Tutorial on Risk Adjusted P-chart Farrokh Alemi, Ph.D.

Slides:



Advertisements
Similar presentations
Estimation of Means and Proportions
Advertisements

Introduction to Control Charts By Farrokh Alemi Ph.D. Sandy Amin Based in part on Amin S. Control charts 101: a guide to health care applications. Qual.
Statistics Review – Part II Topics: – Hypothesis Testing – Paired Tests – Tests of variability 1.
Tutorial on Tukey Charts Farrokh Alemi, Ph.D. Sunday, 11/25/2007.
1 Manufacturing Process A sequence of activities that is intended to achieve a result (Juran). Quality of Manufacturing Process depends on Entry Criteria.
Time Between Charts Farrokh Alemi, Ph.D.. Steps in construction of time in between charts 1. Verify the chart assumptions 2. Select to draw time to success.
S6 - 1© 2011 Pearson Education, Inc. publishing as Prentice Hall S6 Statistical Process Control PowerPoint presentation to accompany Heizer and Render.
Nursing Home Falls Control Chart Problem Statement: Assume that following data were obtained about number of falls in a Nursing Home facility. Produce.
Chapter 18 Introduction to Quality
Introduction to Control Charts.
Confidence Interval and Hypothesis Testing for:
© 2008 Prentice Hall, Inc.S6 – 1 Operations Management Supplement 6 – Statistical Process Control PowerPoint presentation to accompany Heizer/Render Principles.
8-1 Quality Improvement and Statistics Definitions of Quality Quality means fitness for use - quality of design - quality of conformance Quality is.
Chapter Topics Confidence Interval Estimation for the Mean (s Known)
Copyright © 2014 by McGraw-Hill Higher Education. All rights reserved. Essentials of Business Statistics: Communicating with Numbers By Sanjiv Jaggia and.
Inferences About Process Quality
Total Quality Management BUS 3 – 142 Statistics for Variables Week of Mar 14, 2011.
© 2000 by Prentice-Hall Inc Russell/Taylor Oper Mgt 3/e KR: Chapter 7 Statistical Process Control.
Control Charts. On a run chart the centerline is the median and the distance of a data point from the centerline is not important to interpretation On.
The Bell Shaped Curve By the definition of the bell shaped curve, we expect to find certain percentages of the population between the standard deviations.
XmR Chart Farrokh Alemi, Ph.D.. Purpose of Control Chart  Real or random.  Tell a story of changes in outcomes of the process.
Discrete Probability Distributions
Linear Regression Inference
Lesson 7.1 Quality Control Today we will learn to… > use quality control charts to determine if a manufacturing process is out of control.
Go to Index Analysis of Means Farrokh Alemi, Ph.D. Kashif Haqqi M.D.
Inference for a Single Population Proportion (p).
Comparing Two Population Means
Introduction to Control Charts: XmR Chart
Diving into the Deep: Advanced Concepts Module 9.
Basic Business Statistics, 11e © 2009 Prentice-Hall, Inc. Chap 8-1 Confidence Interval Estimation.
Process Capability and Statistical Process Control.
Go to index Two Sample Inference for Means Farrokh Alemi Ph.D Kashif Haqqi M.D.
© 2006 Prentice Hall, Inc.S6 – 1 Operations Management Supplement 6 – Statistical Process Control © 2006 Prentice Hall, Inc. PowerPoint presentation to.
Statistical Process Control (SPC)
Economics 173 Business Statistics Lecture 20 Fall, 2001© Professor J. Petry
Go to Table of Content Single Variable Regression Farrokh Alemi, Ph.D. Kashif Haqqi M.D.
MORE THAN MEETS THE EYE Wayne Gaul, Ph.D., CHP, CHMM Tidewater Environmental Columbia, SC SRHPS Technical Seminar, April 15, 2011.
Relative Values. Statistical Terms n Mean:  the average of the data  sensitive to outlying data n Median:  the middle of the data  not sensitive to.
The Single-Sample t Test Chapter 9. The t Distributions >Distributions of Means When the Parameters Are Not Known >Using t distributions Estimating a.
Statistical Quality Control
1 Six Sigma Green Belt Introduction to Control Charts Sigma Quality Management.
The Single-Sample t Test Chapter 9. t distributions >Sometimes, we do not have the population standard deviation. (that’s actually really common). >So.
Slide 1 Copyright © 2004 Pearson Education, Inc..
Confidence Intervals for a Population Mean, Standard Deviation Unknown.
INFERENCE Farrokh Alemi Ph.D.. Point Estimates Point Estimates Vary.
Time To Missed Exercise Farrokh Alemi, Ph.D.. Why do it? You need to distinguish between random days of missed exercise from real changes in underlying.
Risk Adjusted X-bar Chart Farrokh Alemi, Ph.D. Based on Work of Eric Eisenstein and Charles Bethea, The use of patient mix-adjusted control charts to compare.
MOS 3330 Operations Management Professor Burjaw Fall/Winter
Welcome to MM305 Unit 8 Seminar Prof. Dan Statistical Quality Control.
Statistical principles: the normal distribution and methods of testing Or, “Explaining the arrangement of things”
 List the characteristics of the F distribution.  Conduct a test of hypothesis to determine whether the variances of two populations are equal.  Discuss.
Inference for a Single Population Proportion (p)
Chapter 13 Simple Linear Regression
Probability Distributions for Discrete Variables
Log Linear Modeling of Independence
P-Chart Farrokh Alemi, Ph.D..
Process Capability.
Statistics for Business and Economics
Xbar Chart Farrokh Alemi, Ph.D..
CHAPTER 12 More About Regression
Analysis of Observed & Expected Infections
P-Chart Farrokh Alemi, Ph.D. This lecture was organized by Dr. Alemi.
Xbar Chart By Farrokh Alemi Ph.D
Introduction to Control Charts
Time between Control Chart: When Is Harm Reduction Drug Use?
Risk Adjusted P-chart Farrokh Alemi, Ph.D.
Type I and Type II Errors
Limits from Pre- or Post-Intervention
Time-between Control Chart for Exercise By Farrokh Alemi Ph.D
Tukey Control Chart Farrokh Alemi, Ph.D.
Presentation transcript:

