their generic model, the authors first determined
the components and processes required, then defined the overall structure and bounds of
the model.
Another method used to design generic models is to first build a specific model based
on a single site, then expanding it to be generic. This method was used for the generic
lab model by Berchtold et al. (1994), and a maintenance model by Brown and Powers
(2000). For example, the model of a generic laboratory was developed in two stages
(Berchtold et al., 1994). First, a feasibility model was created based on a single site.
Once it was known that the model was possible and valid, a generalised model was
created that could be used at any laboratory based on the learnings from the first step.
The conceptualisation of the generic framework is based on a set of work cells, each of
which has an input, output and data flow, and can be controlled by outside sources.
When faced with multiple different flows, some models began by mapping the generic
flows of the main flow groups. The flow maps are then used as a basis to build the generic
model. For example, Fletcher et al. (2007) mapped the three main patient types: major,
minor and admitted. Similarly, a generic manufacturing model for the apparel industry
found that it could also model other manufacturing industries such as electronics and
electromechanical (Schroer et al., 1996).
To design their surgical model, Blake et al. (1995) conducted interviews with staff at
four hospitals, then consolidated and validated the patient flow into a set of flow charts.
A DES model was created from the flow charts. To overcome challenges in portability
and ease of use, the authors chose to create a flexible base model that can be quickly
and easily customised for each site. Based on their scope and goals, this was better than
attempting to create a completely general model.
Chapter 3. Background and Literature Review 47
3.5.5 Validation and Verification of Generic Models
Though some articles describing generic models mention that their model was validated
and verified for at least one instance, few discussed issues encountered with verification
and validation of their generic model. Specifically, issues related to the nature of building
a generic model.
A generic model created for reusable launch vehicles was able to compare results to a
previously built specific model (Steele et al., 2002). Steele et al. found that the generic
model yielded the same results as the specific model. While the authors do not specify
how they came to this conclusion, they do mention that this statement is based on a
single test situation. Additionally, analysis of the generic lab model found that when
compared to a fixed-configuration model, that there was no difference in terms of results
and validation (Berchtold et al., 1994).
Few generic models stated that they had been verified and validated across multiple
systems. A generic model for manufacturing processes stated it had been validated in five
different situations, showing that the model can be portable while also being significant
(Nelson, 1983).
The national-level generic ED model (Fletcher et al., 2007) was validated by both
open- and black-box methods. For open box validation, the model’s assumptions, flow
and decisions were confirmed by clinical experts. Data representing the ‘average’ hospital
was used for black box validation to ensure model calibration. When applied locally the
emergency department model by Fletcher et al also validated to multiple sites with some
customisation (Fletcher et al., 2007). Other models that validated at multiple sites were
Paul and Kuljis (1995); Sinreich and Marmor (2004, 2005); Pitt (1997).
Ozdemirel (1991) created a verification tool for a generic manufacturing system model
that could easily be adjusted and applied to other generic models. The purpose of the
verification tool is to measure user acceptance of the model structure and assumptions.
Through testing this system on users with a variety of experience with the system, the
Chapter 3. Background and Literature Review 48
author was able to determine that the generic model proposed had a high acceptance
rating. Further, the author concluded that this helps prove that generic modelling is
“feasible, and its implementation has potential use for simulation modelling in manufac-
turing” (Ozdemirel, 1991, p. 426).
3.5.6 Cost-Benefit Analysis of Generic Models
Many researchers propose creation of generic models to reduce model development time,
reduce expertise required to develop and run models, improve model consistency and
quality, and reduce programming when used multiple times. These are often also touted
as the advantages of a generic simulation model.
There are however some disadvantages to generic simulation modelling. Jain (2008)
noted that generic modelling can lead to a reduction in the flexibility of scope compared
to general purpose simulation languages, object class libraries, etc. Nelson (1983) noted
that the flexibility required of a generic model leads to slower run times. Steele et al.
(2002) noted that in generic simulation modelling, as with most modelling, there can
be a tendency for ‘requirements creep’, leading to increasing costs and complexity. The
authors note that it can be avoided by strictly enforcing that only the requisite level of
detail be included, maintaining scope at a manageable size, and focusing on the important
factors only. Steele et al further conclude that the longer development time required to
initially design and build the generic model is outweighed by the benefit of using the
model across many systems and the shortened time for application and experimentation
on different systems. Moreover, Cope et al. (2007) note that generic models can be
complicated in design and set up, as well as require large amounts of inputs and knowledge
in order to run.
Finally, Jain (2008) discusses a number of trade-offs experienced with generic mod-
elling that should be considered when deciding whether a generic model is worthwhile
over a specific model.
Chapter 3. Background and Literature Review 49
Designing Generic Models
Little work was found suggesting specific principles, guidelines, tools or methods to de-
veloping generic simulation models both during the conceptual modelling and modelling
stages. In a paper speaking to decision support systems and operational research within
the realm of healthcare, Boldy (1987) states that a decision support tool should be
thought of and developed in the mind set that it will be used as a personal tool for
decision makers. He further outlines three features of a decision support system that are
key to the success and use of the tool:
• Emphasis on flexibility and adaptability,
• Ease of use and understanding by non-computer people, and
• Models linked with data sources.
When Brown and Powers (2000) set out out to create a generic model for maintenance
they used a series of questions to help define the structure and bounds of the model.
Examples of questions include ‘what is important to the model?’ and ‘how should it be
flexible, and how flexible?’. The series of questions helped the authors determine model
boundaries, scope, elements to include and exclude and the level of detail. They also
noted that consideration should be given to how the generic model will handle inputs
that can vary in size, and how can it be structured so that inputs can easily be replaced
or connected to the database. The authors also recommend the guidelines proposed by
Law and Kelton (2000): to be as flexible as possible, but to also keep the model within a
reasonable size and scope. When designing simulation models, especially generic model,
a key challenge is making these various design decisions.
Within their study of conceptual modelling of generic simulation studies of security
systems Guru and Savory noted that generic or template-based simulation models often
involve applying a set of “pre-built, ready to use, modelling objectives, modules, or model
Chapter 3. Background and Literature Review 50
of common simulation situations. (Guru and Savory, 2004, p.g 867)” The authors suggest
to develop generic models using switches to turn on or off specific parameters in order to
fit the model to the current system under study.
The authors further noted that there is little in existing literature in the area of generic
or template-based simulations. As such, Guru and Savory set out to develop a framework
to assist a simulation modeller with conceptual model development of physical security
systems. While created specifically for security systems model design, their strategies
can be used in generating frameworks for other generic simulation conceptual models.
Their structure encourages reusability through use of modular and reusable components
and an expandable architecture.
In summary, when building generic simulation models much focus and time should
be dedicated to the design stage. Generic models need to be designed so that they
can be applied to specific real systems that were not known or considered during the
development of the model. Generic model conceptualisation requires the modeller to not
only consider the real system and client needs that he is currently working with, but
he must also look ahead and predict what other clients and systems may look like and
require from the same generic model. Chapter 4 of this thesis reviews in detail the work
and decisions made during the design stage of the proposed generic model.
Chapter 4
The Proposed General Model
This chapter will review the evidence supporting the first contribution of this research
to the fields of healthcare operations research and generic simulation modelling: The
proposed generic model demonstrates that a generic model of a complex healthcare system
process is feasible. Moreover, a generic model can be meaningfully applied to multiple
sites to influence and inform decisions. More specifically, the contribution herein is
the presentation of a generic model that has been successfully applied to multiple sites.
Successful application of the generic model includes implementation of decision options
tested by the model. In addition, the model continues to be applied at the sites and
to new sites, and work is underway to create a commercial version of the generic model
for wider distribution and use. The chapter concludes with providing some tips and
suggested guidelines for generic model based on this research experience.
This chapter will first provide a review of the generic model. Then, the contribution
is demonstrated through discussion of the application of the model at multiple hospitals
of varying sizes and characteristics.
51
Chapter 4. The Proposed General Model 52
4.1 The Generic Model
4.1.1 Model Objective
The literature review demonstrated that there is a need for a decision support tool to
help guide decision makers in tactical decision making. Due to the significant interactions
that occur within the areas within the perioperative process as well as with other areas
of the hospital, it is important that the model considers a holistic view by including more
than just the OR. Thus, the generic model includes the perioperative processes as well
as the post-operative inpatient recovery resources, as depicted in figure 4.1.
The review also indicated a need for generic simulation model solutions to reduce
duplication of effort and cost of creating and applying models. In order to be a generic
model, it must also be flexible to allow for application to numerous hospitals’ periop-
erative service and to be able to consider many different possible decision options that
stakeholders may be interested in evaluating. This flexibility and adaptability will allow
not only application at multiple sites, but also multiple applications at a site over time.
In particular, this generic model is to address tactical decisions affecting the surgical
patient flow of the hospital. These decisions include the availability and scheduling of
the OR resource using the block scheduling method; wait list decisions including order
and length of wait; and the availability of other resources required such as the PACU,
wards, etc.
4.1.2 Model Scope, Level of Detail, Assumptions and Simplifi-
cations
The challenge with designing any model, especially a generic simulation model, is de-
termining the model scope and level of detail - specifically what elements and details to
include and exclude.
Decisions on scope and level of detail, also referred to as simplification and abstraction,
Chapter 4. The Proposed General Model 53
!"#$%&'"()&*$+#&,-+$%&.#$+)$(/0&1$2-+3#$(3%&
4-+5%&,+67/-8&,-+$%&
$3/9&
1$/6:6"(&3"&"2$+-3$&
;+$<=2$+-7>$&
;":3<=2$+-7>$&?$/">$+0&
@##$56-3$&;":3<
=2$+-7>$&=2$+-7>$&
16:/A-+)$&3"&A"#$%&8"()&3$+#&/-+$%&
+$A-B6863-7"(%&$3/9&
;$+6"2$+-7>$&C3-)$&
D"5$8&E"F(5-+0&
Figure 4.1: High level surgical patient flow and generic model boundary shown.
are often a difficult part of conceptual modelling. In modelling, it is not desirable to model
all that is known of the real world, even if it is relevant to the objectives or the problem
situation of the system understudy (Kotiadis and Robinson, 2008).
There are a number of authors who have contributed through anecdotes, methodolo-
gies and rules to help make these decisions. For instance, Robinson (2004, 2007b) suggest
that if the component under consideration is important to the validity, credibility, utility
or feasibility of the model it should be included. When determining what to include and
exclude, the effect of excluding a component on the model’s validity, credibility, utility
and feasibility were considered. Would excluding the component reduce the accuracy of
the model (validity)? Would excluding the component decrease the confidence in the
model’s results in the eyes of the end users (credibility)? Would the exclusion signifi-
cantly increase the complexity of the model or run time (utility)? Is the data required
for the component easily available at most hospitals (feasibility)?
“Perfection is achieved, not when there is nothing more to add, but when
there is nothing left to take away.” Antoine de Saint-Exupery
Numerous authors have proposed principles or guidelines on model development based
on the idea of evolutionary development, meaning to start small and simple, and add as
Chapter 4. The Proposed General Model 54
needed. For example, Pidd (1999) presents six modelling principles that can help guide
conceptual modelling:
• Model simple, think complicated,
• Start small and add,
• Divide and conquer, avoid megamodels,
• Use metaphors, analogies and similarities,
• Do not fall in love with data, and
• Modelling may feel like muddling through
Other authors have also provided insights into abstraction and simplification. For
instance various authors suggest repeatedly removing detail and scope, simplifying com-
ponents, using constants instead of variables or eliminating variables entirely, using linear
relations where possible, strengthening assumptions and restrictions, reducing random-
ness, dropping unimportant components, using a random variable to depict a component
and grouping components. (Robinson, 2007a)
It was with these principles and guidelines in mind that the model scope, detail,
assumptions and simplifications were designed. Figures 4.2 and 4.3 demonstrate which
components are included and excluded in the proposed generic model.
To demonstrate the key scope, detail, assumptions and simplification decisions, a brief
description of the model follows - including notes and discussion on these decisions.
Overview of the Generic Model
The model overview provided herein serves to provide a brief description of the model’s
scope, level of detail, assumptions and simplifications. A detailed description of the
model design is provided in appendix C. A thorough conceptual model description is
Chapter 4. The Proposed General Model 55
•!!"#$%&'()*+,(+-,.-/$0+#,
•!1(/",2(3,14'#"'3,.'"(,
•!.+("%&5"%$(,$+-46)*+,!**/,
•!7!,5*8-$+#,.'"(,
9!:($&,8$%&,/(+(#"/"+&,
9!14'#$6(8,%65"-48$+#,
;'"<7="'()>",
9!7="'()+#,'**/%,
9!?(+6"88()*+%,7="'()>",
9!;*%&<(+"(%&5"%$(,?('",@+$&,
9!?'$)6(8,6('",(+-,%&"=,-*A+,4+$&%,
9! B+=()"+&,A('-%,
;*%&<7="'()>",
•!;'"<*=,68$+$6,%65"-48$+#,(+-,
>$%$&,
•!1(/",-(3,%4'#"'3,
•!C*+<%4'#$6(8,4+$&%,
D8"/"+&%,B+684-"-, D8"/"+&%,DE684-"-,
Figure 4.2: Pictorial summary of perioperative processes included and excluded in the
proposed generic model.
Chapter 4. The Proposed General Model 56
!"#$%"&'()*+',
-..'/0)$',&"#$%"&'()*+',('1"+'(2, !"#$%"&'()*+',('1"+'(2,
3&'()*+',!)*'4$,)((0+)56,('70#$()*"4,)4/,
)/.0##0"4,
89(#0476,:4)'#$;'#0)6,<9(7'"4,
)##'##.'4$#,
=04)5,$'#$#6,/0)74"#*1#6,0.)7047,
!('&)(',&)*'4$,>"(,#9(7'(2,
=04)5,('#"9(1',1;'1?, <9(7'(2,
!('%"&'()*+',
@)0$,50#$,.)4)7'.'4$, <1;'/95',&)*'4$,>"(,&("1'/9(',
<1;'/95',('#"9(1'#,>"(,&("1'/9(',AB'/#6,'C90&.'4$6,'$1DE,
<1;'/95',&)*'4$,>"(,&('%&("1'/9(',+0#0$#,
!('%"&'()*+',150401#6,$'#$#6,/0)74"#*1#6,'$1,
Figure 4.3: A more detailed pictorial summary of components included and excluded in
the proposed generic model.
Chapter 4. The Proposed General Model 57
provided in appendix A. In addition, documentation of decisions regarding level of detail,
assumptions and simplifications are provided in appendices A.3 and A.4.
Patient Types
There are three main types of surgical patients included in the proposed generic
model: elective, urgent/emergent and inpatient. The model allows elective patients to
be further classified by a priority level. This priority allows for further stratification of
elective patients for wait list ordering and priority over elective time. For instance, the
maximum wait time of an elective patient on the wait list can vary by their priority level.
In addition, the model allows for the elective wait lists to be either by surgeon or by
service, to reflect whether wait lists are pooled among surgeons, or as surgeon managed
wait lists.
For urgent/emergent patients, the model allows for the specification of the number
of urgency levels and their maximum allowable waits.
Finally, the model has two options for inpatient flow. One option is to manage these
patients separately using their own waiting lists, with priority levels, maximum allowable
wait times, surgeon scheduling, etc. This option is best suited for hospitals that schedule
an inpatient like an elective patient who must be done within the next week or so by the
surgeon who ordered the procedure. The second option does not categorise any patients
as inpatients. This option is intended for hospitals that consider inpatients as part of the
urgent patient population and follow the same scheduling rules as urgent patients.
Pre-Operative Section
The pre-operative stage of the generic model is concerned with scheduling patients
for surgery and wait list management. The generic model considers block scheduling
schemes and has been designed such that a variety of scheduling rules and procedures
can be considered within the block scheduling practices.
Within the block scheduling scheme, the generic model can place elective patients
on wait lists based on the surgeon assigned or on the service assigned. When wait lists
Chapter 4. The Proposed General Model 58
are surgeon based, the elective patient must be scheduled to OR time assigned to their
surgeon. Alternatively, when wait lists are according to service, patients are scheduled
into time that has been assigned to their service, regardless of which surgeon is assigned.
In this case, the patient is scheduled into the first time slot available within the surgical
service, regardless of the surgeon assigned. The procedure is performed by the surgeon
assigned to that block, not necessarily the surgeon who ordered the procedure. This
second scheduling method can be considered a type of pooled wait list strategy, where
patients are seen by the first available surgeon.
The generic model allows for scheduling rules such as restricting case mix within an
OR block, or the number of a type of case that can be performed across all ORs in a
day. For example, some hospitals choose to restrict the number of admitted procedures
scheduled in a day, or dictate that a certain number of a particular case type must occur
in some ORs in order to meet volume targets. More detail on these rules and how they
can be used can be found in Appendix C.9.
Scheduling the patient for the pre-operative clinic, patient education sessions and di-
agnostic tests can cause delays in the patient’s procedure, especially for elective patients.
However, the reason for these types of delays is largely due to lack of co-ordination rather
than lack of resources. Typically, the pre-operative clinic and patient education sessions
are run by the perioperative department. Often, these visits require access to other hos-
pital resources such as diagnostic imaging and lab. Delays are often due to test results or
consultations not being completed prior to the scheduled surgery date. Better planning
and co-ordination of these activities would help reduce many of the resulting delays and
cancellations.
The activity and processes of these pre-operative activities are not considered by the
generic model. The reasoning for this simplification was twofold. First, it is believed that
the impact of pre-operative activities on the tactical planning and scheduling decisions is
small. As a result, excluding these activities should not change the results of the model
Chapter 4. The Proposed General Model 59
and decisions proposed. The reason for this is because these areas are not the bottleneck
of the system. Often, issues arising from these areas that impact patient flow are due
to poor coordination and planning of resources. For example, a patient is scheduled for
their pre-operative clinic visit only 4 days prior to surgery. Test results conducted at
the visit suggest that a consultation with an internal medicine doctor be performed prior
to being cleared for surgery. Unfortunately there is not enough time to schedule the
visit, and the case is rescheduled for a later date. If, however, the patient’s visit was
scheduled two weeks prior to their scheduled surgery, there would have been sufficient
time to schedule the follow up appointment. Modelling and decision making of these pre-
operative activities can be performed with key information such as the master surgical
schedule as an input rather than a variable.
Second, some of these resources are used by many other hospital departments. For
example, the diagnostic imaging machines are typically a shared resource for both in-
patients and outpatients. The perioperative demand requirements for these resources
represent only a fraction of the total demand. Modelling these resources accurately
would require making assumptions on the capacity available for perioperative, which is
difficult to ascertain. Access is further difficult to quantify as emergent patient access
is often prioritised over elective and scheduled care. Thus, overall, it was determined
that these areas could not be accurately modelled based on the information available
and nature of the resources and their management.
However, development of decision support tools to help these areas better co-ordinate
and plan their demand and capacity would be beneficial.
In order to reflect the fact that some cases are cancelled due to these issues, the model
does include an inputted cancellation rate that can include these cancellations, as well
as other reasons such as medical, or anaesthesia unavailable.
Operative Section
The key resource considered by the model in this stage is the OR. As part of the
Chapter 4. The Proposed General Model 60
operative stage, patients may also spend time at patient registration, same day surgery
beds, OR holding bays and anaesthesia induction rooms prior to going to the OR. The
model does not however consider flow through these pre-OR areas such as admitting,
registration, or same day surgery. Much of the time spent in these areas is simply waiting
for their procedure. Hospitals typically ask elective patients to arrive more than two
hours prior to their procedure, even though there are only typically a few short activities
required such as nurse assessment, surgical site preparation, etc. Better co-ordination
and planning should minimise OR delays and reduce patient waiting. Surgical tactical
decision making assumes that co-ordination and planning of these activities does not
significantly affect overall patient flow.
During this stage of the model, numerous cancellation decisions are made. Reasons
for cancellations include no ICU, ward or SDU bed, not enough scheduled time remaining
to complete a case, and cases bumped due to a more urgent case. The model also allows
for an inputted percentage of cases to be cancelled for other reasons out of its control
such as medical reasons, required tests/clinic visits not completed, etc.
Post-Operative Section
The perioperative process includes immediate post-surgical recovery, typically in the
PACU and other recovery areas such as the day surgery unit.
The patient’s recovery requirement is indicated in the model input and can include
a stay in the PACU. However, the day surgery unit, where patients may continue to
recover post-surgery prior to being discharged the same day, is not included. The reason
for this exclusion is similar to those above for the pre-operative activities and some
operative activities, they have little impact on the intended decisions and results of
tactical planning and scheduling decisions.
Patients who do not require an inpatient recovery post-operatively are discharged
from the model at this stage.
Post-Operative Inpatient Recovery Section
Chapter 4. The Proposed General Model 61
Patients who require admission to the hospital for an extended recovery will enter this
section of the model. The model considers three types of units and allows for multiple
units of each type. Intensive care units (ICU) are intended to represent the most acute
level of care. The lowest level of acuity beds are represented by ward units. The model
also allows for care between the ICU and ward levels through the inclusion of step down
units (SDU). The particular route and length of stay (LOS) each patient takes through
the recovery stage is inputted in the patient file.
When patients complete their stay in these units, the patient is discharged from the
model. Discharge locations, such as rehabilitation and long-term care are not modelled
as they are often separate institutions outside of the acute care setting. The different
types of facilities within discharge locations are numerous and complex, accepting pa-
tients from many different institutions. Modelling this spectrum of care is difficult and
complex within itself. However, this element of the patient’s flow is modelled indirectly
by including the time patients spent in acute hospital beds waiting for access to these
downstream resources.
Two key assumptions were made regarding patient flow though this section. Firstly,
patients can only flow from higher acuity units to lower acuity units, but not the other
way. For example, a patient can go from the ICU to a ward, but not from a ward unit
to the ICU or SDU. This assumption was made to simplify the flow descriptions in the
model and input files. If a patient did actually flow from low acuity to higher acuity
unit, the total length of stay for each unit is inputted into the model. In reality, patients
moving from lower acuity to higher acuity does occur. However, from the experience
of experts at the hospitals tested it is not common in the surgical patient population,
especially for the elective patients, which represent the large majority. By using the total
length of stay for each acuity level, the impact of this simplification is minimised as the
only difference is when during their entire stay the patient spent those days on each unit.
The outputted occupancy of each unit should be fairly representative when the model is
Chapter 4. The Proposed General Model 62
run multiple times for many weeks.
The second assumption is regarding the discharge time of admitted patients. To
facilitate the co-ordination and decisions within the model, patients are discharged at
midnight on the morning of their day of discharge. For instance, if based on the patient’s
inputted length of stay (LOS), they would be ready for discharge at 14:00 on day five,
the model would discharge him at midnight on day five, (the start of day five). This
was done to simplify the flow and co-ordination of new admissions and patients moving
between units. This assumption allows all patient transfers to occur at midnight. This in
turn allows for simplified management of patients to be admitted each day as the exact
number of beds available for the day is known. This simplification does affect outcome
measures concerning transfer delays which will be discussed as a model limitation in
section 4.2.3.
4.1.3 Model Outputs
In order to be meaningful to the decision maker, the outputs of the model must include
the performance indicators that they consider important. The outputs selected were
chosen based on discussions with decision makers at the initial three sites, as well as
common indicators found in the literature. In addition, more detailed drill down of
some outputs is useful to have available when calibrating the model to identify reasons
for variation between model outputs and expected results (i.e. historical values during
validation, or expected results during decision option testing).
• Wait lists - by service, by surgeon, by patient type and by priority within each
patient type
– Average # waiting
– Average waiting time
• Operating Rooms:
Chapter 4. The Proposed General Model 63
– Utilisation as a percentage of time marked as elective used for elective cases
– Utilisation as a percentage of time OR used over 24 hours
– Overtime
– Undertime
• Resources:
– Census at midnight for each ICU, SDU, Ward by day of week
– Census for each PACU by day of week
– Off service rates into each resource unit - the number of surgical patients on
the unit that belong on a unit of a different surgical service
– Off service rates out of each resource unit - the number of surgical patients
that should be on the unit that are on another unit because there wasn’t
enough beds on this unit when needed.
– Hold over in PACU - number of days required to use on-call or closed time for
one or more patients
• Cancellations - by service, by patient type and by reason:
– No ICU bed
– No SDU bed
– No ward bed
– Not enough OR time
– Bumped for more urgent case
– “other” reason
• Throughput -by service, by patient type, and by surgeon
Further detail on the outputs, including definition and how they are collected, are
provided in Appendix C.
Chapter 4. The Proposed General Model 64
4.1.4 Model Inputs
The generic model inputs can be subdivided into three types: hospital structure, schedul-
ing and patient information and flow. Data within the hospital structure group gives spe-
cific information on the resources available at the hospital. The inputs within the schedul-
ing group provide details on the master surgical schedule (MSS) and other scheduling
parameters. Finally, the patient information and flow group provides information on the
characteristics of the hospital’s patients and how they should flow through the system.
A visual summary of the generic model inputs is presented in figure 4.4.
!"#$%&'()*&+,-&,+.)
/ 0)+.#",+-.#1)234)56784)7784)*784)9'+:#)
/ 0)#,+;."<#)/ 0)#.+=%-.#)
*->.:,(%<;)
/ ?("-@)#->.:,(.)/ *.+=%-.)'(("-'A"<)/ *,+;."<)'(("-'A"<)
/ *->.:,(%<;)+,(.#)/ 7'#.)"+:.+%<;)
5'A.<&)B<C"+D'A"<)'<:)E("9)
/ 6++%='()$'F.+<#)/ 5'A.<&):.&'%(#)/ 5'A.<&)$'&>#)/ 2GH#.+=%-%<;)/ I'%&)(%#&)"+;'<%#'A"<)'<:)"+:.+%<;)
Figure 4.4: Summary of model inputs of the proposed generic perioperative simulation
model.
These inputs allows the generic model the capability to test a variety of different
possible decisions (experimental factors):
• Master surgical schedule:
– Elective block lengths (i.e. 8 hr vs. 10 hr, etc)
– Service assignments (i.e. which service has which days)
– Surgeon assignments (i.e. which surgeon has which days)
Chapter 4. The Proposed General Model 65
– Reserved time for urgent/emergent cases (i.e. an OR reserved all day, or for
a few hours within the elective schedule)
– On-call OR time
• Patient demographics:
– Length of stay in any of the post-surgical beds
– Length of the procedure in the OR
– Post-surgical routes i.e. some cases no longer require ICU for a procedure
type
• Scheduling rules:
– Case mix rules within an OR day: dictating the types of cases that can occur
in an elective block
– Case mix rules within a service’s time: specifying the types of cases that can
occur across all ORs of a service in an elective day
– Case mix rules across all ORs: dictating the types of cases that can occur
across all ORs during an elective day
• Schedule ordering:
– Ordering policies
– Anaesthesia type
– Type of case by some category (i.e. joints first)
– Admitted vs. day cases first
– Booked time
– Etc.
• Resources:
Chapter 4. The Proposed General Model 66
– Number of beds in various units
• Emergent/Urgent cases:
– Use of reserved time in elective schedule
– Rules around how and when to schedule
• Inpatient cases:
– Rules around how and when to schedule
• Off-servicing rules (i.e. where can they go if the bed they require is off service):
• Strategic:
– Moving services/surgeons from one site to another
4.2 Application of the Generic Model
The purpose of this section is twofold. First, to review evidence that demonstrates how
the generic model contributes to the generic modelling field through its successful appli-
cation and use at multiple sites. Second, this section reviews the issues faced applying a
generic model to real sites.
4.2.1 Use of the Model at Multiple Sites
To date, the generic model has been applied to six different hospitals and interest for
further application continues. At four of the six applications, results from the model have
been used by decision makers to inform tactical decisions and influence stakeholders. This
is a significant aspect of this model’s contribution to research as much of the literature
in operational research was not able to actually demonstrate that the model was used by
decision makers to inform and influence their decisions. Few proposed generic models,
Chapter 4. The Proposed General Model 67
both in healthcare and other industries, have been successfully applied to multiple sites.
Furthermore, the model has been successfully used by simulation experts not part of the
initial design and applications, demonstrating transportability of the generic model. This
is another feature described in generic modelling research, but has rarely been achieved.
A brief discussion of the implementation at each site is provided below. The applica-
tion at the remaining two hospitals, St. Mike’s and Mt. Sinai, was limited to validation
to historical data only as there was a loss of engagement and interest by the decision
makers at these sites. Thus, full application was not possible.
Application at the Juravinski Hospital
Application of the generic model at the Juravinski Hospital was the most successful of
all sites studied. The generic model was not only used by decision makers to evaluate
tactical decisions, but was also used as a negotiation tool with various stakeholders
including surgeons, unit managers and the medicine department. In January 2012, one
of the options considered by the model was piloted for seven weeks. After analysis of
the pilot period, the changes were implemented permanently. The decision implemented
included changes to the MSS, available ward unit capacity and off servicing policies.
Work is currently underway at Juravinski to update the data in the model to pre-
pare for an upcoming round of tactical decision making. In addition, Hamilton Health
Sciences, has begun to apply the model at their General Hospital site.
The application at Juravinski demonstrates that not only can the generic model be
successfully applied to a hospital, but that the model’s results are considered accurate
for the purposes of influencing stakeholders and making decisions. Furthermore, the use
of the model proved so valuable to the hospital administration that future application at
Juravinski and other sites are underway.
Chapter 4. The Proposed General Model 68
Application to two William Osler Hospitals: Brampton and Etobicoke
As part of a large scale perioperative improvement initiative at William Osler Health
Service, the generic model was applied to both their surgical sites: Brampton Civic
Hospital and Etobicoke General Hospital. Numerous inefficient practices existed at both
sites including late start times for scheduled procedures, underbooking and overbooking
of OR blocks, inaccurate booking times, OR block over runs and lengthy waits for target
cases. The generic model was used as one of the improvement initiatives, with the focus
on adjusting the MSS to co-ordinate resources, accommodate new surgeons hired, increase
volumes and plan for urgent and emergent patients.
Due to the many current inefficient practices at the two hospitals, the generic model
was used to demonstrate what resources, in terms of OR blocks, were required to deliver
their current volumes once their inefficient practices were addressed. The model demon-
strated that OR capacity existed without a need to increase the number of staffed ORs
or the operating hours of the OR. The model was subsequently used to determine how
best to allocate the found OR capacity. Options considered included giving time to the
new surgeons hired, reserving time for urgent patients, and providing additional time to
services to achieve volume targets.
This application of the model contributes to the research field in several ways. Firstly,
the generic model was designed without knowledge of William Osler’s processes and
characteristics, which shows that the generic model is in fact generic. Secondly, the initial
three test hospitals were all academic institutions; application at these two community
hospitals further strengthens claims of the model’s genericity.
Finally, the experience at the two William Osler hospitals revealed a secondary but
important use of the model as a tool that demonstrates what performance could look
like if inefficient and poor practices are resolved. This demonstrative use of a simulation
model is not typically explored. Further detail into this aspect is explored in Chapter 6.
Chapter 4. The Proposed General Model 69
Application at Prince Albert
The Saskatchewan Ministry of Health was concerned with the productivity of the Prince
Albert Hospital, and whether the hospital would be able to achieve the ministry’s imposed
wait time targets. The hospital believed an additional OR and beds would be required.
The ministry wanted to see if efficiencies could be found in lieu of capital expansion.
The generic model was applied to this hospital by a simulation expert who was not
previous involved in the design and research of the generic model. However, support to
the modeller was provided by Daphne Sniekers and Carolyn Busby, the original model
coder.
Similarly to William Osler, validation of the generic model was somewhat challeng-
ing due to existing inefficient processes. The analysis resulting from the generic model
suggested that in order for Prince Albert to reach wait time targets and accommodate
increasing demand, commissioning an additional OR would be required. However, if 17
of the existing 32 beds were reserved for surgical patients (aka ring fenced), no additional
surgical beds would be required. At the end of CRHE’s commitment to the project,
the hospital was addressing their inefficient practices to improve their performance. The
hospital was also in negotiation with the Ministry to implement the proposed capital
expansion.
Future Planned Applications
Besides the planned application at the General Hospital site of Hamilton and the reuse at
Juravinski previously mentioned, the generic model has recently been applied by another
simulation modeller to two other sites in Saskatchewan.
In addition, Visual8 is keen to work with Hamilton, the research team of CRHE to
continue to improve the generic model by improving upon the genericity of the model
based on experience of applying the model to the various hospitals to date, and developing
a user interface that will allow for easier import of hospital data and direct interaction
Chapter 4. The Proposed General Model 70
with the model by end users (hospital decision makers).
4.2.2 Issues Experienced and Changes Required while Imple-
menting
Throughout the implementation of the generic model, notes were made of any changes
and additions required to the generic model, and any issues and limitations faced. The
charges and additions documented can be classified into three categories. First, the
changes that were made during the coding phase to facilitate implementation using the
selected simulation software. Second, those changes and additions required to accurately
portray patient flow in order to make informed tactical decisions within the scope set
out in the generic model. These changes were made not only for the implementation of
the model at a particular hospital, but also to the model’s design. These are considered
changes required for the genericity of the model. Finally, the third type of changes were
made to answer very specific questions posed by hospital decision makers that were not
considered within the scope of the generic model. These changes were made specifically
for the study at a particular hospital and were not made to the generic model framework.
These can be considered as custom, post production charges and additions. A change
database listing all changes made and the reason for the change is included in Appendix
B.
Changes for Implementation into Simul8
These changes were made to simplify coding in the software package Simul8 based on the
characteristics and functionality of the software. These changes do not affect the overall
flow or results of the generic model.
The conceptual model of the proposed generic model includes allowance for an elec-
tive OR block to be shared between two surgeons of either the same surgical speciality,
or different specialities. For the sake of ease of coding when programming the simula-
Chapter 4. The Proposed General Model 71
tion model an assumption was made that an elective OR block would only be shared
between surgeons of the same service. Based on experience to date at the initial three
academic hospitals and large urban community hospitals (i.e. Juravinski, Mt. Sinai, St.
Mike’s, Brampton and Etobicoke), this assumption was thought to be valid as it was
accepted best practice to avoid OR block sharing between services. However, at Prince
Albert, a small rural community hospital, shared OR blocks between services was a real-
ity in their master surgical schedule planning. In order to accommodate their practice, a
workaround was employed by scheduling another unassigned OR for the second service.
This workaround allows for the two services to run, but does not account for any delays
the second service may experience if the first service runs late. This unfortunately re-
duced the credibility of the model in the eyes of the decision makers at Prince Albert. It
is recommended that this coding simplification be addressed in future work. The code
adjustment would not require a significant amount of time or effort to complete.
Changes Required for Model Genericity
There were no changes needed to the generic model itself in order to maintain the gener-
icity of the model, within the scope and objectives laid out.
One limitation found in the generic model was around the assumption that when a
surgeon runs more than one OR, the ORs run parallel with two surgical teams. However,
at William Osler Health Sciences, a non-teaching hospital, the surgeon has neither a team
of residents nor a second OR team to operate parallel ORs. Here they use a second OR to
reduce surgeon idle time between cases for OR cleaning and set up, patient positioning,
etc. Accordingly, activity between the two ORs needs to be better co-ordinated and
scheduled. At William Osler, the only surgeons who were given “swing rooms” were three
orthopaedic surgeons who performed high volumes of total joint replacements (TJR).
These swing rooms allowed these surgeons to perform six TJR cases in a day using a
single OR team, versus four if the surgeon only had a single OR to use.
Chapter 4. The Proposed General Model 72
Since the swing room used here was for a very specific purpose, the existing model
capabilities were used to work around this scheduling method. Scheduling rules were
created to represent these swing rooms. On days when a swing room was scheduled, a
set of scheduling rules was used to dictate how patients should be scheduled. The generic
model would then run the two ORs independently.
As part of any future effort to improve and expand the proposed generic model, it
is recommended that various swing room scheduling schemes be studied further and
functionality added where needed to be more inclusive of swing room scheduling config-
urations.
Changes and Additions for Client Customisation
During the validation at William Osler, the working group identified that the proposed
generic model did not accommodate their specific urgent patient scheduling procedures.
At William Osler, patients were scheduled into available urgent OR block time based
on a reversed priority scheme, giving highest priority to the least urgent patients. The
generic model on the other hand assumes that the most urgent patients will always be
prioritised over less urgent patients. William Osler adopted this scheduling procedure
in order to ensure that wait time targets of their least urgent patients would be met.
Poor planning and scheduling practices in place at the hospitals were causing significant
delays in providing timely care to urgent patients who were not emergent (highly urgent).
In response, they allocated time for the least urgent patients, instead of addressing the
overall issue of urgent patient mismanagement.
This urgent scheduling process mismatch was identified early on in the validation
stage of the modelling, when many working group members were still very sceptical of the
model’s ability accurately model and guide decision making (see Chapter 5). Moreover,
it came to light prior to the development of the idea that the model should be used to
demonstrate potential performance of potential tactical decision options; as opposed to
Chapter 4. The Proposed General Model 73
exactly modelling a hospital’s current processes, good or bad. In an attempt to strengthen
user acceptance at the time, the proposed generic model was customised to their specific
processes.
However, upon reflection and in light of the proposed positioning of the generic model
as a decision and demonstrative tool for tactical decisions (see Chapter 6), this customi-
sation should not have been performed. Customisation of the model to include strange,
poor or inefficient practices detracts from the generic model’s purpose to demonstrate
potential performance of tactical decision options.
4.2.3 Limitations Noted
As a result of the chosen scope, level of detail, assumptions and simplifications of the
generic model, as well as the fact that the proposed model is intended to be generic and
applicable to many hospitals, a number of limitations exist.
First, the generic model assumes that the OR schedule is allocated based on a master
surgical schedule, or OR block schedule. The master surgical scheduling method was
chosen as it is the most commonly used scheduling method for surgical services.
Second, the model does not allow for more than one service to be scheduled into an
elective block. This assumption was made during coding of the model in order to simplify
the scheduling algorithm. Based on experience up to that time, it was assumed that most
hospitals do not schedule more than one service in an OR. A work-around is possible
without changing the code where the shared OR block is entered into the MSS as two half
blocks in two different ORs. This was done at Prince Albert, and the decision makers
agreed that it was an acceptable change. Since this practice appears to be more common
than originally thought, it is suggested that the coded model be adjusted accordingly.
Third, it was found that the assumption regarding how to schedule cases, when a
surgeon is allocated two ORs on the same day, is not reflective of some hospital’s pro-
cesses. The generic model design assumed that scheduling a surgeon in two rooms on
Chapter 4. The Proposed General Model 74
the same day typically occurred in hospitals where residents and surgical assistants are
used to run the ORs in parallel. However, some hospitals will use the second OR as a
“swing room”, that allows the surgeon and the OR team to use the second OR for a
case while the first is being cleaned, effectively reducing the turnaround time for that
surgeon’s schedule to zero. This practice was found at Brampton where some orthopaedic
surgeons are given swing rooms to complete a higher volume of TJR cases. In this case,
the inputted schedule and scheduling rules were adjusted such that the swing rooms were
scheduled appropriately. The adjustment was possible because the swing room was for
a very specific use. However, it may not be possible to use existing inputs to reflect
another hospital’s use of swing rooms. In the future, it would be prudent to adjust this
assumption such that scheduling of the swing room is more reflective of practices found
at hospitals.
The fourth limitation takes note that, based on the model’s design, simplifications
and assumptions, delays within the surgical day are not included nor measured. For
instance, delays experienced from the PACU to the inpatient units or from the OR to
the PACU when a bed is not yet available. Delays often occur when the downstream
resource, the inpatient beds, are not available when needed because the previous patient
has not yet been discharged/left from the bed. As a result, the next patient must wait
in the PACU for their bed to be ready, which may delay another patient in the OR
who requires that PACU bed. On the other hand, the proposed generic assumes that
coordination of admissions and discharges occurs smoothly by performing all discharges
and unit transfers from inpatient beds at midnight, reducing the bottleneck. Based on
the premise of this generic model to demonstrate potential performance, this assumption
was accepted as it should not affect end results and tactical decisions.
Finally, the model is somewhat limited in what type of scheduling rules it can ac-
commodate, even though the idea was to be as flexible as possible. The model is not
able to accommodate a scheduling rule such as “schedule a minimum of four total joint
Chapter 4. The Proposed General Model 75
replacements” when there is an identifier in the patient file for hip replacements and
knee replacements, respectively. For example, say in the same schedule, one wants to
test scheduling rules where on some days only hip replacements are permitted, while on
other days only knee replacements are allowed, and on other days at least four cases must
be completed that are either hips or knees. This type of rule would require an identifier
for cases that are hips and one for knees to handle the first two rules, however, the model
subsequently is unable to quantify the third rule which is a combination.
