A new perspective of survival data on clinical epidemiology: introduction of competitive risk model

Abstract

Competing risks occur frequently in the analysis of survival data that should be dealt with competing risk models. Competing risk is an event whose occurrence precludes the occurrence of the primary event of interest. Previous commonly used Kaplan-Meier method tends to overestimate the cumulative survival functions, while the traditional Cox proportional hazards model falsely evaluates the effects of covariates on the hazard related to the occurrence of the event. There are few domestic reports mentioning the concept, application and methodology of competing risk model as well as the implementation procedures or resolution of model conditions and parameters. The current work aims to explain the core concept and methodology of the competing risk model and to illustrate the process of analysis on cumulative incidence rate, using both the cause-specific hazard function model and the sub-distribution hazard function model. Software macro code in SAS 9.4 is also provided to assist clinical researchers to further understand the application of the model so to properly analyze the survival data.