Logistic regression models for odds of condom use at last sexa.

aAdjusted for partnership clustering using robust standard errors.bThe following variables did not contribute to the multivariable model for women: Knowledge of partner having sex with others in last 6 months and alcohol use at last sex. Relationship quality was not significant when added to the model.cThe following variables did not contribute to the multivariable model for men: partnership status and knowledge of partner's HIV status. Having recently argued or ever used physical violence towards partner were not significant in univariate models.dWald test. In the case of equitable gender norms by age, the Wald p-value is for the interaction term added to a model with gender norms and age as main effects.eSexually active / not sexually active is participant-defined.Logistic regression models for odds of condom use at last sexa.