4.3 Cost Effectiveness of the Generic Model
The costs and benefits of model reuse and generic models are often debated in literature.
Many argue that model reuse and generic modelling often tend to be more work than it
is worth.
In the panel discussion presented in Robinson et al. (2004), Pidd provides a financial
model for cost-benefit analysis of reuse. His model states that if the average cost per
use (KN) is less than the cost of developing the code from scratch (C), then reuse is
worthwhile. Pidd defines the average cost of the N th reuse as the cost of developing the
code the first time plus the cost to adopt the code each of (N − 1) times it has been
reused: KN = C+A(N−1)N
. Pidd’s financial model assumes that the cost of building the
code is shared by all instances of reuse. An alternative would be where the initial cost of
building is paid by the initial developer. Each instance of reuse benefits from the savings
of modifying the code to suit their needs, but do not share in the initial costs.
In the case of a generic model, the entire model is applied to more than one instance.
The hope of a generic model is that after some number of applications, the cost of
creating, modifying and applying the generic model is less than the cost of a specific
model. For the purposes of the analysis here, four costs are considered:
• The cost to create the initial model, whether generic or specific (C)
Chapter 4. The Proposed General Model 76
• Adoption Costs:
– The cost to input the data (I)
– The cost to validate and debug the model (V )
– The cost to add functionality after the initial model creation (F )
TCGn Total cost of the (n)th application of the generic model
TCSn Total cost of the (n)th application of the specific model
CG Cost to build the initial generic model
CSn Cost to build a specific model for the (n)th application
Ijn Cost to input data into model type j for the (n)th application of the generic model
V jn Cost to validate model type j for the (n)th application of the generic model
F jn Cost to add functionality after original build for model type j for the (n + 1)th
application of the generic model
LGn Cost to buy license of existing generic model
j Model type =
G generic model
S specific model
n The number of times the model has been applied
Figure 4.5: Definitions of variables for cost analysis.
To begin, assume that the cost to develop the initial generic model is shared by all
instances of the application of the generic model. The total cost of applying the generic
model n times is the sum of the costs of adoption for each instance, plus the cost of the
initial build, see equation 4.1. Figure 4.5 defines all variables used in the cost analysis
within this section.
TCG(n) = CG +n∑
i=1
(IGi + V G
i + FGi
)(4.1)
Chapter 4. The Proposed General Model 77
The cost of creating a specific model for n instances would simply be the sum of the
costs of the initial model build, plus any other costs of adoption for each instance, as
demonstrated in equation 4.2.
TCS(n) =n∑
i=1
(CS
i + ISi + V S
i + F Si
)(4.2)
In order to justify the cost of a generic model, after some number of applications k,
its total cost should be less than the cost of creating specific models for each application
(equation 4.3).
TCG(n = k) < TCS(n = k)
CG +k∑
i=1
(IGi + V G
i + FGi
)<
k∑i=1
(CS
i + ISi + V S
i + F Si
)(4.3)
For simplicity, assume that for each instance, the costs are the same for the particular
type of model. In other words, the costs to input data, validate and add functionality
are the same for each instance the generic model is applied, and the costs to build, input,
validate and add functionality for each specific model are the same (equations 4.4.
CS1 = CS
2 = . . . = CSn = CS (4.4)
IS1 = IS
2 = . . . = ISn = IS (4.5)
V S1 = V S
2 = . . . = V Sn = V S (4.6)
F S1 = F S
2 = . . . = F Sn = F S (4.7)
Equations 4.1 and 4.2 simplify to equations 4.8 and 4.9. The cost savings of applying
the generic model comes after k applications, when the average cost of applying the
generic model is less than the cost of creating a specific model for the kth instance
(equation 4.10). This result is similar to that of model reuse in Robinson et al. (2004),
except for application of a generic model, in its entirety.
Chapter 4. The Proposed General Model 78
TCG(n) = CG + n(IG + V G + FG
)(4.8)
TCS(n) = n(CS
n + IS + V S + F S)
(4.9)
CG
k+(IGk + V G
k + FGk
)< CS + IS + V S + F S
TCGk < TCS
k (4.10)
In the case of this particular generic model, the initial conceptual design and build
was funded through a research grant. As a result, the initial costs of building the model
CG is not shared by all instances of application. The initial test hospitals, Juravinski,
St. Mike’s, Mt. Sinai, and William Osler contributed to the initial build costs. However,
Prince Albert, and any future tests sites, pay for the application of the generic model to
their site (IGi , V G
i , FGi . In addition, a consultancy type fee may be charged for future
applications: LGi . As a result, the total cost for application to an additional site can be
defined as 4.11.
TCG(c)n = LG
n +(IGn + V G
n + FGn
)(4.11)
Assuming the costs to input data, validate and add functionality are the same in this
instance as prior analysis, then the use of the generic model becomes economical when
for any instance n, the cost of purchasing and applying the generic model is less than
custom building a specific model (equation 4.12).
TCG(c)n < TCS
n (4.12)
Chapter 4. The Proposed General Model 79
4.3.1 Cost analysis at six hospitals
In an attempt to demonstrate the value of using a generic model in terms of cost, the
following section quantifies the costs of designing and using the model at the first six
hospitals in terms of time required (hours). Time is used as a proxy for cost as none of
the hospitals paid for hours worked, but rather provided research stipends or fixed fees
for the research.
Moreover, the actual cost of creating and applying the generic model include more
than time spent. Depending on the situation, costs could also include the cost of pur-
chasing any required software, such as Simul8, the cost of training one or more people
on the software chosen and simulation, consultant firm fees, etc.
The breakdown of the cost (in terms of hours of time) is provided in table 4.1. The
cost of building and using the specific model at Juravinski’s orthopaedic service was 694
hours, which is more than 200 hours greater than the cost to implement the generic
model at any of the six hospitals when the initial cost of creating the generic simulation
model is shared equally.
Assuming that the cost of creating the specific model of orthopaedics at Juravinski
would be about the same as creating a specific model at any of the other instances, the
analysis shows that the cost to apply the generic model is less than the cost of a specific
model. This applies both when the cost to create the generic model was shared among
sites and when a site paid a licensing fee for the model.
A specific model was not built for each individual application. Since each specific
model would have been created by the same person, the cost to build each model would
decrease at each application as code reuse would have occurred. Whereas, if each individ-
ual hospital would have independently created their own specific model, the opportunity
for reuse would not have arisen. Thus, a fair comparison of specific versus generic would
have been possible if these conditions were mimicked using multiple model developers.
Note that the time spent for data input tasks decreased over the course of the study
Chapter 4. The Proposed General Model 80
Tab
le4.
1:C
osts
(in
term
sof
tim
esp
ent)
ofth
esp
ecifi
cm
odel
applied
toth
eJura
vin
ski
orth
opae
dic
serv
ice
and
the
gener
ic
model
applied
tose
ven
inst
ance
sat
six
hos
pit
alsi
tes.
Sp
ecifi
c
Model
Gen
eric
Model
Cos
tJura
vin
ski
orth
o
Bra
mpto
n/
Eto
bic
oke1
Jura
vin
ski
all
serv
ices
St.
Mik
e’s
Mt.
Sin
aiJura
vin
ski
orth
o
Pri
nce
Alb
ert
To
crea
te:
Ci n
174
324
0
Lic
ense
Fee
:L
G(c
)n
n/a
n/a
0
To
input
dat
a:I
i n15
060
7540
4010
2
2103
To
Val
idat
e:V
i n25
019
080
9035
15
To
add
funct
ional
ity:
Fi n
120
510
00
00
Tot
alC
ost4
:T
Ci n
694
409
209
184
129
7921
0
1W
ork
atB
ram
pto
nan
dE
tobic
oke
was
don
esi
mult
aneo
usl
y,th
us
much
ofth
ew
ork
and
cost
sco
uld
not
be
dis
tingu
ished
by
site
.
2N
ote
much
ofth
edat
acl
eanin
gan
dw
ork
was
per
form
edth
rough
the
spec
ific
model
wor
k,
and
was
only
man
ipula
ted
for
the
purp
ose
ofth
ege
ner
icm
odel
.
3A
lot
ofbac
kan
dfo
rth
bet
wee
nad
just
ing
inputs
,ca
libra
tion
and
gain
ing
cred
ibilit
yof
the
dec
isio
nm
aker
sw
asre
quir
ed,
whic
hm
ade
itdiffi
cult
tose
par
ate
tim
esp
ent
toin
put
dat
ave
rsus
validat
ing
the
model
.
4A
ssum
esco
stto
crea
tege
ner
icm
odel
was
shar
edev
enly
bet
wee
nth
efirs
tsi
xap
plica
tion
s.
Chapter 4. The Proposed General Model 81
across the sites. This was due to a learning curve effect from a single person performed
the data analysis, input and generic model validation for multiple sites. With each
implementation, familiarity with hospital data and the model increased, reducing the
time required to complete tasks.
This effect demonstrates additional value of a generic model in two ways. First, reuse
of the generic model can reduce turnaround time for results. As familiarity increases, the
time to implement, validate and test decreases. This benefits the hospitals as they often
require a solution in a short time period and do not want to wait many months for an
answer.
Secondly, it is assumed that since the generic model has been tested many times,
it is likely more robust and bug free than a specific model. As a result, when model
issues arise during future implementations, such as the model producing error messages,
crashing, etc., it is more likely due to an error in the inputs rather than of the model
itself.
These added benefits would be most pronounced if the generic model is provided to
hospitals as part of a consultancy package, or as a standalone, highly automated software
package. The setup, validation, testing and perhaps the first round of decisions would be
performed by an outside group who are experts in the generic model. Once set up, a user
friendly version of the model would remain at the disposal of hospital decision makers to
use to help inform future decisions.
In conclusion, the economic model demonstrates that the development and spread of
this proposed generic model is worthwhile and cost effective.
Chapter 4. The Proposed General Model 82
4.4 Proposed Guidelines for Designing Generic Mod-
els
The purpose of this section is to compile the lessons learned through this research experi-
ence into a set of suggested guidelines for designing generic models. These guidelines can
be used in future generic modelling research projects to help reduce the time required to
design a generic model, and to improve upon the genericity and usability of the model.
The current generic modelling research space is relatively small, particularly in the area
of design methodology. This portion of the research adds to this field by providing some
additional guidelines for designing a generic model.
4.4.1 Clearly Define the Problem Description, Scope, Objec-
tives, etc.
In simulation modelling it is important to begin with understanding the problem, and
defining the scope and objectives of the study (Robinson et al., 2010). When designing
a generic model, this step is a vital step to ensure the successful application to multiple
sites. Having a very clearly laid out set of objectives, scope and problem description
will prove invaluable when making design, level of detail, simplification and assumption
decisions (Steele et al., 2002, e.g.). Furthermore, it will help when applying the model
to different sites as it will help set the scene, manage expectations, and limit the need
for customising or changing the generic model to fit the specific idiosyncrasies of the
particular site.
Upon reflection of this research work, more time and effort should have been spent
defining the problem description, scope and objectives prior to design. Additional time
and effort spent prior to design would likely have saved a significant amount of time
revising and adjusting the conceptual model during the designing phase. In addition,
it would have proved valuable during application at multiple sites when challenges with
Chapter 4. The Proposed General Model 83
achieving validity were faced or when clients made customisation requests.
4.4.2 Study Multiple Sites Prior to Design
When designing a generic model it is obviously essential to have a good understanding
how processes can vary across sites. It is suggested here that in order to ensure the
genericity of the model, numerous sites should be studied prior to design. Moreover,
the sites should represent a wide variety of the types of sites where the model could be
applied. Gaining knowledge and experience of the processes and procedures at a number
of sites helps to ensure that most instances of a process are included in the model.
In addition, if information is available regarding best practices or popular improve-
ments ideas, they should be noted and included in the model.
When studying multiple sites, best practice and the other sources, take some time
to compare processes and look critically at differences that exist. Determine if these
differences are common processes that you are likely to see at other future sites, or if
they are a single case of one site. If they are of the later, you will need to determine it
is a practice that should be included in the generic model, or if it is a poor or inefficient
practice that should be discontinued, or something that should not be considered when
planning and making decisions. For example, William Osler was struggling with starting
their elective OR blocks on time each morning. Though this is somewhat common at
many hospitals, since the goal of the generic model was for tactical level decisions, it was
assumed that they will always start on time. This was appropriate, as at the tactical
level, one should not plan for inefficient practices such as late starts.
4.4.3 Keep it Simple
“Everything should be as simple as possible, but not simpler” Albert
Einstein
Chapter 4. The Proposed General Model 84
As stated in many guidelines on building simulation models, it is best to keep the
model as simple as possible (Pidd, 1999; Robinson, 2007a; Kotiadis and Robinson, 2008,
e.g.). When designing a generic model, simplifications and assumptions play an even
more important role. It s important to document any assumptions and simplifications
made, including the reasoning. In generic modelling, simplifications and assumptions
are necessary to build the model because not every possible instance is known, nor is
it feasible to include a large number of variations. Furthermore, it is important to user
acceptance of the model that decision makers understand where and why simplifications
and assumptions were made, particularly when they are different than their current
processes.
Based on the experience gained in designing a generic model, it is suggested to start
by designing the simplest model possible; adding details only when needed to fulfil the
requirements set out in the problem description, scope and objectives. It is at this stage
where the preliminary work on the model’s objectives, scope and problem description
pays off by guiding modelling decisions. When in doubt, ask whether excluding this
detail would change the decision? If the answer is no, then that detail can be excluded.
Many authors, such as those referred to above, offer useful tips on how to make and justify
simplification and assumption decisions. For example, in the generic model presented, it
was decided early on to exclude the day of surgery activities and resource activity that
occur before the surgery itself, including the same day surgery unit, waiting areas, etc.
“It is not necessary to have a one-to-one mapping between the model and
the real system. Only the essence of the real system is needed.” (Banks and
Carson (1995) as cited in (Lowery, 1998, p. 33))
This rule also applies when applying the model to a site. The nature of generic
modelling entails that not all specific processes and procedures of a site will match the
options available in the model. This can lead to challenges in achieving user acceptance
Chapter 4. The Proposed General Model 85
and statistical validation (discussed in the next chapter). It is important that when im-
plementing the model at a site, to continuously keep in mind the generic model’s intended
scope and objective when considering any differences between the model’s processes and
those found at the site. Before considering changing or adding to the generic model to
appease stakeholders, consider whether it is within the scope and objectives, whether it
would affect any decision outcomes.
4.4.4 Collaborate with End Users
Though in generic modelling, you will never be able to work with every decision maker
during the design phase, it is recommended to work in close collaboration with a number
of decision makers from various initial sites, or if a variety of sites are not available, other
subject matter experts. Allow the decision makers to review the model’s problem de-
scription, scope, objectives, model design, and simplification and assumptions decisions,
and provide valuable insights and feedback.
Chapter 5
Generic Model Validation
Achieving statistical validity in simulation of complex real-world systems is challenging;
convincing end-users that model outputs are reasonable and can be trusted when making
decisions is even more difficult. This section contributes to the large, complex system,
real world model validation literature by providing guidance on how to convince end-
users that results are reasonable and trustworthy, even when statistical prediction error
remains.
In order to demonstrate this contribution, this chapter is composed of three sections.
First, the validation procedures carried out at the test sites are described along with
outcomes achieved. Second, in order to show that use of a generic model did not sig-
nificantly contribute to the validation challenges encountered, the validation efforts are
compared to those of a specific model. The final section provides procedures and tools
recommended for validation of complex, large, real world systems.
5.1 Validation at the Test Sites
The purpose of this section is to describe the steps required to validate the proposed
generic model at the six test sites. The challenges encountered when validating a simula-
tion model of a large, real world, complex system are described. Particular focus is given
86
Chapter 5. Generic Model Validation 87
to reconciling model results to historical values when statistical validation is not achiev-
able, and how to convince decision makers to accept the model’s outputs as acceptable
and reliable for decision making.
Validation of real world simulation models is often considered “an art and a science”
(Martis, 2006) because it can be a long, challenging journey. Classical statistical methods,
including confidence intervals and hypothesis tests, often fail since the model is only
an approximation of the real world, and since these statistical methods assume output
measures are independent and identically distributed (IID), which is often not the case
(Law, 2005).
As a result, (Law, 2005, p. 29) states “it is more useful to ask whether or not
the differences between the model and the system are significant enough to affect any
conclusions derived from the model.” In order for any simulation model, and in particular
a generic simulation model, to be trusted and used for decision making, the end users
must believe that the differences will not affect their decisions. In order to achieve this,
validation must be performed in both a subjective and objective frame. The subjective
frame focuses largely on the use of confidence intervals and prediction error to compare
simulation output results to real results. The objective frame uses face validity, model
behaviour and subject matter experts to determine if the model’s outputs are acceptable.
Each application of the generic model has been validated objectively compared to
historical performance. For those sites with engaged decision makers and stakeholders,
the generic model was also subjectively validated, based on face validity and model
behaviour. The six sites are listed below; those where subjective validity testing occurred
are denoted with an asterisk (*).
1. Hamilton Health Sciences - Juravinski Hospital - all surgical services *
2. St. Michael’s Hospital
3. Mount Sinai Hospital - General surgical service
Chapter 5. Generic Model Validation 88
4. William Osler Health Sciences - Brampton Civic Hospital *
5. William Osler Health Sciences - Etobicoke General Hospital *
6. Prince Albert Hospital *
At each site, models were validated on throughput by case type and service and
cancellation rates by reason.
In the objective validity tests, each individual measure was validated against a 95%
confidence interval of the average simulation output over ten replications. This was
chosen because, in initial validation of the pilot specific model, the working group felt
that an error of ±50 patients on the total yearly orthopaedic throughput was acceptable.
The historical throughput was 1442, so an error or ±50 patients is only a 3.5% error
rate. The specific model only required ten runs to be within this error limit set. A
lower number of runs was desired as the specific model had a lengthy run time: ten
runs of one year duration of the orthopaedic service took on average of 40 minutes. The
improved design and coding of the generic model resulted in shorter run times; the all
services generic model of Juravinski run for a year had an average run time around 25
minutes. As a result, the number of runs could have easily been increased; however, total
throughput results from each site produced confidence interval widths of less than 50
cases, an error rate that decision makers at each site were comfortable with. So for the
purposes of comparing results across all models, the number of runs was kept at ten.
Objective validation efforts performed at three of the six hospitals was relatively easy
and straight forward. The Juravinski, St Mike’s and Mt. Sinai are all high volume
teaching hospitals. A summary of the results from these hospitals are shown in tables
5.1, 5.2, 5.3 and 5.4. However, even at these high volume sites, the validation results
presented include some noteworthy discrepancies where the 95% confidence interval of a
measure from the model’s output does not include the historical value. For example, at
Juravinski, the model’s 95% confidence interval for average throughput for both gynaecol-
Chapter 5. Generic Model Validation 89
ogy/obstetrics (OBGY) and plastic surgery services fall above the historical throughput.
Simultaneously, the number of cases cancelled due to not enough elective time available,
no ward bed and bumped for a more urgent case are also higher than the historical value
(table 5.1). An analysis of the historical data did not reveal significant variation from
the inputs that could explain the difference. Within the three over predicted cancelation
rates, two (not enough time and bumped for a more urgent case) were not significantly
higher than the historical values. Cancellations due to no ward bed however had a much
larger error difference. It was known that there were some patient flow issues at this site,
where medicine patients were regularly off serviced in to the surgical wards. This causes
a high variability in the number of surgical ward beds available for surgical patients. To
help avoid many more cancellations due to no ward beds, the hospital chose to reduce
some off service admissions and open over capacity surgical beds. These operational de-
cisions are not reflected in the generic model. Thus, since the generic model produced
results that were within range or above historical throughput volumes, while simulta-
neously reporting higher than historical cancellation results, it was determined that the
model was accurate.
At the other hospitals (Brampton, Etobicoke and Prince Albert), objective valida-
tion proved to be more challenging. The main challenge was significant differences in
throughput numbers from the simulation model compared to historical, which affects
other measures such as cancellation numbers, OR utilisation and unit census. In some
cases the model would significantly over predict throughput for one or more services,
while under predicting others. Final validated results are provided in tables 5.5 to 5.9
for Brampton, Etobicoke and Prince Albert.
Demonstrating accuracy of large, complex, real systems, such as the perioperative
patient flow, is bound to be challenging if not impossible (Blake et al., 1995). The
experience and results of validating this generic model is no exception. As shown above,
full validation, in the statistical sense, was not achieved at any of the sites. Model results
Chapter 5. Generic Model Validation 90
MeasureHistorical
Value
Model
Result
+/- 95%
CIError
Throughput by Patient Type
Elective 443 453.2 6.8 10.2
Urgent 126 125.0 9.3 -1.0
Throughput by Service
GENL 161 162.0 3.9 1.0
OBGY 38 44.3 4.0 6.3
ORTH 251 254.7 7.9 3.7
PLAS 12 15.8 1.3 3.8
UROL 107 101.4 7.8 -5.6
Cancellations by Reason
No ICU Bed Cancellations 0.0 1.0 1.4 1.0
No Ward Bed Cancellations 1.0 11.4 8.6 10.4
Not Enough OR Time Cancel-
lations10.0 13.2 1.8 3.2
Bumped for More Urgent Case
Cancellations1.0 2.6 0.5 1.6
Other Cancellations 12.0 13.0 2.7 1.0
Table 5.1: Validation results from the Juravinski Hospital’s all services generic model
application. Bold type indicated when the confidence interval from the model results
does not include the historical value.
Chapter 5. Generic Model Validation 91
MeasureHistorical
Value
Model
Result
+/- 95%
CIError
Throughput by Patient Type
All Elective 1763 1748 15.4 -15
All Emergent 625 634.8 16.3 9.8
Throughput by Service
CAR 57 60.1 5.8 3.1
CVS 203 201.1 5.2 -1.9
ENT 146 149 3.2 3
EYE 389 394.5 6.6 5.5
GYN 245 241.3 6.2 -3.7
NS 240 233.8 8.8 -6.2
ORT 495 493.7 8.9 -1.3
PLA 106 102.1 6 -3.9
SUR 331 322.9 9.4 -8.1
URO 72 77.8 4 5.8
VAS 104 106.5 4.7 2.5
Table 5.2: Validation results from St. Mike’s Hospital all services generic model appli-
cation - Throughput. Bold type indicated when the confidence interval from the model
results does not include the historical value.
Chapter 5. Generic Model Validation 92
MeasureHistorical
Value
Model
Result
+/- 95%
CIError
Cancellations by Reason
No ICU Bed Cancellations 1 0.6 1 -0.4
No SDU Bed Cancellations 0 0 0 0
No Ward Bed Cancellations 7 7 6.2 0
Not Enough OR Time Cancel-
lations27 63.9 5.3 36.9
Bumped for More Urgent Case
Cancellations40 36.4 6.7 -3.6
Other Cancellations 240 234.7 13.4 -5.3
Table 5.3: Validation results from St. Mike’s Hospital all services generic model appli-
cation - Cancellations. Bold type indicated when the confidence interval from the model
results does not include the historical value.
Chapter 5. Generic Model Validation 93
MeasureHistorical
Value
Model
Result
+/- 95%
CIError
Throughput by Patient Type
Elective 388 380.3 8.3 -7.7
Urgent 85 85.0 8.4 0.0
Throughput by Service
GENL 473 465.3 11.1 -7.7
Cancellations by Reason
No ICU Bed Cancellations 1.0 0.0 0.0 -1.0
No SDU Bed Cancellations 0.0 0.0 0.0 0.0
No Ward Bed Cancellations 0.0 0.0 0.0 0.0
Not Enough OR Time Cancel-
lations28.0 27.9 4.3 -0.1
Bumped for More Urgent Case
Cancellations5.0 1.3 1.1 -3.7
Other Cancellations 30.0 31.3 4.0 1.3
Table 5.4: Validation results from Mt. Sinai Hospital general surgical service generic
model application. Bold type indicated when the confidence interval from the model
results does not include the historical value.
Chapter 5. Generic Model Validation 94
MeasureHistorical
Value
Model
Result
+/- 95%
CIError
Throughput by Service
General Surgery 412.1 407.6 13.4 -4.4
Obstetrics and Gynaecology 278.0 272.9 8.2 -5.1
Ophthalmology 285.9 299.6 17.4 13.7
Oral Maxillofacial/Dental 25.7 35.6 5.3 10.0
Orthopaedic Surgery 271.8 320.4 8.9 48.6
Otolaryngology 106.8 124.7 3.9 17.9
Urology 119.1 119.9 4.7 0.8
Cancellations by Reason
No ICU Bed Cancellations N/A 0.0 0.0 N/A
No Ward Bed Cancellations N/A 0.0 0.0 N/A
Not Enough OR Time Cancel-
lationsN/A 37.6 3.7 N/A
Bumped for More Urgent Case
CancellationsN/A 11.6 2.7 N/A
Other Cancellations N/A 0.0 0.0 N/A
Table 5.5: Validation results from Prince Albert Hospital generic model application. Bold
type indicated when the confidence interval from the model results does not include the
historical value.
Chapter 5. Generic Model Validation 95
were within the 95% confidence interval range for many of the output measures at most
sites, especially in the first three (Juravinski, Mt. Sinai and St. Mike’s). Despite the
gaps in statistical validity, there is confidence that the generic model is accurate and is
acceptable as a decision tool.
Tactical planning and decisions are made in consideration of the policies and proce-
dures in place or to be amended. Decisions should not be made based on the assumption
that these existing policies and procedures will not be adhered to. For instance, it is
considered inefficient to operate with a low elective OR utilisation rate. However, if a
hospital does not ensure that time is fully and effectively booked, significant under util-
isation will result. Under booking could occur for a number of reasons such as under
filled OR blocks, over estimation of procedure length, and not reassigning part or full
blocks released by a surgeons. It is not practical to make tactical planning decisions such
as the MSS or resource availability assuming under utilisation of an expensive resource.
Tactical decisions should be made on the basis of how processes “should work”; then
operationally, decisions may need to be made to accommodate deviations from proce-
dures and practices. As a result of this notion simulation model results are affected as
the design does not include these poor practices. This does not make the model inac-
curate or meaningless. It is argued here that this property in fact makes the tactical
decision model more powerful as it demonstrates the outcomes of having good practices
and decisions in place, the ideal conditions in which to base tactical decisions.
It is here where an important contribution to the field of validating large, real world
complex system simulation models and specifically generic models, is proposed: how can
one achieve user acceptance when statistical validation is not possible, such that decision
makers are willing to base important decisions on model results. The remainder of this
section will review how user acceptance was achieved at Brampton where initial attempts
at statistical validity met with significant difficulty.
Chapter 5. Generic Model Validation 96
MeasureHistorical
Value
Model
Result
+/- 95%
CIError
Throughput by Patient Type
Elective 2304 2387.1 50.6 83.1
Urgent 447 492.2 14.4 45.2
Throughput by Service
Dentistry 73 70.9 2.5 -2.1
Ear Nose and Throat 223 220 8.9 -3
General 546 595.8 13.1 49.8
Gynaecology 240 246.7 10.3 6.7
Ophthalmology 877 922.1 10.3 45.1
Oral Surgery 40 45 3.9 5
Orthopaedics 428 439.7 11.9 11.7
Plastics 70 75.5 5.1 5.5
Urology 254 274.7 16.8 20.7
Table 5.6: Validation results from Brampton Hospital generic model application -
Throughput. Bold type indicated when the confidence interval from the model results
does not include the historical value.
Chapter 5. Generic Model Validation 97
MeasureHistorical
Value
Model
Result
+/- 95%
CIError
Cancellations by Reason
No ICU Bed Cancellations 0.58 0.5 1.131 -0.08
No SDU Bed Cancellations 0 7.1 8.17 7.1
No Ward Bed Cancellations 0.87 117.4 58.45 116.53
Not Enough OR Time Cancel-
lations23.08 173.8 11.73 150.72
Bumped for More Urgent Case
Cancellations4.62 15.8 3.32 11.18
Other Cancellations 498.75 282.1 12.77 -216.65
Table 5.7: Validation results from Brampton Hospital generic model application - Can-
cellations. Bold type indicated when the confidence interval from the model results does
not include the historical value.
Chapter 5. Generic Model Validation 98
MeasureHistorical
Value
Model
Result
+/- 95%
CIError
Throughput by Patient Type
Elective 1036.0 1022.0 16.6 -14.0
Urgent 227.0 231.9 9.9 4.9
Throughput by Service
Ear Nose and Throat 170.0 162.5 5.6 -7.5
General 279.0 272.3 9.0 -6.7
Gynaecology 267.0 278.4 4.6 11.4
Oral Surgery 38.0 34.2 2.5 -3.8
Orthopaedics 326.0 334.5 16.2 8.5
Plastics 121.0 115.5 6.0 -5.5
Urology 67.0 86.5 6.1 19.5
Table 5.8: Validation results from Etobicoke Hospital generic model application -
Throughput. Bold type indicated when the confidence interval from the model results
does not include the historical value.
Chapter 5. Generic Model Validation 99
MeasureHistorical
Value
Model
Result
+/- 95%
CIError
Cancellations by Reason
No ICU Bed Cancellations 0.0 0.0 0.0 0.0
No SDU Bed Cancellations 0.0 0.0 0.0 0.0
No Ward Bed Cancellations 0.5 3.3 2.7 2.8
Not Enough OR Time Cancel-
lations3.2 17.3 2.7 14.1
Bumped for More Urgent Case
Cancellations0.5 7.6 2.8 7.1
Other Cancellations 223.4 175.1 8.7 -48.3
Table 5.9: Validation results from Etobicoke Hospital generic model application - Can-
cellations. Bold type indicated when the confidence interval from the model results does
not include the historical value.
Chapter 5. Generic Model Validation 100
5.1.1 Achieving Validity and User Acceptance at Brampton
Civic Hospital
A large working group was assembled at Brampton that included the director of pe-
rioperative services, various perioperative and inpatient unit managers, physician and
surgeon leaders and the consulting firm members. Initially, one calendar year of data
was provided for model analysis and implementation. Over the course of the data, three
different master surgical schedules (MSS) were used at Brampton: spring, summer and
fall. Additionally, some new surgeons joined and others left the hospital during this
time. Together, these factors resulted in significant variability over the year of data. In
consultation with the working group, it was determined that the fall schedule would be
used for model validation as it was most current.
The first validation attempt, used the fall MMS as planned. The model outputs (ta-
bles 5.10 and 5.11) contained significant prediction error: the initial schedule significantly
over predicted elective volumes for all services, while under predicting urgent volumes.
These results were brought to the working group for discussion where it was revealed
that the MSS schedule realised is not always reflective of the month’s planned MSS. For
each month, a monthly MSS was produced that took into account holidays and other OR
closures, and any planned days released by surgeons who gave up time due to vacation or
other reasons, and those surgeons who picked up the open slots. The resulting monthly
MSS schedules included a significant number of ORs that were closed because no surgeon
picked up the released time of other surgeons, reducing the total number of elective OR
hours.
As such, the schedule of the month of November was selected by the working group as
a month representative of the fall schedule. When the November schedule was tested in
the model, discrepancies still remained but throughput for some services were predicted at
rates close to the historical values. However, other services’ throughput were predicted at
Chapter 5. Generic Model Validation 101
MeasureHistorical
Value
Model
Result
+/- 95%
CIError
Throughput by Patient Type
Elective 4055.0 4007.1 47.6 -47.9
Urgent 940.0 844.7 19.5 -95.3
Throughput by Service
Dentistry 147.0 147.7 6.0 0.7
Ear Nose and Throat 416.0 447.1 8.6 31.1
General 1051.0 1038.4 19.2 -12.6
Gynaecology 444.0 410.2 20.0 -33.8
Ophthalmology 1430.0 1429.0 31.9 -1.0
Oral Surgery 65.0 66.8 5.3 1.8
Orthopaedics 768.0 638.8 9.0 -129.2
Plastics 182.0 181.1 4.6 -0.9
Urology 519.0 517.2 7.5 -1.8
Table 5.10: First attempt at validating at Brampton Hospital using the Fall 2009 MSS
schedule as planned. Bold type indicated when the confidence interval from the model
results does not include the historical value.
Chapter 5. Generic Model Validation 102
MeasureHistorical
Value
Model
Result
+/- 95%
CIError
Cancellations by Reason
No ICU Bed Cancellations 0.6 0.0 0.0 -0.6
No SDU Bed Cancellations 0.0 0.9 0.6 0.9
No Ward Bed Cancellations 0.9 6.3 5.4 5.4
Not Enough OR Time Cancel-
lations23.1 12.2 4.0 -10.9
Bumped for More Urgent Case
Cancellations4.6 3.5 1.3 -1.1
Other Cancellations 498.8 424.2 18.2 -74.6
Table 5.11: First attempt at validating at Brampton Hospital using the Fall 2009 MSS
schedule as planned - Cancellations. Bold type indicated when the confidence interval
from the model results does not include the historical value.
Chapter 5. Generic Model Validation 103
levels significantly higher or lower than historical. Further analysis of the data and model
results as well as consultation with the working group were performed. This yielded a
number of observations of the current system/process that when inputted improved model
results.
The input changes performed included adjustments to the time booked for surgery,
the composition of the patient input file and the MSS. The data analysis, working group
insight and result model changes are provided.
Booked Time Adjustment: Based on an observation of working group members that
over- and under-booking practices were present and significant, surgical data was analysed
to examine any trends. A report (figure 5.1) was compiled comparing the amount of time
booked to the amount of time used by the surgeon.
It was found that some surgeons and some services consistently under book the time
required for an elective case; allowing surgeons to book more cases in a day. Whereas
in actuality, surgeons were often allowed to run into overtime, despite hospital rules.
The model is not as flexible and forgiving in regards to allowing elective cases to run
into overtime, thus reducing throughput because of a high a number of overtime can-
cellations. The data demonstrated that the under booking was mostly due to surgeons
not accounting for the turnaround time required between cases. To account for this in
the model when scheduling these services and surgeons, the average turnover time was
added to the booked time when scheduling patients to better fill the day, reducing the
cancellation rate. This of course does not help with the fact that the hospital may allow
overtime for surgeons with this practice, hence throughput for some services was still
low, though the cancellation rate was more accurate.
Patient File Adjustments: As described in the previous chapter, the model is sensitive
to the proportion of patients in the patient file compared to validation numbers. The
working group noted that over the course of the year of historical data, the case loads
of surgeons changed significantly due to surgeons joining the hospital, and/or increasing
Chapter 5. Generic Model Validation 104
Figure 5.1: Example of a report provided for validation analysis comparing the amount
of time booked by surgeon versus the amount of time actually used during elective block
time.
Chapter 5. Generic Model Validation 105
their practices, while others were reducing their practices. This can be quantitatively seen
in figure 5.2. The report includes the proportion of patients belonging to each surgeon
within the input patient file used by the generic model to generate patient characteristics;
and the proportion of throughput belonging to each surgeon during the time frame for
validation. The final column highlights the difference in these two proportions. The
patient records were increased for surgeons whose caseloads were higher in the validation
period of fall 2009 than in the historical patient file to be more representative of current
mix.
Further MSS adjustments: Even after the above adjustments, throughput numbers
were still not as close as desired. Analysis showed that even the November MSS was
not entirely reflective of the actual schedule. Moreover, elective OR time was frequently
released or exchanged after the month’s MSS was published, resulting in further changes
to the schedule, including reduction in the elective OR time used. The inputted MSS
was modified to reflect these differences.
In addition, it came to light that some orthopaedic OR blocks were being run as swing
rooms, where the same surgeon and surgical team would have access to two ORs. When
one case was complete in the first OR the team would move immediately to the other
OR and perform another surgery, instead of waiting for the first OR to turnover. These
swing rooms were used to complete total joint replacements. Using a single OR, a high
total joint case surgeon could perform four procedures; using a swing room, the same
surgeon could perform two additional cases. The planned MSS did not clearly indicate
that swing rooms were in use, so were not included in the first round of validation.
After incorporating these changes to the model, the validation results produced
showed improvement on prediction error, as demonstrated in tables 5.12 and 5.13.
At this point, the working group felt that the model was a reasonably good repre-
sentation of their practices, at least in terms of the model design and assumptions. Un-
fortunately, statistical validation was still not achieved on most measures. They noted
Chapter 5. Generic Model Validation 106
Figure 5.2: Example of a report provided for validation analysis comparing the percentage
of patients by surgeon historically and within the input patient file.
Chapter 5. Generic Model Validation 107
that in the last five or so months, many of the issues around the MSS and scheduled
time had improved. Additionally, further changes to the surgeons currently practising at
Brampton, and their proportion of cases within each service had changed. Consequently,
the working group felt that better validity and credibility would be gained if more current
schedules and data were used. The most recent two months (April and May 2010) of OR
data was pulled for another round of validation.
MeasureHistorical
Value
Model
Result
+/- 95%
CIError
Throughput by Patient Type
Elective 4055.0 4394.1 23.6 339.1
Urgent 940.0 923.2 9.5 -16.8
Throughput by Service
Dentistry 147.0 144.5 1.5 -2.5
Ear Nose and Throat 416.0 420.2 4.9 4.2
General 1051.0 1078.4 9.5 27.4
Gynaecology 444.0 483.0 5.3 39.0
Ophthalmology 1430.0 1671.1 12.1 241.1
Oral Surgery 65.0 67.2 1.7 2.2
Orthopaedics 768.0 734.1 9.4 -33.9
Plastics 182.0 218.4 2.2 36.4
Urology 519.0 522.4 4.6 3.4
Table 5.12: Second round at validating at Brampton Hospital using the November 2009
MSS schedule as a representative schedule for Fall 2009. Bold type indicated when the
confidence interval from the model results does not include the historical value.
In order to avoid some of the issues encountered with the MSS in previous validation
Chapter 5. Generic Model Validation 108
MeasureHistorical
Value
Model
Result
+/- 95%
CIError
Cancellations by Reason
No ICU Bed 0.6 0.7 0.4 0.1
No SDU Bed 0.0 11.9 5.0 11.9
No Ward Bed 0.9 57.7 25.4 56.8
Not Enough OR Time 23.1 300.7 13.4 277.6
Bumped for More Urgent Case 4.6 17.1 1.6 12.5
Other 498.8 487.2 7.7 -11.6
Table 5.13: Second round at validating at Brampton Hospital using the November 2009
MSS schedule as a representative schedule for Fall 2009 - Cancellations. Bold type indi-
cated when the confidence interval from the model results does not include the historical
value.
Chapter 5. Generic Model Validation 109
efforts, an eight week MSS was created based on the two new months of data. The
patient file remained the same - based on the original year of data provided. Using the
actual schedule as the MSS helped make initial results significantly closer to historical.
Based on data analysis and expert opinion from the working group only a few additional
minor adjustments were required to further calibrate the model:
• It was observed in the data that many services were running over their elective
blocks. To represent this, OR blocks were extended by 30 minutes for all services
except ophthalmology and plastics.
• Some surgeons were found to consistently under book their OR blocks. These
surgeons were assigned only half days in the model to reflect this trend.
These adjustments brought calibration to a reasonably close level. Overall, the up-
dated analysis on utilisation of elective blocks, including under and over booking, and
trends in booked time compared to actual time demonstrated that though there was some
improvement in these practices since the analysis done on Fall data, there is still room
for improvement. These areas of improvement were part of the overall plan that William
Osler and the consulting firm were implementing to improve perioperative performance.
As a result, the working group felt that the current model at Brampton was sufficiently
accurate and credible. Differences between the model’s results and historical data were
likely due to the variance in practice from what is generally considered best practice as
assumed in the model, and the current practices at William Osler. Therefore, the model
results based on the actual schedule for April-May 2010 as demonstrated in tables 5.6
and 5.7 represented what likely would have been done if practices at Brampton better
conformed to best practices. The working group felt confident that the model’s outputs
were now reflective of their hospital if the identified poor and inefficient practices were
not present at the time. This allowed the application of the model to progress as a
demonstrative and decision tool as outlined in Chapter 6.
Chapter 5. Generic Model Validation 110
5.2 Challenges of Achieving User Acceptance
An important feature when evaluating the value of a generic model is that the use of a
generic model does not induce significant additional challenges. Use of a generic model
should not require significantly more effort to set up, validate and use the model when
compared to a specific model.
