template
TRANSCRIPT
Six Sigma OrientationSix Sigma OrientationPresented By:Presented By:
Joseph DuhigJoseph Duhig
University Medical Center Alliance / Methodist University Medical Center Alliance / Methodist HealthcareHealthcare
November 21, 2003November 21, 2003
The Century of Quality
“We are headed into the next century which will focus on quality… we are leaving one that has been focused on productivity”
Dr. Joseph M. Juran
SIX SIGMASIX SIGMA Sigma, , is a letter in the Greek alphabet. It is used as a symbol to
denote the standard deviation of a process (standard deviation is a measure of variation).
A process with “six sigma” capability means having six standard deviations between the process mean and either specification limit. Essentially, process variation is reduced so that no more than 3.4 parts per million fall outside the specification limits. Hence, as a metric, the higher the number of sigma’s, the better.
The “Six Sigma” term is also used to refer to a:
--philosophy--goal--methodology
to drive out waste, and improve the quality, cost and time performance of any business.
2
3
4
56
308,537
66,807
6,210
SigmaSigma Defects per Million Opportunities
Defects per Million Opportunities
2333.4 .
33 to 6 to 620,000 Times Improvement... A True Quantum Leap20,000 Times Improvement... A True Quantum Leap
What is Six Sigma?What is Six Sigma?
(99.99966% good)
(99.98% good)
(99.4% good)
(93.3% good)
(69.1% good)
Six Sigma BenchmarksSix Sigma Benchmarks
1,000,000
100,000
10,000
1,000
100
3 4 5 6 72
Sigma (Short Term) Scale of Measure
Restaurant Bills
Doctor Prescription Writing
Domestic AirlineFatality Rate(0.43 PPM)
IRS Tax Advice(phone in)
Airline Baggage Handling
AverageAverageCompanyCompany
Best-in-ClassBest-in-Class
1
10
1
De
fect
s p
er M
illio
n
Getting To Six Sigma - Some ExamplesGetting To Six Sigma - Some Examples
Six Sigma99.99966% Good
Six Sigma99.99966% Good
• 20,000 lost articles of mail per hour
• Unsafe drinking water for almost 15 minutes each day
• 5,000 incorrect surgical operations per week
• Two short or long landings at most major airports each day
• 200,000 wrong drug prescriptions each year
• Seven articles lost per hour
• One unsafe minute everyseven months
• 1.7 incorrect operations per week
• One short or long landing every five years
• 68 wrong prescriptions per year
3.8 Sigma99% Good3.8 Sigma99% Good
THE CENTURY OF QUALITYTHE CENTURY OF QUALITY
“We are headed into the next century which will focus on quality… we are leaving one that has been focused on productivity”
Dr. Joseph M. Juran
WHATWHAT IS SIX SIGMA QUALITY?IS SIX SIGMA QUALITY?
Quality
ProductFeatures
Freedom fromDeficiencies
That Customers Want
Design for Six Sigma
At Six Sigma Levels
Improve to Six Sigma
METHODOLOGYMETHODOLOGYDEFINE Identify, prioritize, and
select the right project(s)
MEASURE Identify key product characteristics & process parameters, understand processes, and measure performance
ANALYZE Identify the key (causative)process determinants
IMPROVE Establish prediction modeland optimize performance
CONTROL Hold the gains
INPUT
Project
MissionStatement
DefineDefine
Define customers & CTQ’s
•Prioritized list of customers/segments•Prioritized list of CTQ’s
Define process to be improved
•High level Process Map
Define Project Charter
•Project Charter
MeasureMeasure
Establish Project Y’s
•Performance Measurement Matrix
Identify possible X’s
•Detailed Process Map, C&E Diagram, FMEA
Plan Data Collection
•Data Collection Plan
Validate Measurement System
•Gage R&R, Discrete Data Measure Analysis
Determine Process Capability•Baseline Six Sigma values
AnalyzeAnalyze ImproveImprove ControlControl
Develop and test hypotheses on the sources variation and
cause-effect relationships
•Stated theory (s)•Hypothesis testing results
Develop the list of vital few causes of process
performance
•List of “vital few” variations that account for the majority of variation in performance•Quantified $ Opportunity
•List of possible solutions to test or operating parameters for experimentation
Generate Solution Alternatives
•List of possible risks evaluated for level of seriousness and corresponding abatement actions as needed.
