Applications of Computer Simulation Modeling for Health

Care Process Management

Michael CarterMechanical & Industrial Engineering Mechanical & Industrial Engineering

University of Toronto University of Toronto

IHD 2007 2

Overview of Presentation• Computer Simulation Modelling• CHEO ED – Ottawa• Hamilton Cardiac Surgery• System Dynamics: Toronto Western

and Ottawa Hospital ED• Some Challenges

IHD 2007 3

Computer Simulation• Many packages available• MedModel, Simul8 and Arena• Illustrate with MedModel• Patients arrive randomly • Arrival rate varies over the day• Wait for triage, MD, RN, labs, etc.• Add resources, increase rates, “What if?”

IHD 2007 4

• Location Statistics examples:• Crit exam1: 76.02% util.; • Crit exam1: avg. time: 59.71 min.• Avg. time in waiting room: 107.6 min

• MD: 73.5% util.; • RN: 27.55% util

• Also find the avg. time per patient in the ED: 113.11 min

Resource Statistics

IHD 2007 7

CHEO: Emergency Room

• Children’s Hospital of Eastern Ontario: Ottawa 1993

• Paediatric Teaching Hospital• 50,000 patient visits per year in the ER

IHD 2007 8

CHEO: Waiting Times (1993)

05

101520253035404550

0-1 hr 1-2 hr 2-3 hr 3-4 hr 4 plus

% pat's.

IHD 2007 9

CHEO: Emergency Room• 20 % of patients wait over two hours• Eleven suggestions by staff• Simulation used to evaluate scenarios• Fast track clinic• New Casualty Officer• Staggered start times

IHD 2007 10

Cardiac Surgery Simulation• Hamilton Health Sciences• Opening fourth cardiac OR in Spring

2006• How should OR time be allocated?• How many beds are required in

ICU/ward?• “What if?” Simulation tool

Surgery GroupingCardiac Surgery 2002-2004

N>4000

Redo/Combined

No Redo/Combined

CABGVALVECOTHRCONGD

CAVLVAORTA

CABG 1,2,3TVR,AVRCONGDCOTHR

CABG4,5,6,7MVR

CABGVALVEAORTA

CAVLVCOTHR

IHD 2007 12

Cardiac Surgery 2002-2004

Intermediate322 mins

n=281359

In-btwn284 mins

n=890313

Minor244 minsn=1016

266

Major 1353 mins

n=116

Major 2431 mins

n=60

IHD 2007 13

Surgery Duration Distribution

0

100

200

300

120

180

240

300

360

420

480

540

600

Surgery duration (mins)

0

100

200

300

120

180

240

300

360

420

480

540

600

660

Mor

e

Surgery duration (mins)

020406080

120

200

280

360

440

520

600

680

760

840

Surgery Duration (mins)

0102030

230

320

410

500

590

680

770

860

950

1040

1130

Surgery Duration (mins)

Minor246 minsn=1530

In-btwn285 minsn=1789

Intermediate337 mins

n=499

Major461 mins

n=220

IHD 2007 14

Performance Indicators– Number of cases completed/year– Cancellation rates

• Lack of ICU/ ward bed• Out of scheduled time• More urgent case took precedent

– Operating room utilization• Under-utilization (hours/week)• Overtime (hours/week)

– Ward bed utilization (ICU & CSU)

11 hour OR

0.0

0.5

1.0

1.5

2.0

1 major1 + 1 minor

1 intermediate + 1 in-between

1 intermediate + 1 minor

2 in-between

1 in-between + 1 minor

Combinations

Undertime & Overtime

(hour/day)

0

50

100

Total Cancellations (Cases/year)Undertime

OvertimeTotal Cancellations

Which procedures should be booked together?

IHD 2007 16

Planning ICU and Ward Capacity

0

5

10

15

20

25

30

35

unit/year

Mon Tue Wed Thu Fri Sat Sun

ICUcancel (#cancellations)CSUover (# daysexceeded 30 beds)

IHD 2007 17

General Internal Medicine Resource Allocation at Toronto Western Hospital:

A System Dynamics Approach

Hannah Wong

Healthcare Resource Modelling Lab,

Mechanical & Industrial Engineering, University of Toronto

The ED and GIM units at TWH are functioning over capacity, struggling daily to cope. Decisions that affect the whole hospital are made in silos.