This section will discuss the challenges encountered when applying the generic model
to six hospitals. Many of the challenges experienced implementing the generic model
were similarly encountered when a specific model was built for the initial test site, the
orthopaedics service at Juravinski. This will be demonstrated in the next section.
Master schedule vs. actual schedule - During validation at several sites it was dis-
covered that the planned master surgical schedule was not followed in actuality. While
any hospital will have some minor adjustments to the block schedule over time due to
planned and unplanned OR closures, changes to service or surgeon assignment, etc., sev-
eral hospitals’ actual block schedule differed significantly from planned. If the schedule
inputted in the model was based on the master schedule, and the actual schedule dif-
fered significantly, validation was not successful. In these cases, if the data was used to
generate a schedule based on the actual schedule instead of the master schedule, then
validation was more likely to be achievable.
Arrival rates and waiting list sizes - An important input to any simulation model is
the arrival rate. Unfortunately at the time of data collection, the hospitals studied did
not collect accurate or reliable information on the waiting lists for all surgeons and cases;
arrival rates and waiting list sizes were unavailable. As a result, initial waiting list sizes
and arrival rates were estimated.
Arrival rates were initially estimated based on the actual number of cases performed
during the time period studied. During validation, it was found that these estimates
would generally lead to throughput values being under predicted. This result is intuitive
as the initial estimated arrival rate was based on realised throughput, which ignores
Chapter 5. Generic Model Validation 111
cancellations and other factors that increase the actual arrival rate.
In order to achieve model results within an acceptable range of accuracy compared to
historical values, the arrival rates and initial waiting list sizes were increased incremen-
tally.
Surgeon booking practices - In reality, when surgeons book cases into their OR blocks,
they may adjust the booking time in order to fit cases within the length of the block. For
instance, a surgeon has two hours of time remaining, but the next patient on his waiting
list is usually booked for two and a half hours. If the surgeon feels he can likely speed up
the day’s cases in order to finish on time, he will book that patient for two hours. The
model, however, does not account for this behaviour. Thus it was found that in some
cases throughput and OR utilisation were lower than expected. In these cases, increasing
the OR blocks by 30 minutes resolved the issue.
Overtime allowance - Most hospitals encountered in this study allowed little tolerance
for running past the end of the scheduled block because of staff overtime costs. Usually,
less than 30 minutes of predicted overtime was tolerated by the hospitals studied. The
model has been set up to allow up to a fixed inputted number of minutes of overtime,
based on the remaining time in the block and the time booked for the case. What the
model does not include is the effect of a surgeon who is likely to speed up when faced
with possibly having a case cancelled due to overtime. Additionally, other factors are
often considered in allowing overtime that vary widely by hospital and are difficult to
model. For instance, some will allow overtime if the OR team is willing to stay. Another
example is that they will allow it if they can still do any urgent cases within the time
frame based on the number of ORs that remain open to handle urgent cases.
Since these effects are not considered in the model it sometimes becomes a challenge to
validate the number of cancellations due to overtime. When the rate was predicted by the
model it was significantly higher than actual instances, increasing the inputted allowable
overtime helped decrease the cancellation rate towards a more accurate prediction.
Chapter 5. Generic Model Validation 112
Patient file composition - Another challenge identified during validation at multiple
sites concerned the inputted patient file. If the proportion of patients in the file did
not closely match the proportion of patients seen during the time frame of validation,
the simulation results would not validate well with historical values. For example, at
Juravinski, patient level data for two years was available and inputted into the patient
file. Over the span of these two years, several surgeons joined the program. During the
chosen validation time period, these surgeons had built up their practices. The data
however did not include the current proportion of patients the surgeons see since it is
over a longer time period than the validation figures. This reduced the probability that
a patient of that surgeon would be selected on arrival, which in turn resulted in lower
than expected throughput for these surgeons.
As a demonstration, imagine that within two years of patient records, surgeon A joined
the hospital after the first year of data. When surgeon A first joined, it took six months
to build up his practice, during which time he only performed 10% of cases. In the last
six months of data surgeon A was performing 25% of all cases within his service. Prior to
surgeon A joining, the service’s cases were performed evenly between the other three sur-
geons. Thus, over the two years of data, the proportion of patients belonging to surgeon
A is only [14(6months2years)+ 1
10(6months/2years)+0(1year/2years)] ≤ 25% of patients.
There are two options to overcome this challenge. First, the patient file can be reduced
to only include patients from the time period studied. This ensures that the proportions
are closely related to the throughputs being validated. In the example, the patient file
would be reduced to only include patients operated on in the last six months of the
data pull; thus excluding the first year and a half of data available. This option is fairly
easy to implement, but reduces the variation of possible types of patients in the model.
This in turn may reduce the robustness and reality of the model for scenario testing and
continued use.
A second option is to manipulate the current patient file such that the proportions
Chapter 5. Generic Model Validation 113
better match the current mix. This can be done by duplicating patient records of the case
type that needs to be increased. In our example, one would copy the patients of surgeon
A in the patient file in order to increase the percentage of cases within his service. This
option does not reduce the variability of cases available overall.
Significant deviation from “best” practice - The final challenge faced with demon-
strating the accuracy and validity of the generic model was with hospitals that signifi-
cantly deviated from “best practice”. Hospitals that allowed blocks to regularly be under
booked, or that allowed surgeons to poorly estimate the time required to complete a case,
became especially challenging as the model assumes that best practices are generally ad-
hered to. Despite attempting various calibration techniques, including those described
here, the model was unable to predict historical throughput. Analysis of the data, and
interviews with decision makers and perioperative subject matter experts within the
hospital revealed that poor practices were allowed.
In these cases, a key value of the generic model is not only to be used as a decision tool
for hospitals to explore changes, but also as a demonstrative tool to identify these poor
practices, quantify their effects on outcome measures, and demonstrate what is possible
if these best practices were better followed. More on the value of the generic model as a
decision and as a demonstrative tool is reserved for Chapter 6.
5.3 Compared to a Specific Model
Prior to conceptualising the generic model, the orthopaedic surgical service at Juravinski
was modelled as a specific model; created from scratch, for the sole purpose of being used
for orthopaedic surgery at Juravinski and to gain preliminary insights into modelling
perioperative patient flow for tactical decisions. It is rare to find a generic model in
the literature that is compared to a specific model. In this section, a comparison of the
specific and generic models is made. In conjunction with an engaged working group at
Chapter 5. Generic Model Validation 114
this hospital, this specific model was validated and used a number of times as a decision
tool. The focus here is to demonstrate that a model created specifically for the hospital
studied can experience the same challenges as when applying a generic model. The
accuracy of the generic model is compared to the specific model in terms of the same
four facets: validity, credibility, utility and feasibility.
5.3.1 Validity of the Specific Model
In conjunction with the engaged working group at Juravinski, the process of validation
and achieving model credibility followed a typical reiterative approach. The process
began through a series of onsite interviews and real time observation of the current
processes. Based on the information gathered, detailed process map was created and
verified by the stake-holders. The process map subsequently informed the conceptual
design of the specific model. The model design was also agreed upon by the stake-
holders, including assumptions and simplification decisions. The specific model was then
coded. Any changes to the design made when building and validating the model were
also discussed and approved by the working group.
The validation process, also took an iterative approach by presenting current model
results on a regular basis in order to discuss any results whose 95th percentile confidence
interval did not include the historical value. Together, the team would discuss possible
reasons for any discrepancies, and suggest reasonable and appropriate changes to the
input data and assumptions. This iterative process towards designing, building and
calibrating the model resulted in high credibility of the specific model’s outputs when
conducting scenario tests. The final validated results to the historical data as approved
by the working group is provided in table 5.14.
The model was successfully implemented and used as an interim decision tool for the
orthopaedic service at Juravinski, further demonstrating its validity and credibility in
the eyes of the end users who trusted the model’s results for decision making.
Chapter 5. Generic Model Validation 115
5.3.2 Comparing the Specific and Generic Models
The intention of this section is to demonstrate that use of a generic model does not
necessarily mean one has to sacrifice validity and credibility.
In order to demonstrate that the generic model can produce similarly reliable results
as a specific model, the inputs of the validated specific Juravinski orthopaedic model
were inputted into the generic model. The generic model inputs were then calibrated to
achieve statistically validated results. A comparison of the outputs of the generic and
specific models based on the same inputs are provided in table 5.14.
With the same input values as the specific model, the generic model appears to pro-
duce higher throughput values compared to the specific model. In both applications, the
number of completed urgent cases is significantly higher than historical values. Valida-
tion of the specific model required increasing arrival rates and initial waiting list sizes
in order to achieve desired throughput values. Reducing these input values in the vali-
dated generic model increased the predictive accuracy of the generic model compared to
historical data. For example, the urgent arrival rate was reduced to achieve a confidence
interval range within the historical results. The calibrated generic model however still
results in a lower than historical range of other orthopaedic cases, as well as fewer can-
cellations. The ICU cancellation rate discrepancy is likely due to the fact that in both
models, the full ICU is not modelled as it also services medical patients. Thus, the num-
ber of available beds for surgery is not constant as modelled. A lower cancellation rate
due to more urgent cases is likely a result of modelling assumptions and simplifications.
Cancellations due to no ward bed is higher than expected, likely a result of the fact that
the real life decision to use over capacity beds and push out off-service patients is not
modelled.
Similarly to the successful implementation of the generic models at other sites, the
key factor to the credibility of the specific model is the high and consistent involvement
of decision makers during the entire life cycle of the study. The high involvement of the
Chapter 5. Generic Model Validation 116
Mea
sure
His
tori
cal
Val
ue
Sp
ecifi
cM
odel
Gen
eric
Model
Gen
eric
Model
Cal
ibra
ted
Thro
ugh
put
by
Pat
ient
Typ
e
Ele
ctiv
e14
421418.8±
11.2
1490.3±
12.9
1439
.5±
13.9
Urg
ent
82164.5±
7.8
157.6±
11.3
87.5±
6.4
Thro
ugh
put
by
Pro
cedure
Typ
e
Fra
cture
dhip
s31
66.6±
5.3
60±
7.5
31.1±
2.3
TJR
991
993.
5±13
.1909.1±
11.1
978.
9±
14.9
Oth
er54
5525.1±
13.3
709.8±
11.9
529.4±
15.5
Can
cellat
ions
by
Rea
son
No
ICU
Bed
40.2±
0.30
160±
00±
0
No
War
dB
ed3
3.5±
2.7
16.9±
614.2±
5.1
Not
Enou
ghO
RT
ime
1920
.2±
3.8
28.5±
3.9
21.9±
5
Bum
ped
for
Mor
eU
rgen
tC
ase
21±
0.7
0.3±
0.3
0±
0
Oth
er37
34.8±
4.1
37.8±
3.7
36.3±
3.1
Tab
le5.
14:
Com
par
ing
validat
ion
resu
lts
ofth
esp
ecifi
cm
odel
toth
ege
ner
icm
odel
for
the
Jura
vin
ski
Hos
pit
al’s
orth
opae
dic
serv
ice
applica
tion
.B
old
face
dty
pes
isuse
dto
indic
ate
when
the
his
tori
cal
valu
eis
outs
ide
ofth
eco
nfiden
cein
terv
alpro
duce
d
by
the
model
.
Chapter 5. Generic Model Validation 117
hospital’s decision makers leads to high credibility of the specific model and multiple
rounds of scenario testing. Evaluation of the scenario tests then leads to implementation
of suggested elective block schedules and scheduling rules over the course of three years.
A side benefit noted was that the confidence in the specific model increased the
confidence and credibility of the generic model when it was introduced later to model all
surgical services at the same hospital.
Overall, the generic model can be considered an accurate representation of the peri-
operative process, when compared to historical data and a specific model.
5.3.3 Differences in Model Design
It is important to note that there are some differences between the functionality and
design of the specific model to that of the generic. Figure 5.3 compares the two models
in terms of what is and is not included, and some key assumptions.
Based on the level of detail, scope and objectives chosen during the conceptual mod-
elling of the generic model, a number of patient holding areas were not included: patient
registration, same day surgery waiting area and patient rooms, and anaesthesia induction
rooms. However, the specific model included these areas with the exception of anaesthesia
induction rooms, which Juravinski does not have.
5.3.4 Challenges Compared to a Specific Model
Validating the specific model for the orthopaedic surgical service at Juravinski faced
several challenges, many similar in nature to those faced with validating the generic
model. Table 5.15 compares the challenges faced during validation of the generic and
specific models.
Similar to instances in the generic model, achieving historical volumes in the specific
model required increasing the arrival rates of patients in order to fill the wait lists with
a wide variety of patients to schedule.
Chapter 5. Generic Model Validation 118
Specific Model Generic Model
Pre-Operative Processes
Waiting List ! !
Organisation and Ordering ! !
Waiting time and length ! !
Scheduling ! !
Block and surgeon schedules ! !
Elective patient scheduling rules ! !
Urgent and Emergent patient scheduling rules ! !
Inpatient scheduling rules ! !
Priority/Urgency considerations ! !
Maximum wait time on list ! !
Reserved time for Fractured hips ! !
Sequencing ! !
Pre-Operative Clinic " "
Operative Processes
Registration ! "
Same Day Surgery Unit waiting area and rooms ! "
Anaesthesia Rooms n/a "
Pre-Operative Waiting Area ! "
Check for post-operative recovery bed ! !
Cancellation due to "other" reasons ! !
Cancellation due to overtime or urgent case bump ! !
Procedure ! !
Length of procedure ! !
OR block due to resource unavailable ! !
Turnaround time ! !
Post-Operative
Post-Anaesthesia Care Unit ! !
Nurse and Bed availability ! !
Intensive Care Unit ! !
Step Down Unit ! !
Wards ! !
Off-servicing due to bed unavailability ! !
Bed blocking due to bed unavailability ! !
Process
Figure 5.3: Comparing the specific and generic model in terms of areas included and key
assumptions.
Chapter 5. Generic Model Validation 119
Challenge Specific Generic
Volumes - increased arrival rates and initial waiting list num-
bers to achieve throughput volumes
Y Y
Patient File - Adjustment of proportion of cases to reflect
current reality/validation period
Y Y
Overtime Cancellation - increase fixed rule amount to reduce
model cancellation rate
Y Y
Surgeon booking practices - increase elective block time to
calibrate OR utilisation and throughput
Y Y
Master Schedule vs. actual schedule - ensure inputted sched-
ule reflect practice for validation purposes
Y Y
Deviation from “best practice” N/A Y
Decision Options - Unable to test options under consideration
without changes to model code
Y N
Table 5.15: Comparing the challenges faced when modelling perioperative services, using
the proposed generic model and to using a specific, custom made model.
Chapter 5. Generic Model Validation 120
The specific model also had similar issues in achieving throughput of total joint cases
because the patient file inputted originally had a lower proportion of total joint replace-
ment (TJR) cases compared to current case mix. The patient file was adjusted to reflect
current TJR rates by duplicating a random number of TJR cases to increase the propor-
tion, which yielded better results.
In summary, the specific model experienced similar challenges, demonstrating that
the challenges faced when using the generic model were not due specifically to the use of
a generic model, but rather the nature of the system under study and the data available.
There was one challenge experienced during the implementation of the specific model,
which was not experienced in the generic model. During scenario testing with the specific
model, it was discovered that some desired scenarios were not easily tested because of
hard coded information or processes. This meant that changes to the code itself were
required for some scenarios. The generic model was built on the principle of data-driven
simulation modelling, limiting the number of processes or data that are hard coded into
the model, thus reducing the likelihood that scenario tests will require changes to model
code or structure.
5.4 Guidelines for Achieving User Acceptance
As discussed in this chapter, achieving confidence in this generic model has been fraught
with challenges. Through the successful application of this generic model to numerous
hospitals, a better understanding of how to establish user acceptance, both in terms of the
accuracy of and confidence in the results, as been gained. Herein, steps have been outlined
to calibrate a generic model such that it provides reasonably accurate results and to
convince end users that the model can be trusted to support important tactical decisions.
The guidelines are presented such that they are applicable to simulation modelling in any
industry. They are geared towards generic models of tactical decisions, however can be
Chapter 5. Generic Model Validation 121
adapted for specific models.
The end goal of validation should be when the decision makers and users of the
simulation “accept the model as correct” (Law, 2005).
5.4.1 Step 1: Assemble an Engaged Working Group
As many modellers have noted, including Law (2005), having an engaged group of deci-
sion makers and users involved throughout the phases of a simulation project is key to
achieving model acceptance.
Working closely with the hospitals’ decision makers and subject matter experts was a
key factor in obtaining subjective validity of the generic model. At sites where a working
group was assembled to work with the simulation modelling team on the project, model
validation was able to go beyond objective techniques, building trust in the model such
that decision makers were confident to base decisions on results.
Membership of the working group should be wide and encompass the whole system.
The working group should not be limited only to decision makers, but should also include
subject matter experts, and stakeholders that would be affected by any decision resulting
from the model. In the case of the proposed perioperative generic model, working group
members should include managers, surgeons, and front line staff such as nurses. Mem-
bership should also cover the entire perioperative system, from the OR, to the PACU,
pre-admission clinic, as well as the surgical inpatient units, including stakeholders from
the emergency department and medical inpatient care.
5.4.2 Step 2: Review of the Generic Conceptual Model
One of the disadvantages of using a generic model is that the model design, assumptions,
simplifications, etc., were previously determined without the input or involvement of the
current decision maker. Whereas in the typical specific simulation model study, one of
the first steps to achieving model credibility is to review the conceptual model design with
Chapter 5. Generic Model Validation 122
the decision makers to seek approval and acceptance (Robinson et al., 2010). Typically,
when building a specific model, user input into the conceptual design of the model helps
gain credibility as users are part of the process; take part in decisions about abstractions,
simplifications, scope and level of detail. This needs to be recreated for decision makers
when using a generic model. In order to do so, this paper recommends that a thorough
and purposeful review be conducted of the generic model with the working group of
decision makers and stakeholders.
To begin, it is suggested that a detailed understanding of the site’s processes is gar-
nered by interviewing process owners, including managers and front line staff, and con-
ducting observations of the processes within model scope. This will also aid in getting
buy-in from the decision makers through on-going contact and demonstrating knowledge
of their processes.
Next, a review of the generic conceptual model in terms of its objective, scope, level
of detail, assumptions and limitations should be conducted with the working group.
This is to ensure that model processes, assumptions and behaviours are comparable
to the real system. This review should include comparing the generic model’s design
with the processes and procedures of the hospital, with particular focus on where model
simplifications and assumptions may differ. Where there are differences, it is important
to include a plan of how it would likely affect the model results, and how, if needed the
model will be adjusted to reflect these differences.
5.4.3 Step 3: Calibrate Model
This step can be considered your typical attempt at objective validation: making adjust-
ments to model inputs and design in an attempt to reduce prediction error.
During the validation stage, regular updates on progress, including demonstrating
interim results, will prove invaluable in garnering model credibility. Discussions on re-
sults, inputs and assumptions, and discussion on reasons for validation challenges will
Chapter 5. Generic Model Validation 123
help to include the decision makers and increases their knowledge and understanding of
the generic model, leading to increased credibility.
During this validation step it is recommended to review the current validation re-
sults with the working group, and try to identify likely reasons for difference between
model outputs and historical results. In order to help uncover likely reasons, create var-
ious reports that can help verify model assumptions, and compare model and historical
processes and values.
For example, as described earlier at Brampton, three reports were used to help iden-
tify how inputs could be adjusted to improve variability: analysis on historical booking
practices by surgeon (figure 5.1 on page 104), comparing the surgeon’s case load in the
input file compared to historical performance (figure 5.2 on page 106) and comparing the
realised elective schedule to the planned MSS.
In addition, review again the assumptions and inputs of the model to actual processes.
Determine if there are any notable differences that could be affecting the results. If there
are, determine if the model inputs can be adjusted to reduce the effect of the difference.
5.4.4 Step 4: Reconcile Model Results
When model outputs are “close” to actual, and no reasonable model input adjustments
can be made, focus on explaining the prediction error. This step should be done with
significant involvement with the working group. After all, it is the working group at
the end of the day that must feel that the model’s results can be trusted when making
decisions. Reports, such as those generated in Step 3 should be produced to guide this
reconciliation process.
For example, in the case of Brampton, after the second round of validation, discrep-
ancies still remained. Reports, including surgeon booking practices, late OR block starts,
realised block schedule vs. planned MSS, etc., were all used to demonstrate that current
poor practices in place at Brampton were affecting performance. In this step, the work-
Chapter 5. Generic Model Validation 124
ing group realised that the model’s outputs are based on the assumption that these poor
practices identified are resolved.
5.4.5 Step 5: Repeat and Re-evaluate
As with all validation exercises, it is a reiterative approach. Repeat steps three and four
until model credibility is achieved.
At this step, it would be prudent to re-review model design, assumptions, simplifica-
tions, etc. compared to the processes of the hospital. Any discrepancies identified should
be evaluated to determine their effect on achieving validity and credibility. Explore the
need to make changes to the model’s design to achieve acceptance. However, a well
planned and designed generic model (following the guidelines provided in the previous
chapter) should not require further coding. Changes to the actual design of the generic
model should only be done after careful consideration of the suggested change, and only
if it fits within the model’s problem description, scope and objectives and is common to
multiple potential sites.
5.4.6 Step 6: Accept the Model as Correct
After a series of iterations of steps three to five, the working group should come to
the consensus that the generic model is correct and can be used to guide their decision
making.
5.4.7 Final Notes
The greatest proof of acceptance of a generic model, or any model for that matter, is
the use of its results by the hospital to inform decisions. At the four highly involved and
engaged hospitals, the generic model has been used to inform decisions. For example, at
Brampton and Etobicoke, the model has been used to demonstrate the potential of the
Chapter 5. Generic Model Validation 125
current OR resources when elective booking policies, including accurate booking practices
and high OR utilisation are followed. In addition, the model was used to determine how to
assign elective blocks to accommodate new surgeons and achieve various volume targets.
At Prince Albert, the model was used to justify a request to the Ministry of Health for
commissioning (opening) a fifth OR in order to accommodate increasing demand and to
achieve aggressive provincial wait time targets. Further discussion on using the generic
model as a demonstrative and decision tool is reserved for the following chapter.
Chapter 6
A Decision and Demonstrative Tool:
Case Studies
The generic model has been proven to support decision makers as a decision and ne-
gotiation tool by allowing comparisons of decision options against multiple selected key
performance indicators. In addition, the generic model has proven useful as a tool to
quantify inefficiencies in current practices by demonstrating possible performance if those
practices were improved.
As a decision tool, the generic model was initially intended to answer tactical decisions
such as: “what is the effect of changing the block schedule on the orthopaedic ward?”,
“if we schedule three TJR per orthopaedic OR block, will the throughput of non-TJR
patients be adversely affected?”, and “how can we increase cardiac throughput to meet
our yearly targets?”.
However, during application of the model at various sites, it proved valuable as a tool
to demonstrate the effect of poor practices on a hospital’s performance. This gave rise to
a second use, as a demonstrative tool. The model can be used to answer such questions
as:
• A hospital currently performs poorly in on time starts - what is the effect of this
126
Chapter 6. A Decision and Demonstrative Tool: Case Studies 127
on their productivity?
• Blocks from the MSS are regularly left unfilled or partly unfilled - What is the
effect on volumes?
• What is the effect of the many OR block switches that occur on downstream re-
sources and outcome performance? Ex. Cancelations.
Through two case studies, this chapter will demonstrate the value of the proposed
generic model in terms of these two uses.
6.1 As a Decision Tool at the Juravinski Hospital
6.1.1 Problem Description
The Juravinski Hospital faced significant patient flow issues for both their surgical pa-
tients and medical patients. These issues resulted in elective, non-urgent surgical patients
regularly being delayed or cancelled when no inpatient bed was available, urgent surgical
patients not receiving care in a timely manner, patients being placed in off-program beds,
and medical patients being held in the emergency department as there was no appropriate
bed available.
In addition, the Ontario Ministry of Health and Long-term care (MOHLTC) recently
imposed funded volume targets for key surgical procedures in order to improve access
to care and reduce wait times in the province. The program provided additional fund-
ing to hospitals in exchange for increased volume commitments for specified procedures
including total joint replacements (TJR), cataracts, oncology and cardiology. If, at the
end of the funding year, the hospital was unable to achieve the agreed upon targets, the
hospital would face penalties including returning part or all of the funding provided, and
reduced target volumes and thus funding for future years. The hospital was challenged
to meet their target volumes.
Chapter 6. A Decision and Demonstrative Tool: Case Studies 128
Hospital decision makers were concerned about achieving their target volumes with-
out significantly affecting other procedures, and improving patient flow for all surgical
patients. As a result, they were interested in measuring throughput of cases by service
and for specific targeted procedures (total joint replacements and fractured hip repairs),
the number of cancellations caused by poor patient flow, and the bed resources required.
In addition, the hospital wanted to know what changes in tactical decisions could
be made to their block schedule, scheduling rules, and bed capacities in order to achieve
their targets. The administration was interested in an operations research tool that could
help inform their decisions by enabling them to test different possible solutions virtually
and compare outcomes. The tool would also be used to aid in negotiations with other
hospital stakeholders, including surgeons and the medical program administration, who
were concerned about the consequences of these perioperative decisions on their processes
and work.
They were also concerned with two key patient flow indicators: cancellations and in-
patient unit occupancy levels (census). In terms of cancellations, the administration and
surgeons felt that they were experiencing an unacceptably high number of cancellations
related to poor flow. Reasons for cancellations due to poor flow include cancellations
due to no ICU (Intensive Care Unit) or ward bed, cases cancelled because of block time
overrun, and scheduled elective patients bumped for more urgent patients.
In addition, the hospital regularly maintained a census above budget in order to
maintain procedure volumes and avoid cancellations. One of the reasons for the high
census was that medical patients frequently overflowed into surgical inpatient beds when
there were not enough medical beds available (referred to as off-service or off-program).
Surgical patient census also experienced variability over the week, with the highest census
mid-week, and lowest on Sundays.
As a result, hospital administrators were looking for help in making tough decisions
on how best to allocate limited hospital resources in order to meet their volume targets
Chapter 6. A Decision and Demonstrative Tool: Case Studies 129
and improve surgical patient flow, while not significantly impacting other patients and
services.
6.1.2 Solution
Due to the possible widespread impact of decisions from this project, it had a high profile
within the organisation, including interest from the most senior levels of the administra-
tion. A working team was assembled of key players at the hospital. The team included
the director of perioperative services, the managers of the operating room and the two
orthopaedic inpatient wards, the chief of orthopaedic surgery, the manager of the quality
department, and a decision support (data) expert.
The first task of the working team was to determine the scope and goals of the project,
including key performance indicators. As mentioned previously, the goal was to improve
tactical decisions of their perioperative service. The working group, in consultation with
stakeholders and the administration, defined the following objectives for this improvement
project:
• Achieve their yearly funded target volumes for total joint replacement procedures,
while maintaining current volumes of other procedures
• Complete urgent hip fracture repair procedures within the 48 hour target
• Reduce the daily peak number of beds required to achieve volumes, and if possible
level out the number of beds required over the week
• Reduce, or preferably eliminate, the number of cancellations due to no bed
• Improve the relationship between the hospital and surgeons by working with them
in promoting equitable access to services for patients
Based on these objectives, and the scope of the tactical decisions under consideration,
the working group wanted to focus on three key decision areas:
Chapter 6. A Decision and Demonstrative Tool: Case Studies 130
• The operating room block schedule: including the number and length of blocks
available, and the services and surgeons assigned to the blocks;
• Scheduling rules: such as the type of cases to be scheduled in a block, and how
to schedule emergent and urgent cases;
• Resource availability of post-operative bed resources: such as capacity of
recovery room capacity and inpatient beds, degree of bed blocking due to patients
from other programs occupying surgical beds, and the number of patients waiting
for alternate levels of care in acute beds.
The working group further identified three key outcome measures of interest for eval-
uating possible solutions based on their objectives:
• Throughput by service, and specifically for total joint replacements (TJR) and hip
fracture repairs;
• Number of cancelations by reason, with a focus on cancellations due to no bed;
• Daily midnight census of the inpatient units.
Based on the scope and goals laid out by the working group, it was determined
that the proposed generic, tactical decision simulation model of patient flow through
the perioperative service would meet their needs. The simulation model would allow
for the working team to review numerous possible decision options with stakeholders,
weighing the effects of different the solutions on the identified outcome measures. The
advantage of simulation modelling for decision-making is that decision makers can see
and understand the effect of different solutions on their outcome measures. Furthermore,
it keeps the decision making in the hands of the decision makers, as opposed to an
optimisation model that outputs a solution. Comparing possible solutions based on a
variety of outcomes using simulation allows for consideration of the qualitative objectives
Chapter 6. A Decision and Demonstrative Tool: Case Studies 131
of the stakeholders that cannot be quantitatively defined or measured meaningfully, and
a shared decision-making process.
6.1.3 Application of the Generic Model
The first step was to understand the processes and patient flow specifics of the hospital.
Through process observation and interviews with staff and management from various ar-
eas, a process map of their perioperative patient flow processes was drafted and reviewed
by the working group.
The information was also used to demonstrate to the working group how the generic
model would represent their specific processes and what assumptions and simplifications
were required. This exercise was key in gaining trust and credibility of the generic model
from the working group. Without their trust in the model, successful implementation
of solutions provided would not have been possible. This also helped the working group
understand how the generic model functioned, what it was able to do, and how it could
help them make better decisions.
The next step was to apply the generic model to the site by inputting their specific
information. The input files were based on the information from the process map as well
as one year of historical data. The data consisted of surgical patient records including:
procedure type, length of stay, care path, historical inpatient census data, off-service
rates, and cancellation rates by reason.
The model was validated against two months of actual historical performance. The
results of the validation were reviewed with the working team in order to ensure that
those who would be making decisions based on the results believed that the model was
able to accurately represent their hospital patient flow and patient population. The
validation process was an iterative effort with high involvement from the working team.
The team met regularly to review validation results; discussing the results in terms of
whether the predictive error of the model was acceptable, and how the inputs could be
Chapter 6. A Decision and Demonstrative Tool: Case Studies 132
adjusted in order to improve the models performance.
Once the model was deemed valid and credible by the working team, numerous de-
cision options were run. Options tested varied from small changes to the OR block
schedule, to dramatic and controversial changes to the policies of the hospital around
off-servicing patients and weekend elective blocks. Options were generated from a vari-
ety of sources including ideas from the working team and other hospital staff, literature
and best practice, a bed resource planning model (Liu, 2012) and various combinations
of these. Results were systematically reviewed by the working team, which generated
additional options based on results.
The Juravinski was the third application of the generic model to a hospital, following
the two William Osler sites Brampton and Etobicoke. As detailed in section 4.3, the time
required for data input and validation decreased as some expertise was gained through
multiple applications of the model. In addition, the challenges faced with validation at
the William Osler sites were not present at Juravinski. Testing, evaluating and presenting
numerous decision options was on going over the course of six months, an average a few
hours of work each week. Juravinski benefited from the existing generic model in terms of
time and cost of implementation by not having to design and build a model from scratch.
This allowed for quicker turn around of the project to allow for timely decision making.
6.1.4 Results from the Generic Model
After extensive and iterative testing of numerous possible decision options, the working
team presented a set of seven options to the stakeholders. The stakeholder group included
physicians, unit managers including medicine and hospital administrators. The seven
options were composed of a variety of solutions ranging from some minor adjustments to
the OR block schedule to significant changes to hospital policies that would also affect
medical patient flow.
• TJR LOS Improvement: The current base schedule is unchanged. The length
Chapter 6. A Decision and Demonstrative Tool: Case Studies 133
of stay of total joint replacement patients is adjusted to benchmarked levels.
• Reduce Peak Demand*: This scenario was based on the proposed MSS provided
by a bed resource model optimisation model applied to the hospital (Liu, 2012).
The objective of the model was to reduce the maximum number of inpatient beds
required by making changes to the assignment of OR blocks to surgeons.
• High Volume TJR Surgeons:The MSS is changed such that the orthopaedic
surgeons who perform high volumes of total joint replacements are spread evenly
over the week, as opposed to concentrated earlier in the week.
• Prescribe Case Types: The types of cases (hip replacement, knee replacement,
other orthopaedic cases) that can be performed in each scheduled OR block is
prescribed to manage flow to ward beds.
• Ring Fencing*: This more drastic scenario considers the effect of reducing the
number of surgical beds and changing the off-servicing policy of the hospital by not
allowing medicine patients to use surgical beds.
• Orthopaedic Surgeon Assignment Changes with Ring Fencing: Changes
to the assignment of OR blocks to orthopaedic surgeons is performed to spread case
types over the week based on historical surgeon patterns. The changes are done in
addition to the ring fencing policy.
• Weekend Elective Scheduling with Ring Fencing*: In an attempt to smooth
demand for beds more evenly across the week, it was thought that moving two total
joint replacement (TJR) OR blocks to Saturdays would help increase weekend cen-
sus and reduce weekday peak demand. This scenario was tested with and without
a ring fencing policy in place, the results presented here are with ring fencing.
In order to illustrate the decision making processes that occurred at Juravinski, three
of the seven options are used and are indicated above with an *. The three options were
Chapter 6. A Decision and Demonstrative Tool: Case Studies 134
chosen based on demonstrating the range of decision options considered, and options
that the decision makers found interesting based on the model’s outputted results. For
example, the Director of Perioperative believed that running weekend elective cases (the
seventh option) would significantly improve access to ward beds, reduce cancelations and
improve throughput. Model results, however, indicated that the level of improvement
was not as significant as hoped.
All the options were compared to a base case which was composed of some minor
changes to the current MSS and increases to the minimum number of TJR cases each
surgeon must schedule in their OR blocks. With the adjustment to the scheduling rule,
the base case was able to meet the hospitals yearly TJR volume target. The results from
the three options are presented below in Tables 6.1 and 6.2 and Figure 6.1.
The output analysis shows that reducing peak bed demand resulted in a small increase
in the overall elective throughput. Most of the increase was due to increases in the TJR
and urology volumes. However, this was at the expense of reduced volumes of the plastics
service. These changes were due to the fact that alterations to the schedule opened up
beds on days when urology and orthopaedics were able to take advantage of the available
beds, reducing cancellations. However, plastics was disadvantaged slightly as they lost
some access to beds. Furthermore, the resulting MSS did not significantly reduce the
cancellation rates due to no beds. The graph in Figure 6.1 explains these results; there is
little change to the number of beds required, as any savings in beds were taken advantage
of by other services.
Restricting medicine (off-service) patient access to surgical beds had a noticeable
effect on the average number of beds required in the TJR and general surgery wards. This
indicates that when medicine patients are not occupying surgical beds, the orthopaedic
service can complete a similar number of cases, compared to the base case, using fewer
beds. In fact, there is a total average savings of more than 10 beds over the week that
could be transferred to the medicine service to fund additional beds needed to better meet
Chapter 6. A Decision and Demonstrative Tool: Case Studies 135
Measure - % change com-
pared to Base Schedule
Reduce Bed
DemandRing Fencing
Weekend
Elective
Scheduling with
Ring Fencing
Throughput by Patient Type
Elective 4.9% 1.2% 3.1%
Urgent 0.0% -0.1% -0.1%
Throughput by Service
GENL 1.8% 0.8% 1.5%
OBGY 2.8% 2.1% 2.5%
ORTH 1.8% 0.8% 3.6%
Fract. hip 0.4% -0.4% -2.6%
TJR 3.8% 0.7% 4.4%
PLAS -6.4% 0.6% 0.3%
UROL 13.8% 0.8% 1.2%
Table 6.1: The percent change compared to the base case for three of the scenarios
considered by the Juravinski - Throughput
Chapter 6. A Decision and Demonstrative Tool: Case Studies 136
Measure - % change com-
pared to Base Schedule
Reduce Bed
DemandRing Fencing
Weekend
Elective
Scheduling with
Ring Fencing
Cancellations by Reason
No ICU Bed Cancellations 4.8% -19.4% -26.9%
No Ward Bed Cancellations -4.3% -32.9% -60.4%
Not Enough OR Time Can-
cellations7.3% 0.6% 4.8%
Bumped for More Urgent
Case Cancellations4.6% -2.2% 7.9%
Other Cancellations 8.0% 0.5% 1.5%
Table 6.2: The percent change compared to the base case for three of the scenarios
considered by the Juravinski - Cancellations
Chapter 6. A Decision and Demonstrative Tool: Case Studies 137
!"
#!"
$!"
%!"
&!"
'!"
(!"
)*+" ,-." /.0" ,1-" 234" 567" 5-+"
!"#$%&#'(#)*+*',-'.%/#)0*1'
!"#$%&#'2%345'(#)*+*'6''
7%*#'(%*#'
!"
#!"
$!"
%!"
&!"
'!"
(!"
)*+" ,-." /.0" ,1-" 234" 567" 5-+"
!"#$%&#'2%345'(#)*+*'6'8%098:)'74;<=*'>30?'@3)&'A#)<3)&'
89:" ,;<"=630" >371*"/630" ?.+"5-3@"/630"
!"
#!"
$!"
%!"
&!"
'!"
(!"
)*+" ,-." /.0" ,1-" 234" 567" 5-+"
!"#$%&#'(#)*+*',-'.%/#)0*1'
!"#$%&#'2%345'(#)*+*'6''
@#B+<#B'C#%='2#D%)B'
!"
'"
#!"
#'"
$!"
$'"
%!"
%'"
&!"
&'"
'!"
)*+" ,-." /.0" ,1-" 234" 567" 5-+"
!"#$%&#'(#)*+*',-'.%/#)0*1'
!"#$%&#'2%345'(#)*+*'6''
@3)&'A#)<3)&'
!"
#!"
$!"
%!"
&!"
'!"
(!"
)*+" ,-." /.0" ,1-" 234" 567" 5-+"
!"#$%&#'(#)*+*',-'.%/#)0*1'
!"#$%&#'2%345'(#)*+*'6''
8%098+)'74;<=*'>30?'@3)&'A#)<3)&'
Figure 6.1: Daily midnight census results from selected scenarios presented to Juravinski
Hospital.
Chapter 6. A Decision and Demonstrative Tool: Case Studies 138
their demand. Additionally, there is a significant reduction in the number of cancellations
due to no bed. This was a big selling point to surgeons, who felt it was unfair for their
patients to be cancelled due to no bed, when the beds that should be available are
occupied by non-surgical patients.
Results from the option that considers ring fencing and weekend elective blocks also
showed promising results in terms of bed savings and reduced cancellations. A slight
increase in volumes was experienced in this scenario, particularly the TJR procedures.
However, much of the improvement was due to the ring fencing policy and not the
weekend elective blocks. This result was of particular interest to the director of surgery,
who thought that weekend elective cases would help smooth demand of beds better over
the week. It is possible that if more blocks were moved to the weekends that a more
significant change would be seen. However, running too many weekend elective cases is
not feasible due to the additional cost of paying staff weekend premiums and surgeons
not willing to work weekends.
6.1.5 Implementation of the Decision
Based on the results presented to the stakeholder group at the hospital, the working
group decided to trial the ring fencing scenario in January 2012. The trial was composed
of a number of changes:
• Adjustment of the Master Surgical Schedule as recommended by the working group
based on the ring fencing scenario;
• Instituting a ring fencing policy for surgical beds, thereby restricting non-surgical
patients access to surgical beds (i.e. off-service);
• Closure of ten surgical beds;
• Addition of three inpatient medicine beds;
Chapter 6. A Decision and Demonstrative Tool: Case Studies 139
• Additional funding towards an alternate level of care (ALC) unit that cares for pa-
tients waiting for further care at other facilities in an appropriate and cost effective
care setting.
After seven weeks, the working group evaluated the results of the implemented trial.
The evaluation focused on two measures: daily average census and cancellations due to
no bed.
During the trial period, not a single cancellation occurred due to no bed. While
the simulation model predicted there would still be some cancellations due to no bed,
the working group conceded that in the few times that a cancellation would have been
required, the hospital ensured that every measure possible was taken to avoid the cancel-
lation. This extra effort was done as part of the selling point of the trial to the surgeons.
As part of the trial, some surgical beds were given to medicine as the model predicted that
they would not require them with ring fencing in place. The administration promised
the surgeons that the bed reduction would not cause cancellations. Thus, they needed
to ensure that this was the case. Though it is not possible to measure quantitatively,
the working group agreed that during the trial, the number of times they faced possi-
ble cancellations due to no bed was less frequent than before, as the simulation model
predicted.
In the case of unit census, the model results did differ somewhat to those of the trial.
