Geographic mobility is highly prevalent in Sub-Saharan Africa and challenges HIV prevention and treatment goals: it can break bonds between individuals and care systems, link geographically separate epidemics, and intensify transmission. However, research on links between mobility and HIV has yielded discordant findings, due in part to suboptimal measures that do not capture the full dimensions of mobility, including its gender-specific forms. We therefore sought to characterize relationship between HIV prevalence and mobility in an ongoing HIV test and treat study (SEARCH, NCT01864603).
We examined associations between measures of mobility and baseline HIV prevalence among 134,752 census-enumerated adults aged ≥15 years in 32 rural communities in Uganda and Kenya participating in the ongoing Sustainable East Africa Research in Community Health (SEARCH) trial. Multilevel logistic regression analyses, containing random intercepts for community, were conducted with Stata 13 stratified by gender.
Controlling for region, age, education level, marital status and household wealth index, HIV prevalence was significantly associated with number of nights spent in main household of residence in past month and migrated in past year in both men and women. Relative to those with no past year migration, odds of infection were 35% higher in migrant men (OR 1.36 [95%CI 1.17,1.59]) and women (OR 1.35 [95%CI 1.12,1.63]). Relative to those who spent few or no nights in household in past month, men who spent every night had lower odds of HIV infection (OR 0.79 [95% CI 0.72, 0.88), as did women who spent most nights (OR 0.88 [95%CI 0.78,0.99) or every night (OR 0.79 [95%CI 0.71,0.89]). In addition, the number of months living outside household in past year was associated with HIV prevalence in men only. The adjusted odds of infection were 20% higher in men who had spent 6 or more months away from the household in past year, relative to men who were more residentially stable (OR 1.21 [95%CI 1.04,1.4]).
Gender-specific patterns of mobility were associated with HIV prevalence, controlling for important confounders. Absence from households in past year was a stronger predictor in men, but residence change (migrated in past year) had similar effects in men and women. Even very recent mobility (nights away in past month) was associated with higher HIV prevalence. Causal pathways may be complex and bidirectional, and require further investigation with longitudinal data.