TWH needs a way to explore the interactions of ED & GIM with each other, other departments within the hospital, and with the outside world.

A System Dynamics Approach

ObjectivesGain a better understanding of the influence GIM has on the ED.

Explore the impact of hospital policy on ED-GIM patient flow.

Redesign policies to better allocate resources.

GIM Patient Flow

“Valves” control the rate of inflow and outflow

“Stocks” are where patients accumulate

Patient flow

Source or sink (stocks outside model boundary)

ED Wait GIM/ED GIM ALC LTC

Death

Home

ED

ED Wait GIM/ED GIM ALC LTC

Death

Home

ED

LOS

9.1 hrs

LOS

26.2 days

LOS

15.7 days

LOS

7.5 days

Current Situation

Delay62 min

Delay27.9 hrs

Delay9.4 days

I’d be interested in knowing how much we’d need to decrease patient LOS to impact ED waiting times.

I’m interested in the bed question. E.g. If we add an additional 12 beds, what is the impact on flow?

How is LOS affected if:

5 additional (or any #) of patients are discharged on Friday

All patients are discharged before 3pm weekdays

ED patients move directly to floors and do not spend time in ED.

If we’re able to cut down the time it takes to:

Identify a patient as needing to go to inpatient Rehab/CCC (ALC facility)

Complete and send off applications

Match patients to available ALC beds quicker

What is the impact to GIM LOS, ALC days, etc.?

Questions for the Model to Consider

Using System Dynamics, we can model the effect of:

ED waiting room occupancy on rates of patients who leave without being seen

ED bed occupancy on ED waiting room LOS

GIM bed occupancy on GIM LOS in ED

ALC days on GIM occupancy

Hour of day and day of week trends in GIM discharge…how does this affect the ED?

Changes in the GIM patient population (diagnosis, discharge destination, etc)

IHD 2007 23

Objectives• Help hospital decision makers make more

informed decisions by providing insights into the system.

• Explore relationships between resource levels and bottlenecks in patient flow

• determine best allocation of scarce resources (nurses, beds, etc.) to meet objectives (eg., shorter LOS in ED)

w a it in gr o o m

T r ia g e

w a it in gr o o m 2

fa s t t r a c k

c o r r id o r

o b s e r v a t io n

e v a lu a t io n

r e s u s c ita t io n

p s y c h ia ty

h o m e

lw b s

g e n e r a lm e d ic in e

g e n e r a ls u r g e r y c a r d io lo g y o r th o p a e d ic s

h o m e c a r e

A L C

lo n g te r m c a r e

a r r iv a ls

a w a it E D b e d

to lw b s to ft

to c o r

to o b s

to e v a l

to r e s u s

to p s y c h

ft- hc o r - h

o b s - he v a l- h

r e s u s - hp s y - h

c o r - g m c o r - g s c o r - c a r c o r - o r t

o b s - g m o b s - g s o b s - c a r o b s - o r t

e v a l- g m e v a l- g s e v a l- c a r e v a l- o r t

r e s - g m r e s - g s r e s - c a r r e s - o r t

p s y - g m ,

g m - h g s - h c a r - h o r t- h

g m - h c g s - h c c a r - h c o r t- h c

g m - a lc g s - a lc c a r - a lc o r t- a lc

g m - ltc g s - ltc c a r - ltc o r t- ltc

to tr ia g e

The Ottawa Hospital ED Patient Flows

Rate adjusted by:-hour of day-season

Rates at which pts. get ED bed depends on:-Time of day-# of pts. in waiting room

Admission rate depends on admitting services’:

-Nurse to patient ratio

-nursing shift in

ED LoS depends on:-Area-If consult needed

ED bed type depends on-pt. severity (CTAS) which is dependent on age

Inpt. los depends on:-Service-Discharge destination

Model Validation

Model Accuracy: ED LOS

-100%

-80%

-60%

-40%

-20%

0%

20%

40%

60%

80%

100%

c-no c-yes e-no e-yes o-no o-yes P R F

% d

evia

tion

winter springsummer autumn

yes - with consultno - without consult

% deviation: (real los – model los)

real lose.g.:corridor, no consult c-no):

real: 5.3 hrsmodel: 4.6 hrs

Hence, model has 12% deviation from real

Note: Summer results (June, July, August, 2003) are noticeably different. SARS effect

The model can help answer the following questions:

How will wait times change in the ED if inpatients are directly admitted without a visit from a consult?

