Statistical Analysis
We first examined the distributions of the socioeconomic and demographic variables in the two groups of women and the percentage of missing values for each variable. 37 We used cross-tabular univariate analyses to estimate crude odds ratios for each categorical and ordinal variable. 37 We used cross-tabular stratified analyses to identify confounding and interactions among the variables. 9 We used multiple logistic-regression analyses to estimate adjusted odds ratios. 9,38 We then conducted sensitivity analyses to assess the variability of the estimates of the odds ratios depending on the inclusion and exclusion of different variables and product terms in the models. 9,38 We used odds ratios to estimate relative risks. 9,38 We used residual analyses to assess the fit of the multiple logistic-regression models. 38 The selected model included terms representing all the variables for the women’s characteristics and those of their partners. All P values are two-sided.