Evaluating Failure Outcomes with Applications to Transplant Facility Performance.

Abstract

We develop several methods to evaluate mortality experience of medical facilities with applications to transplant facilities' post-transplant mortality and pre-transplant waitlist mortality. We aim to compare the center-specific outcomes with the standard practice while providing timely feedback to the centers.

In Chapter II, we introduce a risk-adjusted O-E (Observed-Expected) Cumulative Sum (CUSUM) chart along with monitoring bands as decision criterion, to monitor the post-transplant mortality in transplant programs. This can be used in place of a traditional but complicated V-mask and yields a more simply interpreted chart. The resulting plot provides bounds that allow for simultaneous monitoring of failure time outcomes with signals for `worse than expected' or `better than expected'. The plots are easily interpreted in that their slopes provide graphical estimates of relative risks and direct information on additional failures needed to trigger a signal.

In Chapter II, we discuss the construction of a weighted CUSUM to evaluate pre-transplant waitlist mortality of facilities where transplantation can be considered as dependent censoring. Patients are evaluated based on their current medical condition as reflected in a time dependent variable the Model for End-Stage Liver Disease score, which is used to prioritize to receive liver transplants. We assume a ‘standard’ transplant practice through a transplant model, utilizing Inverse Probability Censoring Weights (IPCW) to construct a weighted CUSUM. We evaluate the properties of a weighted zero-mean process as the basis of the proposed weighted CUSUM. A rule of setting control limits is discussed. Case study on regional transplant waitlist mortality is carried out to demonstrate the usage of the proposed weighted CUSUM.

In Chapter III, we provide an explicit road map for using a Cox dependent censoring model in the IPCW approach, complete with details of implementation. In addition, we evaluate an alternative parametric IPCW approach to gain efficiency. Simulation studies and case study on the national liver transplant waitlist mortality are conducted to demonstrate the similarity in estimates between Cox IPCW and PWE IPCW, and the computational savings by the PWE IPCW as compared to the Cox IPCW.

In the last chapter, we discuss the future directions of our work.