Modeling Time in Medical Education Research: The Potential of New Flexible Parametric Methods of Survival Analysis

Abstract

Time – the duration of a certain process or the timing of a specified event – plays a central role in many situations in medical research. Waiting time analysis (“survival analysis”) is a field of statistics providing the tools for solving the unique problems of such studies. In particular, waiting time analysis correctly handles the typical positively skewed distributions of waiting times as well as censored observations on study subjects for whom the target event does not occur before data collection ends. For decades, non-parametric Kaplan-Meier analysis and semiparametric Cox regression despite some inherent limitations have dominated waiting time analysis in medical contexts, while parametric models, although in principle offering important theoretical advantages, were scarcely applied in practice because of lacking flexibility. Recently, however, new flexible parametric methods (Royston-Parmar models) became available offering exciting new research potential. Surprisingly, although medical education research deals with a range of typical problems suited for waiting time analysis, the methods were rarely used in the past. By re-analyzing data from a previous investigation on study dropout of medical students, this is the first study demonstrating the usefulness and practical applications of waiting time analysis with special emphasis on Royston-Parmar models in a medical education research environment.