Figure 6.2 compares the daily average predicted census to the actual average census
during the trial for the three surgical units. As part of the trials evaluation report to
the hospitals administration, the working team analysed the data used for the simulation
and the trial period in order to explain the differences. The team found that they were
reasonably able to explain the causes for the differences.
For instance, despite the small sample size of the trial, the TJR ward census is fairly
similar to that seen in the simulation. Here the working group noted that compared to
the time period of the data used to populate the simulation model, the length of stay of
Chapter 6. A Decision and Demonstrative Tool: Case Studies 140
!"!#
$"!#
%!"!#
%$"!#
&!"!#
&$"!#
'!"!#
'$"!#
()*# +,-# .-/# +0,# 123# 456# 4,*#
78-259-#:-*;,;#<=#>5?-*6;@#
+AB#.52/#:-*;,;#
!"!#
$"!#
%!"!#
%$"!#
&!"!#
&$"!#
()*# +,-# .-/# +0,# 123# 456# 4,*#
78-259-#:-*;,;#<=#>5?-*6;@#
C260)##.52/#:-*;,;#
!"!#$"!#%!"!#%$"!#&!"!#&$"!#'!"!#'$"!#
()*#
+,-#
.-/#
+0,#123#456#4,*#
(-5*#
D2-/3E6-/#78-259-#:-*;,;# FG#.--H#D3I)6#78-259-#J-/#KL,385I-*6;#
!"!#
%!"!#
&!"!#
'!"!#
M!"!#
$!"!#
N!"!#
()*# +,-# .-/# +0,# 123# 456# 4,*#
78-259-#:-*;,;#<=#>5?-*6;@#
O-*#4,29#.52/#:-*;,;#
Figure 6.2: Comparing average daily unit census of the simulation model and the seven-
week trial.
TJR patients had decreased slightly, reducing the demand on beds.
Conversely, the simulation model predicted a lower census than realised during the
trial in the orthopaedic ward, which cares for other orthopaedic patients including frac-
tured hip patients. The working group found that this was due to the fact that the
volume of urgent fractured hip patients experienced was higher than the arrival rate
used in the model; thus, requiring more beds post-surgery. Additionally, the length of
stay of fractured hip patients has been on the rise due to longer waits for alternative level
of care spaces in rehabilitation and other post-acute facilities.
Finally, the working group believed that the key reasons for the higher than expected
census in the surgery ward was owing to three factors. First, the volume of gynaecological
oncology patients has increased compared to what was seen previously in the data used
in the simulation model. Additionally, the hospital recently had the city’s hepatobiliary
surgeons join their team from another hospital. The predicted case volumes from these
surgeons were inputted into the simulation model; however, these volume predictions were
lower than what was realised. Finally, the working group also noted that the waiting
time for surgical patients waiting for alternative level of care, specifically palliative care,
was higher during the trial period than what was considered by the simulation model.
Chapter 6. A Decision and Demonstrative Tool: Case Studies 141
Based on the positive results from the seven week trial, the working group recom-
mended to the hospital administration that the MSS and ring fencing policy be adopted.
The administration agreed and has approved the ring fencing policy.
During the seven-week pilot, the hospital also observed a reduction in the lengths of
stay of emergency patients waiting for inpatient beds. As demonstrated in Figure 6.3,
the average emergency department (ED) length of stay of admitted patients from all
services saw a reduction in wait time for an inpatient bed. The reduction was especially
significant for medicine patients, who constitute the majority of patients waiting in the
ED; their length of stay was reduced to about half in comparison to the same seven weeks
the previous year.
!"!!#
$"!!#
%!"!!#
%$"!!#
&!"!!#
'()*+*,(# -./0123()*+4# 56.7(.8# 9::#23;(,/4#
!"#$%&#'()%*'+,-.$/0'
9<(.37(#=>(.7(,+8#?(23./>(,/#
@(,7/0#1A#5/38B#
C#D((E4#*,#2.(<*164#8(3.# CFD((E#/.*3:#
BG.1>#3)>*44*1,#)(+*4*1,#/1#)(23./6.(#A.1>#/0(#=?#
Figure 6.3: Emergency Department length of stay during the trial compared to the
previous year.
Chapter 6. A Decision and Demonstrative Tool: Case Studies 142
The reduction in ED LOS can be attributed to the addition of the three medicine
and ALC beds. These additional beds created more capacity where needed. The beds
allow the hospital to care for their patients in the most appropriate setting; in units that
are designed for the particular patient care needs and are staffed with knowledgeable
staff. For example, medicine patients are cared for by nurses and other staff trained
specifically in caring for medical diseases, instead of by surgical nurses who specialise in
wound care and surgical recovery. Additionally, the patients are staying in a unit with
the appropriate equipment to manage their care.
The ED data demonstrates that the reduction of surgical beds had a positive effect
on the throughput and cancellation rates of elective surgical patients, but also shows that
emergency surgical patients continued to have access to appropriate care without having
to wait longer in the ED. This was important to Juravinski administrators as they do
not want decisions and changes to improve perioperative care to negatively affect other
patients and areas of the hospital. With any change made in healthcare, it is important
to be monitor the effects on other areas of the healthcare system to ensure that issues
are not simply shifted to another area. The Juravinski should continue to monitor the
impact of these changes to performance across the hospital. Furthermore, it would be
prudent to work closely with up- and downstream partners to monitor possible effects,
both negative and positive to external processes and resources.
6.1.6 Discussion
The working team used the results from the scenarios as a negotiating tool with the
stakeholders of the hospital, including physicians and surgeons, the medicine department
and administration. Based on the results, they were able to convince surgeons to accept
the changes to their MSS schedule on the promise that cancellations due to no bed would
decrease and that medicine would no longer have access to surgical beds. In exchange
for ring fencing surgical beds, surgery agreed to transfer some beds to medicine to help
Chapter 6. A Decision and Demonstrative Tool: Case Studies 143
accommodate their demand. This exchange was part of the negotiation needed to bring
the medicine program on board with the ring fencing policy. They were concerned that
without the bed transfer, their patients would experience further limited access to care
and experience longer waits in the emergency department.
The successful application of the generic simulation model at this hospital can be
largely attributed to the highly engaged and motivated working team. The working
team was composed of a number of key stakeholder representatives, including surgical
program management and the chief of surgery. Having members of the stakeholder groups
in the working group is advantageous as they act as champions of the project and the
proposed solutions, which helps in gaining acceptance from the stakeholders.
In addition, representation from the quality department further contributed to the
successful application of the model. The quality department has had previous experi-
ence in working with operational research tools, including simulation, through previous
graduate student work, as well as a couple of on-staff industrial engineers. Moreover, the
quality team served as the project management and reporting function of the working
team, translating simulation results into presentations geared towards the stakeholders.
Their knowledge and experience with operational research, and with dealing with the
various stakeholders, brought tremendous negotiation power to the project, especially
when implementing the controversial ring fencing policy.
Another success factor was the close tie and involvement of the decision support
department throughout the project. The department provided the electronic data files
required to populate the simulation model and aided in the analysis of the data from the
trial. The decision support department not only had an in depth understanding of the
hospitals data, but also of the current trends, such as higher ALC wait times and shorter
orthopaedic lengths of stay, that were influencing the trial results.
The hospitals corporate initiative structure further helped ensure the success of the
project. For large corporate initiatives, the hospital has implemented a standard plan-
Chapter 6. A Decision and Demonstrative Tool: Case Studies 144
ning and reporting structure that includes quarterly updates on milestones up through
administration, including the CEO, site presidents, vice-presidents and board of direc-
tors. This reporting structure ensures that the working team was accountable to meet
their project goals in a timely manner. Furthermore, when the working team experienced
any challenges or barriers, there was executive support to help address them in order to
keep the project moving forward.
Six months after the trial, the hospital continues to use the proposed MSS and ring
fencing policy. In addition, the working group has re-convened and is using the generic
simulation model for additional tests to determine how they can meet their revised volume
targets, taking into consideration recent changes to the perioperative service such as case
mix, length of stay and ALC waiting times.
6.2 As a Demonstrative and Decision Tool at William
Osler Health Sciences
6.2.1 Problem Description
The perioperative service at William Osler Health Sciences is composed of the surgical
resources available at the Brampton Civic Hospital and the Etobicoke General Hospital.
Based on internal analysis and benchmarking, it was determined that the perioperative
service was significantly under performing in efficiency measures including OR utilisation
and throughput. In addition, William Osler was unable to achieve any of their Ministry
waiting time surgical volume targets, including orthopaedic total joint replacements and
some oncology surgical procedures.
Though William Osler experienced few cancellations due to no bed, the perioperative
service was challenged with balancing the need for timely access to surgical care for urgent
patients with the needs of the elective case load. This resulted in expensive overtime costs
Chapter 6. A Decision and Demonstrative Tool: Case Studies 145
in the OR, lengthy case delays, and dissatisfaction among staff and surgeons.
Other issues identified in the analysis and benchmarking study included:
• Poor on time start performance: Elective OR block times were often not started as
scheduled due to delays in equipment, OR staff (including surgeons and anaesthe-
siologists), and patient readiness.
• Inaccurate booking times: The amount of time a surgeon scheduled for a proce-
dure often differed significantly from the time the case actually took, resulting in
requiring overtime to complete cases if under-predicted, or under-utilisation of the
OR if over-predicted.
• Master Surgical Schedule: William Osler was struggling to meet demand for OR
block time for current volumes, targeted volumes and planned predicted increases
without additional funding to run additional or longer OR blocks.
• Policies and their adherence: Many policies were not well adhered to, while other
policies typically found in perioperative services were out of date or non-existent.
In response, William Osler hired a consulting firm to develop solutions and imple-
mentation plans to address these issues. As part of their work, it was determined that
changes to the surgical schedule and resources were needed.
The generic model was proposed as a tool to help inform v on tactical decisions,
specifically:
• The planned master surgical schedule (MSS)
• Scheduling rules
• Resource capacity available
• Consolidation of a surgical speciality to a single site
Chapter 6. A Decision and Demonstrative Tool: Case Studies 146
The two key objectives set out by the William Osler administration for this project
was to determine how to achieve the targets set by the Ministry of Health and how
to increase their service provision to meet the demands of their growing and ageing
populations.
6.2.2 Application of the Generic Model
A working group of key decision makers and stakeholders was assembled for this sub-
project. This working group reported to the larger perioperative improvement initiative.
The working group included the director of perioperative services at William Osler, the
OR managers at both sites, the chief of surgery, and members from the consulting team.
The first step in this project was to introduce the generic model to the working
group. Once the working group agreed that the generic model would satisfy their needs,
interviews were conducted with managers and nursing leaders of the perioperative and
surgical inpatient areas to gain an understanding of the specifics of surgical patient flow
at the two William Osler sites. The information gathered was compared to the design
and assumptions of the generic model in order to determine if the model would be able
to adequately describe William Osler patient flow. This was presented to the working
group for discussion. The resulting diagrams are provided in Appendix H.
Since the generic model had been originally designed based on a set of academic
teaching hospitals, attention to any differences was important. Furthermore, this process
of studying and documenting their processes and demonstrating that the conceptual
model was able to capture their processes not only helped prove that the model was
generic, but also helped gain credibility and trust from the decision makers and users of
the tool at William Osler.
Along with one year of patient record data, the information on patient flow at William
Osler was inputted into the generic model. Two models were populated, one for each
hospital, Brampton and Etobicoke. The work at both sites was being done in tandem in
Chapter 6. A Decision and Demonstrative Tool: Case Studies 147
a single working group. The remainder of this case study will focus on the application of
the model at the Brampton site; a similar experience occurred at Etobicoke.
6.2.3 The Generic Model - Results as a Demonstrative Tool
As described in Chapter 5, the validation process at Brampton was met with many
challenges. Statistical (objective) validation was not possible due to numerous poor
practices at Brampton. However, through detailed analysis of the data and the model and
a close working relationship with the working group, the model was validated subjectively
through a reconciliation of the model outputs to historical results.
It was within this validation effort that a second use of the generic model as a demon-
strative tool was revealed. As a demonstrative tool, the generic model illustrated what
performance Brampton could have expected if:
• Surgeons’ booked time estimates were more in line with the actual time to complete
a case and turnover the OR. Methods for more accurate prediction based on statis-
tical analysis were being implemented concurrently as part of the overall William
Osler perioperative plan.
• Surgeons efficiently used the elective block time assigned by booking the entire time
available, or releasing unused time in a timely manner such that it could be used
by another surgeon or service.
• William Osler more strictly followed elective overtime tolerances by not allowing
surgeons to overbook their ORs and by cancelling cases when elective time would
be exceeded, as per policy.
• Elective OR blocks were started on time.
These findings were consistent with an efficiency analysis and observations performed by
the consulting firm as part of their diagnostics and recommendations. The benefit of the
Chapter 6. A Decision and Demonstrative Tool: Case Studies 148
generic model as a demonstrative tool was that it quantified the effect of poor practices
on perioperative performance.
To demonstrate the effect of their poor practices on their throughput performance,
a report was used to compare the number of OR blocks used to achieve their current
volumes to the number of blocks the model required to achieve the same level of service
(table 6.3). Columns A and C show the difference in the planned MSS (column A) and the
actual realised block schedule (column C). The outputted volumes of the model, based on
the realised block schedule are provided in column D. Compared to the historical volumes
in column B, there are some notable differences in the outcomes. Column F calculates,
based on the simulation model results, the number of blocks that would be required
by each service to achieve their current volumes if inefficient practices were addressed.
Column G is the difference between the number of OR blocks each service was assigned
in the MSS (column A) and the number required (column F).
From this report, a number of conclusions were derived. First, nine of the eleven
services at Brampton could have achieved their volumes as performed in the eight week
period with fewer elective OR blocks than assigned in the original MSS schedule. This
conclusion is based on comparing the throughputs from the model to actual as well as
the originally planned MSS compared to the actual block schedule. Table 6.4 compares
the MSS as planned to the actual elective schedule for eight weeks in April-May 2010.
The table also demonstrates that not all of these changes were planned ahead of time as
the number of planned block releases and pick ups from the MSS does not account for
all the differences between the MSS and the actual schedule.
On the other hand, the two other services, oral and orthopaedics, require more time
than originally allocated in the MSS to achieve their current volumes. Both these services
picked up released time from other services during the two month study period.
As a demonstrative tool, the generic model was used to substantiate not only what
would have been achieved based on the actual elective schedule (as discussed above) but
Chapter 6. A Decision and Demonstrative Tool: Case Studies 149
A:
B:
C:
D:
E:
F:
G:
Ser
vic
eB
lock
Sch
edule
Act
ual
Cas
es
Rea
lise
d
Blo
ck
Sch
edule
Sim
ula
tion
Res
ult
Cas
es
Sim
ula
ted
Cas
es/b
lock
Blo
cks
Req
uir
edto
do
Act
ual
Vol
um
e
Blo
cks
Ove
r
i.e.
Req
uir
ed
vs.
Ass
igned
(=D
/C
)(=
B/
E)
(=A
-F
)
Den
tist
ry24
7317
70.9
4.17
17.5
-6.5
EN
T40
223
3822
05.
7938
.52
-1.4
8
Gen
eral
118
546
111
595.
85.
3710
1.72
-16.
28
Gynae
colo
gy36
240
3224
6.7
7.71
31.1
3-4
.87
Ophth
alm
olog
y60
877
6192
2.1
15.1
258
.02
-1.9
8
Ora
l6
4010
454.
58.
892.
89
Ort
hop
aedic
s92
428
104
439.
74.
2310
1.23
9.23
Pla
stic
s28
7018
75.5
4.19
16.6
9-1
1.31
Uro
logy
5225
456
274.
74.
9151
.78
-0.2
2
Tot
alB
lock
s45
644
742
5.48
Tab
le6.
3:D
emon
stra
tive
Tool
Rep
ort
-Il
lust
rati
ng
the
effec
tof
poor
pra
ctic
eson
thro
ugh
put
volu
mes
atB
ram
pto
n.
Chapter 6. A Decision and Demonstrative Tool: Case Studies 150
Service
Planned
Block
Schedule
Picked up* Released*Actual
Schedule
Difference
to planned
Dentistry 24 3 0 25.5 1.5
ENT 40 4 3 39.5 -0.5
General 118 1 9.5 105.5 -12.5
Gynaecology 36 1 2.5 35 -1
Ophthalmology 60 1 0 61 1
Oral Surgery 6 4 2 10 4
Orthopaedics 92 2 0 87.5 -4.5
Plastics 28 1 5 21 -7
Urology 52 0 4 42 -10
*Changes according to utilisation sheet (planned and known in advance). It
was found that not all changes were reflected in the utilisation sheet).
Table 6.4: Differences in elective OR time by service from the MSS to actual in April-May
2010.
Chapter 6. A Decision and Demonstrative Tool: Case Studies 151
also what the MSS would need to look like to achieve exactly what they had achieved
over the period of study. The results in tables 5.6 and 5.7 are based on the actual elective
schedule. However, the model based on that schedule was able to complete more cases
than required for general, ophthalmology, plastics and urology. This indicates that a
further reduction in OR blocks assigned should result in the same throughput, if these
services were to follow better booking practices as previously described. Conversely, the
model was unable to complete the same volume of dentistry and ear, nose and throat
(ENT) cases as completed historically based on the current actual block schedule. This
demonstrates that these two services, if following better booking practices would require
more OR time than currently allocated to maintain current volumes. Finally, gynaecology
and orthopaedics can, according to the model, maintain current volumes based on the
block schedule as carried out.
Note that oral surgery was excluded from the lists above. During the time period
studied, oral surgery was the only service to have used more block time slots than origi-
nally assigned in the MSS. At the time of this analysis, the decision makers at William
Osler did not place emphasis on increasing volumes for oral surgery, as it was not a cur-
rent priority. Even though the model shows that oral surgery could complete more cases
than actually completed in the same number of blocks, the decision was made not to
increase the service above the current allocation according to the MSS. As a result, the
volume of cases for oral surgery will be less for any schedule proposed from this analysis.
This was known and understood by the decision makers and working group at William
Osler, and accepted based on their priorities and objectives.
Based on this analysis, a proposed base MSS was provided to William Osler for
Brampton. The proposed MSS assumes that the objective is to maintain current volumes
for all services, except oral surgery as discussed above. The proposed base MSS assumes
that the perioperative service achieves its goal of following best practices for scheduling
and operating the ORs. The proposed base MSS is presented in table 6.5. Expected
Chapter 6. A Decision and Demonstrative Tool: Case Studies 152
volumes from the base MSS are provided in table 6.6 as generated from the generic
model. Notice that the base MSS uses only 208 out of the total 228 OR blocks available
in the four week MSS. This gives decision makers 20 OR blocks to reassign based on
priority areas as needed.
ServiceMaster Schedule
Allocation
Proposed 4
week schedule
with 1 stat day
Difference
Dentistry 12 9 -3
Ear Nose and Throat 20 19 -1
General 58 51 -7
Gynaecology 17 15 -2
Ophthalmology 30 30 0
Oral Surgery 3 3 0
Orthopaedics 46 46* 0
Plastics 12 9 -3
Urology 26 26 0
Total Assigned 224 208 -16
Total Available 228 228
Unassigned Blocks 4 20
*Includes two swing rooms per week
Table 6.5: The proposed base MSS for Brampton based on the demonstrative results
from the generic model.
Based on the results of using the generic model as a demonstrative tool, the working
group was ready to move forward with the model and the base MSS to study future
scenarios using the generic model as a decision tool.
Chapter 6. A Decision and Demonstrative Tool: Case Studies 153
Measure Average95% CI
Range
Throughput by Patient Type
Elective 2511.9 23.5
Urgent 489.9 15.7
Throughput by Service
Dentistry 77.2 2.2
Ear Nose and Throat 258.2 9.3
General 596.4 14.8
Gynaecology 249.0 9.3
Ophthalmology 979.0 12.2
Oral Surgery 29.5 2.9
Orthopaedics 474.2 10.0
TJR 127.7 4.6
Plastics 75.2 3.4
Urology 274.3 23.6
Cancellations by Reason
No ICU Bed Cancellations 0.0 0.0
No SDU Bed Cancellations 2.2 1.8
No Ward Bed Cancellations 27.8 12.7
Not Enough OR Time Cancellations 156.7 24.0
Bumped for More Urgent Case Cancellations 16.0 2.9
Other Cancellations 274.7 16.6
Table 6.6: The predicted throughput and cancellation rates achievable from the proposed
base MSS.
Chapter 6. A Decision and Demonstrative Tool: Case Studies 154
6.2.4 The Generic Model - Results as a Decision Tool
A new base schedule centred on results from the application of the generic as a demon-
strative tool could then be used to help support William Osler in making short and long
term decisions. In the short term, decisions on how best to allocate some emergent time
to general surgery and accommodate a new gynaecological surgeon were needed. These
short term decisions were required immediately based on various existing agreements. In
the long term, William Osler decision makers wanted to determine how to create capac-
ity and an OR block schedule to achieve their volume targets in total joint replacements
(TJR), cataracts and oncology.
Short Term Decisions
Decision makers chose to focus on two key changes in the short term. First was to carry
out their plans for allocating emergent time for general surgery and accommodating the
arrival of a new gynaecological surgeon at Brampton. Secondly, they were interested in
exploring how to reduce the peak demand of surgical ward beds through changes in the
MSS assignment.
In negotiation with the surgical services, the hospital had agreed to assign general
surgery two half-day blocks a week for emergent cases. This time would be reserved
out of the elective schedule for general surgery to perform emergent cases in order to
reduce wait times of these patients and reduce the use of after hour OR time to complete
these cases. Prior to the results above, William Osler believed that in order to do this,
the emergent time would be taken from existing elective time assigned, reducing general
surgery elective throughput at the expense of emergent wait times. However, the process
produced a base MSS with unassigned budgeted OR blocks. As a result, decision makers
chose to evaluate the result of assigning the two emergent half blocks per week, plus an
additional OR block per week for general surgery to replace the elective time taken to
open the emergent blocks. The total amount of elective time for general surgery remains
Chapter 6. A Decision and Demonstrative Tool: Case Studies 155
the same.
Additionally, in response to increasing demand for gynaecological cases, William Osler
had hired a new surgeon. The question unanswered prior to this analysis was how to
accommodate the additional surgeon within the MSS schedule and downstream inpatient
resources, without increasing resource availability of the OR or inpatient units. Fortu-
nately, the demonstrative analysis performed above revealed that there was enough unas-
signed budgeted OR block time at Brampton to accommodate the new surgeon, without
reducing time assigned to other surgeons. Accordingly, the only remaining question was
if the current inpatient resources would be able to absorb the extra volume.
Based on these two needs identified by the hospital, the first scenario tested by the
generic model was to determine if the emergent time for general surgery and the new
gynaecology surgeon could be accommodated within the current perioperative resources.
The outputs of the generic model demonstrate that the proposed adjustments to the
base MSS could be accommodated. Throughput results in table 6.7 demonstrate that
the changes do not significantly affect the predicted volumes of other services. The
addition of the general service emergent time and additional elective block maintains
current elective volumes. The additional gynaecological surgeon increases gynaecological
throughput, but not at the expense of another service. Average census of the units
increases slightly due to increased gynaecological volumes, as demonstrated in figure 6.4.
The impact of this scenario can also be seen in table 6.8, showing that there is little
effect on cancellation rates; notably, cancellation due to no ward bed is not impacted
significantly more than expected from the changes.
William Osler was also interested in reducing the peak demand for beds required
across the organisation, including surgical beds. A Monte Carlo simulation model (Liu,
2012) that was being used for this analysis at the hospital level was leveraged to determine
if any changes to the proposed short term MSS, could be made to reduce the peak number
of surgical beds required. Based on the proposed MSS with the short term changes,
Chapter 6. A Decision and Demonstrative Tool: Case Studies 156
Measure - % change to Base Schedule Short Term
Short Term
with Urology
Adjustment
Throughput by Patient Type
Elective 1.4% 2.2%
Urgent 0.5% 0.3%
Throughput by Service
Dentistry 0.6% -3.2%
Ear Nose and Throat 0.8% 0.4%
General 1.4% 2.0%
Gynaecology 13.2% 18.2%
Ophthalmology 0.0% -0.3%
Oral Surgery 2.4% -1.4%
Orthopaedics -0.6% -1.2%
TJR 0.1% -0.7%
Plastics 0.8% 2.0%
Urology -2.3% 3.0%
Table 6.7: Model throughput results from short term MSS changes.
Chapter 6. A Decision and Demonstrative Tool: Case Studies 157
!"
#"
$!"
$#"
%!"
%#"
&!"
&#"
'!"
'#"
()*" +,-" .-/" +0," 123" 456" 4,*"
78-259-":-*;,;"<=">5?-*6;@"
!"#$%&#'(#)*+*','-%*#'(%*#'
!"
#"
$!"
$#"
%!"
%#"
&!"
&#"
'!"
'#"
()*" +,-" .-/" +0," 123" 456" 4,*"
78-259-":-*;,;"<=">5?-*6;@"
!"#$%&#'(#)*+*','./0$1'2#$3'
!"
#"
$!"
$#"
%!"
%#"
&!"
&#"
'!"
'#"
()*" +,-" .-/" +0," 123" 456" 4,*"78-259-":-*;,;"<=">5?-*6;@"
!"#$%&#'(#)*+*','./0$1'2#$3'451/'
6$070&8'!9:+*13#)1'
!"
#"
$!"
$#"
%!"
%#"
&!"
&#"
'!"
'#"
()*" +,-" .-/" +0," 123" 456" 4,*"
A:B" 4CB" 4,293D5E"5*/"40)26"465F" C260)>-/3D;"
Figure 6.4: Model average census results from short term MSS changes.
Measure - % change to Base Schedule Short Term
Short Term
with Urology
Adjustment
Cancellations by Reason
No ICU Bed Cancellations N/A N/A
No SDU Bed Cancellations 63.6% 154.5%
No Ward Bed Cancellations 115.8% 39.9%
Not Enough OR Time Cancellations 8.2% -4.8%
Bumped for More Urgent Case Cancellations 8.7% 0.0%
Other Cancellations 2.9% 3.0%
Table 6.8: Model cancellation rate results from short term MSS changes.
Chapter 6. A Decision and Demonstrative Tool: Case Studies 158
the Monte Carlo bed model recommended that the current urology block assignment of
Mondays, Wednesdays and Thursdays, is a main contributor to the peak bed demand level
on Thursdays. However, if urology blocks are changed to be on Mondays, Tuesdays and
Fridays, the peak demand in beds will be reduced. Figure 6.5 demonstrates the predicted
reduction in peak bed demand according to the bed model. This recommended change
to the MSS was tested in the generic model for more detailed analysis of the effect on
throughput, census and cancellation rates.
The results from the generic model in tables 6.7 and 6.8 and figure 6.4 show the effect
on throughput, cancellation rates and census. These results show that this additional
short term change to the MSS does not significantly affect throughput, but does smooth
ward census better over the weekdays, as well as slightly reduce cancellations due to no
ward bed.
!"#$#%&'()*#"+(,&"-(.))( !"#$#%&'()*#"+(,&"-(.))(/0+*(1"#2#34(5*673&(
Figure 6.5: Total average census by service as predicted from the Monte Carlo bed model.
Chapter 6. A Decision and Demonstrative Tool: Case Studies 159
Long Term Scenarios
In addition to the short term solutions outlined, William Osler wanted to explore schedule
changes that would help achieve their yearly volume targets in total joint replacements,
cataracts and oncology procedures. For each of these three procedure categories, William
Osler is accountable to reach these targets or face reduction in funding and future targets.
In previous years, William Osler struggled to achieve all of these targets. At the time,
current projections were indicating that William Osler would likely fail to achieve volumes
in TJR, but exceed in volume of cataract. Oncology targets overall were projected to be
met, but not when stratified by disease site. As a result, there was a desire at William
Osler to develop a plan that will achieve their targets the following year.
In order to achieve targets, the working group was willing to consider a number of
options if warranted, including:
• Assigning additional time to services to achieve volumes,
• Reserving time within the MSS specifically for targeted cases,
• Increasing ward beds available for targeted cases.
In order to evaluate these various solutions, the generic model was used as a decision
tool to compare possible solutions, and provide a recommendation to William Osler
executive to implement for the following year. For each of the targets, the model was
used in tandem with data analysis to determine the recommended course of action.
Total Joint Replacement Target: To achieve the TJR target, William Osler was willing
to consider any combination of the following options:
• TJR swing rooms: Three of the surgeons at Brampton were able to perform up
to six TJR cases a day if given access to two OR rooms. With a single OR team,
the surgeon and team would switch between the two OR rooms to maximise their
time, by not experiencing idle time due to room turnover.
Chapter 6. A Decision and Demonstrative Tool: Case Studies 160
• TJR dedicated rooms: A single OR dedicated to TJR, where a high volume surgeon
could perform four TJR cases in an elective day.
• Non-TJR rooms: An OR where cases other than TJR can be performed. These
would be used in tandem with swing rooms and TJR dedicated rooms to complete
other orthopaedic cases.
• Open orthopaedic rooms: Where surgeons can perform cases without restrictions.
Table 6.9 and figure 6.6 provide an example of some of the scenario results provided to
the working group for determining how to achieve the TJR targets. The table compares
the simulation results from the MSS schedule, as proposed from the short term solutions,
to results from running two swing rooms per week plus one, two or three TJR dedicated
rooms per week. As the number of TJR dedicated rooms increase, the number of com-
pleted TJR cases increases, at the expense of a reduction of total number of orthopaedic
cases completed. This decrease occurs because the increase volume of TJR cases reduces
access to ward beds for other orthopaedic cases, increasing their cancellation rate. In-
creasing the number of orthopaedic ward beds available was found to help reduce this
impact (not shown).
Based on the yearly target, and planned summer OR slowdowns and OR closures
throughout the year, it was determined that between 17 and 19 TJR cases per week were
required to achieve their yearly target. Over the eight week simulated time in the results
in table 6.9, a target of 125 cases would meet their yearly goal. The scenario with two
TJR reserved blocks achieves this volume with less effect on other orthopaedic cases than
reserving three ORs.
Cataract Target: The William Osler working group was also interested in knowing
how many ophthalmology blocks to reserve in order to achieve the cataract volume target.
Reserving five blocks per week for cataracts achieves the yearly target by completing 752
cases over the eight week simulation run.
Chapter 6. A Decision and Demonstrative Tool: Case Studies 161
MeasureOne TJR
Room
Two TJR
Rooms
Three TJR
Rooms
Throughput by Patient Type
Elective 0.93% 0.26% -0.22%
Urgent 2.98% 3.2% 3.07%
Throughput by Service
Dentistry 7.46% 5.89% 2.9%
Ear Nose and Throat 2.28% 1.93% 2.32%
General 2.38% 0.71% -0.76%
Gynaecology 5.08% 4.19% 7.24%
Ophthalmology 2.09% 2.38% 1.58%
Oral Surgery 8.05% 10.72% 5.93%
Orthopaedics -0.12% -0.55% -4.12%
TJR -11.95% 3.95% 13.17%
Plastics 5% 3.2% 3.44%
Urology 11.19% 5.21% 4.03%
Table 6.9: Results from comparing orthopaedic scenarios varying the number of TJR
reserved blocks.
Chapter 6. A Decision and Demonstrative Tool: Case Studies 162
!"
#"
$!"
$#"
%!"
%#"
&!"
&#"
'!"
'#"
()*" +,-" .-/" +0," 123" 456" 4,*"
!"#$%&#'(#)*+*',-'.%/#)0*1'
!"#$%&#'(#)*+*'2'345$0'6#$7'
!"
#"
$!"
$#"
%!"
%#"
&!"
&#"
'!"
'#"
()*" +,-" .-/" +0," 123" 456" 4,*"
789" 4:9" 4,2;3<5="5*/"40)26"465>" :260)?-/3<@"
!"
#"
$!"
$#"
%!"
%#"
&!"
&#"
'!"
'#"
()*" +,-" .-/" +0," 123" 456" 4,*"
!"#$%&#'(#)*+*',-'.%/#)0*1'
!"#$%&#'(#)*+*'2'8'69:':557'
!"
#"
$!"
$#"
%!"
%#"
&!"
&#"
'!"
'#"
()*" +,-" .-/" +0," 123" 456" 4,*"
!"#$%&#'(#)*+*',-'.%/#)0*1'
!"#$%&#'(#)*+*'2';'69:':557*'
!"
#"
$!"
$#"
%!"
%#"
&!"
&#"
'!"
'#"
()*" +,-" .-/" +0," 123" 456" 4,*"
!"#$%&#'(#)*+*',-'.%/#)0*1'
!"#$%&#'(#)*+*'2'<'69:':557*'
Figure 6.6: Difference in average census results from long term orthopaedic MSS changes.
Chapter 6. A Decision and Demonstrative Tool: Case Studies 163
Long Term Plan Selected: The following schedule changes were accepted by the
working group:
• Base MSS schedule: Including short term changes in general emergent time, an
additional gynaecological surgeon and adjustments to urology assignment based on
the bed model.
• Orthopaedic TJR Swing Rooms: Schedule two swing rooms per week to be shared
between the three surgeons who manage these swing rooms.
• Orthopaedic TJR Reserved Rooms: Reserve two orthopaedic blocks per week for
TJR cases.
• Cataract Reserved Rooms: Reserve five ophthalmology blocks per week for cataract
cases.
• Oncology: Current performance indicates that the capacity to achieve target exists,
but there is a need to focus on the mix of cases to better align with targets.
Monitoring the mix of oncology cases across all services to ensure meet all targets
within oncology was recommended. Remaining unassigned OR block time can be
assigned as needed to increase volumes of specific oncology procedures.
6.2.5 Discussion
At William Osler the generic model served two purposes: as a demonstrative tool and as
a decision tool.
As a demonstrative tool, the model illustrated that there was an opportunity for
William Osler to improve their scheduling and booking practices and other poor practices
which would release OR time for other priority activities. Practices to be improved
included fully booking allotted OR block time and accurately booking procedure times.
Chapter 6. A Decision and Demonstrative Tool: Case Studies 164
As a decision tool, the model was able to help decision makers determine how best to
use the unassigned blocks to achieve short term and long term objectives. In the short
term, the model was able to help determine how best to schedule the new gynaecological
surgeon and when to allocate general emergent time. In conjunction with a Monte Carlo
simulation model of bed use, a small schedule change for urology helped reduce the peak
number of surgical beds required and smooth average census over the week.
The model was used for longer term planning to determine how William Osler could
achieve yearly volumes of targeted cases. The model determined that a combination of
TJR swing rooms and reserved OR blocks for TJR and cataract cases was a realistic
solution to achieving their targets.
The recommendations were presented to the larger perioperative improvement initia-
tive team for final decision making. Since the proposed MSS require significant improve-
ments to their current practices, the implementation of the proposed MSS was delayed
until after headway was made on improving their numerous poor practices.
Chapter 7
Conclusions and Future Work
Perioperative patient flow is a high priority in many hospitals in Ontario and around the
world because of the drive to reduce waiting time for elective and urgent patients, and the
effect it has on the flow of the entire hospital. Furthermore, operating rooms are one of
the most expensive services in the hospital due to the staff, space and equipment required.
Despite this, tactical perioperative decision making is typically based on aggregate data,
historical performance and decision maker intuition. Basing these important decisions
on statistical rigour through simulation modelling will help improve performance of the
perioperative service and other areas of the hospital.
The research presented herein proposes a generic simulation model of perioperative
patient flow to inform tactical decision making for improved performance. The model is
successfully applied to six hospitals; demonstrating that not only is the model generic
but also that it can be a valuable and cost effective tool in decision making. This research
is characterised by three contributions summarised below.
A Generic Perioperative Simulation Model for Tactical Decisions
First, the design, development and implementation of the generic model itself is pre-
sented. As demonstrated in the literature review (chapter 3) a significant amount of
165
Chapter 7. Conclusions and Future Work 166
work has been done in regards to improving surgical patient flow using operational re-
search methodology. However, no generic tool was found that addressed the needs for
tactical decision making, specifically for developing the master surgical schedule (MSS)
and resource availability while taking into account multiple key performance indicators.
Further, the literature pointed to a need for a generic tactical decision tool that could
help inform these decisions. The need for a generic tool versus a specific tool is based on
constrained hospital budgets, lack of in-house expertise at many hospitals, and the need
to have a tool that is robust, flexible and user friendly.
The first contribution presents a generic model that was created based on typical pe-
rioperative flow processes and best practices. The generic model is data-driven, allowing
for a significant amount of flexibility to accommodate specific characteristics of a hospital
and their patients.
The proposed model was successfully tested at six different hospitals of various sizes
and processes. At four of the hospitals application of the model was successful as not only
was the model validated against historical performance, but testing of various possible
decision alternatives was performed and decision recommendations were made based on
their analysis. Furthermore, the recommendation was applied to real operations to at
least one of the sites, Juravinski.
To further strengthen the first contribution, the generic model continues to be used
by simulation modellers who were not part of the design - demonstrating that the model
has potential for wide spread use.
Finally, based on the experience of designing and implementing a generic simulation
model, a series of guidelines are provided. Based on simulation conference programs, the
interest and research in the field of generic simulation modelling is growing. However,
to date, little has been written on how to successfully design a generic model. These
guidelines add to this area of research by providing counsel to future endeavours in
hopes that a clear and concise standard methodology for generic simulation design be
Chapter 7. Conclusions and Future Work 167
developed.
A Process forAchieving User Acceptance
Secondly, the research provides valuable insight into achieving user acceptance of generic
tactical decision model. As is typical in large system modelling, validation of the model
to a hospital’s historical performance was fraught with challenges. These challenges
were most frequently experienced when a hospital’s processes were counter to model
assumptions based on best practice guidelines. As a result of the challenges experienced,
a method is proposed to help guide the validation process with the end users (decision
makers) in order to achieve user model acceptance and trust for decision implementation.
In a classical simulation modelling application, the current processes are modelled
as is. Then various solution options are tested to help decision makers choose. It is
proposed here that when considering a tactical decisions and generic models, exact details
of processes do not need to be modelled. In contrast, it is proposed that tactical decisions
should be made assuming that efficient and effective processes and procedures are in place
and followed. This however complicates user acceptance as the model does not exactly
reflect their current reality, both in model outputs and processes. In response, a set
of guidelines are provided to help achieve user acceptance of generic tactical decision
models.
In an era where demand for improved healthcare system planning and shrinking bud-
gets, generic tactical simulation provides a cost effective solution for decision support.
As discussed in the literature review, successful application and use of simulation models
in healthcare, yet alone generic models, is limited. The guidelines proposed are intended
to help improve the success rate of simulation models in affecting actual decisions by
achieving user acceptance through validation and building credibility.
Chapter 7. Conclusions and Future Work 168
A Decision and Demonstrative Tool
The third contribution demonstrates that the proposed generic model is a useful applica-
tion as a decision and demonstrative tool - to tease out inefficiencies in current processes
and to compare and evaluate various possible new processes, schedules and resources.
The primary intent of the generic model was as a decision tool such that comparisons
of different decision options on a number of performance indicators can be made for
improved decision making. However, during application at the sites a second use of the
generic model was found: as a demonstrative tool. When current practices at a hospital
are counter to best practice assumptions, the model can be used to present decision
makers and stakeholders the effect of these practices on performance. Model output
reporting can be used to demonstrate what performance could be achieved if practices
were improved. Similarly, it can be used to illustrate the effect on utilisation of resources
of these poor practices by reporting what resource capacity is required to achieve current
volumes. This typically results in freeing current capacity for additional volume.
The use of this generic model as a demonstrative tool adds to the simulation research
field by proposing and successfully applying an alternative use for simulation modelling.
Healthcare processes and systems are complicated by variation in practice. Furthermore,
the complex and interrelatedness of processes makes it difficult for decision makers to un-
derstand and quantify the effect of poor or varying practice on outcome measures. Using
simulation as a tool to demonstrate and quantify these effects has not been previously
done successfully. The case study presented demonstrates that not only can simulation
be used for this demonstrative purpose, but can also be used as a catalyst to help decision
makers account for improvements in these processes when making important mid and
long term decisions. The concept of simulation as a demonstrative tool can and should be
applied to other areas within the healthcare system, and other industries where inefficient
and poor practices are affecting outcomes.
Chapter 7. Conclusions and Future Work 169
7.1 Future Work
Based on the overreaching goal of the research and the limitations noted, some future
work is proposed below.
7.1.1 Additional Functionality and Addressing Limitations
From the limitations noted in section 4.2.3 some changes to the model may be required
to allow for additional flexibility. However, these should only be addressed if it is found
that the limitations are hindering model credibility at future hospitals.
