2Agenda Features of Run Charts Interpreting Run Charts A quick mention of variationFeatures of Shewhart ChartsInterpreting Shewhart Charts
3Displaying Key Measures over Time - Run Chart Data displayed in time orderTime is along X axisResult along Y axisCentre line = medianOne “dot” = one sample of data
4Three Uses of Run Charts in Quality Work Determine if change is an improvementMedian 429The Data Guide, p 3-184
5Three Uses of Run Charts in Quality Work 2. Determine if improvement is sustainedMedian 429The Data Guide, p 3-185
6Three Uses of Run Charts in Quality Work 3. Make process performance visibleMedian 429The Data Guide, p 3-186
7How Do We Analyze a Run Chart? Visual analysis firstIf pattern is not clear, then apply probability based rulesThe Data Guide, p 3-107
8Non-Random Signals on Run Charts A Trend5 or moreA Shift:6 or moreToo many or too few runsAn astronomical data pointEvidence of a non-random signal if one or more of the circumstances depicted by these four rules are on the run chart. The first three rules are violations of random patterns and are based on a probability of less than 5% chance of occurring just by chance with no change.The Data Guide, p 3-118
9Source: Swed, Frieda S. and Eisenhart, C Source: Swed, Frieda S. and Eisenhart, C. (1943) “Tables for Testing Randomness of Grouping in a Sequence of Alternatives.” Annals of Mathematical Statistics. Vol. XIV, pp , Tables II and III.The Data Guide, p 3-149
13Interpretation? There is a signal of a non-random pattern There is less than 5 % chance that we would see this pattern if something wasn’t going on, i.e. if there wasn’t a real change
14Plain Language Interpretation? There is evidence of improvement – the chance we would see a “shift” like this in data if there wasn’t a real change in what we were doing is less than 5%
15Two few or too many runs. - 1. bring out the table 2 Two few or too many runs?- 1. bring out the table 2. how many points do we have (not on median?) 3. how many runs do we have (cross median +1) 4. what is the upper and lower limit?
16Two few or too many runs. -. new slide 1. bring out the table 2 Two few or too many runs?- **new slide 1. bring out the table 2. how many points do we have how many runs do we have (cross median +1) what is the upper and lower limit?
17Two few runs? Plain language interpretation There is evidence of improvement – our data only crosses the median line twice – three runs. If it was just random variation, we would expect to see more up and down.
18What if we had too many runs? Plain language interpretation There is evidence of a non-random pattern. There is a pattern to the way the data rises and falls above and below the median. Something systematically different. Should investigate and maybe plot on separate run charts.
21Understanding Variation Walter Shewhart(1891 – 1967)W. Edwards Deming( )The Pioneers of Understanding Variation
22Understanding Variation: Intended and Unintended Variation Intended variation is an important part of effective, patient-centered health care. Unintended variation is due to changes introduced into healthcare process that are not purposeful, planned or guided.Walter Shewhart focused his work on this unintended variation. He found that reducing unintended variation in a process usually resulted in improved outcomes and lower costs.(Berwick 1991)Health Care Data Guide, p. 107
23Shewhart’s Theory of Variation Common Causes—those causes inherent in the system over time, affect everyone working in the system, and affect all outcomes of the systemCommon cause of variationChance causeStable processProcess in statistical controlSpecial Causes—those causes not part of the system all the time or do not affect everyone, but arise because of specific circumstancesSpecial cause of variationAssignable causeUnstable processProcess not in statistical controlCould insert “a” gameHealth Care Data Guide, p. 108
24Health Care Data Guide, p. 113 Shewhart ChartsThe Shewhart chart is a statistical tool used to distinguish between variation in a measure due to common causes and variation due to special causes(Most common name is a control chart, more descriptive would be learning charts or system performance charts)Health Care Data Guide, p. 113
25Control Charts – what features are different than a run chart?
26Control Charts/Shewhart Charts upper and lower control limits to detect special cause variationExtend limits to predict future performanceNot necessarily ordered by timeadvanced application of SPC – is there something different between systems26
27Health Care Data Guide, p. 114 Example of Shewhart Chartfor Unequal Subgroup SizeHealth Care Data Guide, p. 114
29Adapted from Health Care Data Guide, p. 151 & QI Charts Software Some things such as UTIs can be U chart… you can get two UTIs in one case. Note for KimberlyAdapted from Health Care Data Guide, p. 151 & QI Charts Software
41Case Study #1bPercent of cases with urinary tract infection
42Case Study #1c Percent of cases with urinary tract infection
43Case Study #1dPercent of cases with urinary tract infection
44Case Study #1ePercent of cases with urinary tract infection
45Case Study #1fPercent of cases with urinary tract infection
46Case Study #2aPercent of patients with Death or Serious Morbidity who are >= 65 years of age
47Case Study #2bPercent of patients with Death or Serious Morbidity who are >= 65 years of age
48Case Study #2cPercent of patients with Death or Serious Morbidity who are >= 65 years of age
49Case Study #2dPercent of patients with Death or Serious Morbidity who are >= 65 years of age
50ReferencesBCPSQC Measurement ReportLangley GJ, Moen R, Nolan KM, Nolan TW, Norman CL, Provost LP (2009) The Improvement Guide (2nd ed).Provost L, Murray S (2011) The Health Care Data Guide.Berwick, Donald M, Controlling Variation in Health Care: A Consultation with Walter Shewhart, Medical Care, December, 1991, Vol. 29, No 12, page****CHELSEA, can you add Lloyd’s run chart article reference from R2?Associates in Process Improvement website