Evergreen A
Accounting for Informative Sampling in Estimation of Associations Between Sexually Transmitted Infections and Hormonal Contraceptive Methods (306523)
Jennifer E. Balkus, Fred Hutchinson Cancer Research Center/University of WashingtonElizabeth R Brown, Fred Hutchinson Cancer Research Center/ University of Washington
Petra Buzkova, University of Washington
*Anu Mishra, University of Washington
Keywords: informative sampling, recurrent events, time-varying covariates, sexually-transmitted infection, Andersen-Gill model
Many studies implement routine screening for diseases that are not the primary outcome. The resulting information can be useful for identifying potential risk factors for these diseases. However, the complex nature of longitudinal observational data can lead to challenges when estimating of associations of the hypothesized risk factors and the outcome of interest. In particular, if screening is not random, estimates of association can suffer from informative sampling bias. Time-varying predictors can further contribute to difficulty in generating unbiased estimates. We propose a method for estimation of hazard ratios when the exposure of interest is time-varying and the data are subject to informative sampling using a two-step procedure to jointly model the non-random screening process and the outcome. Associations between hormonal contraception and risk of sexually transmitted infections among women participating in a HIV-1 prevention trial are estimated using the proposed method and compared to estimates using conventional approaches that do not account for potential informative sampling bias.