Assess Risk
•Results of DOE and/or pilot and/or simulation
Test Solution Alternatives. Select Solution(s) to optimize
performance
•SPC charts in place•Feedback mechanisms and Mistake Proofing devices implemented
Design and implement sustainable feedback
mechanisms and methods to achieve self control for dominant variables.
•Updated Standard Operating Procedures (SOP), Process Maps, FMEA•Preventative Maintenance Plans•Personnel trained
Control Plans and Documentation.
•Final project report•Audit plan
Document Project work. Close Project
Module
•Deliverables
•Possible Tools
•VOC Continuum, Surveys, •Interviews
•List of Possible Xs
•List of Project Ys
•Reliable Measurement System
•ANOVA, tests for equal variance, regression, t-tests, tests for proportions, contingency tables, non-parametric tests, Detailed Process Map, C&E Diagram, FMEA, Pareto
•Designed Experiments, Pilots, Simulations
SIX SIGMA TOOLBOXSIX SIGMA TOOLBOXAnalysis of Variance (ANOVA)Box Plots BrainstormingCause-effect Diagrams Correlation & RegressionDesign Of ExperimentsEvolutionary Operation (EVOP)FMECA Graphs and ChartsHistogramsHypothesis TestingLean Manufacturing (Lean Enterprise)Measurement System AnalysisMistake ProofingPareto AnalysisProcess Capability StudiesProcess Control PlansProcess Flow DiagramsQuality Function DeploymentResponse Surface MethodsScatter DiagramsStandard Operating Procedures (SOPs)Statistical Process ControlStratification
Why We Need Six Why We Need Six Sigma in HealthcareSigma in Healthcare
Presented By:Presented By:
Joseph DuhigJoseph Duhig
University Medical Center Alliance / Methodist University Medical Center Alliance / Methodist HealthcareHealthcare
November 21, 2003November 21, 2003
The Century of Quality
“We are headed into the next century which will focus on quality… we are leaving one that has been focused on productivity”
Dr. Joseph M. Juran
GOOD NEWSGOOD NEWS Incredible Advances in Medicine 2 Million Articles/20,000 Journals/Year Applying this knowledge is like:
“Trying to drink water from a fire hose”
BAD NEWSBAD NEWSThe average time from discovery of knowledge until thatknowledge is in wide-spread use is over 17 years
The IOM RoundtableThe IOM Roundtable
“…Serious and widespread quality problems exist throughout American medicine. These problems… occur in small and large communities alike, in all parts of the country, and with approximately equal frequency in managed care and fee-for-service systems of care. Very large numbers of Americans are harmed as a result….”
Source: 2002 Institute for Healthcare Improvement
The call to action...
What is Wrong??What is Wrong?? OVERUSE (of procedures, medications, visits
that cannot help)
UNDERUSE (of procedures, medications, visits that can help)
MISUSE (errors of execution)
Source: 2002 Institute for Healthcare Improvement
Examples of OVERUSEExamples of OVERUSE 30% of children receive excessive antibiotics for
ear infections
20% to 50% of many surgical operations are unnecessary
50% of X-rays in back pain patients are unnecessary
Source: 2002 Institute for Healthcare Improvement
Examples of UNDERUSEExamples of UNDERUSE 50% of elderly fail to receive pneumococcal vaccine
50% of heart attack victims fail to receive beta-blockers
27% of high blood pressure is adequately treated
Source: 2002 Institute for Healthcare Improvement
Examples of MISUSEExamples of MISUSE
7% of hospital patients experience a serious medication error
44,000-98,000 Americans die in hospitals each year due to injuries in care
Source: 2002 Institute for Healthcare Improvement
What the IOM Said….What the IOM Said…. The patient safety problem is large.
It (usually) isn’t the fault of health care workers.
Most patient injuries are due to system failures.
