Uni-Logo
You are here: Home Competing Risks
Document Actions

Competing Risks

(Full day, 18 March, 9:00 – 17:00)

Jan Beyersmann1,2 and Arthur Allignol1,2

1 Institute of Medical Biometry and Medical Informatics, University Medical Center Freiburg

2 Freiburg Centre for Data Analysis and Modelling, University of Freiburg, Germany

 

Summary:

The analysis of competing risks data is an extension of standard survival analyses. Survival analysis studies the time until some endpoint. The endpoint often combines different event types, like different causes of death or disease occurrence and death without prior disease, into one single endpoint. Competing risks models distinguish between different endpoint types. These models have applications in fields such as medicine, biology, economics, demography, social science and reliability theory.

Survival analysis is based on hazards. Hazard-based analyses are still available for competing risks. However, the interpretation becomes more challenging, as there are as many hazards as there are possible competing events in the model. These hazards are often called cause-specific hazards. E.g., the cumulative incidence function, i.e., the expected proportion of individuals with a certain competing event over the course of time, is an involved function of all cause-specific hazards. This requires customized estimation techniques and careful interpretation of results from regression modeling.

This course explains hazard-based analyses of competing risks with R. Special emphasis is placed on the interpretation of the results. Participants will analyze real data examples from studies in clinical medicine where the authors have been involved. A further topic is simulation of competing risks data based on the cause-specific hazards only. The simulation point of view is helpful for interpretation and for extending the competing risks techniques of this course to studying more complex event patterns. Competing risks are approached from a multistate perspective throughout. That is, the occurrence of a competing risk is modeled as the transition from an initial at risk state into one of the competing event states. The multistate approach disposes of the difficulties that arose from the more classical competing risks model based on hypothetical risk-specific event times.

The course is aimed at data analysts with a background in standard survival analysis, who wish to understand, analyze and interpret more complex event histories with R.

 

Prerequisites:

Experience with survival analysis (Kaplan-Meier, Cox), ideally some experience with R, laptop with R installed, including the following packages: survival, mvna, etm, cmprsk, mstate, kmi

 

Course materials:

Beyersmann, J., Allignol, A., and Schumacher, M. (2012): Competing Risks and Multistate Models with R. Springer, New York (included in course fee). Presentation and R solution files will be sent ahead.

 

Course fees:

Regular: 95 EUR

Student: 75 EUR

Personal tools