Concordance for prognostic models with competing risks

• DOI: 10.1093/biostatistics/kxt059
• Publication Date: 2014-02-04T03:11:09Z
• Authors: M. Wolbers; P. Blanche; M. T. Koller; J. C. M. Witteman; T. A. Gerds
• Keywords: Censoring (clinical trials); Concordance; Inverse probability

The concordance probability is a widely used measure to assess discrimination of prognostic models with binary and survival endpoints. We formally define the for model absolute risk an event interest in presence competing risks relate it recently proposed time-dependent area under receiver operating characteristic curve measures. For right-censored data, we investigate inverse censoring weighted (IPCW) estimates truncated index based on working distribution. demonstrate consistency asymptotic normality IPCW estimate if correctly specified derive explicit formula variance independent censoring. small sample properties estimator are assessed simulation study also against misspecification model. further illustrate methods by computing coronary heart disease (CHD) events non-CHD death.