Applying Competing Risks Analysis Methods to a Population-Based Study

• Activity Number: 190
• Type: Contributed
• Date/Time: Monday, August 1, 2011 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Hyun Ja Lim*+ and Xu Zhang and Roland Dyck and Nathaniel Osgood
• Companies: University of Saskatchewan and Georgia State University and University of Saskatchewan and University of Saskatchewan
• Keywords: Competing Risks Analysis ; cause-specific hazards model ; subdistribution hazards model ; Lunn-McNeil model ; Diabetes Study

Abstract

In medical research, each patient can experience one of several different types of events over the follow-up period. Survival times are subject to competing risks if the occurrence of one event prevents the other event types from occurring. The cause-specific hazards and the cumulative incidence functions (CIF) are the most appropriate approaches to analyse the competing risks data. In this study, we compare three models: the cause-specific hazards model, the subdistribution hazards model, and the Lunn-McNeil unstratified model. A population based diabetes study is used to demonstrate the three methods. Our analyses show that the three methods yield different results with regards to the effects of the covariates. The CIF of the cause-specific hazards model reveals a higher CIF curve than the subdistribution hazards model while the CIF of the unstratified Lunn-McNeil model is the lowest. For a dominant risk, one can either use the cause-specific hazards model or the subdistribution hazards model. For a minor risk, the cause-specific hazards model is more appropriate. The Lunn-McNeil method is applicable only when the baseline hazards for the different risk types are proportional.