Colorado 5M WebEx Variation, Run Charts, and Control Charts

Colorado 5M WebEx Variation, Run Charts, and Control Charts Beth A. Katzenberg, EdM, MBA, CPHQ Director, Corporate Quality & Compliance Colorado Foundation for Medical Care 1 Types of variation Common cause  Always present  Inherent in process  Can predict performance with a range of variation  Cannot tell what specifically causes variation Special cause  Abnormal, unexpected  Due to causes not inherent in process  Can be identified (e.g., change in shift, weather, process) 2 You must understand the type of variation that is occurring as this will determine how you address the problem. 3 Variation Type of variation Appropriate action to take Common cause  Change the process (predictable, stable,  Do not react to individual in control, inherent differences or try to explain in process) differences between high and low numbers Special cause (unpredictable, unstable, out of control) Identify and study special cause  If negative, minimize or prevent  If positive, build into process  4 Pitfalls  If only common cause variation and treat as special cause (tampering), leads to greater variation, mistakes, defects  If common cause and special cause, and change the process, leads to wasted resources because the change won’t work 5 Tools to identify variation 6 Run charts 7 Run chart Run Chart 1.07 - 12.07 50 Number 40 30 20 10 0 1.07 2.07 3.07 4.07 5.07 6.07 7.07 8.07 9.07 10.07 11.07 12.07 Tim e Fram e (Month.Year) Median Graph of data over time Track performance Display & identify variation 8 Run chart analysis: Common cause variation only 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Tim e Common cause variation around the median: Only common cause variation present. Output may or may not meet customer/ patient requirements 9 Run chart analysis: Runs   Run = one or more consecutive data points on the same side of the median Excludes data points on the median 12 10 8 6 4 2 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 10 11 runs Expected number of runs # data pts not on median 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 Smallest Largest run run count count 3 8 3 9 3 10 4 10 4 11 4 12 5 12 5 13 6 13 6 14 6 15 7 15 7 16 8 16 8 17 9 17 # data pts not on median 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 Smallest Largest run run count count 9 18 9 19 10 19 10 20 11 20 11 21 11 22 11 22 12 23 13 23 13 24 13 25 14 25 14 26 15 26 16 26 11 High probability ( = 0.05) of special cause variation: Too few runs Too many runs 12 Run chart analysis: Run length 8 6 4 2 0 1 2 3 4 5 6 7 8 Time 9 10 11 12 13 14 15 Special cause—run length: <20 data points (not on median): A run of 7 data points on one side of the median (either above or below) 20+ data points (not on median): A run of 8 data points on one side of the median 13 Run chart analysis: Trends 8 6 4 2 0 1 2 3 4 5 6 7 8 Time 9 10 11 12 13 14 15 Special cause—trends: Consecutive points all going up or all going down. May cross the median. Ignore 2+ consecutive points that are the same. # Consecutive points all increasing or decreasing 4 5 6 7 Total # data points on chart 5 to 8 9 to 20 21 to 100 101 or more 14 (Pyzdek, 2003) Run chart analysis: Freaks 10 8 6 4 2 0 1 2 3 4 5 6 7 8 Time 9 10 11 12 13 14 15 Freaks: The presence of more than one or two dramatic spikes suggests the process is out of control. Run charts not as sensitive in identifying, thus may fail to detect. 15 Run chart analysis: Cycling 10 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 Cycling: A zigzag or saw-tooth pattern with 14+ points in a row alternating up or down. 16 Run charts tips  How many data points?  15-20 minimum is preferable  Median = 50%/50% split point  Precisely half of the data set will be above the median and half below it 17 Control charts 18 Control chart High An indication of a special cause UCL Quality Characteristic X LCL Low Time Run chart with control limits Determines type of variation Is process stable? Predictable? 