Statistical process monitoring to improve quality assurance of inpatient care

BMC Health Serv Res. 2020 Jan 7;20(1):21. doi: 10.1186/s12913-019-4866-7.

Abstract

Background: Statistical Process Monitoring (SPM) is not typically used in traditional quality assurance of inpatient care. While SPM allows a rapid detection of performance deficits, SPM results strongly depend on characteristics of the evaluated process. When using SPM to monitor inpatient care, in particular the hospital risk profile, hospital volume and properties of each monitored performance indicator (e.g. baseline failure probability) influence the results and must be taken into account to ensure a fair process evaluation. Here we study the use of CUSUM charts constructed for a predefined false alarm probability within a single process, i.e. a given hospital and performance indicator. We furthermore assess different monitoring schemes based on the resulting CUSUM chart and their dependence on the process characteristics.

Methods: We conduct simulation studies in order to investigate alarm characteristics of the Bernoulli log-likelihood CUSUM chart for crude and risk-adjusted performance indicators, and illustrate CUSUM charts on performance data from the external quality assurance of hospitals in Bavaria, Germany.

Results: Simulating CUSUM control limits for a false alarm probability allows to control the number of false alarms across different conditions and monitoring schemes. We gained better understanding of the effect of different factors on the alarm rates of CUSUM charts. We propose using simulations to assess the performance of implemented CUSUM charts.

Conclusions: The presented results and example demonstrate the application of CUSUM charts for fair performance evaluation of inpatient care. We propose the simulation of CUSUM control limits while taking into account hospital and process characteristics.

Keywords: CUSUM; Hospital performance; Quality assurance; Statistical process control.

MeSH terms

  • Germany
  • Hospitalization*
  • Humans
  • Models, Statistical
  • Quality Assurance, Health Care / methods*