CHAPTER 9 2Nd Edition, Jones & Bartlett Learning, 2019
Total Page：16
File Type：pdf, Size：1020Kb
Source:Lloyd. R. Quality Health Care: A Guide to Developing and Using Indicators CHAPTER 9 2nd Edition, Jones & Bartlett Learning, 2019. Understanding Variation with Shewhart Charts e details related to these dierences are dis- ▸ Run Charts versus cussed in the remaining sections of this chapter. Shewhart Charts For many teams just beginning their quality mea- ▸ What Is a Shewhart surement journey (QMJ) the run chart provides an excellent starting point. It is easy to construct Chart? with paper and pencil, it does not require a so- Like run charts, Shewhart charts are graphic ware package in order to make one, and it can displays of process variation as it lays itself out be used with any type of data (i.e., time, money, over time. FIGURE 91 shows the basic elements of counts of errors, percentages, rates, scores, or days a Shewhart chart and one of the tests to identify between adverse events). Also, the four run chart a special cause (i.e., a data point exceeded the rules are easy to understand and apply. So, why upper control limit [UCL], signaling too much would I want to use a Shewhart chart instead of a variation in the data, which, by the way, you run chart?1 ere are basically three reasons why should recognize as an astronomical data point Shewhart charts are preferable over run charts: on the run chart). A run chart and a Shewhart 1. Shewhart charts are more sensitive chart look similar in that the indicator of interest than run charts. and its values are plotted on the vertical or y 2. Shewhart charts have the added axis and the chronological order of the data are feature of control limits and zones, organized by what are called subgroups (e.g., which run charts do not have. by individual patients, by day, week, or month) 3. Shewhart charts allow us to more along the horizontal or x axis. e data points accurately predict process behavior, are then connected by a line and the mean of the future performance, and process data points is then plotted as the centerline (CL) capability than do run charts. on the Shewhart chart. e presence of control © Michal Steovic/Shutterstock 211 9781284023077_CH09_211_258.indd 211 31/07/17 5:59 PM 212 Chapter 9 Understanding Variation with Shewhart Charts Signal of a Upper Control special Limit 60.0 50.0 UCL=46.910 40.0 Data are plotted in time order 30.0 CL=23.381 20.0 Centerline (the mean) 10.0 Number of Patient Complaints Number of Patient 0.0 LCL=0.148 Lower Control -10.0 Limit 12345678910 11 12 13 14 15 16 17 18 19 20 21 Week The unit of time is plotted along the horizontal axis FIGURE 91 Elements of a control limits on Shewhart charts are major points that data point actual values. By using the mean we separate it from a run chart. are ensuring that the absolute value and the Shewhart charts are more sensitive than distance of each data point from the CL will be run charts because the run chart cannot detect considered in determining the variation in the special causes that result from point-to-point indicator and if special cause variation exists. variation. is is because the CL on the run Another reason why Shewhart charts are chart is the median (i.e., the 50th percentile). more sensitive than a run chart is that Shewhart e run chart basically allows you to classify charts have the added feature of control limits, the data points as being only above or below which run charts do not have. e control the median. e actual distance a data point limits are properly referred to as the UCL and is from the CL is not an issue on a run chart. the lower control limit (LCL). ey are also erefore, if one data point is 2 units above referred to as sigma limits. You will probably the median and another point is 22 units hear someone refer to control limits, however, above the median, they will both be treated as condence intervals, condence limits, or the same because they are both on the same even standard deviation (SD) limits, which they side of the median. e logic for this decision are not (Blalock, 1960; Carey, 2003; Daniel & is related to the denition of the median and Terrell, 1989; Provost & Murray, 2011). of a run (i.e., one or more data points on the e UCL and LCL basically dene the same side of the median). If these same two boundaries of process variation around the data points (i.e., 2 and 22) were placed on a mean. e developer of the chart does not set or Shewhart chart, however, you would notice a dene the UCL and LCL. ese are determined discernable dierence because the CL on the by mathematical formulae and the width of these control chart is the mean or average of all the limits is dependent on the inherent variation 9781284023077_CH09_211_258.indd 212 31/07/17 5:59 PM What Is a Shewhart Chart? 