Additional functionality can be added to the model to develop its use as a demon-
strative tool and to further improve user acceptance testing. Various reports based on
historical data and model outputs should be standardised and automated to facilitate
discussions regarding the effects of poor practices at the hospital. These reports should
provide data at various levels of detail down to the surgeon or operating room where
applicable.
7.1.2 Development of a Commercial Product
One of the aspirations of this research was to develop a generic model that could be
transformed into a commercial product, which can be easily applied to many hospitals in
Ontario and around the globe. A key requirement for this is to improve the usability of
the model in the hands of other hospitals and decision makers. This involves two aspects;
one to improve the ease of data entry and updating such that data from other hospitals
can be easily entered. The second involves the design of a user interface that allows
non-simulation experts to “play” with the model, observe outcomes of various options
and make informed decisions. Some further detail into these two aspects is provided in
the following subsections.
Chapter 7. Conclusions and Future Work 170
Improved Data Entry and Robustness of the Model
The current simulation model requires that the inputs be formatted in a very specific
way. The spread sheet used is adopted for each instance to incorporate the number of
ORs, surgeons, resources, etc. of the current hospital under study. This is currently a
manual process, requiring time and attention to detail to ensure that the file is updated
properly and completely. This manual process often leads to errors in the input file, which
translates to errors in the simulation model, leading to time wasted on debugging the
model and input file to find the error. Through implementation at six hospitals, many of
the possible data errors are now known, and thus easier to either check before importing
or identification when model errors occur. However, this knowledge and expertise is only
known to those who have experience with implementing the model.
To improve this process, a more automated generation of the spread sheets, including
a user interface and error checks, would reduce the number of mistakes as well as the
time to import and test data sets and scenarios.
Improved data entry and error checking will help improve the robustness of the cur-
rent model as implemented in Simul8. This would lead to decreased implementation
time for each subsequent implementation. It will also help bring the model towards a
commercialised product that can be used by hospitals with little support.
Design of a User Interface
Designing a user interface would be invaluable to the usability of the model across hospi-
tals. Most perioperative decision makers do not have strong data analysis or simulation
skills. Thus, for the generic model to be truly usable, it must cater to the typical skill
set and interest of a hospital decision maker.
The user interface should include the ability to make some changes to model param-
eters in an easy and visual fashion. This would allow the decision maker to “play” with
the model on their own to see the effect of different ideas for change.
Chapter 7. Conclusions and Future Work 171
Further, the user interface should include a set of default reports that display the key
outputs of the model that are understandable and usable to the decision maker. The
reports should display data using summary statistics, charts and graphs to allow for easy
analysis and comparison of decision options.
Overall, the user interface would allow limited ability for the decision maker to analyse
and compare scenarios on their own through simplified input parameter screens and a
set of easy to understand and analyse reports. A second layer of the user interface would
be for the simulation expert to make more complicated input changes and to set up and
calibrate the model.
There has been on-going discussion with Visual8, a research partner and simulation
consulting firm, around these future work topics. In collaboration with the Centre for
Research in Healthcare Engineering and test hospitals, Visual8 plans to further test the
generic simulation model and move the model towards a commercial product.
7.1.3 Additional Uses of the Model
With some minor adjustments the proposed generic model can provide a number of
additional uses, as discussed below.
As a Teaching Tool
The generic model would be a valuable teaching aid in healthcare systems and manage-
ment courses to demonstrate the effect of tactical perioperative decisions on key perfor-
mance indicators. The model could be presented as part of a case study of a hospital.
Students would study the case study information and propose one or more possible de-
cision options. The model would be used to study the effect of their proposed ideas and
guide their final recommendation. Alternatively, the model can be used as part of a
game, where teams of students compete against each other to make the best decisions
for the case study hospital.
Chapter 7. Conclusions and Future Work 172
A user interface would be required that allows students to adjust input parameters
according to their proposed solutions. One or more case studies would also be required,
including data sets, for use with the model.
As a Research Tool
A second potential use for the model would be as a research tool. The model can
be used by operational researchers to test out proposed decision options derived from
other mathematical models such as alternative scheduling algorithms, optimal MSS and
resource capacity levels. For example, a researcher may have a developed a scheduling
algorithm that he believes will improve the hospital’s throughput. The researcher can
code the algorithm into the model and input hospital specific data to test whether the
algorithm results in any unforeseen consequences on aspects of the perioperative system.
This additional use has in fact already been done during the application at Brampton
and Juravinski as described in chapter 6. At each site, a simulation optimisation bed
resource model proposed a MSS that would reduce the peak demand for inpatient beds
(Liu, 2012). The proposed MSS was inputted into the generic model to determine how
it compared to other options under consideration in terms of throughput, cancellation
rates and inpatient census.
As an Operational Decision Tool
The intent of the generic model presented was to help inform tactical decisions. However,
every day hospitals face difficult online operational decisions such as whether to cancel
an elective case when no bed is available. Typically, this decision is made the day the
case is scheduled, resulting in unused elective OR time, as either the time is used for
urgent cases or the OR is closed early. The cancellation is also an inconvenience for the
patient who is likely already at the hospital ready for surgery. The patient will have to
reschedule their procedure and re-experience the mental and physical duress of preparing
Chapter 7. Conclusions and Future Work 173
for surgery. If the hospital was able to better forecast bed availability, then cancellation
decisions can be made one or more days ahead of time. This would allow the hospital to
adjust the surgical schedule such that the elective time does not go unused and patients
would be less inconvenienced.
It is proposed that the generic model formulation could be used as a base to develop an
operational decision model. The model would allow hospital decision makers to predict
bed and other perioperative resource availability based on patients currently admitted,
the current elective patient schedule and predicted urgent cases. Decision makers would
interact with the model to determine if cancellations can be avoided and/or inpatient bed
census can be smoothed by adjusting the elective schedule. For instance, if the model
predicts that inpatient census will be high on Thursday likely resulting in cancellations,
decision makers can reschedule some patients who require admission in advance. Further,
the newly available elective OR time could be used to schedule upcoming cases that do
not require admission. Alliteratively, the decision makers may decide to temporarily
increase the capacity of a ward to allow for the scheduled cases to proceed as scheduled.
In this case, the decision makers can use the model to determine how long the additional
beds would likely be required.
Bibliography
Adan, Ivo J.B.F, Jos A Bekkers, Nico Dellaert, Jos Vissers, Xiaoting Yu. 2009. Patient
mix optimisation and stochastic resource requirements: A case study in cardiothoracic
surgery planning. Health Care Management Science 12 129–141.
Banks, J, John S Carson. 1995. Descrete-event system simulation. Englewood Cliffs, NJ:
Prentice-Hall, Inc. .
Belien, Jeroen, Erik Demeulemeester. 2007. Building cyclic master surgery schedules
with leveled resulting bed occupancy. European Journal of Operational Research 176
1185–1204.
Belien, Jeroen, Erik Demeulemeester. 2008. A branch-and-price approach for integrating
nurse and surgery scheduling. European Journal of Operational Research 189 652–668.
Belien, Jeroen, Erik Demeulemeester, Brecht Cardoen. 2006. Visualizing the demand for
various resources as a function of the master surgery schedule: A case study. Journal
of Medical Systems 30 343–350.
Belien, Jeroen, Erik Demeulemeester, Brecht Cardoen. 2009. A decision support system
for cyclic master surgery scheduling with multiple objectives. Journal of Scheduling
12 147–161.
Berchtold, Gunter, Helga Blaschke, Friedrich Hanssmann, Franz Liebl. 1994. Simula-
174
BIBLIOGRAPHY 175
tion modeling as a tool to evaluate alternative configurations of clinical laboratories.
Simulation 63(2) 108–120.
Blake, John T, Michael W Carter. 1997. Surgical process scheduling: a structured review.
Journal of the Society for Health Systems 5(3) 17–30.
Blake, John T, Michael W Carter, Linda O’Brien-Pallas, Linda McGillis-Hall. 1995. A
surgical process management tool. Medinfo 8(1) 527–31.
Blake, John T, Joan Donald. 2002. Mount sinai hospital uses integer programming to
allocate operating room time. Interfaces 32(2) 63–73.
Boldy, Duncan. 1987. The relationship between decision support systems and operational
research: Health care examples. European Journal of Operational Research 29 128–134.
Brown, Nancy, S Powers. 2000. Simulation in a box: a generic reusable maintenance
model. Proceedings of the 2000 Winter Simulation Conference 1056.
Calichman, Murray V. 2005. Creating an optimal operating room schedule. AORN
journal 81(3) 580–588.
Cardoen, Brecht, Erik Demeulemeester, Jeroen Belien. 2010. Operating room planning
and scheduling: A literature review. European Journal of Operational Research 201(3)
921–932.
Carter, Michael W, John T Blake. 2005. Using simulation in an acute-care hospital:
easier said than done. Operations Research and Health Care 70 191–215.
Centeno, Martha A, Ronald Giachetti, Richard Linn, A Ismail. 2003. A simulation-ilp
based tool for scheduling er staff. Proceedings of the 2003 Winter Simulation Confer-
ence 1930–1937.
BIBLIOGRAPHY 176
Cope, Dayana, Mohamed Sam Fayez, Mansooreh Mollaghasemi, Assem Kaylani. 2007.
Supply chain simulation modeling made easy: an innovative approach. Proceedings of
the 2007 Winter Simulation Conference 1887–1896.
Costa, A.X, S.A Ridley, A.K Shahani, Paul R Harper, V De Senna, M.S Nielson. 2003.
Mathematical modelling and simulation for planning critical care capacity. Anaesthesia
58 320–327.
Dexter, Franklin, Rodney D Traub. 2002. How to schedule elective surgical cases into
specific operating rooms to maximize the efficiency of use of operating room time.
Anesthesia & Analgesia 94 933–42.
Drake, Glenn R, Jeffrey S Smith. 1996. Simulation system for real-time planning, schedul-
ing, and control. Proceedings of the 1996 Winter Simulation Conference 1083–1090.
Fletcher, Adrian, D Halsall, S Huxham, Dave Worthington. 2007. The dh accident and
emergency department model: a national generic model used locally. Journal of the
Operational Research Society 58 1554–1562.
Fletcher, Adrian, Dave Worthington. 2009. What is a ‘generic’ hospital model?—a com-
parison of ‘generic’ and ‘specific’ hospital models of emergency patient flows. Health
Care Management Science 12(4) 374–391.
Guerriero, Francesca, Rosita Guido. 2011. Operational research in the management of
the operating theatre: a survey. Health Care Management Science 14(1) 89–114.
Gupta, Diwakar. 2007. Surgical suites’ operations management. Production and Opera-
tions Management 16(6) 689–700.
Guru, Ashu, Paul Savory. 2004. A template-based conceptual modeling infrastructure
for simulation of physical security systems. Proceedings of the 2004 Winter Simulation
Conference 866–873.
BIBLIOGRAPHY 177
Hamilton Health Sciences. 2011. Juravinski hospital. URL
http://www.hamiltonhealthsciences.ca/body.cfm?ID=229. Retrieved July
26, 2011.
Hans, Erwin W, Mark Van Houdenhoven, Peter J.H Hulshof. 2012. A framework for
healthcare planning and control. Handbook of Healthcare System Scheduling Springer
internaltional Series in Operations Research and Management Science 168 301–320.
Harper, Paul R. 2002. A framework for operational modelling of hospital resources.
Health Care Management Science 5(3) 165–173.
Houdenhoven, Mark Van, Jeroen M van Oostrum, Gerhard Wullink, Erwin W Hans,
Johann L Hurink, Jan Bakker, Geert Kazemier. 2008. Fewer intensive care unit refusals
and a higher capacity utilization by using a cyclic surgical case schedule. Journal of
Critical Care 23(2) 222–226.
Jain, Sanjay. 2008. Tradeoffs in building a generic supply chain simulation capability.
Proceedings of the 2008 Winter Simulation Conference 1873–1881.
Kaylani, Assem, Mansooreh Mollaghasemi, Dayana Cope, Mohamed Sam Fayez, Ghaith
Rabadi, Martin J Steele. 2008. A generic environment for modelling future launch
operations—gem-flo: a success story in generic modelling. Journal of the Operational
Research Society 59 1312–1320.
Kharraja, Said, Pascal Albert, Sondes Chaabane. 2006. Block scheduling: Toward a mas-
ter surgical schedule. 2006 International Conference on Service Systems and Service
Management 429–435.
Kim, Seung-Chul, Ira Horowitz. 2002. Scheduling hospital services: the efficacy of
elective-surgery quotas. Omega 30 335–346.
BIBLIOGRAPHY 178
Kim, Seung-Chul, Ira Horowitz, Karl K Young, Thomas A Buckley. 2000. Flexible
bed allocation and performance in the intensive care unit. Journal of Operations
Management 18(4) 427–443.
Kotiadis, Kathy, Stewart Robinson. 2008. Conceptual modelling: knowledge acquisition
and model abstraction. Proceedings of the 2008 Winter Simulation Conference 951–
958.
Kuo, Paul C, Rebecca A Schroeder, Samuel Mahaffey, R. Randall Bollinger. 2003. Opti-
mization of operating room allocation using linear programming techniques. Journal
of the American College of Surgeons 197(6) 889–895.
Kuzdrall, Paul J, N.K Kwak, Homer H Schmitz. 1981. Simulating space requirements
and scheduling policies in a hospital surgical suite. Simulation 36 163.
Kwak, N.K, Paul J Kuzdrall, Homer H Schmitz. 1976. The gpss simulation of scheduling
policies for surgical patients. Management Science 22(9) 982–989.
Law, Averill M. 2005. How to build valid and credible simulation models. Proceedings of
the 37th conference on Winter simulation 24–32.
Law, Averill M, W. David Kelton. 2000. Simulation modeling and analysis. McGraw-Hill
Third Edition.
Lebowitz, Philip. 2003. Schedule the short procedure first to improve or efficiency. AORN
78(4) 651–659.
Liu, Tian Mu. 2012. A generic bed planning model. Masters Thesis - Department of
Mechanical and Industrial Engineering, Univeristy of Toronto .
Lowery, Julie C. 1992. Simulation of a hospital’s surgical suite and critical care area.
Proceedings of the 24th conference on Winter simulation 1071–1078.
BIBLIOGRAPHY 179
Lowery, Julie C. 1996. Introduction to simulation in health care. Proceedings of the 1996
Winter Simulation Conference 78–84.
Lowery, Julie C. 1998. Getting started in simulation in healthcare. Proceedings of the
1998 Winter Simulation Conference 31–35.
MacLeod, Hugh, Alan Hudson, Sarah Kramer, Murray Martin. 2009. The times they
are a-changing: What worked and what we learned in deploying ontario’s wait time
information system. Healthcare Quarterly 12(Special Issue) 8–15.
Marcon, Eric, Franklin Dexter. 2006. Impact of surgical sequencing on post anesthesia
care unit staffing. Health Care Management Science 9 87–98.
Marcon, Eric, Said Kharraja, Nicole Smolski, Brigitte Luquet, Jean Paul Viale. 2003.
Determining the number of beds in the postanesthesia care unit: a computer simulation
flow approach. Anesthesia & Analgesia 96 1415–23.
Martis, Morvin Savio. 2006. Validation of simulation based models: a theoretical outlook.
The Electronic Journal of Business Research Methods 4(1) 39–46.
McAleer, W.E, J.A Turner, D Lismore, I.A Naqvi. 1995. Simulation of a hospital’s theatre
suite. Journal of Management in Medicine 9(5) 14–26.
Miller, Martin J, David M Ferrin, Marcia G Messer. 2004. Fixing the emergency de-
partment: a transformational journey with edsim. Proceedings of the 2004 Winter
Simulation Conference .
Modernisation, NHS. 2001. Tackling cancelled operations: iterim guidance from the nsh
modernisation agency theatre project. NHS Modernisation London .
Moreno, L, R.M Aguilar, C.A Martın, J.D Pineiro, J.I Estevez, J.F Sigut, J.L Sanchez,
V.I Jimenez. 1999. Patient-centered simulation tool for aiding in hospital management.
Simulation Practice and Theory 7 373–393.
BIBLIOGRAPHY 180
Mount Sinai Hospital. 2011. Fast facts. URL http://www.mountsinai.on.ca. Retrieved
July 26, 2011.
Nelson, Elaine A. 1983. The use of a generic simulation model in evaluating manufac-
turing processes. Computers & Industrial Engineering 7(3) 217–224.
Nguyen, J.M, P Six, D Antonioli, P Glemain, G Potel, P Lombrail, P Le Beux. 2005.
A simple method to optimize hospital beds capacity. International Journal of Medical
Informatics 74 39–49.
Ozdemirel, Nur E. 1991. Measuring the user accepatance of generic manufacturing sim-
ulation models by review of modeling assumptions. Proceedings of the 23rd conference
on Winter simulation 419–427.
Paul, Ray J, Jasna Kuljis. 1995. A generic simulation package for organising outpatient
clinics. Proceedings of the 1995 Winter Simulation Conference 1043–1047.
Pidd, Michael. 1992. Guidelines for the design of data driven generic simulators for
specific domains. Simulation 59 237–243.
Pidd, Michael. 1999. Just modeling through: A rough guide to modeling. Interfaces 29
118–132.
Pitt, Martin A. 1997. A generalised simulation system to support strategic resource
planning in healthcare. Proceedings of the 1997 Winter Simulation Conference 1155–
1162.
Ramis, Francisco J, Jorge L Palma, Felipe F Baesler. 2001. The use of simulation for pro-
cess improvement at an ambulatory surgery center. Proceedings of the 33nd conference
on Winter simulation 1401–1404.
Ramis, Francisco J, Jorge L Palma, Vıctor F Estrada, Gloria Coscolla. 2002. A simulator
BIBLIOGRAPHY 181
to improve patient’s service in a network of clinic laboratories. Proceedings of the 2002
Winter Simulation Conference 1444–1447.
Robinson, Stewart. 2004. Simulation: The practice of model development and use. John
Wiley & Sons .
Robinson, Stewart. 2007a. Conceptual modelling for simulation part i: definition and
requirements. Journal of the Operational Research Society 59 278–290.
Robinson, Stewart. 2007b. Conceptual modelling for simulation part ii: a framework for
conceptual modelling. Journal of the Operational Research Society 59 291–304.
Robinson, Stewart, Roger Brooks, Kathy Kotiadis, Durk-Jourke van der Zee. 2010. Con-
ceptual modeling for discrete-event simulation. CRC Press .
Robinson, Stewart, Richard E Nance, Ray J Paul, Michael Pidd. 2004. Simulation model
reuse: definitions, benefits and obstacles. Simulation modelling practice and theory 12
479–494.
Santibanez, Pablo, Mehmet Begen, Derek Atkins. 2007. Surgical block scheduling in a
system of hospitals: an application to resource and wait list management in a british
columbia health authority. Health Care Management Science 10 269–282.
Saremi, Alireza, Payman Jula, Tarek ElMekkawy, G Gary Wang. 2013. Appointment
scheduling of outpatient surgical services in a multistage operating room department.
International Journal of Production Economics 141(2) 646–658.
Schroer, Bernard J, Phillip A Farrington, James J Swain, Dawn R Utley. 1996. A generic
simulator for modeling manufacturing modules. Proceedings of the 1996 Winter Sim-
ulation Conference 1155–1160.
Sciomachen, Anna, Elena Tanfani, Angela Testi. 2005. Simulation models for optimal
schedules of operating theatres. International Journal of Simulation 6(12-13).
BIBLIOGRAPHY 182
Sinreich, David, Yariv Marmor. 2004. A simple and intuitive simulation tool for ana-
lyzing emergency department operations. Proceedings of the 2004 Winter Simulation
Conference .
Sinreich, David, Yariv Marmor. 2005. Emergency department operations: the basis for
developing a simulation tool. IIE Transactions 37(3) 233–245.
Sobolev, Boris G, David Harel, Christos Vasilakis, Adrian Levy. 2008. Using the state-
charts paradigm for simulation of patient flow in surgical care. Health Care Manage-
ment Science 11 79–86.
Sobolev, Boris G, V Sanchez, Christos Vasilakis. 2011. Systematic review of the use
of computer simulation modeling of patient flow in surgical care. Journal of Medical
Systems 35 1–16.
St. Michael’s Hospital. 2011. Programs and focus areas - emergency department. URL
http://www.stmichaelshospital.com/programs/emergency/index.php. Retrieved
July 26, 2011.
Steele, Martin J, Mansooreh Mollaghasemi, Ghaith Rabadi, Grant Cates. 2002. Generic
simulation models of reusable launch vehicles. Proceedings of the 2002 Winter Simu-
lation Conference 747–753.
Tanfani, Elena, Angela Testi. 2010. A pre-assignment heuristic algorithm for the master
surgical schedule problem (mssp). Annals of Operations Research 178(1) 105–119.
Testi, Angela, Elena Tanfani. 2008. Tactical and operational decisions for operating room
planning: Efficiency and welfare implications. Health Care Management Science 12(4)
363–373.
Testi, Angela, Elena Tanfani, Giancarlo Torre. 2007. A three-phase approach for oper-
ating theatre schedules. Health Care Management Science 10 163–172.
BIBLIOGRAPHY 183
van Oostrum, Jeroen M, Eelco Bredenhoff, Erwin W Hans. 2008a. Managerial impli-
cations and suitability of a master surgical scheduling approach. Erasmus School of
Economics (ESE) 1–22.
van Oostrum, Jeroen M, Mark Van Houdenhoven, Johann L Hurink, Erwin W Hans,
Gerhard Wullink, Geert Kazemier. 2008b. A master surgical scheduling approach for
cyclic scheduling in operating room departments. OR spectrum 30(2) 355–374.
VanBerkel, Peter T, Richard J Boucherie, Erwin W Hans, Johann L Hurink, Nelly Lit-
vak. 2010. A survey of health care models that encompass multiple departments.
International Journal of Health Management and Information 1(1) 37–69.
VanBerkel, Peter T, Richard J Boucherie, Erwin W Hans, Johann L Hurink, Wineke
A. M van Lent, Wim H van Harten. 2011a. Accounting for inpatient wards when
developing master surgical schedules. Anesthesia & Analgesia 112(6) 1472–1479.
VanBerkel, Peter T, Richard J Boucherie, Erwin W Hans, Johann L Hurink, Wineke
A. M van Lent, Wim H van Harten. 2011b. An exact approach for relating recovering
surgical patient workload to the master surgical schedule. Journal of the Operation
Research Society 62(3) 1851–1860.
Vasilakis, Christos, Boris G Sobolev, L Kuramoto, Adrian Levy. 2007. A simulation
study of scheduling clinic appointments in surgical care: individual surgeon versus
pooled lists. Journal of the Operational Research Society 58(2) 202–211.
Vissers, Jos, Ivo J.B.F Adan, Jos A Bekkers. 2005. Patient mix optimization in tactical
cardiothoracic surgery planning: a case study. IMA Journal of Management Mathe-
matics 16(3) 281–304.
Wachtel, Ruth E, Franklin Dexter. 2008. Tactical increases in operating room block time
for capacity planning should not be based on utilization. Anesthesia & Analgesia 106
215–26.
BIBLIOGRAPHY 184
William Osler Health System. 2012. Hospitals. URL
http://www.williamoslerhc.on.ca/body.cfm?id=8. Retrieved December 22,
2012.
Wright, M.B. 1987. The application of a surgical bed simulation model. European Journal
of Operational Research 32 26–32.
Zellermeyer, Valerie. 2005. Report of the surgical process analysis and improvement
expert panel. health.gov.on.ca 1–72.
Zhang, B, P Murali, M Dessouky, D Belson. 2009. A mixed integer programming approach
for allocating operating room capacity. Journal of the Operational Research Society
60 663–673.
Appendices
185
Appendix A
Conceptual Model Tables
Table A.1: Model scope: components identified within
the boundary of the study.
Component Include/
Exclude
Justification
Pre-Operative
Wait list man-
agement
List ordering Include Experimental factor, required for
evaluating scheduling rules.
Schedule proce-
dure
Include Experimental factor, required for
evaluating scheduling rules.
Pre-op clinic and
other visits
Scheduling Exclude Clinic visit scheduling is done
independently of OR schedul-
ing, though coordinated with the
scheduled OR date.
Continued on next page . . .
186
Appendix A. Conceptual Model Tables 187
Table A.1 – Continued
Component Include/
Exclude
Justification
Clinic visit Exclude Clinic visit is independent of OR
schedule. Though there is a
chance of cancellation of OR date,
the cancellation is incorporated in
the model elsewhere. Data on vis-
its/tests completed and times not
usually available.
Operative
Before Entering
the OR
Registration and
Admitting
Exclude Does not impact patient flow at
tactical level. Data for time and
flow not usually available.
Patient Prepara-
tion
Exclude Does not impact patient flow at
tactical level. Data for time and
flow not usually available.
Tests and Diag-
nostics
Exclude Does not impact patient flow at
tactical level. Data for time and
flow not usually available.
Nurse Assess-
ment
Exclude Does not impact patient flow at
tactical level. Data for time and
flow not usually available.
Anaesthesia As-
sessment
Exclude Does not impact patient flow at
tactical level. Data for time and
flow not usually available.
Continued on next page . . .
Appendix A. Conceptual Model Tables 188
Table A.1 – Continued
Component Include/
Exclude
Justification
Surgeon Assess-
ment
Exclude Does not impact patient flow at
tactical level. Data for time and
flow not usually available.
Anaesthesia
Room
Exclude With proper coordination and
planning, the anaesthesia room
should not greatly affect flow at
the tactical level. Including this
component would also lead to
increased and unnecessary com-
plexity. Data is also not always
available.
Operating Room Include Required component to overall
flow of patients. Experimental
Factor.
Post-Operative
Immediate
Post-Operative
Recovery
Post-
Anaesthesia
Care Unit
Include Required component to overall
flow of patients. Experimental
Factor.
Same Day
Surgery Recov-
ery Unit
Exclude Does not impact patient flow at
tactical level. Only used by pa-
tients discharging home same day.
Continued on next page . . .
Appendix A. Conceptual Model Tables 189
Table A.1 – Continued
Component Include/
Exclude
Justification
Post-Operative
Recovery
Critical Care
Units
Include Required component to overall
flow of patients. Experimental
Factor.
Step Down
Units
Include Required component to overall
flow of patients. Experimental
Factor.
Inpatient Ward
Units
Include Required component to overall
flow of patients. Experimental
Factor.
Appendix A. Conceptual Model Tables 190
Tab
leA
.2:
Lev
elof
det
ail:
det
ail
incl
uded
for
each
com
-
pon
ent
incl
uded
inth
em
odel
.
Com
pon
ent
Det
ail
Incl
ude/
Excl
ude
Com
men
tJust
ifica
tion
Pre
-Op
erat
ive
Pat
ient
Arr
ival
Incl
ude
Model
led
asan
ex-
pon
enti
aldis
trib
uti
on.
Eac
hse
rvic
ean
dpa-
tien
tty
pe
(ele
ctiv
e,
inpat
ient,
urg
ent)
has
it’s
own
arri
val
rate
.
Exp
onen
tial
isea
syto
under
stan
dan
dca
lcu-
late
bas
edon
aver
age.
Dat
aon
actu
alar
riva
l
pat
tern
sis
usu
ally
not
avai
lable
todet
erm
ine
dis
trib
uti
on.
Con
tinued
onnex
tpag
e..
.
Appendix A. Conceptual Model Tables 191
Tab
leA
.2–
Con
tinued
Com
pon
ent
Det
ail
Incl
ude/
Excl
ude
Com
men
tJust
ifica
tion
Wai
tlist
man
-
agem
ent
Ele
ctiv
ew
ait
list
sor
ganis
a-
tion
By
serv
ice
orby
surg
eon
Incl
ude
Model
allo
ws
choi
ceto
model
wai
tlist
by
ser-
vic
e(p
ool
ed)
orby
surg
eon.
Sch
eduling
and
pa-
tien
tflow
are
affec
ted
by
the
orga
nis
atio
nof
the
wai
tlist
.In
par
-
ticu
lar
wai
ting
tim
e
and
the
elec
tive
sched
-
ule
.
Lis
tor
der
ing
Ele
ctiv
ew
ait
list
order
ing
Incl
ude
Model
led
assi
mple
or-
der
ing
rule
s(F
IFO
,
pri
orit
y,dea
dline)
.
Sim
ple
tom
odel
and
under
stan
d.
Indiv
id-
ual
surg
eon
wai
tlist
pra
ctic
esar
eva
ried
and
diffi
cult
tode-
scri
be.
Con
tinued
onnex
tpag
e..
.
Appendix A. Conceptual Model Tables 192
Tab
leA
.2–
Con
tinued
Com
pon
ent
Det
ail
Incl
ude/
Excl
ude
Com
men
tJust
ifica
tion
Inpat
ient
wai
t
list
order
ing
Incl
ude
Model
led
assi
mple
or-
der
ing
rule
s(F
IFO
,
pri
orit
y,dea
dline)
.
Sim
ple
tom
odel
and
under
stan
d.
Urg
ent
wai
tlist
order
ing
Incl
ude
Model
led
assi
mple
or-
der
ing
rule
s(F
IFO
,
pri
orit
y,dea
dline)
.
Sim
ple
tom
odel
and
under
stan
d.
Sch
edule
pro
ce-
dure
Mas
ter
OR
Sch
edule
Incl
ude
OR
sched
ule
,in
dic
at-
ing
oper
atin
gti
mes
ofO
R,
and
funct
ion
(ele
ctiv
eblo
ck,
ur-
gent,
emer
gent,
on
call).
Key
com
pon
ent
ofpa-
tien
tflow
isth
eop
er-
atin
gsc
hed
ule
.
Con
tinued
onnex
tpag
e..
.
Appendix A. Conceptual Model Tables 193
Tab
leA
.2–
Con
tinued
Com
pon
ent
Det
ail
Incl
ude/
Excl
ude
Com
men
tJust
ifica
tion
Ele
ctiv
esc
hed
ul-
ing
rule
s
Incl
ude
Sch
eduling
rule
sin
-
cludin
gty
pe
ofca
ses
tob
esc
hed
ule
din
OR
sched
ule
,bas
ed
onca
sety
pe,
pro
ce-
dure
typ
e,ad
mis
sion
,
oran
aest
hes
iaty
pe.
Key
com
pon
ent
ofpa-
tien
tflow
isth
eop
er-
atin
gsc
hed
ule
.
Inpat
ient
sched
uling
pro
cedure
s
Incl
ude
Rule
onhow
tohan
-
dle
inpat
ient
sched
ul-
ing
-as
par
tof
urg
ent
sched
ule
orel
ecti
ve.
Key
com
pon
ent
ofpa-
tien
tflow
isth
eop
er-
atin
gsc
hed
ule
.
Con
tinued
onnex
tpag
e..
.
Appendix A. Conceptual Model Tables 194
Tab
leA
.2–
Con
tinued
Com
pon
ent
Det
ail
Incl
ude/
Excl
ude
Com
men
tJust
ifica
tion
Urg
ent
sched
ul-
ing
pro
cedure
s
Incl
ude
Rule
son
how
tom
an-
age
urg
ent/
emer
gent
pat
ient
list
and
how
tosc
hed
ule
pro
cedure
s
wit
hin
the
OR
sched
-
ule
.
Key
com
pon
ent
ofpa-
tien
tflow
isth
eop
er-
atin
gsc
hed
ule
.
Pro
cedure
Ord
erIn
clude
Sim
ple
order
ing
rule
s
bas
edon
pro
cedure
lengt
h,an
dad
mis
sion
,
anae
sthes
iaan
dpre
vi-
ousl
yca
nce
lled
flag
s.
Key
com
pon
ent
ofpa-
tien
tflow
isth
eop
er-
atin
gsc
hed
ule
and
the
order
itis
per
form
ed.
Op
erat
ive
Con
tinued
onnex
tpag
e..
.
Appendix A. Conceptual Model Tables 195
Tab
leA
.2–
Con
tinued
Com
pon
ent
Det
ail
Incl
ude/
Excl
ude
Com
men
tJust
ifica
tion
Can
cellat
ion
Dec
isio
ns
Due
tono
bed
Incl
ude
Can
cel
pat
ients
on
day
ofsc
hed
ule
d
pro
cedure
ifno
pos
t-
oper
ativ
eb
edis
avai
lable
.D
ecis
ion
det
erm
ined
onin
-
putt
edru
les
bas
edon
pat
ient
typ
ean
dif
pre
vio
usl
yca
nce
lled
.
Key
com
pon
ent
of
pat
ient
flow
and
thro
ugh
put.
Model
outp
ut.
Con
tinued
onnex
tpag
e..
.
Appendix A. Conceptual Model Tables 196
Tab
leA
.2–
Con
tinued
Com
pon
ent
Det
ail
Incl
ude/
Excl
ude
Com
men
tJust
ifica
tion
Due
toO
Rblo
ck
tim
eov
erru
n
Incl
ude
Can
cel
pat
ients
on
day
ofsc
hed
ule
dpro
-
cedure
ifco
mple
ting
the
case
will
mea
n
exce
edin
gal
low
able
over
tim
eof
the
OR
blo
ck.
Dec
isio
nde-
term
ine
onin
putt
ed
rule
sbas
edon
pat
ient
typ
e.
Key
com
pon
ent
of
pat
ient
flow
and
thro
ugh
put.
Model
outp
ut.
Due
tom
ore
ur-
gent
case
Incl
ude
Can
cel
pat
ients
on
day
ofsc
hed
ule
d
pro
cedure
ifa
mor
e
urg
ent
case
requir
es
the
OR
tim
e.
Key
com
pon
ent
of
pat
ient
flow
and
thro
ugh
put.
Model
outp
ut.
Con
tinued
onnex
tpag
e..
.
Appendix A. Conceptual Model Tables 197
Tab
leA
.2–
Con
tinued
Com
pon
ent
Det
ail
Incl
ude/
Excl
ude
Com
men
tJust
ifica
tion
Due
toot
her
rea-
sons
Incl
ude
Can
cel
pat
ients
on
day
ofsc
hed
ule
dpro
-
cedure
bas
edon
the
inputt
edpro
bab
ilit
y
ofa
pat
ient
bei
ng
can-
celled
for
are
ason
not
list
edab
ove.
Rea
sons
can
incl
ude
pat
ient
not
read
yfo
rsu
rger
y,
staff
/surg
eon/a
nae
sthes
ia
not
avai
lable
,et
c.
Key
com
pon
ent
ofpa-
tien
tth
rough
put
and
flow
.
Con
tinued
onnex
tpag
e..
.
Appendix A. Conceptual Model Tables 198
Tab
leA
.2–
Con
tinued
Com
pon
ent
Det
ail
Incl
ude/
Excl
ude
Com
men
tJust
ifica
tion
Op
erat
ing
Room
Pro
cedure
tim
eIn
clude
Tot
alti
me
pat
ient
inO
Rfo
rpro
cedure
,
from
pat
ient
ente
red
room
unti
lpat
ient
read
yto
be
tran
s-
ferr
ed.
Does
not
incl
ude
del
ayti
me.
Key
com
pon
ent
ofpa-
tien
tflow
.
Del
ayIn
clude
Del
ayof
pat
ient
tran
s-
fer
top
ost-
oper
ativ
e
unit
sdue
toca
pac
ity.
Key
com
pon
ent
ofpa-
tien
tflow
.
Turn
arou
nd
tim
e
Incl
ude
Ave
rage
turn
arou
nd
tim
eby
serv
ice
be-
twee
nca
ses.
Key
com
pon
ent
ofpa-
tien
tflow
.
Pos
t-O
per
ativ
e
Con
tinued
onnex
tpag
e..
.
Appendix A. Conceptual Model Tables 199
Tab
leA
.2–
Con
tinued
Com
pon
ent
Det
ail
Incl
ude/
Excl
ude
Com
men
tJust
ifica
tion
Pos
t-O
per
ativ
e
Rec
over
y
Cri
tica
lC
are
Unit
s
Rec
over
yti
me
Incl
ude
Med
ical
lynec
essa
ry
reco
very
tim
e.
Key
com
pon
ent
of
pat
ient
flow
asaf
-
fect
sav
aila
bilit
yof
reso
urc
es.
Ste
pD
own
Unit
s
Rec
over
yti
me
Incl
ude
Med
ical
lynec
essa
ry
lengt
hof
stay
.
Key
com
pon
ent
of
pat
ient
flow
asaf
-
fect
sav
aila
bilit
yof
reso
urc
es.
Inpat
ient
War
d
Unit
s
Rec
over
yti
me
Incl
ude
Med
ical
lynec
essa
ry
lengt
hof
stay
.
Key
com
pon
ent
of
pat
ient
flow
asaf
-
fect
sav
aila
bilit
yof
reso
urc
es.
Con
tinued
onnex
tpag
e..
.
Appendix A. Conceptual Model Tables 200
Tab
leA
.2–
Con
tinued
Com
pon
ent
Det
ail
Incl
ude/
Excl
ude
Com
men
tJust
ifica
tion
Del
ayfo
rot
her
leve
lsof
care
outs
ide
ofm
odel
scop
e
Wai
tti
me
in
bed
for
other
reso
urc
es(A
LC
)
Incl
ude
Tim
epat
ient
must
wai
tin
am
odel
led
bed
for
adow
nst
ream
reso
urc
eth
atis
not
incl
uded
inth
em
odel
.
Key
com
pon
ent
of
pat
ient
flow
asaf
-
fect
sav
aila
bilit
yof
reso
urc
es.
Bed
Man
agem
ent
Rec
over
yti
me
Incl
ude
Med
ical
lynec
essa
ry
lengt
hof
stay
.
Key
com
pon
ent
of
pat
ient
flow
asaf
-
fect
sav
aila
bilit
yof
reso
urc
es.
Off
serv
icin
gIn
clude
Model
offse
rvic
ing
of
pat
ients
toot
her
unit
s
ifb
edin
unav
aila
ble
.
Bas
edon
inputt
edoff
serv
icin
gru
les.
Key
com
pon
ent
of
pat
ient
flow
asaf
-
fect
sav
aila
bilit
yof
reso
urc
es.
Con
tinued
onnex
tpag
e..
.
Appendix A. Conceptual Model Tables 201
Tab
leA
.2–
Con
tinued
Com
pon
ent
Det
ail
Incl
ude/
Excl
ude
Com
men
tJust
ifica
tion
Nurs
ing
rati
oIn
clude
Model
will
adju
st
nurs
ing
rati
ow
hen
pat
ient
isoff
serv
iced
tore
flec
tdiff
eren
t
leve
lof
care
requir
ed
when
offse
rvic
ed.
Key
com
pon
ent
of
pat
ient
flow
asaf
-
fect
sav
aila
bilit
yof
reso
urc
es.
Del
ayIn
clude
Del
ayof
pat
ient
tran
s-
fer
tonex
tunit
due
to
capac
ity.
Key
com
pon
ent
of
pat
ient
flow
asaf
-
fect
sav
aila
bilit
yof
reso
urc
es.
Appendix A. Conceptual Model Tables 202
Table A.3: Model Assumptions.
Assumption Description Reason
Pre-
operative
Activities
Activities that occur before the
day of surgery, such as the pre-
operative clinic appointment, is
not included
Activities may often cause an op-
erational issue, in terms of coordi-
nation with OR schedule, and run
largely independent of the periop-
erative patient flow.
Day of pre-
operative
activities
Activities that occur on the day
of surgery, prior to entering the
OR are not included. Activi-
ties and areas include registra-
tion/admission, same day surgery
area, OR holding, etc.
Activities may often cause an op-
erational issue, in terms of coor-
dination with OR schedule.
Elective
patient
waiting
lists
Simple wait listing included al-
lows for ordering by FCFS, pri-
ority and deadline. Does not at-
tempt to capture details of indi-
vidual surgeon practices.
Eery surgeon manages his/her
wait list in a very specific manner,
which they can not often quantify
in a meaningful way for the pur-
poses and ease of modelling.
Length of
Stay in
long stay
units
LOS in long stay units (ward,
ICU, SDU) are rounded down to
the nearest day such that dis-
charges occur at midnight of the
day of discharge. LOS less than
one day are rounded up to mid-
night.
Ease of modelling and handling
patient flow and bed decisions.
Continued on next page . . .
Appendix A. Conceptual Model Tables 203
Table A.3 – Continued
Assumption Description Reason
Anaesthesia
Induction
Room
Anaesthesia induction Room, as
seen at SMH) is not included.
