Statistical Process Control in Quality Assurance of Inpatient Care

Statistical Process Control (SPC) in hospital benchmarking using control charts is a common instrument for monitoring clinical performance and early detection of quality deficits. The external quality assurance program (EQA) of German hospitals does not yet employ SPC. Previous work has failed to come up with suggestions for efficient application of SPC. There is also a lack of focus on the importance of preventing false positive signals.
In this contribution we study control limits for defined false signal probability and their dependence on specific features such as hospital volume, risk score and patient mix. We also determine the detection quality of specific control switches. We conduct simulation studies in order to investigate optimal designs for crude and risk-adjusted performance indicators of the log-likelihood CUSUM chart of Steiner et al. (Biostatistics 1.4 (2000), pp. 441-52). Examples are taken from the EQA in Bavaria, Germany.
Focusing on signal probability instead of average run length allows control of the false signal probability and performance evaluation of control charts. Thus it was possible to construct CUSUM charts for different hospital volumes and failure probabilities. We gained better understanding of the influence of control switches in constructing CUSUM charts. We also compare our results to run-length based control strategies.
The presented results are useful for regulatory decision making and help to implement CUSUM charts within EQA. We expect application of CUSUM control charts to significantly improve early detection of quality deficits with appropriate adjustment for different case mix and hospital volume.