Phase I Risk-Adjusted Control Charts for Monitoring Surgical Performance by Considering Categorical Covariates

In recent years risk-adjusted control charts have been increasingly studied for monitoring surgical outcomes by accounting for patients' health conditions prior to surgery. However, most existing research focuses on phase II monitoring, and very little work has been done on phase I control of surgical outcomes. In this paper, a general phase I risk-adjusted control chart is proposed for monitoring binary surgical outcomes based on a likelihood-ratio test derived from a change-point model. Different from the existing methods, this paper further shows that the binary surgical outcomes depend on not only the patient conditions described by the Parsonnet scores but also on other categorical operational covariates, such as different surgeons. The proposed risk-adjustment model is fitted by incorporating dummy variables to reflect different surgeon groups' performance. The analysis of a case study and simulation results show that the inclusion of surgeon groups as a categorical covariate in the risk-adjustment model can effectively reflect the inherent data heterogeneity, thus improving the estimation of the risk-adjustment model parameters. As expected, the proposed risk-adjusted control charts can achieve a better detection performance by including the surgeon covariate.