Coordination of activity through
the anaesthesia induction rooms
is an operational problem, and
does not affect patient flow at the
tactical level of detail.
Surgeons
running
more than
one OR at
once
Surgeons who are assigned to
more than one OR at one time are
assumed to use residents and in-
terns to run both ORs at once,
and thus largely independently.
Rather than using the OR as a
swing room to save turn around
time.
Simplification in scheduling pa-
tients and coordinating day of ac-
tivity.
Number
of non-
modelled
patients
inputted
by month
The inputted number of non-
modelled patients per unit can
vary by month only. Variability
by day of week is not considered.
Simplification to reduce input re-
quirements and coding. Rate
of non-modelled patients is often
not well documented nor under-
stood to provide valuable input
data.
Scheduling
Functions
Urgent and emergent time - can
not interrupt an elective block,
but can be scheduled immediately
before or after elective time, or in
an OR on its own
To simplify model coding logic.
Continued on next page . . .
Appendix A. Conceptual Model Tables 204
Table A.3 – Continued
Assumption Description Reason
On Call time will be used before
closed OR time to schedule emer-
gent patients.
For the purposes of results collec-
tion in terms of OR utilisation.
Surgeon
assignment
Surgeons only assigned to elective
time blocks.
Reflective of usual practice.
Model assumes that ur-
gent/emergent patients are
not assigned a particular sur-
geon, as they are performed by
the on-call surgeon.
Up to two surgeons can split an
elective block.
To simplify model coding and re-
flective of usual practice.
Surgeon can not be assigned to
two half blocks in two different
ORs. Surgeons can be assigned
to both halves of the same OR if
probability allows.
To simplify coding. Considered
to not occur regularly unless sur-
geon picks up time released by
other surgeons.
Service as-
signment
No more than one service per
elective block.
Simplify model coding.
Elective time must be assigned a
service. Other OR functions do
not require assignment, though
inpatient and emergent time can
be assigned, or left open to all ser-
vices.
Assumed to be common practice
in hospitals.
Appendix A. Conceptual Model Tables 205
Table A.4: Model Simplifications.
Assumption Description Reason
SimplificationDescription Reason
Single OR
and PACU
group
Model allows only a single set of
ORs and PACU. If a hospital,
such as SMH has more than one
set of OR suites or PACU units,
they should be modelled together
in a single set. Scheduling rules
can be used to control the types
of patients allowed in each set of
ORs.
Simplification in terms of patient
routing though the model.
Patient
flow
though
post-
operative
beds oc-
curs only
in ”down-
ward”
direction
Flow occurs from higher levels of
care to lower. i.e. ICU to SDU.
Never from ward to SDU or to
ICU.
Reduction of allowable patient
flow routes, and reduces num-
ber of bed resource checks re-
quired. If patient did move back
and forth, length of stay in each
unit should be equal to the total
amount of time spent in the unit.
Continued on next page . . .
Appendix A. Conceptual Model Tables 206
Table A.4 – Continued
Assumption Description Reason
Inpatient
unit capac-
ity
Inpatient unit (wards, SDU, ICU)
capacity can be adjusted between
weekdays and weekends, but not
by each day of the week nor by
time of day.
Unit capacity at hospitals at the
inpatient level is usually budgeted
based on weekend and weekday
levels, and does not vary by time
of day or day of week.
OR Equip-
ment (e.g..
Scopes)
OR equipment is not included. Reserving, use and availability of
OR equipment is excluded as de-
lays related to equipment avail-
ability is an operational issue that
can be evaluated independently.
If some consideration is needed
when scheduling cases, a schedul-
ing rule can be used to limit the
number of cases that requires a
particular piece of equipment.
Continued on next page . . .
Appendix A. Conceptual Model Tables 207
Table A.4 – Continued
Assumption Description Reason
Inpatient
equipment
Equipment used during the hos-
pital stay in not included. Eg.
Therapeutic beds, monitors.
Equipment is often used across
man units and patients. An inde-
pendent study on use and avail-
ability, that included all services
and patient types is sufficient and
useful. Historical cancellation
rates due to equipment can be in-
cluded in the “other cancellation”
rate to reflect equipment avail-
ability.
Appendix B
Change Database
The following table (B.1) is the change database of changes made to the initial generic
conceptual model during coding in Visual8 and implementing at the test sites. Client
Request refers to changes made to the generic model to customise for a specific client
to achieve credibility at the particular site. Implementation refers to changes made to
simplify coding and implementation within the Visual8 coding environment.
208
Appendix B. Change Database 209
Tab
leB
.1:
Chan
gedat
abas
e.
Chan
geT
yp
eT
he
chan
geW
hat
itw
asR
easo
n
Clien
tR
eques
tU
rgen
tblo
ckti
me
-al
low
to
spec
ify
order
ofpri
orit
ypat
ients
tofill
tim
e.
Sea
rched
thro
ugh
urg
ent
wai
t
list
inor
der
asdefi
ned
inin
put
file
.
Cust
omis
atio
nfo
rW
illiam
Osl
er.
Imple
men
tati
onA
singl
epat
ient
input
file
in-
stea
dof
one
per
pat
ient
typ
e
Thre
eI:
Pat
ient
Input
file
,on
e
for
each
pat
ient
typ
e(e
lect
ive,
urg
ent,
inpat
ient)
Eas
ier
toco
de
and
man
age
in
Sim
ul8
Imple
men
tati
onA
dd
anin
put
for
the
min
imum
OR
LO
S
n/a
For
ease
ofco
din
gof
sched
uling
funct
ion.
Imple
men
tati
onU
rgen
tar
riva
lra
teby
serv
ice
only
,not
serv
ice
and
pri
orit
y.
Was
by
bot
hse
rvic
ean
dpri
or-
ity.
Eas
eof
codin
gan
db
ecau
sear
-
riva
lra
tew
hen
by
serv
ice
and
pri
orit
yis
ofte
nlo
w.
Imple
men
tati
onSep
arat
equeu
efo
rurg
ent
pa-
tien
tsw
ho
are
tob
edon
enex
t
from
other
pat
ients
wai
ting
to
be
called
.
Sin
gle
wai
ting
list
.Sim
plify
codin
gin
model
.
Con
tinued
onnex
tpag
e..
.
Appendix B. Change Database 210
Tab
leB
.1–
Con
tinued
Chan
geT
yp
eT
he
chan
geW
hat
itw
asR
easo
n
Imple
men
tati
onR
esou
rce
shif
ts-
up
tofo
ur
avai
lable
ina
day
.M
ust
be
order
edfr
omm
idnig
ht
tom
id-
nig
ht,
wit
hou
tcr
ossi
ng
over
nig
ht.
On-c
all
shif
tsar
ein
di-
cate
din
ase
par
ate
input
table
.
Allow
edfo
rth
ree
shif
ts,
plu
san
on-a
lsh
ift,
but
assu
med
tim
e
could
cros
sov
erm
idnig
ht.
Eas
eof
codin
gan
dunder
stan
d-
ing.
Imple
men
tati
onT
ime
rese
rved
for
urg
ent
and
emer
gent
pat
ients
isas
sum
edto
occ
ur
bef
ore
oraf
ter
anel
ecti
ve
blo
ck,
but
will
not
cut
anel
ec-
tive
blo
ckin
two.
Allow
edfo
rurg
ent/
emer
gent
tim
eto
cut
into
anel
ecti
ve
blo
ck.
Eas
eof
codin
g.
Imple
men
tati
onIn
put
requir
edfo
rto
tal
num
-
ber
ofsc
hed
uling
rule
cate
gori
es
tohel
pm
odel
det
erm
ine
outp
ut
file
.
Not
incl
uded
.E
ase
ofco
din
g.
Con
tinued
onnex
tpag
e..
.
Appendix B. Change Database 211
Tab
leB
.1–
Con
tinued
Chan
geT
yp
eT
he
chan
geW
hat
itw
asR
easo
n
Imple
men
tati
onA
dded
ata
ble
for:
Gen
eral
Char
acte
rist
ics
ofSurg
ical
Pop
-
ula
tion
.(t
otal
num
ber
ofse
r-
vic
es,
max
imum
pri
orit
yle
vels
for
each
pat
ient
typ
e).
n/a
Eas
eof
codin
g.
Imple
men
tati
onA
dded
pat
ient
file
index
shee
t
tosh
oww
hat
line
each
serv
ice
star
tsan
den
ds
at.
n/a
Eas
eof
codin
g.
Imple
men
tati
onA
dded
Tot
alnum
ber
ofR
ule
sto
I:Sch
eduling
Rule
sL
ist
n/a
Eas
eof
codin
g.
Imple
men
tati
onA
dded
anIn
put
for
min
imum
OR
LO
Sfo
rea
chse
rvic
e.
n/a
So
that
model
willnot
try
tofill
rem
ainin
gti
me
that
isle
ssth
en
this
amou
nt
Con
tinued
onnex
tpag
e..
.
Appendix B. Change Database 212
Tab
leB
.1–
Con
tinued
Chan
geT
yp
eT
he
chan
geW
hat
itw
asR
easo
n
Imple
men
tati
onA
dded
asp
read
shee
tto
trac
k
star
tan
den
dti
me
ofea
chfu
nc-
tion
inea
chO
Rfo
rth
atday
-so
that
we
know
when
toca
llth
e
firs
tpat
ient
and
when
over
tim
e
beg
ins
n/a
Eas
eof
codin
g.
Imple
men
tati
onA
dded
M:
Pos
tO
pR
esou
rces
whic
htr
acks
star
ting
row
inP
a-
tien
tE
xit
spre
adsh
eet
for
each
pos
tO
punit
and
poi
nts
toap
-
pro
pri
ate
LO
Slo
cati
ons
inP
a-
tien
tF
ile
(bas
edon
the
typ
eof
unit
)
n/a
Eas
eof
codin
g.
Con
tinued
onnex
tpag
e..
.
Appendix B. Change Database 213
Tab
leB
.1–
Con
tinued
Chan
geT
yp
eT
he
chan
geW
hat
itw
asR
easo
n
Imple
men
tati
onA
dded
M:O
Rsp
lit
tim
esto
reco
rdw
hen
OR
mov
esfr
om
one
surg
eon
toth
eot
her
ona
split
OR
day
.
n/a
Eas
eof
codin
g.
Imple
men
tati
onA
dded
M:P
ost
Op
Pat
ient
Exit
Tim
esto
trac
kw
hic
hb
eds
are
take
nby
who,
unti
lw
hen
,an
d
how
man
yre
sourc
es(b
eds)
are
requir
ed.
n/a
Eas
eof
codin
g.
Appendix C
Generic Model Detailed Description
C.1 Introduction to the Proposed Framework
The purpose of this appendix is to describe in detail the structure and components of the
proposed generalised framework for a perioperative simulation model. The perioperative
process can be sub-divided into three main processes: pre-surgical day, the surgical day
and post-surgical. Consequently, the proposed framework has been subdivided into the
same three sub-processes. The following three chapters will describe the framework’s
structure and components regarding each of the main sub-processes. Before the sub-
process can be addressed, the general set-up of the framework is addressed below.
C.2 Overall Framework Structure
The perioperative process is a complex system of numerous possible routes, decisions,
and interactions, all which can significantly affect a patient’s path through the system.
In order to capture this complexity in a manner that will be intuitive to the user, and
also take advantage of concepts from existing framework methodologies, the framework is
built on a “module” like basis, where processes can be further broken down into smaller,
more detailed sub-modules.
214
Appendix C. Generic Model Detailed Description 215
Based on the three sub-processes of the perioperative patient flow, the framework
is divided into three modules. Each of these three modules contains a number of sub-
modules, processes and decisions. For the purpose of this framework, any module that
is a component of a larger module will be called a sub-module. Sub-modules can also be
composed of other processes and sub-modules.
Pre-Operative
Module
Post-
Operative
Module
Surgical Day
Module
Start of the
Day Module
framework Aug 24 2009.vsd Overall View
Friday, August 28, 2009 2 of 10
Figure C.1: The overall view of the framework, showing the three modules’ patient flows.
Figure C.1 is the overall view of the framework, including the three modules, and
the patient flow (solid arcs) between the three modules. There are three types of shapes
used for the pictorial framework views, as shown in figure C.2. The rectangles are used
to show a single process, where only a single activity occurs. The rectangles with double
vertical lines represent modules and sub-modules that can be further broken into a set of
processes and decisions. These sub-modules will either have an accompanying diagram,
showing it’s internal processes, or will have a set of decisions, represented by a decision
tree or algorithm. The cylindrical shape represents queues where patients wait for the
next process. The solid arcs represent patient flow between processes. The dashed arcs
represent the modules that the Start of the Day Module communicates with, and affects,
other processes when run.
Appendix C. Generic Model Detailed Description 216
Process
(Sub-)Module
Patient
Flow
Wait – Queue
or delay
Affects
Patient Flow
framework Aug 24 2009.vsd legend
Friday, August 28, 2009 1 of 10
Figure C.2: Legend of shapes used in the pictorial framework.
C.2.1 Flow of Patients and Information
In order to allow for flexibility of the model and its processes, the flow of information and
patients are handled differently. The main flow element of the framework is the patients.
When two modules are connected by a flow, it is by the flow of patients moving from one
process to the other. Throughout this presentation, this type of flow will be represented
by solid arrows connecting two points.
Information is passed to and from processes to specified data elements. The flow of
information can be a result of a number of factors:
• A model process requires a piece of input information, such as a scheduling rule,
resource capacity, etc.
• A model process requires information about a patient in the system, such as the
time that is required to perform surgery, or where the patient is to go to next.
• A model process has made an adjustment to a resource measure, or a patient’s
information, such as the current location of a patient, and needs to update the
value in the data.
There are three types of data files that are used in the framework:
Appendix C. Generic Model Detailed Description 217
1. Input Data Files: These sheets are used to input data into the model. They are
used to initially build the model specifics, as well as input variables into processes,
etc.
2. Intermediate Data Files: These sheets are used by the model while running to store
pertinent information about the patients, resources, etc.
3. Output Data Files: These sheets collect various outputs of the model, and are used
to create the final model report.
For the remainder of this discussion, when referencing a specific data file, it will first
indicate what type of data file through a capital letter (I for input, M for intermediate,
and O for output), followed with a colon and the name of the data file. For example,
I: File Name refers to the input data file called filename. If a specified field of the
file needs to be referred to, it will follow the file name within square brackets, ex. I:
File Name[Field Name]. The structure and content of the data files are also described
throughout this document.
Further, each patient in the model is assigned a unique identifying number that can
be used in reference to that patient throughout the model. For the remainder of the
discussion, this will be referred to as the patient ID.
C.2.2 Modelled Patients
There are three main types of surgical patients whose journeys through the system differ
significantly. The first type of patient is scheduled for surgery during the regular operating
room hours, and is often referred to as an elective patient. These patients generally visit
a surgeon at a clinic, where the surgeon determines that they require surgery. These
patients can often wait several weeks to months for their procedure. Depending on the
patient’s health, health policies, and other influences, these patients can wait indefinitely
for their procedure, or up to a specified amount of time, such as 180 days (6 months); the
Appendix C. Generic Model Detailed Description 218
model allows for either case as indicated in I: Wait List Expiry Information - Elective
(see table C.1). For example, SMH currently has no specific rules on how long an elective
patient can wait before being scheduled, thus he could wait a long time on the elective
waiting list. Urgent patients, however, require surgery within a specified amount of
time, depending on their urgency level. Priority 1-A patients must have surgery within
two hours, while priority 1-B patients can wait up to eight hours. Thus, 1-B patients
can wait six hours (the difference between eight hours and two hours) before he would
require surgery in the same timeline as patients of priority 1-A. Inpatients at SMH are
to be scheduled within seven days, if they are not, they must be added to the surgeon’s
elective list. Thus, in this case, elective patients would have infinite waiting times,
regardless of priority. Urgent patients would have maximum wait times of two, six, 46
and 126 hours, for priority 1-A, 1-B, 1-C and 1-D respectively. Inpatients would have an
inputted wait of seven days.
Table C.1: I: Wait List Expiry Information - Elective/ Urgent/ Inpatient. Details when
each patient will need to be taken off the wait list and scheduled immediately, or as
another type of patient. One file for each patient type. Urgent is given in hours, while
elective and inpatient is given in days.
Number of Priority Levels
Maximum Time on list Priority level
Service Level 1 . . . level n
Service 1
. . .
Service n
Further, elective patients remain at home prior to their procedure, and arrive at the
hospital a few hours prior to their scheduled procedure start time. They are also often
Appendix C. Generic Model Detailed Description 219
referred to as same day patients, as they arrive at the hospital the same day as their
procedure. Some hospitals further classify elective patients into various priority levels,
which may indicate their place on the wait list, their priority, or the maximum time they
can wait for their scheduled procedure. The model allows for priority among elective
patients, and to stipulate how they should be ordered in the wait list (I: Patient Input
File[Priority Level], and I: Wait List Rules [Wait List Ordering] shown in tables C.2 and
C.3 respectively). The elective patient wait list can be by surgeon or service, to reflect
whether wait lists are pooled among surgeons, or no. The choice is noted in I: Wait List
Rules[wait list organisation] in table C.3.
The second group of patients are emergent and urgent patients. These patients often
present themselves at the emergency department, and require surgery within a few hours
to a few days. These patients often require a procedure on a life-or-death basis, while
others are stable enough to wait a day or two. These patients are often already admitted
to the hospital through the emergency department. Their level of urgency determines
how long they can wait before their procedure must begin. How many levels of urgency,
and their times can vary by hospital, thus the framework allows for the specification of
the number of urgency levels and their maximum allowable waits (I: Wait List Expiry
Information - Urgent as shown in table C.1).
A third common patient type is often referred to as inpatients. These patients have
already been admitted to the hospital, and require surgery, but are not considered ur-
gent. These patients often have been admitted for some other reason, and now require
additional care.
Some hospitals treat inpatients as part of the urgent population, but in their least
urgent level, as they can wait up to a week or so for their procedure. In this case, these
patients would follow similar scheduling rules as the urgent patient population, including
the use of after elective time OR time, etc. Other hospitals treat them more like an
elective patient who must be done within the next week or so, but should be scheduled
Appendix C. Generic Model Detailed Description 220
Table C.2: I: Patient Input File - Elective/Inpatient/Urgent
Column Details
Service ID
Surgeon ID #
Patient type code Elective, urgent, inpatient
Priority level For wait list management (all case types)
Booking time Time to book in OR
OR LOS In minutes
PACU LOS In minutes
ICU LOS In days
Ward LOS In days, includes ALC days
Step Down Unit LOS in days
SDS post Op LOS
ICU flag Indicates if the ICU visit is planned or unplanned
ICU divert flag Indicates if the patient can be diverted to the PACU
Post surgical route code
Scheduling rule category
Arrival time of day Only for the urgent patient input file
On service ward
Anaesthesia flag
Admit or non-admit flag
Appendix C. Generic Model Detailed Description 221
Table C.3: I: Wait List Rules
Wait List Organisation Choices: by surgeon or by
service
Service Wait List Ordering: (Select
one option: by priority, by
FIFO, by time until waited
too long)
Priority when returning to
wait list after being can-
celled: (Select one option:
to front of list, same as
original, emergent of lowest
level)
Service 1
. . .
Service n
within elective time whenever possible.
Regardless of how the patients are scheduled, inpatients most often have a surgeon
assigned prior to scheduling, thus cannot simply be lumped in with the urgent patients.
In order to accommodate both methods of scheduling inpatients, and allowing for a pre-
assigned surgeon, the framework employs a third patient group, with its own scheduling
algorithm and rules. As with the two other patient types, the details of any priority
levels and maximum waiting times within the inpatient group is details in I: Wait List
Expiry Information - Inpatient as shown in C.1.
C.3 General Model Set Up Information
There are a number of general input parameters required by the model for set up:
• Run time of the simulation - the length of time, in days to run the model. This
Appendix C. Generic Model Detailed Description 222
should usually be several months to a year or more. This time does not include
the warm up period. (I: General Simulation Information[Run time of simulation],
table C.4)
• Length of warm up period - the length of time, in minutes, to run the model before
collecting any output results. I: General Simulation Information[Length of Warm
up period])
• Initial waiting list sizes - here the user can specify the starting number of patients on
each waiting list, whether by surgeon or pooled by service. (I: Model Initialisation
Wait List Sizes, table C.5
.
Table C.4: I: General Simulation Information
Run time of simulation (in
minutes)
Length of Warm up period
(In minutes)
Table C.5: I: Model Initialisation - Wait List Sizes
Surgeon/Service ID # on Wait List
1
. . .
n
Appendix C. Generic Model Detailed Description 223
C.4 Pre-surgical Processes
Pre-surgical processes include all activities prior to a patient’s arrival in the perioperative
unit on the day of their surgical procedure. This generally includes wait list management
processes, scheduling processes, and pre-surgical clinic scheduling and visits.
C.4.1 Pre-Surgery Module
The Pre-Surgery module captures the patient flow, information and decision making
that occur prior to the day of surgery. The key function of the pre-surgery module is to
manage the wait lists and schedule patients for surgery. This module’s discussion will
start with a detailed, step-by-step look into the processes within the module, focusing
on the flow of patients, and the module’s interaction with the information files. Figure
C.3 shows the processes and sub-modules of the pre-surgery module.
Decision to
Perform Surgery is
made
Assign Patient
Characteristics
Elective
Patient
Management
Sub-module
Urgent
Patient
Management
Sub-module
Urgent/Emergent
Patients
Elective Patients
Elective Patients who have Waited too long
Inpatient
Patient
Management
Sub-module
Inpatients
Decision to
Perform Surgery is
made
Assign Patient
Characteristics
Urgent patients for
day are created
Wait for Time
of Arrival
Assign Patient
Characteristics
framework Aug 24 2009.vsd Patient Arrives in Queue
Friday, August 28, 2009 3 of 10
Figure C.3: The pre-surgery module.
Appendix C. Generic Model Detailed Description 224
For both elective and inpatients, the module begins with the decision to perform
surgery process. Here, patients are generated based on a specified arrival process distri-
bution for each service (I: Arrival Patterns, as shown in table C.6). The arrival pattern
is assumed to follow an exponential distribution. Thus for each service, the user needs
to specify the average number of patients that enter the list per day. Using the exponen-
tial distribution is easy for hospital users to understand, and to quantify for the input,
without having extensive knowledge of statistical distributions, or when arrival patterns
are difficult to determine as the data is not available. Further, the use of an exponential
arrival pattern has been shown to be realistic in many cases within health care (Lowery,
1996).
Table C.6: I: Arrival Patterns - Elective/Inpatient
Service # Arrival Rate (# per day)
1
2
. . .
# services
Each service has its own arrival pattern process. This allows hospitals to measure
the effect of changing demand volumes of specific services. The hospital can also adjust
volumes of specific procedure types, surgeons, etc. or other characteristics through the
inputted patient characteristics as described below.
Each patient is assigned a unique identification number as his patient ID. The elective
or inpatient then moves to the assign patient characteristics process, where the charac-
teristics of the patient, such as procedure length, patient type, length of stay in ward, etc.
(see I: Patient Input File - Elective/Inpatient/Urgent table C.2 on page 220) are deter-
mined. The characteristics chosen for a patient are based on the set of characteristics of
Appendix C. Generic Model Detailed Description 225
an actual patient. Each service has its own set of patient characteristics. As opposed to
other models, this model does not sample from distributions to determine characteristics
such as procedure length. Creating a realistic set of characteristics of a patient based on
a number of distributions would be difficult to guarantee. Further, to help ensure that at
least some realism is maintained when using distributions, a large number of conditional
probabilities and distributions would likely be required. This would not only be difficult
to determine, but would also make the model difficult to use, update and understood by
the targeted users, hospital administrators.
As such, the framework uses an inputted set of possible patients and their character-
istics, which is easy for the targeted end users to work with and understand. This input
file can be based on historic patient records, but also allows decision makers to adjust
specific characteristics of the file to represent changes to methodology, surgeons, patient
mix, etc. For instance, if the hospital would like to implement a new technology that
will reduce some procedures’ length by 10%, the data file can be adjusted to reflect this
improvement. Alternatively, if the hospital would like to increase the incoming volume
of a type of procedure, the file can be adjusted such that there are more cases of that
type to randomly select. A further discussion of the use and possibilities of this patient
file is discussed later in the model development and testing sections as performed at the
three test sites.
The assign patient characteristics process randomly selects one of the rows of the
patient input file for the specified service (I: Patient Input File). This process will then
copy over data of the selected patient row to the intermediate data file M: Current
Patient File (as shown in table C.7), along with the patent ID. This intermediate file
will be accessed often throughout the remainder of the model whenever information is
required about a specific patient.
Appendix C. Generic Model Detailed Description 226
Table C.7: M: Current Patient File
Column Details
Patient ID #
Input file ID #
Time Entered WL (for first time)
Scheduled date (blank if none yet)
# times cancelled
Current location
Route code
Step # in route
Next Location assigned
Router
Three Patient Streams
Elective patients are managed within the elective patient management sub-module. Here,
patients wait until their scheduled day of surgery, or until they have waited too long and
require surgery immediately, changing their status to an urgent patient. When elective
patients enter, they are slotted into their appropriate wait list, are scheduled and then
wait for the scheduled procedure date to arrive. For further detail of this sub-module
please see section C.4.1.
Urgent and emergent patients enter the Urgent patient scheduling sub-module, which
manages the urgent wait list based on how long the patient can wait. The sub-module
also attempts to schedule the patients according to their urgency and various scheduling
and decision rules. See section C.4.1 for more details.
Finally, the inpatient management sub-module attempts to schedule inpatients for
surgery based on the specified scheduling rules. Please refer to section C.4.1 for further
Appendix C. Generic Model Detailed Description 227
details.
Elective Patient Management Sub-Module
The elective patient management sub-module manages the wait lists as shown in figure
C.4. When a patient first enters this module, the process set queue characteristics de-
termines the patient’s priority in the queue, and the maximum time that the patient
can wait to be scheduled, according to the patient’s characteristics (M: Current Patient
File) and the inputted queuing rules (I: Wait List Rules [Wait List Ordering], as shown
in table C.3 on page 221).
Set queue
characteristics
Waited too long
Move to Surgical Day Module
Wait list
Queues (By
surgeon or
service)
...
Elective Patient Management Sub-Module
Route to Urgent Patient Management
framework Aug 24 2009.vsd Elective Patient Queue Management
Friday, August 28, 2009 6 of 10
Figure C.4: The elective patient management sub-module.
The maximum time a patient can wait to be scheduled is based on his priority level
and service. The maximum time for each type and priority is provided in the input data
file I: Wait List Expiry Info (table C.1 on page 218). The process will determine where
in the queue the patient is to go, based on the inputted wait list ordering rules (I: Wait
List Rules [Wait List Ordering]) and the characteristics of the patients already in queue.
There are three wait list ordering schemes that hospitals can select from:
Appendix C. Generic Model Detailed Description 228
• First In First Out (FIFO): Patients are simply ordered in the order in which
they arrived.
• Priority: Patients are ordered strictly on their priority (from their patient charac-
teristics). If a patient of a higher priority enters the wait list, patients with lower
priority will be pushed further back in the queue to accommodate this patient.
• By deadline: Patients are ordered based on the amount of time they have left
before they are declared to have waited to long. Thus the patient who is nearing
the end of their maximum wait will be near the front of the wait list. Thus a lower
priority patient would be ahead of a higher priority if the former has been waiting
for long enough that he must be completed in less time than the newly arriving
patient of a higher priority.
Patients are placed in their respective queues, and wait until they are scheduled,
or until they have waited too long, whichever comes first. If a hospital wants to test a
pooled waiting list system, this can be done by using a single surgeon for each service, and
adjusting the patient input files accordingly. Scheduling of elective patients is performed
by the Start of the New Day module, details to follow in section C.9.
The process waited too long will remove a patient from his queue when he has reached
his maximum waiting time without being scheduled, according to I: Wait List Expiry
Information - Elective (table C.1, page 218). This process will move the patient to the
urgent patient queue, as an urgent patient of the least urgent priority level.
Urgent Patient Management Sub-Module
Due to the nature of urgent patients, there are many different ways a hospital can manage
their urgent patients, and many different decisions to be made. The how and what
decisions can also differ within a hospital depending on the urgency of the case, the time
of day, and the availability of ORs and staff. Independent of the hospital, the general
Appendix C. Generic Model Detailed Description 229
objectives are the same:
1. Minimise disruptions to the elective patient schedule.
2. Minimise the time used overnight, on weekends and of the on-call staff.
3. Perform the case within a reasonable amount of time and /or achieve wait time
targets.
When an urgent patient enters the Urgent Patient Management sub-module, the
patient is routed to either the Urgent Patient Waiting List queue or the Schedule Im-
mediately process. This route decision is based on the patient’s level of urgency and the
inputted rule from I: Urgent Patient Scheduling Rules (table C.8).
Urgent Patient
Waiting List
Waited too long
Urgent Patient Management Sub-Module
Route -based on
urgent scheduling
rules
Move to Surgical Day Module
Move to Surgical Day Module
Schedule
Immediately
framework Aug 24 2009.vsd Urgent Patient Queue Management
Friday, August 28, 2009 7 of 10
Figure C.5: The urgent patient management sub-module.
Patients placed in the Urgent Patient Waiting List wait there until they are scheduled
through the Determine next OR Activity Algorithm (see section C.10.1). When scheduled,
Appendix C. Generic Model Detailed Description 230
Table C.8: I: Urgent Patient Scheduling Rules
Maximum urgency level where patient must be scheduled immedi-
ately, without waiting for another option
Urgent patient waiting list ordering preference (choose from FIFO,
Strict Priority or By Deadline)
For urgent patients who need to be scheduled immediately, order the following
options in order of preference:
Schedule to an OR dedicated to emergent and urgent cases.
Schedule to an OR that has time reserved for urgent cases within
its elective schedule, where the service is currently running.
Schedule to an OR that has time reserved for urgent cases within
its elective schedule, regardless of service.
Schedule to an OR that has time remaining at the end of the elective
OR time that was not filed by elective patients, service specific.
Schedule to an OR that has time remaining at the end of the elective
OR time that was not filed by elective patients, regardless of service.
Schedule to an OR that has finished their elective case list, service
specific.
Schedule to an OR that has finished their elective case list, regard-
less of service.
Schedule to an OR that is currently closed, service specific (i.e.
the service does operate in that room sometime during the regular
elective schedule).
Schedule to an OR that is currently closed, regardless of service.
Appendix C. Generic Model Detailed Description 231
the patient is moved to the Surgical Day module. Similar to the elective patients queues,
this queue is ordered based on one of the three ordering criteria as inputted in I: Urgent
Patient Scheduling Rules.
If a patient has been waiting on the urgent patient waiting list for too long, and is
close to reaching the maximum time he can wait for surgery based on his priority and
maximum allowed wait time as indicated in I: Wait List Expiry Information - urgent
(table C.1), the process waited too long will take the patient off the list and into the
Schedule Immediately.
Patients who enter the Schedule Immediately process will be scheduled into the next
available OR, with consideration of the preferences indicated in I: Urgent Patient Schedul-
ing Rules. When the patient is scheduled, he will move to the Surgical Day Module to
wait for his procedure. Patients scheduled by this process do not require a bed resource
check as it is assumed that the patient must have surgery immediately, and the hospital
will find a bed as he can not wait any longer.
Inpatient Management Sub-module
As mentioned previously, inpatients are patients who have already been admitted but
are not part of the emergent/urgent patient group. Since they are already in the hospital
using resources, hospitals do not lump them with elective patients who wait at home.
Inpatients are also unique because unlike urgent patients, they have already been assigned
a surgeon, thus can not be treated as an urgent patient, whose surgeon is unassigned.
When an inpatient enters the Inpatient Management sub-module, he is placed on
the Inpatient Waiting List queue. This queue is ordered based on the same choices
for criteria as urgent patients and is inputted in I: Inpatient Patient Scheduling Rules
(table C.9). Normally, the inpatient waits here until the Determine next OR Activity
Algorithm schedules him. However, if the patient has not been scheduled within the
allowable maximum wait time for the patient, the patient will be moved to the Move
Appendix C. Generic Model Detailed Description 232
Inatient Waiting
List
Waited too long
Inpatient Management Sub-module
Move to Surgical Day Module
Move to
Appropriate
Queue
framework Aug 24 2009.vsd Inpatient Queue Management
Friday, August 28, 2009 10 of 10
Figure C.6: The inpatient management sub-module.
to Appropriate Queue process. This process will route the patient to either the Urgent
Patient Management sub-module, or to the Elective Patient Management sub-module,
as indicated in I: Inpatient Scheduling Rules (table C.9). If the patient is moved to the
Urgent Patient Management sub-module, he will be placed as an urgent patient with the
least urgent level, and will wait to be scheduled by any surgeon. If, on the other hand,
the patient is moved to the Elective Patient Management sub-module, he will be placed
at the front of his assigned surgeon’s queue, so that he will be scheduled the next time
the surgeon is given elective OR time.
Table C.9: I: Inpatient Scheduling Rules
Inpatient waiting list ordering preference
(choose from FIFO, Strict Prioirty or By
Deadline)
When waited too long, move to Urgent or
Elective queue?
Appendix C. Generic Model Detailed Description 233
C.5 The Surgical Day Processes
Day of surgery processes include all activities that occur on the day of surgery, from
arrival into the perioperative area transfers to a hospital bed, home, or some other level
of care. This includes any day-of decisions such as cancellation decisions, use of overtime,
etc.
C.5.1 Surgical Day Module
This module follows patients through their surgical day, and is represented in figure C.7.
The module tracks patients through the day of surgery processes, specifically their flow
through the OR. It does not however, follow in detail the patient flow through the various
waiting and preparation rooms, such as the same day surgery area, anaesthesia room,
and registration. The module makes decisions, as needed, such as cancelling due to bed
availabilities or use of OR overtime, urgent patient scheduling, etc.
Surgical Day Module
OR LOS
Determine
Next OR
Activity
OR Activities
Called by OR OR ActivitiesWait until called
Hold Until
Route Clear
Determine First
actual route
Route Cleared –
Proceed to Post-
surgical module
Cancel
Patient
framework Aug 24 2009.vsd Surgical Day
Friday, August 28, 2009 4 of 10
Figure C.7: The surgical day module.
Any patient scheduled for their procedure on the current day, whether elective, urgent
or inpatient, will wait in the wait until called queue. When called, the patient will move
into another queue, Called by OR, where he will wait until the OR is cleared and ready
to accept the patient.
Appendix C. Generic Model Detailed Description 234
Surgical Activities
The patient enters the OR when it is ready to accept the next patient. Within the OR
Activities sub-module, the first process is the procedure, for the inputted length of time
from M: Current Patient File[OR LOS]. Once the procedure is completed, Determine
Next OR Activity Algorithm determines what to do next in the OR and call the next
patient to proceed. The details of this decision are described later in section C.10.1.Surgical Day Module
OR LOS
Determine
Next OR
Activity
OR Activities
Called by OR OR ActivitiesWait until called
Hold Until
Route Clear
Determine First
actual route
Route Cleared –
Proceed to Post-
surgical module
Cancel
Patient
framework Aug 24 2009.vsd Surgical Day
Friday, August 28, 2009 4 of 10
Figure C.8: The OR Activities sub-module.
The model then determines in Determine first actual route if the first location of the
route is clear for the patient to proceed based on their post-operative route (M: Current
Patient File[Post Surgical Route]).
If the patient is to go to the PACU, where LOS is less than a day, but no bed is
currently available, the model determines how long until the first PACU bed should
become available. If the wait for a PACU bed is less than the allowable wait inputted in
I: Allowable wait for PACU bed (table C.13), the patient will block the OR until the bed
is available.
If the patient is to go directly to an inpatient bed (ICU, SDU, ward) the model will
check if an on service bed is available. If so the patient can be routed immediately. IF
the on service bed is not available, the model will find an available off-service bed based
on the off-servicing rules (I: Off-servicing Rules table C.14). Since the model does bed
checks at midnight, and cancels any scheduled cases if an on or off service bed will not be
available, there should be a bed available for each patient who undergoes surgery. The
only exception is urgent and emergent patients with high priority who must be done that
Appendix C. Generic Model Detailed Description 235
day in order to meet the maximum wait time set. If no bed is available on or off service,
these patients will use a “flex bed”. These beds can be thought of as those beds the
hospital must find for the emergent patient, whether it is off-serviced to a non-surgical
bed, or the hospital opens another bed over budget (planned).
If the bed and nursing resources are available in the off-servicing unit, the model ad-
justs the actual route choice to go to the PACU. The model will also note that the patient
is to have a different nurse-to-patient ratio based on the inputted information found in I:
Off-Servicing Nurse Ratios (table C.10). When the bed and the nurse resources are both
available, the patient will leave the OR. If a patient is to be off-serviced, the off-service
bed must become available before the on-service bed, otherwise the patient should go to
the on-service bed.
Table C.10: I: Off-Servicing Nurse Ratios
Sent to:
Where supposed to be: Resource 1 . . .
Resource 1 -
. . . -
-
The Route Clear - proceed to post-surgical module process routes the patient to the
post-surgery module, updates the patient’s data file to mark the actual location he will
be going (M: Current Patient File[router]), and marks the step in recovery route to zero
(M: Current Patient File[step #]).
The process then turns over the OR, before accepting the next patient, or closing for
the day. As a simplifying assumption, the turnover time is dependent only on the patient’s
service as described in I: Turnover Time (table C.11). Actual turnover can depend on
many factors including the preceding and following procedures, the availability of cleaning
Appendix C. Generic Model Detailed Description 236
staff and the timeliness of the patient and surgical staff. Due to the tactical purpose of
this framework, this simplification was found acceptable based on the validation of the
models created for this study.
Table C.11: I: Turnover Times
Service # Turnover Time (in minutes)
1
2
. . .
# services
C.6 Post Surgical Processes
After completing their procedure patients will enter the post-surgical modules for any
post-surgical recovery, i.e. PACU, ward, ICU stays.
C.6.1 Post-surgery Module
The post-surgery module is designed using a flexible flow pattern to allow for the various
routes that a patient can take through his post-surgical recovery, as shown in figure C.9.
When a patient enters the router - by route and step # process, the router will increase
the step number in M: Current Patient File[step #] by one to indicate that the patient
is entering into the next step of his post-surgery route. The router will then route the
patient to the appropriate area. To note, the patient will only ever enter this router if
the location he is to go to is free. The patient will never wait at the router for the bed
to become available.
Appendix C. Generic Model Detailed Description 237
Router – by
lbl_route and
step#
PACU LOSHold Until
Route Clear
ICU LOS
SDU LOS
Ward LOS
SDS LOS
Discharge
Determine
Next Location
Determine
Next Location
Determine
Next Location
Determine
Next Location
Determine
Next Location
framework Aug 24 2009.vsd Post-Op Activities
Friday, August 28, 2009 5 of 10
Figure C.9: The post-surgery module.
Appendix C. Generic Model Detailed Description 238
In terms of the framework construction, the recovery areas are all organised in a sim-
ilar fashion. The first process is the medically necessary length of stay for that patient,
which is read from the patient’s information (M: Current Patient File). Medically nec-
essary length of time refers to the recovery time that is required medically, excluding
any delays due to downstream resource availability, etc. When the LOS has elapsed, the
model then determines the next location based on the patient’s route and the current
step within the route (M: Current Patient File[Post Surgical Route] and [step #]). This
process is organised as a decision tree to determine where to send the patient next, based
on bed availability and off servicing rules. These decisions are represented by the diagram
in figure C.10.
First, the decision tree determines what the next step should be based on the patient’s
recovery route (M: Current Patient File[Post Surgical Route]), his current step in that
route (M: Current Patient File[step #], see table C.7 on page 226) and his next assigned
location (I: Post OR Routes, table C.12). The decision tree then determines if there is
a bed available in that area for the patient. If there is, the patient will be routed there
immediately through the router.
For patients requiring a PACU bed, if a bed is not available, the model then determines
when the next available bed becomes available. If the bed will become available within
the allowable amount of time (I: Allowable Wait For PACU, see table C.13), the patient
will still be routed next to the PACU, but will wait in his current location (likely the
OR) until the bed becomes available.
If an on-service bed is not available, the model will then determine if it can off-
service the patient to another area. This is done by systematically looking at the bed
availability in the areas where the patient is allowed to be off-serviced (I: Off-Servicing
Rules, as shown in table C.14), beginning with the most preferred. If a bed is found in
one of these areas, the patient will to go to that area. If no suitable bed is found, the
model will keep the patient where he is and try again to move him to his next destination
Appendix C. Generic Model Detailed Description 239
Determine where
supposed to go
next
Is a bed
available
there?
