Factors predicting instances and rates of condom breakage: An exploration of categorical data methods

Abstract

In this report, I conduct a secondary analysis of the RESPECT-2 study data in order to identify individual and partner characteristics related to instances and rates of condom breakage in main partnerships. The data used in the analysis were collected on 2,090 study participants involved in 9,111 partnerships over the course of one year. I will analyze a binary response variable reflecting whether there was an instance of breakage using both random-intercept logistic regression and marginal logistic regression. I will analyze a categorical response variable reflecting breakage rate using a mixed-effects model and a generalized estimating equation model. The models were all similar and yielded length of time in the study, age, whether the participant was Asian, whether the couple had sex while impaired, the number of sex acts, number of acts with a condom, and the risk of STD posed by the partner to be the relevant predictor variables.