Source: 2002 Institute for Healthcare Improvement
The Situation – Health Care CostsThe Situation – Health Care Costs
0
2
4
6
8
10
12
141
99
1
19
92
19
93
19
94
19
95
19
96
19
97
19
98
19
99
20
00
20
01
Large Firms
All Firms
How Hazardous is Health Care?How Hazardous is Health Care?((Leape)Leape)
1
10
100
1000
10000
100000
1 10 100 1000 10,000 100,000 1,000,000 10,000,000
Tot
al l
ives
lost
pe
r ye
ar
DANGEROUS(>1/1000)
REGULATED ULTRA-SAFE(<1/100K)
HealthcareDriving
Scheduled Airlines
CharteredFlights
ChemicalManufacturing
MountainClimbing
BungeeJumping
European Railroads Nuclear Power
Number of encounters for each fatalitySource: 2002 Institute of Healthcare Improvement
Core ConclusionsCore Conclusions
There are serious problems in quality and safety.--Between the health care we have and the care we
could have lies not just a gap but a chasm.
The problems come from poor systems…not bad people--In its current form, habits, and environment, American
health care is incapable of providing the public with the quality health care it expects and deserves.
We can fix it…but it will require changes.
Source: 2002 Institute for Healthcare Improvement
““The First Law of Improvement”The First Law of Improvement”
Every system is perfectly designed to achieve exactly the results it gets
Source: 2002 Institute for Healthcare Improvement
Quality is a system property
Why Six Sigma?Why Six Sigma?
Safe Timely Efficient Effective Equitable Patient-centered
Variation is the Key:Six Sigma is all about understanding variation in providing
care that is:
How is Six Sigma different from traditional How is Six Sigma different from traditional Performance Improvement ApproachesPerformance Improvement Approaches
Strategically Deployed Financially Focused Trained Professionals vs. Good Intentioned
Amateurs Statistically Based Y = f(x) Project Management is Built-in Measurement System is Validated Focus on Mistake Proofing – Failure Modes and
Effects Analysis (FMEA)
The Business Case – Doing Well by Doing GoodThe Business Case – Doing Well by Doing Good
Six Sigma Impact on Net Income
Six Sigma Results
Discounted FFS
Per Case Per Diem Shared Risk
Decreased cost/unit
Decreased # units/case
Decreased LOS
Decreased # of cases
PROJECT FOCUSPROJECT FOCUS
Process Problems and
Symptoms Process outputs Response variable, Y
Independent variables, Xi
Process inputs The Vital Few determinants Causes Mathematical relationship
Y
X’s
Measure
Analyze
Improve
Control
Pro
cess
Cha
ract
eriz
atio
nP
roce
ss
Opt
imiz
atio
n
Goal: Y = f ( x )
Define The right project(s), the right team(s)
PROCESS CONTEXT FOR PROCESS CONTEXT FOR MEASUREMENTMEASUREMENT
Y = f(X1, X2,... , Xn)
MeasuresMeasuresMeasuresMeasures
PPSS II OO CC
ProcessMap
Suppliers Inputs Process Outputs Customers
CTQs
MeasuresMeasuresMeasuresMeasures
MeasuresMeasuresMeasuresMeasures
AHRQ Medicare SMR vs. Standardised Charge, AHRQ Medicare SMR vs. Standardised Charge, 1997 1997 (Random Sample 250 Hospitals Plotted)(Random Sample 250 Hospitals Plotted)
0
20
40
60
80
100
120
140
160
180
0 5000 10000 15000 20000 25000
Source: 2002 Institute for Healthcare Improvement
$ 3,922$ 4,439$ 4,940$ 5,444$ 6,304
Per-capitaMedicare Spending1996 2000
The cohorts had The cohorts had similarsimilar baseline baseline health across quintileshealth across quintilesBut were But were treated differentlytreated differently. .