Route patient to
go immediately
Determine when a
bed will be
available
Will there be a bed
available within time
allowed?
Route patient, but
hold until clear
Off service check
i = 1
Bed available in ith
preferred location?Route patient here
Will it be available
within allowed wait?
Route patient
here, but hold until
clear
Increment i
i !!"!#$$%&'(!%))!
*'+,-.'!#+'#*
Hold for 24 hours
and try again
Yes
No
Yes
No
Yes
No
Yes
No
No
Yes
Determine Next Location
framework Aug 24 2009.vsd Determine Next Location
Friday, August 28, 2009 9 of 10
Figure C.10: The determine next location decision tree.
Appendix C. Generic Model Detailed Description 240
Table C.12: I: Post OR Routes
Number of Different Routes
Max number of steps in any route
* Do not specify specific Ward or Step Down Unit or ICU
Route Code Description (for ref) # Steps 1st Step 2nd Step . . .
1 SDH with PACU 3 OR PACU
2 SDA to ward 3 OR PACU Ward
3 SDA ICU-SDU-ward 5 OR PACU ICU SDU Ward
. . .
# of Routes
Table C.13: I: Allowable Wait for PACU
Next location:
Current location Resource 1 . . .
Resource 1 -
. . . -
-
Appendix C. Generic Model Detailed Description 241
the following day. This is because the model knows that no suitable bed will become
available that day, so the patient must try tomorrow, when the next day’s discharges are
accounted for.
Table C.14: I: Off-servicing Rule
Send to:
Where supposed to be: Resource 1 . . .
Resource 1 -
. . . -
-
Determining the Patient’s Length of Stay
When a patient enters a ward, ICU or SDU bed, the framework determines his medically
necessary length of stay (LOS) based on his inputted time, his route information and the
inputted discharge information. The framework makes a simplifying assumption when
determining LOS: patients will have a LOS that will discharge him from his current unit
at midnight on the day of his inputted LOS (M: Current Patient File).
This simplification is done for a number of reasons. For transferring patients, the
framework uses midnight in order to give priority for beds to patients already admitted
over incoming surgical patients. This simplifying assumption does pose a minor limitation
as some delay results may be underestimated by patients who don’t have to wait for the
bed as the patient preceding him was already transferred at midnight. Due to this
simplification, transfer delays are not considered within the outputs. Additionally, the
model assumes that the LOS of patients in ICU, SDU and ward should be a multiple
of days. The least amount of time a patient spends will be until midnight the day he
entered. I.e. if a patient enters the ICU after surgery at 10AM, a one day LOs will
Appendix C. Generic Model Detailed Description 242
discharge him 14 hours later at midnight. If the patient has a two day LOS, he will
discharge the next day at midnight, after 38 hours.
As for discharging patients, there are many factors affecting their actual discharge
time, including the hospital’s discharge policy, the timeliness of the patient’s pickup
(whether family or a service), the timeliness of the surgeon’s rounds, the timeliness and
availability of hospital staff, etc. Since the objective of this decision support tool is
for tactical decisions, and not the day of decision around how to juggle incoming and
discharging patients, this assumption is acceptable.
Within the wards, the inputted LOS can also include the patient’s LOS waiting for
alternative care, which is due to external, post-acute wait for resources within the surgical
resources. This wait time should be included in the patients ward LOS value.
C.7 Hospital Resources
The framework allows for a number of resources to be included in the perioperative
patient flow:
• Operating Rooms
• Bed Resources
– Post-Operative Care Units (PACU) beds
– Intensive Care Unit (ICU) beds
– Ward Beds
– Step Down Unit (SDU) beds
– Flex beds
• Equipment
Appendix C. Generic Model Detailed Description 243
For each of these resources, a number of input values are required to populate the
model. For each resource type, this chapter will cover what information the framework
requires, and how the framework uses the data.
C.7.1 The Operating Room Resource
Depending on the size and organisation of the hospital, there may be a single set of
ORs (such as JH and MSH) or multiple units (such as the main and ambulatory ORs
at SMH). However, as a simplification, the model assumes only one OR unit that can
have any number of OR rooms. The number of rooms is indicated in I: Resources to
Model [Operating Rooms], table C.15.
For example, SMH would indicate in I: Resources to Model [Operating Rooms] that
they run 22 OR rooms, to account for their 16 main ORs and six ambulatory ORs. JH
would indicate that they run eight ORs.
C.7.2 Bed Resources
For each type of bed resource (ICU, SDU, ward) considered by the framework, the user
can specify the number of units and the number of beds in that unit. The number of
units is inputted into I: Resources to Model (table C.15). For each unit the user is able
to specify the number of beds available based on the resource’s shift patterns. They can
also specify if the bed resource is affected by the type of patient that is occupying it.
The model is also able to consider the fact that there are non-surgical patients who may
also occupy the same bed resources. These three aspects of bed resources are described
in detail in the following sections.
For PACU, the same simplification as the ORs is made to simplify the flow from
the OR to the PACU, thus only a single PACU unit is modelled. In I: Resources to
Model[PACU] indicates whether to model a PACU or not (if no PACU LOS data is
available, then there is no need to model the PACU). The number of bed available by
Appendix C. Generic Model Detailed Description 244
Table C.15: I: Resources to Model
Resource Name Number of OR Rooms
Operating
Rooms
Resource Name Number of PACU Beds
PACU
Resource Name
Number of Units to model
(not number of beds)
ICU
Ward
Step Down
Units
Regional Block
Units
Flex Beds
Appendix C. Generic Model Detailed Description 245
shift is considered in I:Resource PACU-weekday/weekend as explained in the following
section.
Resource Availability Affected by Time
The PACU is most often scheduled to be open and staffed only during specific times of
the day. However, it may have the option of running past its operating hours in special
circumstances using on-call staff. The framework allows for this change of availability
through the use of shift-based resources. The user can stipulate up to four different
shift-associated times (I: Resource PACU - weekday/weekend, table C.16). in order to
accommodate different shifts on the weekends versus the weekdays, the user can specify
weekday and weekend resource shifts.
1. First shifts - their regular, fully operational/staffed times.
2. Second shift - if a second shift is used with a different resource level than the regular
shift.
3. Third shift - if a third shift is required to reflect another change in the resource
level.
4. Forth shift - if for some period of time, the resource is only available on an on-call
basis.
The PACU can change its resource level up to three times each day. For instance, the
PACU on weekdays, runs at full capacity of ten beds from 8 AM to 5 PM, then two beds
remain available until midnight for urgent cases. After midnight no beds are available.
This configuration would use the first shift to indicate that from midnight to 8AM, there
are no beds available. The second shift indicates the operating hours of 8AM to 5PM of
the fully operational time with ten beds available. From 5 PM to midnight, only 2 beds
would be available using the third shift.
Appendix C. Generic Model Detailed Description 246
Table C.16: I: Resource PACU - weekday/weekend
Shift Details PACU
First Shift Start time (HH:MM)
End Time
Number available
Second Shift Start time (HH:MM)
End Time
Number available
Third Shift Start time (HH:MM)
End Time
Number available
On-call Shift Start time (HH:MM)
End Time
Number available
Table C.17: I: PACU on call flag - weekday/weekend
Table C.18: I: Bed Resources
Number of Beds Unit 1 . . . Unit n
On weekdays
On weekends
Appendix C. Generic Model Detailed Description 247
In some hospitals the PACU is not open 24 hours a day, however, if a patient still
waiting in the PACU for a downstream bed, the PACU will remain staffed until that
patient can be moved to the next location. A specialised marker is used here so that the
model knows that it is not normally available. The model can then access the results
from this resource to measure the use of overtime of the resource. For example, for the
same PACU example as above, if during the first and third shift, when it is closed, the
PACU will actually remain open if required, the on-call shift marker would be used in
I: PACU on call flag - weekday/weekend (table C.17), to indicate that from 12 AM to 8
AM, on-call staff is available if required.
To simplify resource checks for longer stay units, the framework assumes that ward,
ICU, and SDU units (i.e. those with stays f one or more days) only change the number of
resources available on weekdays and weekends, and cannot change mid-week or mid-day.
The capacity of each of these units is indicated in the table I: Bed Resources (table C.18.
Resource Availability Affected by Type of Patient
In some cases the type of patient occupying a bed will affect the resource availability in
terms of nurses. For example, patients who require more critical care require a smaller
nurse-to-patient ratio than less critical patients. This often occurs when a patient is
staying in a unit not usually meant to care for his type. A common example is when a
patient requiring an ICU bed is held in the PACU until an ICU bed becomes available.
In this example, the patient requires a ratio of 1:1 vs. a PACU patient who requires
1:2. This ICU patient would require a fully dedicated PACU nurse, leaving one PACU
bed, that the nurse would normally care for her second patient, open during his stay. To
reflect this nurse level change, the framework allows for patients to dictate how many
resources they require in the location they are being off-serviced to, through the inputted
table I: Off-Servicing Nurse Ratios (table C.10). This is only required when a patient
is off-serviced, as the model assumes that the patient requires only one nurse resource
Appendix C. Generic Model Detailed Description 248
(some fraction of a nurse) in the area where he is meant to go. For example, if the PACU
usually uses a 1:2 ratio, the PACU nurse resource would represent half a nurse. When
an ICU patient is in a PACU bed, it would require two half nurses in order to meet his
care ratio of 1:1.
Non-surgical (medical) Patients
Units are often shared between medical and surgical patients, or in times of bed shortages,
medical patients who usually belong in another ward will be off-serviced into a surgical
patient ward. Additionally, there are patients who are admitted under a surgeon, often
for diagnosis likely requiring surgery, who never end up having surgery. These patients
are often referred to as “non-surgical surgical” patients.
In order to accurately reflect the number of available beds in wards, SDU and ICU,
the model must account for these non-surgical (medical) patients. Since the number
of these patients occupying surgical beds can vary day to day, with seasonality effects,
etc., the model tries to emulate some of this variability without explicitly modelling these
patients. To do this, the model creates the number of non-surgical patients to occupy beds
on a daily basis, based on inputted monthly average occupancy rates of these patients
by ward, SDU and ICU assuming a normal distribution (I: Non-surgical Patient Bed
Occupancy, table C.19). These non-surgical patients are modelled to occupy the bed
until midnight the following night, when a new set is created (for more information on
the timing of creating non-surgical patients, see section C.8). Thus, the model does not
have to attempt to predict arrival rates and length of stays of non-surgical patients, but
can still show the effect of them on the availability of surgical beds.
Appendix C. Generic Model Detailed Description 249
Table C.19: I: Non-surgical Patient Bed Occupancy - Ward/SDU/ICU
Average number of non-
surgical patients.
Month #
Resource 1 (Jan.) . . . 12 (Dec.)
1
. . .
Resource n
C.8 Start of the Day Module
The Start of the Day module serves to update various data files. These processes are set
to occur every day at midnight, and perform a variety of tasks ranging from determining
the day’s schedule, the number of beds available in each ward, SDU and ICU, etc. This
is done through the following series of steps:
1. Generate the OR schedule - assigns services and surgeons based on schedule related
inputs.
2. Schedule patients.
3. Bed Management:
(a) Discharge patients.
(b) Perform any inter-unit transfers.
(c) Generate day’s non-surgical patients.
(d) Count number of beds available in each area.
(e) Generate day’s urgent patients.
(f) Perform bed resource checks.
Appendix C. Generic Model Detailed Description 250
C.8.1 Step 1: Generate the OR schedule
Based on the schedule information given, and the current day in the simulation run, the
model will assign the day’s services and surgeons to the ORs. For details on the creation
of the schedule, see section C.9 on scheduling.
C.8.2 Step 2: Schedule patients
Here, the scheduling algorithm is performed to schedule the day’s patients as per the
inputted scheduling rules. Please refer to the section on scheduling for further details,
section C.9.
C.8.3 Step 3: Bed Management
Step 3.1: Discharge all patients who are discharging today
The time element discharges patients from the wards, SDU and ICU. Recall that any
surgical patient scheduled to be discharged from the hospital today will discharge at
midnight. This is a simplifying assumption as patients actually discharge during the day,
after physician rounds, and other formalities. However, since this model is not concerned
with the day-to-day operational decisions of patients moving in and out of beds at specific
times, the framework model clears these patients at midnight so that the bed is available
when needed.
Step 3.2: Perform any inter-unit transfers
The model then allows any inter-unit transfers to occur, i.e. from ICU to SDU bed, ICU
to ward, etc. If there is no bed available in the destination resource area, the framework
will try to off-service the patient to another area, as per the off-service information.
However, on-service patients will get priority over beds, so the off-service decision will
be done after all on-service transfers are complete. If there is no on- or off-service bed
Appendix C. Generic Model Detailed Description 251
available, the patient will have to wait in the area where he currently holds a bed until
the next day at midnight.
Any patients who are off services in the PACU will also be transferred to an on-service
bed if available to complete their stay. Patients who were off-serviced to any other unit
will remain off-service for the duration of that step in their route.
Step 3.3: Generate day’s medical patients
Next, the model generates a new set of non-surgical patients to occupy some of the beds
for the day, in each bed resource area. There is a chance that there will be no beds
available for some of these patients, as the number generated may exceed the number of
beds available. In this case, the patient will be removed from the model.
Step 3.4: Count number of beds available in each area
The model counts the number of beds that are available for new patients to occupy in
each ward, ICU and SDU unit. The model stores this information in M: Beds Available
Today [# Available Start] (see table C.21 on page 253). This is the number of beds
available for the day for all incoming surgical patients.
Step 3.5: Generate day’s urgent patients
This step will first generate, according to the inputted distribution information I: Urgent
Patient Arrival Information (table C.20), the number of urgent patients of service, based
on a Poisson arrival process. For each generated patient, the model will also randomly
choose their characteristics from I: Patient Input File - Urgent (table C.2 on page 220,
which lists the same information as did I: Patient Input File - elective/Inpatient for
elective and inpatients. However, the I: Patient Input File - urgent also includes a column
indicating the arrival time of the patient, i.e. the time of the day that the decision to
perform surgery was made.
Appendix C. Generic Model Detailed Description 252
Table C.20: I: Urgent Patient Arrival Information
Service # Arrival Rate (# per day)
1
2
. . .
# services
For any urgent patient created, or already waiting, who requires surgery within 24
hours and requires a post-surgical bed, the framework will increment the count of these
patients arriving to the on-service bed area in M: Beds Available Today [# Incoming -
Urgent]. Urgent patients that can wait more than one day for surgery will be accounted
for bed-wise once scheduled, when the model knows which day the surgery will occur.
All urgent patients are then placed in the Urgent Patient Wait Until Arrival queue in
the Pre-Surgical module, where they will wait until their decision to perform surgery
has been made, at which time they will proceed through the Pre-Surgical module to be
scheduled.
Step 3.6: Perform bed resource checks
The model now determines which available beds to assign to which incoming patients
using the following steps:
• Assign beds (on or off-service) to all urgent patients who are arriving today with an
urgency that requires surgery within 24 hours. Also, include any urgent patients
who arrived previously who are still waiting to be scheduled, that have waited
enough time, based on their urgency, that they now must have surgery today.
Additionally, include any urgent patients who are scheduled for surgery today.
Appendix C. Generic Model Detailed Description 253
• Assign beds (on or off service) to all scheduled inpatients.
• If there are not enough beds for the above urgent and inpatients, their cases will
not be cancelled, but rather a “flex bed” will be used.
• For each unit, if there are enough beds remaining for all the incoming on-service
electives, assign accordingly.
• For units that have more incoming elective patients than beds, assign the available
beds to the patients based on their scheduled start time.
• Assign any remaining incoming elective patients to any remaining beds, maintaining
off-servicing rules.
• Any incoming elective patient who did not have a bed assigned, will be cancelled
due to no bed.
Table C.21: M: Beds Available Today - Ward/ICU/SDU
Resource # available # in-
coming -
Urgent
# in-
coming -
inpatient
# in-
coming -
elective
Resource 1
. . .
Resource n
C.9 Scheduling Details
This section focuses on the scheduling aspects of the model. It includes the set up of
the schedule based on a number of input files, as well as the scheduling algorithms for
Appendix C. Generic Model Detailed Description 254
elective, urgent and inpatients.
C.9.1 Schedule Inputs
The schedule inputs can be considered as two sets of input files. The first set details
the actual schedule, such as service and surgeon assignments and OR availabilities. The
second set of input files describe the various scheduling rules that need to be considered.
Schedule Details
There are five input files that together describe the actual OR schedule, each of which
will be described below.
• Schedule - Details
• Schedule - Function
• Schedule - Service
• Schedule - Surgeons
• Schedule - Surgeon # ORs
This first input file describes a number of scheduling input parameters for the model
(I: Schedule - Details, table C.22.
• Schedule length: The number of days in the inputted schedule (i.e. if a schedule is
on a two week rotation, thus is the same every other week, it would be 14 days).
• Min/Max Scheduling Rules used: To indicate whether there are any min/max rules
to be used while scheduling. More details to come in the scheduling rules section
below.
Appendix C. Generic Model Detailed Description 255
• Scheduling Preferences: The user is to rank in order of preference what to do with
any remaining elective time that is not booked by the surgeon. Further details are
given below in the section on the elective scheduling algorithm.
Table C.22: I: Scheduling Details - various input parameters about scheduling.
Schedule length (including weekends) in days
Combination scheduling? (0 = No, 1 = Yes)
Scheduling Preferences -
Rank the following five options in order of preference:
Open unscheduled elective time to other surgeon’s
within the service
Open unscheduled elective time to other services
Open unscheduled elective time to inpatient scheduling
Open unscheduled elective time to urgent patient
scheduling
Close the OR to further scheduling
The next two input files allow the user to enter the rotating OR schedule in terms of
its function and service. Both files are laid out for every OR for every day of the schedule
length. For each day, in each OR, the user enters the function and service assigned in
30 minute time intervals. The I: Schedule - Function (table C.23) allows the hospital to
assign time in the ORs as one of the following:
• Elective - time allocated for elective patient scheduling.
• Urgent - time allocated to scheduling urgent patients. This time can be scheduled
prior to the day of.
Appendix C. Generic Model Detailed Description 256
• Emergent - time allocated to urgent and emergent patients that is only to be
scheduled on the day of, based on the urgent and emergent list.
• On-call - similar to emergent, except it is only staffed if required to complete
urgent/emergent patients in a timely manner, using on-call staff.
• Closed - if the OR is to be closed to any patient scheduling.
Table C.23: I: Schedule - Function and Service
Day and OR
Day 1 Day 2 . . . Day n
OR 1 OR 2 . . . OR n OR 1 . . . OR n
00:00 On Call
00:30 On Call
01:00 On Call
. . .
08:00 Elective
08:30 Elective
. . .
17:00:00 Emergent
. . .
23:30:00 Emergent
The service input schedule (I: Schedule - service, table C.24) allows the user to assign
specific services to the ORs. The user can either specify a service to a time block, or keep
it unassigned. If left unassigned, this indicates that no preference is given to a particular
service for the use of the OR during that time. This is only appropriate when the OR’s
Appendix C. Generic Model Detailed Description 257
function is not elective. The framework assumes that during an elective block, only one
service is assigned, i.e. an OR block is not split between two services on the same day.
Table C.24: I: Schedule - Service
Day and OR
Day 1 Day 2 . . . Day n
OR 1 OR 2 . . . OR n OR 1 . . . OR n
00:00 0
00:30 0
01:00 0
. . .
08:00 ORTH
08:30 ORTH
. . .
17:00:00 ORTH
. . .
23:30:00 ORTH
As an example, the two tables shown here stipulate that on day 1 of the schedule OR
1 is assigned as follows:
• From midnight to 8 a.m. the OR is on-call for all patients.
• From 8 a.m. to 5 p.m., the OR is for orthopaedic elective patient scheduling.
• From 5 p.m. to midnight, the OR is open for emergent cases of the orthopaedic
service.
There are two input files for the OR schedule in the surgeon’s schedule (I: Schedule
- Surgeons and I: Schedule - Surgeons # ORs, see tables C.25 and C.26 respectively).
Appendix C. Generic Model Detailed Description 258
Surgeon assignment is often done by each service, and depends on the preferences of the
surgeons, holidays, time off, surgeon-patient load, etc. In order to simulate this assign-
ment, the framework assigns surgeons to OR time according to the inputted probability
of a surgeon being assigned on a particular day in a particular OR (I: Schedule - Sur-
geons). If a particular surgeon is always assigned a specific day and OR, then the user
simply gives that surgeon a probability of one for that day and OR.
The framework determines a day’s surgeon assignment by selecting a surgeon for each
OR, based on these inputted probabilities. The framework assumes that surgeons are
only assigned elective time. The other functions are not assigned for a specific surgeon’s
list, but rather a surgeon is assigned to perform any cases required during that time. For
example, during an Urgent assigned slot, the surgeon will perform any cases assigned to
the OR within his specialty during that time.
Some hospitals schedule more than one surgeon during an elective block of time in
the same OR. Thus, the normal eight to ten hour elective block is split between two
surgeons. To allow for this, the input sheet is composed of two probability assignment
matrices. One for the first assignment, and the other for the second. The second matrix
will have a zero for any surgeon-OR-day combination where a second surgeon is not to
be assigned. Probabilities greater than zero are assigned to any combination where a
second surgeon has a chance of being assigned as the second surgeon on the same day in
the same OR. The framework assumes that a split OR is split into two blocks of equal
time, i.e. a morning and an afternoon block, each with a single surgeon assigned.
The final surgeon scheduling input is to accommodate hospitals that allow surgeons
to run two operating rooms at the same time. This is sometimes used to increase the
utilisation of the surgeon as it decreases his downtime between surgeries, and also gives
more time for his interns to perform part or all of some procedures.
The framework assumes that the surgeons schedule ORs independently, instead of
staggering such that the ORs do not have the actual procedure for both patients at the
Appendix C. Generic Model Detailed Description 259
Table C.25: I: Schedule - Surgeons
First Slot
Day and OR
Day 1 Day 2 . . . Day n
OR 1 OR 2 . . . OR n OR 1 . . . OR n
Surgeon 1 0
Surgeon 2 0 0.5 0.2
. . . 0.7
Surgeon n 0.3 0.2
Second Slot
Day and OR
Day 1 Day 2 . . . Day n
OR 1 OR 2 . . . OR n OR 1 . . . OR n
Surgeon 1 0
Surgeon 2 0 0 0 0.6 0/8
. . . 0.7
Surgeon n 0 0
Table C.26: I: Schedule - Surgeons # ORs
Day 1 Day 2 . . . Day n
Surgeon 1 1 1 . . . 1
Surgeon 2 1 2 . . . 2
. . . . . . . . . . . . . . .
Surgeon m 2 1 . . . 2
Appendix C. Generic Model Detailed Description 260
same time. Essentially it assumes a system where the surgeon uses his interns, as opposed
to the possibility of parallel ORs, where the surgeon’s idle time is reduced.
The input file I: Schedule - Surgeons # ORs allows the user to enter for each surgeon,
for each day of the rotating schedule, if he is to run a single or two ORs simultaneously.
When assigning surgeons, the model will ensure that the surgeon is not already scheduled
the maximum number of ORs he may run a day, as inputted in I: Schedule - Surgeons #
ORs (table C.26). Further, if a surgeon is assigned a half block in one OR, this counts
as one OR, thus a surgeon restricted to one OR a day, will not be assigned two half day
blocks in two different ORs. The surgeon can however be scheduled in the morning and
afternoon slot of the same OR if the surgeon assignment probability allows.
When the framework creates the schedule, it uses these four files to build each day.
The framework uses a set of intermediate files to build the schedule information and to
schedule procedures. Each intermediate file a contains scheduling information for the
current day. The current schedule’s assigned functions, by time are maintained in M:
Schedule - Function (table C.27), which may be updated after the elective scheduling
process. For example, after the scheduling algorithm has scheduled the day’s elective
patients, the intermediate file will adjust that day’s function to reflect the fact that the
time can no longer be used for elective patient scheduling or to close the OR.
There is also an intermediate file used to track the current day’s service and surgeon
assignments (M: Schedule - Service and M: Schedule - Surgeon, tables C.28 and C.29
respectfully). Both of these files track for each 30 minute time interval through each OR
the service and surgeon assigned.
Additionally, for the purposes of scheduling patients, an intermediate file is used
to track the amount of time remaining on any particular day in an OR that is still
unscheduled, for elective, and urgent function scheduling (M: Schedule - Time Remaining,
table C.30). As patients are scheduled during an elective or urgent time slot, the model
updates the amount of time remaining that can be scheduled according to the schedule
Appendix C. Generic Model Detailed Description 261
Table C.27: M: Schedule - Function
OR 1 OR 2 . . . OR n
00:00 On Call
00:30 On Call
01:00 On Call
. . .
08:00 Elective
08:30 Elective
. . .
17:00:00 Emergent
. . .
23:30:00 Emergent
Table C.28: M: Schedule - Service
OR 1 OR 2 . . . OR n
00:00 0
00:30 0
01:00 0
. . .
08:00 ORTH
08:30 ORTH
. . .
17:00:00 ORTH
. . .
23:30:00 ORTH
Appendix C. Generic Model Detailed Description 262
Table C.29: M: Schedule - Surgeon
OR 1 OR 2 . . . OR n
00:00 0
00:30 0
01:00 0
. . .
08:00 Surgeon 1 Surgeon 16
08:30 Surgeon 1 Surgeon 16
. . .
17:00:00
. . .
23:30:00
M: Schedule - Function.
Table C.30: M: Schedule - Time Remaining
Time Remaining OR 1 OR 2 . . . OR n
Elective
Urgent
The final intermediate file used for scheduling is the list of patients scheduled for each
OR, in order. The file is simply an ordered list of patient IDs for each OR, as seen in M:
Schedule - Patients (table C.31).
Appendix C. Generic Model Detailed Description 263
Table C.31: M: Schedule - Patients
OR 1 OR 2 . . . OR n
1234 5493 590
5940 5435 657
. . .
C.9.2 Scheduling Rules
The rules involved in scheduling patients varies widely across hospitals. The framework
uses two sets of scheduling rules for elective patients: rules for the types of patients to
be scheduled, and patient ordering rules.
Patient Related Scheduling Rules
These rules are used to allow hospitals to implement rules such as “no more than two
planned elective ICU admissions per day” and “at least three total joint procedures per
orthopaedic OR per day”. These rules are all structured in a similar way, by specifying a
maximum or minimum number of patients by a specified patient classifier. The patient
classifier used for a rule can either be the ICU flag, the Scheduling Rule category, the
anaesthesia type flag or admit/non-admit flag. These rules are inputted into the I:
Scheduling Rules List (table C.32).
The I: Scheduling Rules List table shows an example of each type of rule allowed:
• Rule # 1 states that in an orthopaedic room, at least three total joint replacement
patients must be scheduled.
• Rule # 2 states that across all ORs and services within the elective day, no more
than two planned admissions to the ICU can be scheduled.
• Rule # 3 states that across all the general surgery OR rooms scheduled in the
elective day, only three patients of category type 4, say some scope procedure are
Appendix C. Generic Model Detailed Description 264
Table C.32: I: Scheduling Rules List
Rule
ID
Rule Scope Service Classifier Classifier
Value
Min
/
Max
Number Relax
(Y/N)?
1 OR ORTH Scheduling rule category 1 Min 3 Y
2 Day N/A ICU Flag planned Max 2 N
3 Service GEN Scheduling rule category 4 Max 3 N
. . .
allowed. This rule should be used for example when there is limited equipment,
such as a scope, and scheduling more than 3 procedures a day for this instrument
would be impossible as there would not be enough time to clean and sterilise it
between procedures.
The framework also allows for different rules to be applied on different days of the
schedule. For each day and OR of the schedule, the user may indicate which set of rules
must be followed, and which set of rules must have at least one obeyed. This is indicated
in I: Scheduling Rules Schedule (table C.33). For example, the table shows that for day
1 of the rotating schedule, in OR 1, rules one and two must be followed. Thus, for this
particular day in this OR, there must be at least three total joints scheduled, as well
across all ORs, no more than two ICU patients can be scheduled. Note that rule two
would need to be included in the first row for all ORs for that day to ensure it is followed.
In OR 2, rules two and three must be followed, meaning that no more than three general
patients of category four can be scheduled across any of the ORs assigned to the general
service that day. As with all ORs for this day, no more than two planned ICU cases can
be scheduled. Additionally, OR 2 has two rules listed in the “must obey at least one of”
Appendix C. Generic Model Detailed Description 265
row. This row allows for either-or type rules to be set up. For instance, perhaps rule
five states that at at least five cases of category seven must be scheduled, while rule six
states that at least five cases of category six must be scheduled. Thus, for day one of the
schedule in OR 2 either at least five cases of type seven, or five cases of type six must be
scheduled.
Table C.33: I: Scheduling Rules Schedule
Day 1 Day 2 . . . Day n
OR 1 OR 2 . . . OR n OR 1 . . . OR n
Must obey all of 1, 2 2, 3 2
Must obey at least one of 5, 6
While scheduling elective patients, the framework tracks these rules through the M:
Schedule Rules Tracking sheet as shown in table C.34. Here, each time a patient is going
to be scheduled, the algorithm will refer to the sheet to check if any rules will be helped
or violated based on the patients that have already been scheduled. If a patient scheduled
the algorithm will update this sheet accordingly. More details on how this is taken into
account within the scheduling algorithm to follow in the section C.9.4
Table C.34: M: Schedule Rules Tracking
Rule ID Day Service OR 1 . . . OR n
1
2
3
Appendix C. Generic Model Detailed Description 266
Patient Ordering Rules
In order to schedule patients according to the ordering preferences of the hospital, the
input file I: Scheduling Ordering (table C.35) is used to rank the following schedule
ordering options:
• Longest procedures first (LPF).
• Shortest procedures first (SPF).
• Admitting patients first.
• Admitting patients last.
• Non-admitting patients first.
• Non-admitting patients last.
• No anaesthesia patients first.
• No anaesthesia patients last.
• Previously cancelled patients first.
• ICU patients first.
• ICU patients last.
The user ranks the above choices in order of importance. Any ordering option not used
by the hospital should be left unranked, indicating to the model that it does not need to
consider that ordering rule. At midnight on the day of, the Start of New Day Element
logic will order the day’s patients according to the inputted ranked order preferences.
Appendix C. Generic Model Detailed Description 267
Table C.35: I: Scheduling Ordering
Ordering Method Rank
Longest procedures first (LPF).
Shortest procedures first (SPF).
Admitting patients first.
Admitting patients last.
Non-admitting patients first.
Non-admitting patients last.
No anaesthesia patients first.
No anaesthesia patients last.
Previously cancelled patients first.
ICU patients first.
ICU patients last.
Appendix C. Generic Model Detailed Description 268
C.9.3 Daily Schedule Creation
Within the Start of the Day module, the day’s schedule is generated based on the inputs,
and the various intermediate files are updated accordingly. This is done by the four steps,
as described below.
1. Update the variables for tracking the day of the week, and what day within the
inputted schedule. As an example, if the inputted schedule is two weeks long, the
day within the inputted schedule is the day of the two week cycle that we are
currently on, expressed as a number between one and 14.
2. Read in the day’s function and service schedules from the input files (I: Schedule
- Function and I: Schedule - Service) into the respective intermediate files, M:
Schedule - Function and Schedule - Service.
3. Generate the surgeon assignments for the day’s elective slots based on the proba-
bility of a surgeon being assigned to a specific OR on that day, as inputted in I:
Schedule - Surgeon. Allow for a surgeon running multiple ORs if indicated in I:
Schedule - Surgeon # ORs. Mark the assigned surgeons as appropriate into the
intermediate file M: Schedule - Surgeon.
4. Update the other intermediate files as appropriate, M: Schedule - Time Remaining,
and M: Schedule Rules Tracking.
C.9.4 Elective Patient Scheduling Algorithm
For each OR assigned elective time today, the Elective Patient Scheduling Algorithm will
attempt to schedule elective patients from the assigned waiting list using the following
algorithm. When the chosen wait list organisation is by surgeon, patients are scheduled
into OR time assigned to that surgeon. When organised by service, patients are scheduled
into OR time assigned to the service, regardless of the surgeon assigned. Whenever a
Appendix C. Generic Model Detailed Description 269
patient is booked during this algorithm, the M: Schedule - Patients, M: Schedule - Time
Remaining, and M: Schedule - Rule Tracker are updated appropriately.
Step 1: Schedule any inpatients who are at the front of the elective wait-
ing list : This step will schedule, regardless of rules, any inpatients who are at the front
of an elective waiting list. This is done because inpatients are only placed at the front of
the queue if they are to be scheduled in the surgeon’s/service’s elective time.
Step 2: Schedule patients who fulfil at least one of the minimum rules :
This step will search through the surgeon’s/service’s elective waiting list in an attempt to
find patients to schedule within the remaining elective time available that will contribute
to meeting at least one of the “minimum” scheduling rules required. The algorithm will
only schedule a patient whose booking time (I: Current Patient File[Booking Time])
fits within the remaining elective time, as tracked in M: Schedule - Time Remaining.
Additionally, the algorithm will only schedule a patient at this stage if he has at least
one characteristic of one of the inputted scheduling rules that are “minimum” rules. The
patient must not break any of the “maximum” rules.
Step 3: Schedule patients, without breaking any of the maximum rules :
This time algorithm will schedule any patient whose booking time fits within the remain-
ing unscheduled time, and does not break any of the “maximum” rules.
Step 4: Schedule patients, relaxing allowed rules : In this step, the algorithm
searches a final time through the surgeon’s/service’s elective waiting list to schedule
patients into the remaining unscheduled time. This time however, the algorithm will
allow some rules to be broken, if they have been flagged as relaxable, i.e. I: Schedule -
Rules [Relax (Y/N)?] is set to “Y”.
Step 5: Order patients according to inputted ordering rules : Now that the
surgeon’s/service’s elective list has been set for the day, the patients are ordered based
on the inputted ordering preference as indicated in I: Scheduling Ordering (table C.35).
When an OR is split between two surgeons, each surgeon’s list is ordered as per the rules,
Appendix C. Generic Model Detailed Description 270
independent of the other.
Step 6: Schedule elective patients from other surgeons, services, and
urgent and inpatients as per inputted preferences: This final step attempts to fill
any remaining unscheduled time at the end of the elective day according to the inputted
preferences in I: Schedule - Details (table C.22). Here, any “maximum” rules that are
not allowed to be relaxed must be upheld for patients to be scheduled. These cases will
be performed in the order booked after the end of the assigned surgeon’s elective list.
When an OR is shared by two surgeons, additional patients are scheduled at the end of
the day into unbooked time, after both surgeons have completed their scheduled lists.
Step 7: Cancel elective patients due to non-modelled reasons : Some pa-
tients are cancelled on the day of surgery because they are not cleared medically, do not
show up, etc. that is out of the control of the framework model. This type of cancel-
lation will be referred to “non-modelled” or “other” reason within the context of this
framework. The final step of the algorithm determines if any scheduled patient is to be
cancelled due to a non-modelled reason. See section C.10.2 for more details on this type
of cancellation.
Patients that are flagged as cancelled are routed to the Cancelled Patient Management
sub-module, where their status is updated, and they are returned to the appropriate
waiting list as per the rules indicated in I: Wait List Rules (table C.3 on page 221).
C.10 Additional Framework Details
C.10.1 Determine next OR Activity Decision Tree
When a procedure in the OR is close to completion, a number of decisions are made
by the hospital staff concerning the operating room. They must determine, based on
the amount of regular time remaining, whether to continue with the elective schedule,
perform an urgent case, or to close the OR. The following decision tree represents the
Appendix C. Generic Model Detailed Description 271
generalised decision making process to determine what should occur after the current
procedure is complete (figure C.11).
Is there a next
case waiting?Yes
No
Can it be completed with
less than X Overtime?
Call Patient to OR
HoldingYes
Cancel
Patient – due
to OT
No
Check
Additional List
Determine Next OR Activity
framework Aug 24 2009.vsd Closeup and reversal
Friday, August 28, 2009 8 of 10
Figure C.11: The Determine Next Activity Process.
First, the model checks if there is another patient scheduled for that operating room
(which includes the elective patients of the day and any urgent or inpatients who have
been scheduled). If there is a patient scheduled, the model determines if there is enough
time remaining in the scheduled time for his booked procedure length. The model de-
termines the estimated time for completion of that next case based on the current time,
the turnover time, and the booked length of the waiting case less turnover time. If the
case would require to run past the scheduled current function’s end time, the amount
of overtime required must be less than the inputted overtime decision in I: Overtime
Allowance (table C.36) based on the patient’s type and priority.
If the model determines that the patient passes the overtime rule, the process moves
the patient from the Wait until called queue to the called by OR queue. If the model
Appendix C. Generic Model Detailed Description 272
Table C.36: I: Overtime Allowance - Elective/Urgent/Inpatient
Priority level
Service 1 . . . Priority
level n
1
. . .
Service n
determines that the patient must be cancelled due to not enough time, the process
moves the patient into the Cancel Patient process to be cancelled and returned to the
appropriate waiting list. The process will then start this decision tree again to see if
another patient can be completed.
If there is no patient currently waiting for that OR, the model will then enter the check
additional list sub-module, to determine if there is an unscheduled urgent or inpatient
who can be done.
Check Additional List Algorithm
This algorithm determines if there is an urgent or inpatient waiting who is eligible to be
scheduled in the current operating room. If it is during elective time, the algorithm will
first search through the Inpatient Waiting List for a patient to be scheduled. Preference
is given to a patient who is of the same surgeon, then same service, then any service. In
order to be scheduled, the patient must also fit within the allowed overtime amount as
inputted.
If no inpatient is found, or the OR is currently assigned as an Urgent or Emergent
OR, the algorithm will search the Urgent Patient Waiting List. Here, preference is given
to a patient who is of the same service first. The patient selected must also meet the
Appendix C. Generic Model Detailed Description 273
maximum overtime allowance.
If the patient being scheduled is of an urgency level where he can wait more than 24
hours to be scheduled, a check for bed availability of any post-operative admitting bed
must be performed prior to scheduling the patient. If a bed is not available, the patient
can not be scheduled.
C.10.2 Further Information about Cancellations
Cancel for other reasons
There are many reasons why a patient can be cancelled on the day of surgery. Some
of these reasons are within the control of the model, such as no bed available, out of
regular time, more urgent patient, etc. However, there are a number of reasons that are
not specifically controlled by the model, such as patient no show, incomplete test results,
etc. To account for these types of reasons, the model will randomly select patients to
be cancelled for “other” reasons, based on an inputted service-specific probability of this
occurring (I: Other Cancellations, table C.37). Patients that are cancelled for an “other”
reason are routed through the cancel patient process (see section C.10.2 for details on
this process).
Cancelled Patient Process
This section describes the processes that occur in the cancelled patient sub-module.
First, the patient information is updated to reflect that he has been cancelled. This
is done by increasing the number of times the patient has been cancelled by one in M:
Current Patient File[# times Cancelled]. Similarly, the cancellation report data file is
also updated for this patient’s cancellation, by reason (O: Cancellation Report).
The model then routes that patient to the appropriate wait list to be rescheduled.
If the patient is an elective patient, he will return to his surgeon’s wait list. His place
Appendix C. Generic Model Detailed Description 274
Table C.37: I: Other Cancellations - Elective/Inpatient/Urgent
Percentage of patients
who cancel due to other
reasons
Priority level
Service ID 1 . . . priority n
1
. . .
Service n
in the wait list is determined based on the inputted rules in I: Wait List Rules [Priority
When Returning] (table C.3 on page 221).
On the other hand, if the patient is an inpatient, the patient will be placed on the
least urgent wait list, as this patient already has a bed, and needs to be cleared as soon
as possible. Urgent patients are returned to the urgent queue using the same ordering
rules as when they first entered, with consideration for how long the patient has already
waited.
This sub-module then conducts administrative tasks to update the files, specifically
removing the patient from the current day’s schedule (M: Schedule - Patients) so that
the patient is not chosen later as the next patient in the schedule.
Since these cancellations occur on the day of, the hospital does not have a chance to
adjust the elective schedule. Thus no new elective patients can be added on such short
notice.