19 Dividing control chart into zones Zone A Each zone is 1 sigma wide UCL Zone B Zone C Zone C Zone B Zone A X LCL 20 Identifying special causes  Apply independently to each side of the center line: point outside the 3 sigma limit  2 out of 3 consecutive points in zone A or beyond  4 out of 5 consecutive points in zone B or beyond  <20 total data points: 7 consecutive points in zone C or beyond on one side of center line  20+ total data points: 8 consecutive points in zone C or beyond on one side of center line (continued) 21 1 Identifying special causes, cont.  Apply this test to entire chart:  <21 total data points: 6 or more points in a row steadily increasing or decreasing  21+ total data points: 7 or more points in a row steadily increasing or decreasing  14 consecutive points alternating up and down in saw-tooth pattern  15 consecutive points in zone C (above and below center line) 22 Deciding which control chart to use Decide on type of data Attributes (count, discrete) data (Values in discrete categories; e.g., % waste, # falls, # errors, % incomplete charts) Continuous (Variables, measurement) data (Values on continuous scale; e.g., time, temperatures, cost) Yes More than one observation per sub-group? No Yes Can both occurrences and non-occurrences be counted? No Yes Fewer than 10 observations per sub-group? No Yes Are the subgroup sizes equal? No Yes Are there equal areas of opportunity? No X -R chart Average & range chart X -S chart Average & standard deviation (sigma) chart XmR chart Individual & moving range chart np-chart p-chart c-chart u-chart Source: Carey, R. C. and Lloyd, R. C. Measuring Quality Improvement in Healthcare: A Guide to Statistical Process Control Applications, 1995 23 Types of data Count/attribute Count observations or incidents falling into categories Whole numbers only Cannot be converted to measurement  Yes/no  Dead/alive  Infected/not infected  On time/late  % c-sections  % incomplete charts  # pt falls  # medication errors Measurement/continuous Take on values on a continuous scale Whole numbers and decimals Can be converted to count       Time in minutes or hours Weight in grams Length of stay Blood sugar levels Costs Temperature 24 Control chart example 1 CBC Turn Around Time 120 CBC Turn Around Time (Minutes) UCL = 114.6 110 100 90 80 70 60 50 40 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 LCL = 51.9 X = 83.3 Day (Not Counting Weekends) Common cause variation only 25 Control chart example 2 new hire snowstorm 26 Control chart example 3 Net Operating Margin for Hospital A 1/05-9/06 14 12 10 8 Percent UCL = 12.1 6 4 2 0 -2 -4 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 Month LCL = -2.9 X = 4.6 (From: Carey, R. G. & Lloyd, R. C. Measuring Quality Improvement in Healthcare Common cause variation only; can predict will stay within control limits, if no changes 27 Control chart example 4 Net Operating Margin for Hospital B 1/92-9/93 12 10 8 Percent UCL = 9.25 6 4 2 0 -2 -4 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 Month LCL = -.04 X = 4.60 (From: Carey, R. G. & Lloyd, R. C. Measuring Quality Improvement in Healthcare Out of control, unpredictable 28 Just because a process is under control (common cause variation only), it does not mean that the process is meeting expectations. It just means that the process is predictable and you are getting consistent performance. 29 Control charts tips  Control limits are not specifications limits (specification limits related to customer requirements) After removing special causes and recalculating chart, continue to plot new data on this chart, without recalculating control limits.   Recalculate control limits only when a permanent, desired change has occurred in the process and only using data after the change occurred 30 Share the data     Team meetings Post in break-rooms Newsletters Intranet 31 Examples of Software QI Macros www.qimacros.com  StatSoft www.statsoft.com  Minitab www.minitab.com  32 References Carey, R.G. & Lloyd, R.C. Measuring Quality Improvement in Healthcare: A Guide to Statistical Process Control Applications, Quality Resources, 1995. Pyzdek, R. The Six Sigma Handbook: A Complete Guide for Green Belts, Black Belts, and Managers at All Levels, 2003. The Six Sigma Memory Jogger II, GOAL/QPC, 2002. 33 Beth Katzenberg, EdM, MBA, CPHQ Director, Corporate quality & compliance Colorado Foundation for Medical Care bkatzenberg@cfmc.org 34