Space-time migration patterns and risk of HIV acquisition in rural South Africa

OBJECTIVE To quantify the space-time dimensions of human mobility in relationship to the risk of HIV acquisition. METHODS We used data from the population cohort located in a high HIV prevalence, rural population in KwaZulu-Natal, South Africa (2000-2014). We geolocated 8006 migration events (representing 1 028 782 km traveled) for 17 743 individuals (≥15 years of age) who were HIV negative at baseline and followed up these individuals for HIV acquisition (70 395 person-years). Based on the complete geolocated residential history of every individual in this cohort, we constructed two detailed time-varying migration indices. We then used interval-censored Cox proportional hazards models to quantify the relationship between the migration indices and the risk of HIV acquisition. RESULTS In total, 17.4% of participants migrated at least once outside the rural study community during the period of observation (median migration distance = 107.1 km, interquartile range 18.9-387.5). The two migration indices were highly predictive of hazard of HIV acquisition (P < 0.01) in both men and women. Holding other factors equal, the risk of acquiring HIV infection increased by 50% for migration distances of 40 km (men) and 109 km (women). HIV acquisition risk also increased by 50% when participants spent 44% (men) and 90% (women) of their respective time outside the rural study community. CONCLUSION This in-depth analysis of a population cohort in a rural sub-Saharan African population has revealed a clear nonlinear relationship between distance migrated and HIV acquisition. Our findings show that even relatively short-distance migration events confer substantial additional risk of acquisition.