The Aggregate Point Rule for Identifying Shifts on P Charts and U Charts

Many valuable tools currently used in health care quality improvement (QI) have been borrowed from industry, initially from nuclear power and commercial aviation, where high reliability is essential.1–4 QI data require statistical analysis to determine if interventions have caused a system change that improved performance. A common statistical approach to accomplish that involves a concept borrowed from manufacturing: statistical process control (SPC).5–8 Data are gathered regarding specific processes and the points are plotted on an SPC chart, also known as a control chart, a process-behavior chart, or a Shewhart Chart—after its inventor, Shewhart.9 In a large majority of cases, more than 99% of observations of a stable process will fall within 3 SDs of the mean, although the exact percentage is variable for the types of charts addressed in this article, dependent upon actual process mean and sample sizes. Thus, any points outside of those limits suggest a “special cause” indicating that the process at that point in time is not stable—something has changed significantly,10 which, depending on the process, could be good or bad