Random Survival Forest With Multiple Imputation Analysis for Case‐Cohort and Generalized Case‐Cohort Studies

Case-cohort and generalized case-cohort study designs are widely used in epidemiological studies as a cost-effective alternative to much larger cohort studies while achieving similar efficiency. Current research mostly focuses on the estimation and inference of semiparametric survival models, with limited attention directed towards nonparametric survival predictions within these study designs. In this paper, we propose to use the random survival forest with multiple imputation by chained equation (RSF-MICE) and the random survival forest with substantive model compatible imputation (RSF-SMCI) for survival prediction in case-cohort and generalized case-cohort studies. Simulation studies demonstrate superior finite-sample performance for both proposed methods in various situations. We also apply our new approach to predict the incident diabetes in an analysis of a generalized case-cohort study conducted as a part of the Atherosclerosis Risk in Communities Study.