Appendix C. Generic Model Detailed Description 275
C.11 Framework Output Files
This section will go over the various output files that are used by the framework to track
andreport and a number of measures of interest for perioperative services. For each file,
an explanation of it’s purpose and its updating procedures will be covered. The output
files are:
• Wait list Report
• Cancellation Report
• OR Reports
– OR Utilisation
– Over and under time
– Delays
• PACU Report
• Ward Report
• ICU Report
• SDU Report
• Throughput Report
• Census Report
Additionally, there are also a number of other measures that a hospital may be in-
terested in that the framework assumes can be collected through automatic functions
within the simulation model. These measures include:
• Queue Statistics - including average size, average wait, etc. for urgent and emergent
patients only.
Appendix C. Generic Model Detailed Description 276
• Bed Resource Utilisation - including average number of patients in each area.
C.11.1 Wait List Report
The M: Wait List Tracking (table C.38 tracks all patients waiting list details. For each
patient, the file records as they enter the wait list, their service, patient type and service-
based category type, and the time entered. When a patient successfully begins his pro-
cedure, the file is updated to reflect the time he has left the waiting list and the number
of times he was cancelled before surgery.
Table C.38: M: Wait List Tracking
Patient
ID
Service
ID
Patient
Type
Category
ID
Surgeon
ID
Time
entered
list
Time
left
list
#
times
Can-
celled
. . . . . . . . . . . . . . . . . . . . . . . .
At the end of the simulation, the framework then aggregates the data from the M:
Wait List Tracking table and summarises the waiting information by service, patient
type, category and surgeon. The O: Wait List Report (table C.39) gives the following
output measures:
• Average time patients spent on the waiting list.
• Average time patients currently on the waiting list have waited.
• Average number of patients on the list.
• Average number of times patients are cancelled before receiving surgery.
The M: Wait List Tracking can also be analysed to find the maximum, minimum and
percentiles of the various measurers stated.
Appendix C. Generic Model Detailed Description 277
Table C.39: O: Wait List Report
Service
ID
Patient
Type
Category
ID
Surgeon
ID
Ave
Wait-
ing
time
Current
Ave
waiting
time
Ave.
size
Ave.
times
can-
celled
1 1 1 1
1 1 1 1
. . . . . . . . . . . .
C.11.2 Cancellation Report
The O: Cancellation Report (table C.40) tracks the cancellation statistics. Each time a
patient is cancelled, this output report is updated accordingly. It tracks by service and
patient type, the number of cancellations that occur for the following reasons:
• Cancelled due to no available ICU bed.
• Cancelled due to no available Ward bed.
• Cancelled due to no available SDU bed.
• Cancelled due to no available ward resource.
• Cancelled due to OR overtime.
• Cancelled due to being bumped for a more urgent case.
• Cancelled due to an “other” reason.
Appendix C. Generic Model Detailed Description 278
Table C.40: O: Cancellation Report
Reason
Service
ID
Patient
Type
# No
ICU
# No
Ward
# No
SDU
# No
Re-
source
# Over-
time
#
Bumped
for more
Urgent
# Other
1 1 1
1 1 1
. . .
C.11.3 OR Reports
There are a number of reports to track the various key measures around the OR. The
first report tracks the utilisation of the various time designations (elective, urgent, on-call,
emergent, and closed). At the completion of each surgery, the O: OR Report - Utilisation
(table C.41) report is updated to reflect the time used for that procedure during the
assigned OR function, based on I: Schedule - Function from table C.23. Additionally,
the time assigned for each service is updated at the start of each day to be used as the
denominator of the utilisation calculation. At the end of the simulation run, a utilisation
ratio can be calculated and reported. Turnover time is not considered in the utilisation
calculation. If a procedure spanned across more than one assigned OR function, the
utilisation will be divide accordingly to the different assigned functions.
The O: OR Report -Over and Under Time (table C.42) tracks for each OR and as-
signed function the amount of time and the number of times it experiences either overtime
(runs past the assigned end time of the function) or under-time (the last procedure of the
function is completed before the end of the assigned time). If at the end of a procedure,
Appendix C. Generic Model Detailed Description 279
Table C.41: O: OR Report - Utilisation
OR
ID
Elective
Time
Used
Elective
Given
Urgent
Time
Used
Urgent
Time
Given
OnCall
Time
Used
OnCall
Time
Given
Emergent
Time
Used
Emergent
Time
Given
Close
Time
Used
Close
Time
Given
1
1
. . .
OR
n
the case ran past the assigned function the amount of overage is updated in the report.
If at the end of a procedure, there are no more cases to follow, the amount of under time
is calculated and entered in to the report. At the end of the simulation run, the average
over- and under-time for each OR and each function can be computed.
Each time a patient is held in the OR past their procedure LOS, the information is
recorded in the O: OR Delay Report (table C.43). At the end of the simulation, the
average delay time can be reported.
C.11.4 PACU Report
For each PACU unit, O: PACU Report (table C.44) tracks the number of patients who
used the PACU, as well as the length of time they spent in the PACU. The report also
tracks the number and length of stay of any patients that are off-serviced to the PACU.
The number of patients delayed in the PACU waiting past their LOS to transfer to
another area is also recorded along with the amount of time they were delayed. At the
end of the simulation run, the average LOS, average off-service LOS and average delay
Appendix C. Generic Model Detailed Description 280
Table C.42: O: OR Report -Over and Under Time
OR ID OR function # times in OT Total OT time # times in UT Total UT Time
1 Elective
1 Urgent
. . .
2 Elective
. . .
OR n Closed
Table C.43: O: OR Delay Report
OR Unit # Patients de-
layed waiting in
OR
Total Delay
Time
1
2
. . .
Appendix C. Generic Model Detailed Description 281
time are reported.
Table C.44: O: PACU Report
PACU
Unit
# Patients Total LOS # off-
service
Total Off-
service
LOS
# Delayed Total De-
lay Time
1
2
. . .
The O: PACU Report - Diverts (table C.45) file tracks in more detail the number of
times there is a patient diverted from another area into the PACU. The report records
each time a patient who is supposed to go to either an ICU, ward or SDU bed is diverted
into the PACU because no bed is available when needed.
Table C.45: O: PACU Report - Diverts
PACU
Unit
Total #
Off-service
# ICU Di-
verts
# Ward
Diverts
# SDU Di-
verts
1
2
. . .
C.11.5 Ward, ICU and SDU Reports
Similar to the O: PACU Report, the O: Ward/ICU/SDU Report (table C.46) tracks
the LOS of patients in the wards/SDU/ICU units, including off-serviced and delayed
Appendix C. Generic Model Detailed Description 282
patients.
Table C.46: O: Ward/ICU/SDU Report
Ward
Unit
# Pa-
tients
Total
LOS
Total
Off-
service
LOS
Total
ALC
LOS
# De-
layed
Total
Delay
Time
Total
# Off-
service
1
2
. . .
C.11.6 Throughput Report
The O: Throughput Report, as shown in table C.47, counts the number of completed
cased by service, patient type, category and surgeon. The report is updated each time a
patient leaves the system after post-surgical recovery.
Table C.47: O: Throughput Report
Service ID Patient Type Category Surgeon ID Throughput
1 1 1 1
1 2 1 1
. . .
C.11.7 Census Report
Daily unit census is calculated based on the O: Census Report. At midnight each day,
this report counts the number of patients in each unit (ICU, SDU, ward). The census
Appendix C. Generic Model Detailed Description 283
count is done prior to patients being discharged or transferred. The template of this
report is presented in C.48.
Table C.48: O: Census Report
Unit ID Sunday Monday Tuesday Wednesday Thursday Friday Saturday
ICU 1
ICU 2
. . .
ICU n
SDU 1
. . .
SDU n
Ward 1
. . .
Ward n
Appendix D
Generic Model Implemented in
Simul8
D.1 Screen Shots of the Generic Model Implemented
in Simul8
284
Appendix D. Generic Model Implemented in Simul8 285
Figure D.1: Screen shot of generic model in Simul8.
Appendix D. Generic Model Implemented in Simul8 286
Figure D.2: Screen shot of generic model in Simul8 - focused in on arrival processes and
waiting lists.
Figure D.3: Screen shot of generic model in Simul8 - focused in on post-surgical units
(ICU, SDU, ward and flex beds)
Appendix D. Generic Model Implemented in Simul8 287
Figure D.4: Screen shot of generic model in Simul8 - focused in on operative day processes.
Appendix E
Juravinski Process Map
!"#$#%&'()*+&)#
,)-'&)#./0#'-#1+&2)&0#
Decision to
perform surgery Wait for day
of surgery
Schedule Patient
- OR +pre_Op
Wait for pre-
op Clinic
Appt
Wait to be
schedule Pre-Op Clinic
Wait for
surgery
Decision to
perform surgery Enter Hospital
Wait for
surgery
Decision to
perform surgery
Already Admitted
to Hospital
./0#'-#1+&2)&0#
Patient Reports
to Same Day
desk
Patient Registers
at Admitting SDS Waiting
area Enter SDS room Wait in SDS
until called SDS activities
Called to OR
holding
OR Holidng
activites
Wait in OR
Holding until
called
Called to OR
holding
OR Holidng
activites
Wait in OR
Holding until
called
Called to OR
holding
OR Holidng
activites
Wait in OR
Holding until
called
%&'()*)#3456#(/7)#
Figure E.1: The pre surgical patient flow at Juravinski.
288
Appendix E. Juravinski Process Map 289
!"#$%&'()*+
OR Booking
Office:
- inputs info
Surgeon selects
date and sends
request to
Booking office
Check resource
availability (staff,
rooms,
equipment)
Make Change
and confer with
surgeon
No
Pre-OP Clinic Booking
Office:
- schedules pre-op visit
(1 -4 wks prior)
Yes
Notify ICU Need ICU bed? Yes
If patient
cancelled due to
bed shortage (2
times for ICU or
1 time for ward)
then try to book
as first case of
day
No
,-./01+"23435+
Surgery confirmed Pre-Op Clinic :
ready for
surgery?
Return to Surgeon's
Office for further consult
No
Yes
PRE_OP VISIT
Patient Reports
to Pre-Op Desk
Nurse gets info,
height, weight,
bloodwork
Anaesthesia
consult (if
necessary)
Internal consult (if
necessary) Figure E.2: The scheduling details at Juravinski.
!"#$%&'()*+
OR Booking
Office:
- inputs info
Surgeon selects
date and sends
request to
Booking office
Check resource
availability (staff,
rooms,
equipment)
Make Change
and confer with
surgeon
No
Pre-OP Clinic Booking
Office:
- schedules pre-op visit
(1 -4 wks prior)
Yes
Notify ICU Need ICU bed? Yes
If patient
cancelled due to
bed shortage (2
times for ICU or
1 time for ward)
then try to book
as first case of
day
No
,-./01+"23435+
Surgery confirmed Pre-Op Clinic :
ready for
surgery?
Return to Surgeon's
Office for further consult
No
Yes
PRE_OP VISIT
Patient Reports
to Pre-Op Desk
Nurse gets info,
height, weight,
bloodwork
Anaesthesia
consult (if
necessary)
Internal consult (if
necessary)
Figure E.3: The pre-op clinic details at Juravinski.
!"#$#%&'()*+&)#
,)-'&)#./0#'-#1+&2)&0#
Decision to
perform surgery Wait for day
of surgery
Schedule Patient
- OR +pre_Op
Wait for pre-
op Clinic
Appt
Wait to be
schedule Pre-Op Clinic
Wait for
surgery
Decision to
perform surgery Enter Hospital
Wait for
surgery
Decision to
perform surgery
Already Admitted
to Hospital
./0#'-#1+&2)&0#
Patient Reports
to Same Day
desk
Patient Registers
at Admitting SDS Waiting
area Enter SDS room Wait in SDS
until called SDS activities
Called to OR
holding
OR Holidng
activites
Wait in OR
Holding until
called
Called to OR
holding
OR Holidng
activites
Wait in OR
Holding until
called
Called to OR
holding
OR Holidng
activites
Wait in OR
Holding until
called
%&'()*)#3456#(/7)#
Figure E.4: The surgical day patient flow at Juravinski.
Appendix E. Juravinski Process Map 290
!"#$%&'()*#"%%+#,-./(.01#232#"%%+#,-./(.01#
30-('0#(4#56%-0'#7(89#-:10#;#(<0<#601%=6-0#-90->#
Yes
Ward bed
available (if
required)?
If cancelled due to bed
shortage 2 times prior
(ICU) or 1 time prior
(Inpatient), procede?
Cancel and re-
book
Yes
No
Wait until next Bed
meeting to re-
evaluate.
HOLD
First case of
day?
No
ICU bed
available (if
required)?
No
Yes
Yes
Procede with case
Nurse checks
chart, NPO,
alergies
Patient brought to
holding room
Final meeting with
surgeon and
anaesthesiologist
Surgeon IDs site
and side of surgery,
marks limb Figure E.5: The decision tree on whether to proceed with a case, at Juravinski.
!"#$#%&'('))#*+,#,'(-)-.+)#
Emergency
Elective
Enter OR
7 (8) Patient Prep and Anaesthesia
Close up and
reversal of
aneasthesia
Surgery
Room
Changeover
Call to Prep
patient
Patient taken to
OR - by nurse
Is there
enough time
to complete?
Add surgery to
next day's
additional list
Is there a
procedure
following? List re-
evaluation.
Yes
Yes
No No
Is there a procedure on
the additional or
emergency lsits that can
be performed? Yes
OR closed for day
No
Emergency or
Elective Status?
Run OR
into OT?
Cancel
Yes
No
Can
procedure
can wait?
No Yes
Figure E.6: The patient flow and decisions made during the procedure at Juravinski.
Appendix E. Juravinski Process Map 291
!"#$%&'%'()"*(+,%
Admit to ward
bed
Discharge
Block OR until
bed ready
ICU or PACU
or SDS?
Go to PACU bed (13
bays)
ICU bed
ready?
PACU bed
ready?
Can they stay in
PACU until
ready?
PACU stay
Is ward bed
ready?
Wait in PACU/
ICU until ready
Return to
Sameday
surgery Room
Go to ICU
bed ICU stay
Goto PACU bed
until ICU bed
ready
Inpatient
Outpatient
Yes
No
Yes PACU
Yes
No ICU
No
Yes
No
Block OR until
PACU/ICU
bed ready
Ward stay
Go to SDS room (??)
SDS
SDS stay
If telemetry is
required, is it
available?
Yes
No
Figure E.7: The post surgical patient flow at Juravinski.
Appendix F
St. Mike’s Process Map
Decision to
Perform Surgery
Surgeon – Set
Surgical date
Schedule PAF
(1 to 35 days prior
to surgery date)
Wait for PAF PAF visitPAF – ready
for Surgery?
Return to Surgeon
for cosult
No
Elective
Wait
Schedule with
Booking
office (min 7
days prior)
Urgent (P1-A/B/CWait to be
scheduled
Wait for
scheduled time
Yes
Urgent P1-D and Inpatients
Wait to be
booked by
booking office
Scheduled by
booking office
Wait for
scheduled time
Wait for
scheduled time
OR Desk to
Schedule
Urgent Pt
St. Michael's Hospital pre surgical
Thursday, June 11, 2009 1 of 7
Figure F.1: The pre surgical patient flow at St. Mike’s.
292
Appendix F. St. Mike’s Process Map 293
Booking Office - Elective
Receive request
Resources
avaialble (staff,
equipment, rooms)
Make change and
confer with
surgeon
Surgery confirmedYes
No
OR Desk – Urgent 1A,B,C Scheduling
1- If P1-C: try to book into next day’s Scheduled Urgent room if possible.
2- Try to schedule in unscheduled urgent OR if time and resources permitted.
3- Bump into scheduled list. If P1A, into first OR avaiable, but preference to service
given if possible.
4- can book into overnight if need to based on urgency (P1-A/B only)
Booking Office – Urgent 1D/Inpatients
1- Attempt ot book into unused/released scheduled time of service, or any OR.
2- P1-C, attempt to schedule into next day’s scheduled urgent room.
3- Send patient ot service to schedule within scheduled time by adjusting schedule.
St. Michael's Hospital Booking Office
Thursday, June 11, 2009 2 of 7
Figure F.2: The elective patient scheduling process map at St. Mike’s.
Booking Office - Elective
Receive request
Resources
avaialble (staff,
equipment, rooms)
Make change and
confer with
surgeon
Surgery confirmedYes
No
OR Desk – Urgent 1A,B,C Scheduling
1- If P1-C: try to book into next day’s Scheduled Urgent room if possible.
2- Try to schedule in unscheduled urgent OR if time and resources permitted.
3- Bump into scheduled list. If P1A, into first OR avaiable, but preference to service
given if possible.
4- can book into overnight if need to based on urgency (P1-A/B only)
Booking Office – Urgent 1D/Inpatients
1- Attempt ot book into unused/released scheduled time of service, or any OR.
2- P1-C, attempt to schedule into next day’s scheduled urgent room.
3- Send patient ot service to schedule within scheduled time by adjusting schedule.
St. Michael's Hospital Booking Office
Thursday, June 11, 2009 2 of 7
Figure F.3: The inpatient patient scheduling process map at St. Mike’s.
Appendix F. St. Mike’s Process Map 294
Booking Office - Elective
Receive request
Resources
avaialble (staff,
equipment, rooms)
Make change and
confer with
surgeon
Surgery confirmedYes
No
OR Desk – Urgent 1A,B,C Scheduling
1- If P1-C: try to book into next day’s Scheduled Urgent room if possible.
2- Try to schedule in unscheduled urgent OR if time and resources permitted.
3- Bump into scheduled list. If P1A, into first OR avaiable, but preference to service
given if possible.
4- can book into overnight if need to based on urgency (P1-A/B only)
Booking Office – Urgent 1D/Inpatients
1- Attempt ot book into unused/released scheduled time of service, or any OR.
2- P1-C, attempt to schedule into next day’s scheduled urgent room.
3- Send patient ot service to schedule within scheduled time by adjusting schedule.
St. Michael's Hospital Booking Office
Thursday, June 11, 2009 2 of 7
Figure F.4: The urgent patient scheduling process map at St. Mike’s.
Elective Patients
Morning of
Surgical Day
- Decisions
Patient Arrives
Register at
Surgical Desk
Amb Pod (20)
FDS Pod (13)
Nurse ActivitiesWait to be
called
Final Check and
Anaesthetist visit
Amb OR
Activities
Nurse ActivitiesWait to be
called
Pre-Op Room
(6 stretchers + 8
chairs)
Pre-Op
Assessments
Wait to be
called to OR
CORE OR
Activities
Wait for
Regional Block
procedure
Regional Block
Procedure
Wait to be
called by OR
Reg’l block
admisnistered
within time limit?
CORE OR
Activities
Cancel (rare)
No
Yes
St. Michael's Hospital Day of -elective
Thursday, June 11, 2009 3 of 7
Figure F.5: The surgical day process map at St. Mike’s.
Appendix F. St. Mike’s Process Map 295
Scheduled ICU or
Ward – is there a bed
available?
Continue with
ProcedureYes
Are the patients
who can be
discharged?
No
Discharge PtYes
Cancel or
Delay?
Cancel – return
booking office to
schedule for next
day
Cancel
Hold Pt.
Check next pt
Delay
Try again
No
First Pt of day?Require Ward
bed?Yes Yes
Continue whether
bed avaialble or
not.
No
St. Michael's Hospital Morning of Decisions
Thursday, June 11, 2009 6 of 7
Figure F.6: The decisions made the morning of surgery at St. Mike’s.
Pt Prep
Anaesthesia
induction (If
required)
Procedure
Close patient and
initial post-
anaesthesia
Hold Pt until
route clear
Patient moves to
next stage
Determine
next OR
activity
Turnover
St. Michael's Hospital OR Activities
Thursday, June 11, 2009 4 of 7
Figure F.7: The process map within the OR at St. Mike’s.
Appendix F. St. Mike’s Process Map 296
Is there another
procedure
waiting?
EnterIs there enough
time to complete?Yes
Check Additional and
Emerg List for a patient
No
Can the
procedure
wait?
NoElective or
UrgentYes
Return to
additional listUrgent/inpt
Cancel – return to
surgeon wait list
Elective
Call patient
Yes
Run OR into
OT?
No No
Yes
Yes
Close OR
No
Return to start with
next in line
St. Michael's Hospital Determine Next OR Activity
Thursday, June 11, 2009 5 of 7
Figure F.8: The decision tree to determine the next OR activity at St. Mike’s.
ICU, PACU or
Pod?
PACU bed
avialable?PACU PACU stayYes
Hold in OR
until bed
avialable
No
ICU
ICU bed
available?
No
Can the patient
stay in PACU until
ICU is ready?
Hold in OR
until bed
avialable
No
ICU StayYes
Wait in PACU
(1:1 Nurse
Ratio)
Yes
Pod Stay (Amb or
CORE)Pod
outpatient
Ward bed
ready?Admitted
Discharge
Wait in current
bed
No
Ward StayYes
St. Michael's Hospital Post-Op Activiites
Thursday, June 11, 2009 7 of 7
Figure F.9: The post-surgical day patient flow at St. Mike’s.
Appendix G
Mt. Sinai Process Map
297
Appendix G. Mt. Sinai Process Map 298
!"#$%&#'(%%('
()*+%,&
-)*+%,&'
.%#%*/$&%('
()*+%*0'*%1)$*%.
-)*+%,&'
.%#%*/$&%('2"(%'
#03%'45-6
!"#$%&#'"**$7%('8,*'
!9:'"33#;
9((%((/%&#'.,&%<('<=:>-5:'
?%.'*%1)$*%.@A%(
B,
B,#$80'CD'
?,,E$&+('"&.'
()*+%,&F(',88$2%',8'
2G"&+%
!"#$%&#'+,%('#,'
!9=:'G,H.$&+'
"*%"
!"#$%&#F('2G"*#'$&8,'
*%7$%I%.'?0'&)*(%
!9=:'&)*(%'
2G%2E('3"#$%&#'$&
9*%'#G%*%'"&0'
$(()%(@
B,
A%(
JG%&'CD'*%".0K'
3"#$%&#'?*,)+G#'
$&#,'CD'?0'CD'
&)*(%
B)*(%'2"HH('
()*+%,&>"&%(;
B)*(%'.,%('8$&"H'
2G%2E(
9&"%(;'5,%('
$&.)2#$,&
9&"%(;'2)%('
()*+%,&
-)*+%,&'?%+$&('
()*+%*0K'$&2$($,&'
/".%
-)*+%*0'$('
2,/3H%#%K'$&2$($,&'
2H,(%.
!"#$%&#'*%7$7%.
!"#$%&#'#"E%&'#,'
!9=:'"&.'
*%2,7%*%.
A%(!"#$%&#'.$(2G"*+%.'
L,/%
5-'M NO;PN;PQ
-%%'-59'8H,I
5,%('*%2,7%*0'
+,'I%HH@
="&'3*,?H%/'
?%'8$R%.@A%(
-)*+%*0'
*%(2G%.)H%.
B,
!*,?H%/'8$R%.
!"#$%&#'"./$##%.
B,
-)*+%*0'(2G%.)H%.
!9:'"33#;'
(2G%.)H%.'I$#G$&'O'
/,&#G',8'()*+%*0
!"#$%&#'I"$#('8,*'
3*,2;'."#%
5"0',8'3*,2;'
!"#$%&#'*%+$(#%*('$&'
"./$##$&+
Figure G.1: The process flow map for DS patients at Mt. Sinai.
Appendix G. Mt. Sinai Process Map 299
!"#$%&#'(%%('
()*+%,&
-)*+%,&'
.%#%*/$&%('
()*+%*0'*%1)$*%.
-)*+%,&'
.%#%*/$&%('2"(%'
#03%'4-567
-)*+%*0'(28%.)9%.
!6:'"33#;'
(28%.)9%.'<$#8$&'='
/,',>'()*+%*0
!"#$%&#'"./$##%.'#,'
)&$#?@%.'4=ABC'
==B7
6*%'#8%*%'"&0'
$(()%(D
E,
F%(
G8%&'3*%2%.$&+'
()*+%*0'@%$&+'
29,(%.')3C'HI'
&,#$>$%(')&$#
-%*J$2%'"(($(#"&#'
,*'8,)(%K%%3$&+'
+,%('#,'2,99%2#'
!"#$%&#
!"#$%&#'#"K%&'#,'
HI'
6&"%(;'L%+$&('
$&.)2#$,&
6&"%(;'2)%('
()*+%,&
5,%('3"#$%&#'
&%%.'MN:'@%.D
E,
6&%(;'6&.'HI'
E)*(%',*'3,*#%*'
#*"&(3,*#('3"#$%&#'
#,'MN:
!"#$%&#'*%2,J%*(
F%(
E,
!"#$%&#'
#*"&(3,*#%.'#,'
!6N:'"&.'
*%2,J%*%.
-)*+%,&'@%+$&('
()*+%*0C'$&2$($,&'
/".%
-)*+%*0'$('
2,/39%#%C'$&2$($,&'
29,(%.
!"#$%&#'<,K%&')3
!"#$%&#'#"K%&'#,'
!6N:'"&.'
*%2,J%*.
F%(
5,%('3"#$%&#'
&%%.'-5:'
@%.D
!"#$%&#'#*"&(>%**%.'
#,'G"*.
!"#$%&#'
#*"&(3,*#%.'#,'
-5:
!"#$%&#'.$(28"*+%.'
<8%&'*%".0
-56'O P=;QP;QR
!"#$%&#'@*,)+8#'#,'
S#8'>9,,*'3*%T,3'
<"$#$&+'*,,/',*'
8,9.$&+'"*%"
HI'&)*(%'"&.'
"&%(;'.,'3*%T,3'
28%2K
5"0',>'()*+%*0C'
3"#$%&#'"./$##%.'
M('@%.'
"J"$9"@9%DE,
F%(
5,%('3"#$%&#'
&%%.'-5:D
E,
F%(
M('@%.'*%".0D
F%(E,
N"&'3*,@9%/'
@%'>$U%.D
!*,@9%/'>$U%.;
F%(
E)*(%'2"99('
()*+%,&?"&%(;I%(28%.)9%. E,
!"#$%&#'8"('!6:'
"33#;
-)*+$2"9'#%"/'
&,#$>$%.
N"&'"'@%.'@%'
/".%'*%".0D
-)*+%*0'
*%(28%.)9%.'@0'
()*+%,&
E,
F%(
N"&'!6N:'
#"K%'3"#$%&#D
E,
F%(
Figure G.2: The process flow map for SDA patients at Mt. Sinai.
Appendix G. Mt. Sinai Process Map 300
!"#$%&#'(%%('
()*+%,&
-)*+%,&'
.%#%*/$&%('
()*+%*0'*%1)$*%.
-)*+%,&'
.%#%*/$&%('2"(%'
#03%'4522)#%6
-)*+%*0'(27%.)8%.!"#$%&#'"**$9%(':,*'
!5;'"33#<
=('"&'=>;?-#%3@
.,A&'B%.'*%1)$*%.'
3,(#',3C
D%(
E,
E,#$:0'()*+%,&F('
,::$2%'"&.'GH'
B,,I$&+(
J"0',:'"./$(($,&K'
3"#$%&#'"./$##%.'#,'
)&$#?B%.'4LMNK'
LLN6
=&3"#$%&#'O PL<QP<QR
E)*(%'"&.'
"&%(#7%#$(#'.,%('
3*%@,3'27%2I
5*%'#7%*%'"&0'
$(()%(C
E,
D%(
-%*9$2%'"(($(#"&#'
+,%('#,'2,88%2#'
!"#$%&#
!"#$%&#'#"I%&'#,'
GH'
5&"%(<'S%+$&('
$&.)2#$,&
5&"%(<'2)%('
()*+%,&
J,%('3"#$%&#'
&%%.'=>;'B%.C
E,
5&%(<'5&.'GH'
E)*(%',*'3,*#%*'
#*"&(3,*#('3"#$%&#'
#,'=>;
!"#$%&#'*%2,9%*(
D%(
E,
!"#$%&#'
#*"&(3,*#%.'#,'
!5>;'"&.'
*%2,9%*%.
-)*+%,&'B%+$&('
()*+%*0K'$&2$($,&'
/".%
-)*+%*0'$('
2,/38%#%K'$&2$($,&'
28,(%.
!"#$%&#'A,I%&')3
!"#$%&#'#"I%&'#,'
!5>;'"&.'
*%2,9%*.
D%(
J,%('3"#$%&#'
&%%.'-J;'
B%.C
!"#$%&#'#*"&(:%**%.'
#,'T"*.
!"#$%&#'
#*"&(3,*#%.'#,'
-J;
!"#$%&#'.$(27"*+%.'
A7%&'*%".0
!"#$%&#'B*,)+7#'#,'
U#7':8,,*'3*%@,3'
A"$#$&+'*,,/',*'
7,8.$&+'"*%"
GH'&)*(%'"&.'
"&%(<'.,'3*%@,3'
27%2I
J,%('3"#$%&#'
&%%.'-J;C
E,
D%(
=('B%.'*%".0C
D%(E,
>"&'3*,B8%/'
B%':$V%.C
!*,B8%/':$V%.<
D%(
E)*(%'2"88('
()*+%,&?"&%(<
H%(27%.)8%.
E,
T7%&'3*%2%.$&+'
()*+%*0'B%$&+'
28,(%.')3K'GH'
&,#$:$%(')&$#
Figure G.3: The process flow map for inpatients at Mt. Sinai.
Appendix G. Mt. Sinai Process Map 301
!"#$%&#'$('
")*$##%)+',-'$('
&.#$/$%)
!"#$%&#'"00$1%('$&'
2*%03
40%'#5%0%'"&6'
$((7%(8
9.
:%(
;5%&',-'0%")6+'
<"#$%&#'=0.735#'
$&#.',-'=6',-'
&70(%
970(%'>"??('
(703%.&@"&%(A
970(%').%('/$&"?'
>5%>B(
4&"%(A'C.%('
$&)7>#$.&
4&"%(A'>7%('
(703%.&
D703%.&'=%3$&('
(703%06+'$&>$($.&'
*")%
D703%06'$('
>.*<?%#%+'$&>$($.&'
>?.(%)
!"#$%&#'0%1$1%)
:%( !"#$%&#')$(>5"03%)C.%('0%>.1%06'
3.'E%??8
F"&'<0.=?%*'
=%'/$G%)8
:%(
D703%06'
<.(#<.&%)
9.
!0.=?%*'/$G%)
!"#$%&#'")*$##%) 9.
H('=%)'
0%I7$0%)8:%(
9.
H('=%)'
"1"$?"=?%8
:%(
9.
4((%((*%&#').&%
2*%03%&>6'J KLAMNAMO
C.%('<"#$%&#'
0%I7$0%'
(703%068
!"#$%&#'#0%"#%)'$&'
2*%03
9.
:%(
!"#$%&#'E"$#('/.0'
=%)'#.'/0%%'7<
H('(703%06'
0%I7$0%)'E$#5$&'
K'5.70(8
:%(
9.
!"#$%&#P('
%*%03%&>6'?%1%?'
)%#%0*$&%)'Q!R'4'
J'!R'CS'
!"#$%&#'#0"&(/%00%)'
#.'=%)'$&'"1"$?"=?%'
&70($&3'7&$#
!"#$%&#'E"$#('/.0'
(703%06'QE"$#'#$*%'
)%<%&)('.&'!R'
?%1%?S
C.%('<"#$%&#'
&%%)'HFT'=%)8
4&%(A'4&)',-'
970(%'.0'<.0#%0'
#0"&(<.0#('<"#$%&#'
#.'HFT
!"#$%&#'0%>.1%0(
:%(
!"#$%&#'
#0"&(<.0#%)'#.'
!4FT'"&)'
0%>.1%0%)
!"#$%&#'#"B%&'#.'
!4FT'"&)'
0%>.1%0)
C.%('<"#$%&#'
&%%)'DCT'
=%)8
!"#$%&#'
#0"&(<.0#%)'#.'
DCT
C.%('<"#$%&#'
&%%)'DCT8
:%( H('=%)'0%")68 :%(
9.
9.
:%(
9.!"#"$%&'(&)*+,"-+&./"01"-23
)4&56&%701"03&0"87,0"9&,-&:&;<
)4&=6&%701"03&0"87,0"9&,-&;>?<
)4&@6&%701"03&0"87,0"9&?>A?<
)4&B6&%701"03&0"87,0"9&,-&;>C&9*3%
!"#$%&#'$('
")*$##%)+',-'$('
&.#$/$%)
Figure G.4: The process flow map for urgent patients at Mt. Sinai.
Appendix H
William Osler Process Map
302
Appendix H. William Osler Process Map 303
!"#$%&#'()%#*+','(-#,./#-)%&#'0".1
!"#$%&#','
2#$343.*'+.'
/#-5.-6
'47-8#-9'
:7-8#.*'
:$;#<7"#4'
:7-8#.*':#*<4'
=..>3*8'+.'?@'
A..>3*8'?B$#'
A..>3*8'?B$#'
<#+#-6
3*4-#4.7-$#4'
)&)3")=3"#C')*<'5."".14'
=..>3*8'-7"#4'
@#+7-*'+.'
47-8#.*D4'
.B$#'+.'=#'
-#4$;#<7"#<'
A..>3*8'
E//-.&#<'
:$;#<7"#'(-#,
E<63443.*'F343+'
G)3+'7*%"'(-#,
E<63443.*'
F343+'
(-#,
E<63443.*'
F343+'
G)3+'7*%"'
=..>#<'
/-.$#<7-#'
<)+#'
2)9'=#5.-#'4$;#<7"#<'$)4#'
G3""'+;#-#'"3>#"9'=#'
#*.78;'=#<4'5.-'
*#H+'<)9D4'$)4#4I'
@#&3#1'$)4#4'
13+;'47-8#.*4'
E-#'+;#-#'
/)%#*+4'1;.'
$)*'=#'2J'
5-.6'7*3+I'
E-#'+;#-#'$)4#4+;)+'$)*'
=#'6
)<#'2)9':7-8#-9''
3*4+#<'.5')<63K#<I'L.-'
1)-<'&4'MJNC'.-'1)-<'&4O'
:?NP'
J)*$#"'$)4#L4P'
)4'*##<#<'
Q.'13+;'
4$;#<7"#'
@#+7-*'+.'
47-8#.*D4'
.B$#'+.'=#'
-#4$;#<7"#<'
R)>#'$;)*8#4'
+.'/)%#*+L4P'
=..>3*8'
G-3+#'2J'
.-<#-4'
:$;#<7"3*8'!"#$%&#'/)%#*+4'
,'!)$;'47-8#.*'4$;#<7#"4'+;#3-'.1*'/)%#*+4')$$.-<3*8'+.+';#'G
SM:'/-3.-3+9'
"#&"#'83&#*O'
,':7-8#.*4'=..>'13+;3*'+;#3-')"".+#<'%6#'.*'+;#'?@'=".$>'4$;#<7"#'=)4#<'.*'
+;#'=..>#<'%6#'L=)4#<'.*';34+.-3$)"T#H/#-3#*$#P'
?-<#-'.5'$)4#4U'
,'V.*8#-'/-.$#<7-#4'+.'=#'4$;#<7"#<'W-4+'
,'$)4#4'/-#&3.74"9'$)*$#""#<'+.'=#'4$;#<7"#<'W-4+'
,'/-#5#-)="9'=..>'*.*,)<63K#<'/)%#*+4'W-4+'
,'$;3"<-#*'W-4+'X'
,')"#-83#4')*<'$.6/"3$)%.*4','W-4+X'
,'3*5#$%.*'$.*+-."'/)%#*+4','")4+'X'
X'*.+')&)3")="#'3*'6
.<#"'
Figure H.1: The pre-operative process flow map for elective patients at William Osler.
Appendix H. William Osler Process Map 304
!"#$%#&'()*+#&'(,()$#,-.#$*+/#(01-2
3&.*+#&'(,()$#,-.#$*+/#(01-2
!"#$%#&'()*+#&'(,(
4#5676-&('-(.#$8-$"(
79$%#$:(
;9$%#-&(5*117(
<=(>#7?('-(
*>>('-(167'(
@*6'(9&+1(
;5A#>91#>(
@*6'(9&+1(
;5A#>91#>(
+"#(
;5A#>91#(
#"#$%#&'(
.*+#&'(
3&.*+#&'(,(4#5676-&('-(
.#$8-$"(79$%#$:(
;9$%#-&(5*117(
<=(>#7?('-(*>>(
'-(167'(
@*6'(9&+1(
;5A#>91#>(
@*6'(9&+1(
;5A#>91#>(
+"#(
;5A#>91#(
6&.*+#&'(
;5A#>91#(#"#$%#&'(.*+#&'(
,(38(1#/#1(BC(,(D--?(6&'-(E$7'(*/*6*1D1#($--"(F.$#8#$*D1:(-8('A*'(7#$/65#G(
,(68(1#/#1(BH(,(A-1>(9&+1('A#(#&>(-8('A#(#1#5+/#(>*:I((
,(68(1#/#1(BJ(,(A-1>(9&+1('A#(#&>(-KA#(#1#5+/#(>*:L(5*&(2*6'(9&+1(&#M'(>*:(68(&-'(
#&-9%A(+"#(
,(68(1#/#1(B4(,(75A#>91#(6&'-('-(<=(+"#(>#76%&*'#>(*7('$*9"*($--"(
,(<&(2##?#&>7(,(*11(#"#$%#&'(5*7#7(FBC,JG(*$#(D--?#>(D*7#>(-&(9$%#&5:(*&>(
'A#(*"-9&'(-8(+"#($#"*6&6&%(D#8-$#('A#6$('*$%#'I(
;5A#>91#(6&.*+#&'(
,(7A-91>(D#(75A#>91#>(>9$6&%(#1#5+/#(+"#L(68(&-&,#"#$%#&'I(
Figure H.2: The pre-operative process flow map for urgent and in- patients at William
Osler.
!"#$%&'())*+$,'
*%'-"./,0)1$).'
!"#$%&'2"33$4'
&5'67'8"*&'95)'
67'
!"#$%&'$%&$),'
67'!)5:$40)$'
!"#$%&'3$"+$,'
67'8"*&'95)'
;$4'
<,'<2='5)'!(2='
;$4')$"4.>'
<2='?"#$%&,':"%'@"*&'
*%'!(2='95)';$4''*9'%5'
<2=';$4'*,'"+"*3";3$A'
6+$)#B$'2"%:$33"#5%C'
/'<9'$3$:#+$'@",';0B?$4'
;$:"0,$'59'DB$)1$%&':",$'/'45'
%5&':"%:$3'
/'E"+$',&"F%1'0?'&5'G'67,'0%#3'
HIJJ'95)'6K'$3$:#+$'"%4'0)1$%&A'
D3$:#+$',:L$403$C'JMJJ'/'HNOJ'
6KP=)1$%&C''
''''''HNGJ'/''HQJJC'G'67'RSET'OUGV'DSE'
'''''''HQJJ'/'OGJJC'H'67'RSE'
'''''''HQ//'/'OGGJC'H'67'DSE'
6%'2"33'"W$)'L50),''H'67'
8$$X$%4,C'H'67'RSET''H'67'5%:"33'DSE'
Figure H.3: The operative (day of surgery) process flow map for patients at William
Osler.
Appendix H. William Osler Process Map 305
!"#$%&'(%
)#$%&'(%
('$%&'(%
*+,-%&'(%
(./,0%(0+1%
$230%&'(%
)#$%45-%
+6+37+475%
!+8520%2/0%,5+-1%
0/%9/,%:#%/2%
(+0;,-+1%<.52%
;230%=7/>5>?%
:#%
@/7-%;287%
25A0%45-%
9,55%
BC@D%
%%%%%%C52%(;,E%*+,-%F%GH%
%%%%%%%',0.%(;,E%<+,-%F%GI%
JC@D%
%%%%%%%C52%(;,E%*+,-%F%GK%
%%%%%%%%('$%F%L%
%%%%%%%(./,0%(0+1%%F%MH%
%%%%%%%%%%%%%%%%NOIKIIF(MGIIP%
%%%%%%%',0./%(;,E%<+,-%F%QG%
BC@D%IRQIFGQQID%MI%NSQP%%
%%%%%%%%%%%TGQQI%U<V52->%;W0/%L%
JC@DIKIIFGLIID%MX%NSLP%
%%%%%%%%%%TGLII%U%<V52->%;W0/%L%
BC@D%O+A%M%4//V5-%W5,%-+1%
%%%%%%%%%%MG%)#$%S%H%##$%Y%MK%
JC@D%O+A%G%4//V5-%W5,%-+1%
%%%%%%%%%%GL%)#$%
(+Z5%:+1%&'(%
Figure H.4: The post-operative process flow map for patients at William Osler.