All Times ET
Keywords: time-to-event, proportional hazards, complex surveys, data linkage
When analyzing time-to-event data, the Cox proportional hazards model is often used to compare the relative risk of the outcome (e.g., mortality) for various characteristics. One of the main assumptions of this model is that of proportional hazards; the hazard ratio of the outcome between two groups is constant over time. The proportional hazards assumption is validated in a variety of ways including the use of Kaplan-Meier curves, interactions between time and the covariate, or Schoenfeld residuals. While these approaches were initially developed for non-survey data, some survey data sources, such as the National Health Interview Survey (NHIS) and the National Health and Nutrition Examination Survey (NHANES) conducted by the National Center for Health Statistics, Centers for Disease Control and Prevention, also have a time-to-event component when linked with mortality data from the National Death Index (NDI). While some of these approaches have been developed for survey data, a recent search of PubMed® indicated that the validation of the proportional hazards assumption is rarely addressed in manuscripts analyzing time-to-event survey data via Cox proportional hazards regression. The goal of this presentation is to compare the various approaches for evaluating the assumption of proportional hazards for time-to-event survey data using public-use NHIS and NHANES data each linked with the NDI. Because the public-use linked NHIS-NDI files have discrete mortality outcomes (i.e., mortality occurred within a calendar quarter) and the public-use linked NHANES-NDI files have mortality status with person-months of follow-up, the comparison of approaches will be conducted for discrete and continuous time, respectively.