Tutorial on Risk Adjusted P-chart Farrokh Alemi, Ph.D.

Have Changes Led to Improvement? Common cause variation (changes in outcomes because of chance) is everywhere. Decision makers often mistakenly attribute positive outcomes to their own skills and negative outcomes to others, while in reality both could be a chance outcome

Why Chart Data? To discipline intuitions To communicate data in vivid graphical ways

Purpose of Risk Adjustment Purpose of p-chartPurpose of risk- adjusted p-chart To detect if the process has improved beyond historical levels. Assumes historically we have been serving the same type of patients as now To detect if the process has improved beyond what can be expected from patient conditions

Data Needed Data collected over time Risk (expected outcomes) for each patient Outcomes for each patient The purpose is to improve not to get so lost in measurement to loose sight of improvement.

What Is Risk? A patient’s condition or characteristics that affects the expected outcomes for the patient A severity index used to predict patient outcomes Clinicians’ consensus regarding expected outcomes Patient’s self rating of expected outcomes

MI Patients Over 8 Months in One Hospital Observed mortality during this time period Number of cases Expected probability of mortality for case 8 in time period 1. Estimated from severity indices or experts’ consensus.

Elements of a Control Chart X axis shows time Y axis shows probability of adverse events Observed rates are plotted against time sequence Upper control limit is a line drawn so that points above it are rare to be observed by mere chance Lower control limit is a line drawn so that points below it are rare to observe by mere chance Lets take a look

An Example of P-chart Upper control limit Lower control limit Observed rate

Steps in Creating P- chart for Mortality 1. Check assumptions 2. Calculate observed rates and plot them 3. Calculate expected rates and plot them 4. Calculate expected deviation 5. Calculate control limits and plot them 6. Interpret findings 7. Distribute chart and interpretation

Step One: Check Assumptions We are examining discrete events that either happen or do not happen, e.g. mortality among MI patients, falls among nursing home patients, error in medical record entry, etc. The event is not rare, meaning the probability of it occurring exceeds 5% for each time period. Observed events are independent from each other. The probability of the event occurring does not change over time. This assumption is violated if one patient’s outcomes affects the outcomes for others, e.g. when dealing with infectious diseases.

Step Two: Calculate Observed Rates and Plot P i = Mortality rate in time period “i” O i = Mortality in period “i” n i = Number of cases in time period “i” P i = O i / n i

Observed Mortality Rates for All Time Periods Plot of mortality rates

Plot of the Observed Rates Time period 7 and 3 seem different but don’t rush to judgment. Wait, until you see control limits of what could have been expected.

Step Three: Calculate Expected Mortality E ij = Expected mortality of case ‘j’ in time period “i” E i = Expected mortality for time period “i” E i = (  j=1,…,n i E ij ) / n i Sample calculation: E 1 = ( )/8

Expected Mortality Rates for All Time Periods

Plot of Expected Mortality Plotting expected mortality helps interpret the observed rates but does not settle the question of whether differences are due to chance.

Step Four: Calculate Expected Deviation E ij = Expected mortality of case ‘j’ in time period “i” D i = Standard deviation of expected mortality in time period “i”, called by us as the Expected Deviation D i = (  j=1,…,n i E ij (1-E ij )) 0.5 / n i See sample calculation

Expected Deviation for Time Period 1 Expected deviation A B C D

Results: Expected Deviations for All Time Periods

Step Five: Calculate Control Limits UCL i = Upper control limit for time period “i” LCL i = Lower control limit for time period “i” t = Constant based on t-student distribution UCL i = E i + t * D i LCL i = E i - t * D i Where for 95% confidence intervals: Degrees of freedom 95% student t value

Calculation of Control Limits for Time Period 1 UCL = *.13 =.59 LCL =.28 –2.37 *.13 = -.03 t-value Negative limits are set to zero as negative probabilities are not possible

Results: Control Limits for All Time Periods

Plot Control Limits UCL LCL

Step Six: Interpret Findings There are no points above UCL. There is one point below LCL. In time period 3, mortality is lower than what can be expected from patient’s conditions. All other time periods are within expectations, even time period 7 with its high mortality rate is within expectation.

Step Seven: Distribute Control Chart Include in the information: How was severity measured and expected mortality anticipated? Why are assumptions met? What does the control chart look like? What is the interpretation of the findings?

Index of Content Click on the Slide You Wish to Review 1. Check assumptions Check assumptions 2. Calculate and plot observed mortality Calculate and plot observed mortality 3. Calculate expected mortality Calculate expected mortality 4. Calculate expected deviation Calculate expected deviation 5. Calculate and plot control limits Calculate and plot control limits 6. Interpret findings Interpret findings 7. Distribute control chart Distribute control chart