Multilevel Logistic Regression

The null model found that 17.5% of variance in reported barriers was attributable to unobserved community differences. The full model had the best fit but still left significant residual cluster variance. The median odds ratio of 2.22 indicated substantial between-community heterogeneity in healthcare access barriers. The full combined model was retained because it offered the most comprehensive explanation of individual and community influences. The analysis used multilevel logistic regression with random intercepts at the cluster level. The full adjusted model included individual, household, and community factors. Community variation was evaluated with intraclass correlation coefficient, median odds ratio, and proportional change in variance. The study used mixed-effects logistic regression because the survey data were clustered. Adjusted odds ratios with 95% confidence intervals were used to report fixed effects.