Gráficos CUSUM ajustados ao risco para monitoramento de tempos de sobrevivência com fração de cura

Abstract

In this work we study the use of techniques of Statistical Process Control (SPC) for monitoring survival times. Unlike applications in the industrial area, where the study population is considered homogeneous, the SPC in healthcare admits heterogeneity and takes into account particular characteristics of patients who, before undergoing a medical procedure, may present di erent risks of death. In this perspective, some authors propose the use of a risk-adjusted survival times CUSUM chart (RAST CUSUM) to monitor clinical outcomes in which the response is the time until the occurrence of an event and is subject to right censoring. However, the models used do not consider the possibility of cure fraction. In this study we propose to extend this approach considering a survival model with a cure fraction. To do so, we assume the log-logistic and Weibull distributions as examples. Finally, we conducted a simulation study with the Weibull distribution to obtain optimum control limits and evaluate the performance of the chart we suggest compared to the RAST CUSUM without cure fraction. As a result, we note that the RAST CUSUM chart without cure fraction is inappropriate to apply to data with cure fraction, but the RAST CUSUM chart with cure fraction seems to have a similar performance when applied to data without cure fraction.