Optimal Detection of Health Care Adverse Event Rate Changes

Abstract

The overall purpose of this research area is to develop and apply quality engineering methods that can help accelerate healthcare improvement. Within this context, current projects are focusing on two general areas: (1) improved statistical quality control methods that detect process changes faster for risk-adjusted, rare event, care compliance, and under-reported data; and (2) the use and extension of industrial type experimental design methods to help identify how to improve processes faster. Shewhart p and g statistical control charts often are used to detect special cause variation in healthcare adverse event (AE) rates and other performance measures. When change detection run lengths for either chart are still too slow, optimizing sample sizes or the use of exponentially weighted moving average (EWMA) version can further improve performance. We investigate the relative performance (average run lengths) of these charts across a range of different scenarios. Shewhart x ? and s chart is widely used to monitor a continuous data. However, it is often encountered that no mean and standard deviation are given or the actual observation values are not known. Instead, the continuous observations are classified into groups which is the case in some of Veteran’s Affairs (VA) data. We propose a method for this grouped data, namely quantile chart, which includes different approaches to estimate the parameters and compare the performances of these approaches. We also present a baseline template which is created for VA to be used as a tool for monitoring specialty care data.