Monitoring patient survival times in surgical systems using a risk-adjusted AFT regression chart

Abstract

Monitoring surgical processes has gained prominence by accounting for patients’ health condition prior to surgery in recent years. However, most of previous researchers have focused on Phase-II monitoring based on binary outcomes, while very little attention has been paid to Phase-I monitoring procedures, especially when the outcomes are continuous. In this paper, a general Phase-I accelerated failure time-based risk-adjusted control chart is proposed to monitor continuous surgical outcomes based on a likelihood-ratio test derived from a change-point model. Different from the existing models, this paper shows that continuous outcomes depend not only on the patient conditions described by the Parsonnet score, but also is dependent on other categorical operational covariates such as surgeons. The proposed risk-adjustment model is fitted by incorporating dummy variables to reflect different surgeon groups as a categorical covariate. We show that the categorical covariate can effectively reflect the inherent data heterogeneity, thus can improve the estimation of the parameters of the risk-adjustment model and hence can enhance the detection power of the proposed risk-adjusted control chart.