Multilevel modeling of HIV positive status disclosure to the sexual partner among persons enrolled on HIV care in greater Rakai

Abstract

Introduction: As an HIV prevention strategy, HIV sero status “disclosure” (HSSD) is defined as the process of revealing HIV positive status to sexual partner(s), relatives, or others in their social circle. Literature on HSSD is majorly analysed at individual level, yet respondents are usually nested within higher levels such as households, or community that may account for the variability that remains unexplained at the individual level. Objective: To determine the multilevel variability in HIV positive status disclosure to the sexual partner accounted for at individual and community levels, and identify the individual and community level factors associated with HIV positive status disclosure among persons aged 15-49 enrolled in HIV care in greater Rakai, between 2016 and 2018 to inform potential interventions at such a higher level. Methodology: Data from the Rakai Community Cohort Study Round 18 was used to analyse the effect of individual and community factors on HIV positive status disclosure. Predictor variables with the p<0.05 in the multivariable random intercept logistic model were considered statistically significantly associated with HIV positive status disclosure. The ICC was used to assess the effect size for the random effects. The ROC curves, AIC and BIC of the successive models were compared for model fitness. Data were analysed using STATA version 15 and Ms Excel 2016. Results: The prevalence of HIV positive status disclosure to the sexual partner was 87.4%. There was explicitly very small variation in the outcome across communities. Level of education, gender, awareness of partner’s HIV sero status, ever received HIV test results with partner and community duration of stay were statistically significant variables associated with HIV positive status disclosure. The AUC for the combined model denoted better predictive ability (0.892). Conclusion: There was no evidence of variability of HIV positive status disclosure to the sexual partner across communities. There was an interaction between awareness of partner’s status and sex of the respondent. Recommendation: Datasets with cluster size less than the minimum recommended cluster size may not be appropriate for multilevel. Community level variables should be aggregated from the entire study population in the community and not from just the target population