Ratio: High to Low: 1.61 1.58
$ 5,229$ 5.692$ 6,069$ 6,614$ 8,283
Glucose Levels of Diabetic Cardiac Surgery PatientsGlucose Levels of Diabetic Cardiac Surgery Patients
0 100 200 300 400
LSL USL
Process Capability Analysis for Blood Sugar
USL
Target
LSL
Mean
Sample N
StDev (Within)
StDev (Overall)
Cp
CPU
CPL
Cpk
Cpm
Pp
PPU
PPL
Ppk
% < LSL
% > USL
% Total
% < LSL
% > USL
% Total
% < LSL
% > USL
% Total
150.000
*
80.000
193.386
329
30.8392
55.9094
0.38
-0.47
1.23
-0.47
*
0.21
-0.26
0.68
-0.26
0.30
76.60
76.90
0.01
92.03
92.04
2.13
78.11
80.24
Process Data
Potential (Within) Capability
Overall Capability Observed Performance Exp. "Within" Performance Exp. "Overall" Performance
Within
Overall
OR First Case Start TimeOR First Case Start Time
0 10 20 30 40 50 60 70 80
USLUSL
Process Capability Analysis for Case Rolled
USL
Target
LSL
Mean
Sample N
StDev (Within)
StDev (Ov erall)
Cp
CPU
CPL
Cpk
Cpm
Pp
PPU
PPL
Ppk
PPM < LSL
PPM > USL
PPM Total
PPM < LSL
PPM > USL
PPM Total
PPM < LSL
PPM > USL
PPM Total
30.0000
*
*
37.1050
238
8.4564
10.0149
*
-0.28
*
-0.28
*
*
-0.24
*
-0.24
*
710084.03
710084.03
*
799601.67
799601.67
*
760977.44
760977.44
Process Data
Potential (Within) Capability
Ov erall Capability Observ ed Perf ormance Exp. "Within" Perf ormance Exp. "Ov erall" Perf ormance
Within
Overall
SOURCES OF VARIATIONSOURCES OF VARIATION
People
ProcessProcess
Place Procedure
Provisions Measurement Patrons
“Y”
5 P’s + 1 M
Measurements
Commonor
Special?
CommonCausesMEASURE
ANALYZE
Investigate all of the variation
Develop solutions forthe “vital few” process
and input XsIMPROVE
Develop solutions forspecial causes and
implement asappropriate
IMPROVE
SpecialCausesMEASURE
COMMON vs. SPECIAL CAUSESCOMMON vs. SPECIAL CAUSES
CONTROL
Sustain The Improvements
Investigate specific data points
ANALYZE
Measurements
Two Types Of Mistakes
How you treat variation . . .
Common Causes Special Causes
Common
Causes
Special
Causes
Mistake 1Tampering
(increases variation)
Focus on fundamentalprocess change
Mistake 2
Underreacting(missed prevention)
Focus on investigatingspecial causes
What the
variation
really is...
COMMON vs. SPECIAL CAUSESCOMMON vs. SPECIAL CAUSES
Suppose we say that there are 4 key characteristics which must be executed (without error) in order to par the hole. In this case, what is the probability of accomplishing the task error free?
Rolled Yield .7581 .9999864
33 66With
Shifting
Tee Shots .9331 .9999966
Fairway Shots .9331 .9999966
Chipping .9331 .9999966
Putting .9331 .9999966
Rolled Throughput Yield
CALCULATING SIGMA - YIELDCALCULATING SIGMA - YIELD
Remedy 1: Remedy 1: Reduce Parts/StepsReduce Parts/Steps
Remedy 2:Remedy 2:Improve Sigma per Part/StepImprove Sigma per Part/Step
Yields thru Multiple Steps/Parts/ProcessesZst
(distribution shifted 1.5)
# of parts, steps, or
processes3 4 5 6
1 93.32% 99.38% 99.9767% 99.99966%
5 70.77% 96.93% 99.88% 99.9983%
10 50.09% 93.96% 99.77% 99.997%
20 25.09% 88.29% 99.54% 99.993%
50 3.15% 73.24% 98.84% 99.983%
100 53.64% 97.70% 99.966%
200 28.77% 95.45% 99.932%
500 4.44% 89.02% 99.830%
1000 0.20% 79.24% 99.660%
2000 62.79% 99.322%
10000 9.76% 96.656%
Yields thru Multiple Steps/Parts/ProcessesZst
(distribution shifted 1.5)
# of parts, steps, or
processes3 4 5 6
1 93.32% 99.38% 99.9767% 99.99966%
5 70.77% 96.93% 99.88% 99.9983%
10 50.09% 93.96% 99.77% 99.997%
20 25.09% 88.29% 99.54% 99.993%
50 3.15% 73.24% 98.84% 99.983%
100 53.64% 97.70% 99.966%
200 28.77% 95.45% 99.932%
500 4.44% 89.02% 99.830%
1000 0.20% 79.24% 99.660%
2000 62.79% 99.322%
10000 9.76% 96.656%
CALCULATING SIGMA - YIELDCALCULATING SIGMA - YIELD
YIELD DECREASES WHEN COMPLEXITY INCREASESYIELD DECREASES WHEN COMPLEXITY INCREASES
“Is the variation (spread) of my measurement system too large to study the current level of process variation?”