What will happen if we add/reduce # of GM beds? GS beds?

If adding a nurse, which service should one be added to in order to have greatest impact on ED los?

What will happen to ED LoS if there is a x% increase in patients 90 years old or older??

Scenario Testing

IHD 2007 27

Generalized Simulation of Ontario Emergency Departments:

The CROWDED study

MW Carter1, DJT Fernandes1,2, MJ Schull2, GS Zaric3, G Geiger4

1 Healthcare Productivity Laboratory, Mechanical & Industrial Engineering, U of Toronto; 2 Institute for Clinical Evaluative Sciences;

3 Richard Ivey School of Business, University of Western Ontario4Sunnybrook and Women’s College Health Sciences Centre

IHD 2007 28

Background• ED overcrowding and waiting - major problem• Most analysis based on ED-LOS data • A few simulation models - typically model ED-

LOS• Does not help us analyze improvements• Wanted to understand what happens in an

ED

IHD 2007 29

Potential Opportunities• Evaluation of alternatives (e.g., CTAS 4, 5)• Best practices• Identify bottlenecks• Impact on wait times of:

– faster consults– improved MRI process– additional nurses, doctors, treatment rooms,

etc.– reduction in average time to admit

IHD 2007 30

Opportunities (cont.)• Evaluate various policy scenarios:

– Alternate staff schedules

– Focused care-areas (e.g., for Cardiac).

– Impact of changing demographics (e.g., study role of non-emergent pts on LOS)

– Analysis of Fast Track clinic option

IHD 2007 31

• CROWDED study: Causes and Relationships of Overcrowding and Waiting in Different Emergency Depts.

• Two full time research assistants for one year• One full time PhD student: Dominic Fernandes• Cross section of hospitals in type & geography

– 3 rural, 4 community and 3 teaching from each geographic area of Ontario

IHD 2007 32

The Hospital Partners• Academic

– Kingston General– Sunnybrook &

Women's– London HSC

• Rural– Quinte Health Corp– Stevenson

Memorial– South Muskoka

• Community– Royal Victoria -

Barrie– Sudbury Regional– Markham-

Stouffville– Windsor Regional

IHD 2007 33

Generalized Model• Build ten individual ED models• Use components to build a general model• Design an interface to allow user to create

their own model

IHD 2007 34

Data Collection• Three visits to each hospital:

– pre-visit: three days– two data collection visits: seven days each

• Followed individual patients and tracked around 60 different processes– included triage, nurse- and physician-assessments,

lab and imaging activities, consults and disposition– Patient data include presenting problem, diagnosis,

mode of arrival, age, sex, and disposition

IHD 2007 35

Data Analysis

• Approx. 2200 patients tracked • Data is being used:

– for process time distributions– to determine patient pathways

• Approximately 25 patient-types identified ~90% of pts.

• Pareto Rule: 80% of patients fall under 20% of possible diagnoses

MD-Tracking• MD-activity characteristics will be used to define resource availability in simulation model.

• Different trends can be observed among hospital types.

Pt. Ass't

Documentation

Disc'n w/ Nurse

Review Chart

Follow-Up Ass't

Phone Consult

Pt. Treatm

ent

Review Image/Rpt

General Teaching

Break

Other RuralCommunity

Academic

0

5

10

15

20

25

30

35

Perc

enta

ge o

f Tim

e

Physician Activity Hospital Type

IHD 2007 37

Challenges• Doctors are hard to track (multi-tasking)• Missing data (long LOS; trauma)• Layout issues • Off-site fast-track clinic • Wait time before triage• Unplanned critical events (SARS, Norwalk)• Preemption• Administrative issues (suspicion, turnover)

IHD 2007 38

Conclusions• Simulation is a powerful tool for analyzing

emergency departments• It is not as simple as it looks: Most of my

projects took over a year– I wrote a book chapter on Challenges

• It is worth the time and effort

IHD 2007 39

Readings• “Operations Research and Health

Care: A Handbook of Methods and Applications” International Series in Operations Research and Management Science, Vol. 70, Brandeau, Sainfort & Pierskalla (Eds.) 2004, 872 p.Chapter 8: Carter & Blake “Using

Simulation in an Acute Care Hospital: Easier Said Than Done”