Issue link: https://www.e-digitaleditions.com/i/341135
32 July 2014 Tablets & Capsules called "run rules." Run rules—commonly the Western Electric or Nelson rules—help you distinguish "common- cause" variation from "special-cause" variation. The differ- ence is critical because common-cause variation deter- mines your process capability: whether it can do what it's designed to do. While there is some overlap in what each of these charts expresses, they all tell a different part of the whole story. For example, the data in Figure 10 about tablet weights may look pretty good to you, depending on your process. It shows that the mean tablet weight is 135.8 milligrams; the confidence level is 95 percent; and the standard deviation is 2.37. Now look at the same data in a run chart that shows the tablet weights over time (Figure 11), and the story changes: Weights have been creeping up. What else might be happening? Are the dosages getting more vari- able? Have the densities changed? What are the potential effects on safety, identity, strength, purity, and quality? To divine the cause of unwanted variation, begin with the presumption that willful ignorance or neglect is rare. That means taking a sober approach to systems design, one that recognizes that most errors at the interface of people and equipment likely stem from a lack of under- standing or clarity. Sometimes it's a failure to measure what's important or even obvious. At Stanford University's hospitals and clinics, for example, wait times in the emergency room decreased once managers started measuring and reporting them [8]. Such visible accountability mirrors what is known as the Hawthorne effect: Performance improves solely because the subject knows that he or she is being studied [9]. In addition, reporting wait times should be a sustainable improvement step because what's measured (incoming Figure 9 Mean of weight Control chart 23 22 21 20 19 Sample UCL = 22.01 Avg – 20.40 LCL = 18.78 1 2 3 4 5 6 7 8 Figure 10 Histogram of tablet weights Mean Median 132 134 136 138 140 134.5 135.0 135.5 136.0 136.5 137.0 95% confidence intervals Anderson-Darling normality test A-squared 0.55 P-value 0.141 Mean 135.81 Standard deviation 2.37 Variance 5.63 Skewness 0.6222829 Kurtosis -0.027441 N 33 Minimum 132.00 1st quartile 134.00 Median 135.00 3rd quartile 137.00 Maximum 141.50 95% confidence interval for mean 134.97 136.65 95% confidence interval for median 134.45 136.90 95% confidence interval for standard deviation 1.91 3.14 f-Locwinart_24-35_Masters 7/2/14 1:29 PM Page 32