Evaluating Methods for Dealing with Missing Outcomes in Discrete-Time Event History Analysis: A Simulation Study

Background: In discrete-time event history analysis, subjects are measured once each time period until they experience the event, prematurely drop out, or when study concludes. This implies measuring status of a subject in determines whether (s)he should be subsequent periods. For that reason, intermittent missing causes problem because, unlike other repeated measurement designs, it does not make sense to simply ignore corresponding from analysis (as long as dropout is ignorable). Method: We used Monte Carlo simulation evaluate and compare various alternatives, including occurrence recall, (non-)occurrence, case deletion, single multiple imputation methods, deal with status. Moreover, we showed methods’ performance an empirical example on relapse drug use. Result: The strategies assuming (non-)occurrence recall strategy had worst because substantial parameter bias sharp decrease coverage rate. Deletion methods suffered either loss power undercoverage issues resulting biased standard error. Single recovered issue but estimate. Multiple imputations performed reasonably negligible error leading gradual power. Conclusion: On basis results real example, provide practical guidance researches terms best ways data.