+ =
(Observed Variability)
Total VariabilityProduct VariabilityProcess Variability
Variationin the
measurementprocess
THE FUNDAMENTAL MSA QUESTIONTHE FUNDAMENTAL MSA QUESTION
To address actual process To address actual process variability, the variation due to the variability, the variation due to the measurement system must first be measurement system must first be identified and separated from that of identified and separated from that of the process. the process.
Observed Process Variation
Actual Process Variation
Measurement Variation
Long-term Process Variation
Short-term Process Variation
Repeatability
Accuracy
Stability
Linearity
POSSIBLE SOURCES OF VARIATIONPOSSIBLE SOURCES OF VARIATION
Reproducibility
LEVELS OF ANALYSISLEVELS OF ANALYSIS
Measure 1 Individual Experience
Measure 2 Group Experience
Analyze 3 Graphical Interpretation of Observed Data
Analyze 4 Statistical Interpretation of Observed Data
Improve 5 Graphical Interpretation of Experimental Data
Improve 6 Statistical Interpretation of Experimental Data
“ Think Directional”
Continuous DataContinuous Data Discrete DataDiscrete Data
DiscreteData
ContinuousData
X
Y
THE ANALYSIS TOOL DEPENDS ON THE THE ANALYSIS TOOL DEPENDS ON THE QUESTION AND THE DATA TYPEQUESTION AND THE DATA TYPE
Variance Different ?
Statistical: Test of Equal Variances
Graphical: Stratified Box Plots
Variance Different ?
Statistical: Test of Equal Variances
Graphical: Stratified Box Plots
Means Different ?
Statistical: t-test; ANOVA
Graphical: Histogram(s)
Means Different ?
Statistical: t-test; ANOVA
Graphical: Histogram(s)
How does change in X affect change in Y ?
Statistical: Correlation /Regression
Graphical: Scatter Plots
How does change in X affect change in Y ?
Statistical: Correlation /Regression
Graphical: Scatter Plots
How does change in X affect change in Y ?
Statistical: Logistic Regression
How does change in X affect change in Y ?
Statistical: Logistic Regression
Are the outputs different ?
Statistical: Chi Square, Proportion tests
Graphical: Stratified Pareto Diagrams
Are the outputs different ?
Statistical: Chi Square, Proportion tests
Graphical: Stratified Pareto Diagrams
Allows us to answer the practical question:“Is there a real difference between Dr. A and Dr. B ?”
A practical process problem is translated into a statistical hypothesis so that we may answer the question above.
In hypothesis testing, we use relatively small samples to answer questions about large populations. There is always a chance that we selected a sample that is not representative of the population - a “weird” sample. Therefore, there is always a chance that the conclusion obtained is wrong.
With some assumptions, inferential statisticsnferential statistics allows us to estimate the probability of getting a “weird” sample. Hypothesis testing quantifies the probability (P-Value) of a wrong conclusion.
Data vs. Gut Feeling
HYPOTHESIS TESTING DESCRIPTIONHYPOTHESIS TESTING DESCRIPTION
TruthTruth
HoHo HaHa
Fail to Reject HoFail to Reject Ho
Reject HoReject Ho
Type IError
Type IIError
Correct
Decision
CorrectDecision
Also called: Type II error Consumers’ Risk
Also called: Type I error Producers’ Risk
is the risk of finding a difference when there really isn’t one.
is the risk of not finding a difference when there really is one.
ALPHA & BETA RISKALPHA & BETA RISK