213 that lives within the data. e only thing the UCL is 47 minutes, the lower control limit is developer of the chart can place on the Shewhart 23 minutes and there are no special causes chart is a target or goal and annotations as to detected. is means that the process is a stable when improvements were introduced. and predictable. erefore, if we do nothing e control limits enable the Shewhart to change how this process works we can charts to have increased precision over the run predict that patients will wait on the average chart. A run chart will miss certain nonrandom 35 minutes with the possibility that the wait patterns that would be detected on a Shewhart time could go up as high as 47 minutes or as chart as special causes. According to Perla, low as 23 minutes. In light of the target of Provost, and Murray (2011, p. 47), “e three having all patients seen by their doctor within probability-based (run chart) rules are used to 20 minutes or less, however, you can see that objectively analyze a run chart for evidence of we have our work cut out for us!” non-random patterns in the data based on an is scenario provides a summary of how α error of p < 0.05.” is means that run charts process capability for the wait time in a clinic could miss a nonrandom pattern in the data can be based on the parameters calculated for a approximately 5% of the time. Shewhart chart Shewhart chart (i.e., the UCL, LCL, and mean). rules, on the other hand, will not miss detecting Classically, process capability is dened as, “e a special cause. is is why it is recommended calculated inherent variability of a characteristic that the terms special and common cause as well (indicator) of a product or service. It represents as stable or unstable should be reserved for use the best performance of the process over a period only with Shewhart charts and that the terms of stable operation” (ASQ, 2005, p. 78). Process random and nonrandom patterns be applied capability is essentially aimed at determining to run charts. whether under current operating conditions the Shewhart charts also allow us to more process can meet the predetermined specications accurately predict process behavior and future or achieve the target or goal we have established performance than do the run charts. On a (Blank, 1998; Carey, 2003; Kume, 1985; Provost & run chart, if the variation is random the best Murray, 2011; Western Electric, 1985; Wheeler & prediction of the future performance of an Chambers, 1992). indicator is the median value. For example, Besides a verbal summary of the Shewhart if a team is trying to improve the wait time chart parameters using the UCL, LCL, and mean to see a doctor and have plotted the data on as described previously, process capability can a run chart the median is the best estimate of also be dened statistically by “a single number future performance. Let’s say that the median assessment of the ability of the process to meet wait time is 27 minutes. If you were present- specication limits on the quality characteris- ing this data to a team or a committee all you tic(s) of interest (ASQ, 2005, p. 78). When you could say would be, “Ladies and gentlemen, the move to the statistical indices that capture pro- median wait time is 27 minutes. e process cess capability it is necessary to have an upper reects only random variation. erefore, if specication limit (USL) and a lower speci- we do nothing to change the current process cation limit (LSL), which are then compared we can expect to have patients wait about 27 to the performance of the process as dened minutes to see the doctor.” On a Shewhart by the UCL, LCL, and the mean.2 Although chart, however, because we have the UCL, these indices have not been used extensively LCL, and the mean as the CL, we have more in healthcare settings I believe that they have precision. In this case, when you present the great utility. We have many physiological tests data to the team or a committee you would be that have upper and lower preferred levels of able to say, “Ladies and gentlemen, the average performance (i.e., specication limits). ese wait time to see a doctor is 35 minutes, the include such indicators as temperature, blood 9781284023077_CH09_211_258.indd 213 31/07/17 5:59 PM 214 Chapter 9 Understanding Variation with Shewhart Charts pressure, hematocrits, neutrophils, white and the interpretation of the chart and what can be red blood cell counts, platelets, and clotting learned from it must come from the dialogue factors.3 that emerges when people with subject matter ere are many useful books and articles on knowledge interpret the chart.