Association of life-course events with predictors of HIV risk behaviour and HIV acquisition in Manicaland, Zimbabwe: a longitudinal analysis of a population cohort

Abstract

Background: An individual's HIV risk, and consequently their HIV prevention needs, change over time. In this study we aimed to quantify these changes, examine which life-course events were associated with them, and investigate the extent to which those life-course events were associated with HIV acquisition.

Methods: We used longitudinal data from eight rounds of a general population cohort in Manicaland province, eastern Zimbabwe, on sociodemographic and HIV risk behaviours, as well as HIV serostatus from the first seven rounds. We first visualised how HIV risk behaviours, comprised of having multiple, concurrent, non-regular, or transactional partners, condom non-use, drug use, and visiting bars, changed for individuals over time using Sankey diagrams. We then examined whether logistic regression models incorporating life-course events-namely, changes in marital or employment status, in-migration, or birth of a child-were more strongly associated with changes in HIV risk behaviour than models using only sociodemographic variables. Finally, we compared how well sociodemographic, HIV risk behaviour, and life-course events were associated with the person's risk of HIV acquisition as follows: we used logistic regression to identify which states (divided into sociodemographic, HIV risk behaviour, and life-course events) were most strongly associated with risk of HIV acquisition; based on this we use three models (corresponding to the three divisions) to identify the top 20% of individuals predicted to be at risk of acquiring HIV by each model, and computed what proportion of the actual HIV infection events occurred in that group.

Findings: Between 1998 and 2021, 21 213 individuals were interviewed at least twice, contributing a total of 34 212 participant observations. In this setting, individuals had periods of HIV risk lasting less than 3 years; only 12·3% (102 of 831) of those reporting transactional sex had also reported this in the previous round. We found that life-course events such as changes in marital status, employment status, and in-migrant status were associated with these changes in HIV risk behaviour. Using life-course events, particularly ones related to changes in marital status, 23% and 30% more HIV acquisitions were identified than using HIV risk behaviours or sociodemographic information, respectively.

Interpretation: HIV risk changes dynamically in this population, and life-course events could be a powerful way to understand changes in HIV risk behaviour and risk of HIV acquisition.

Funding: Bill and Melinda Gates Foundation, UK Medical Research Council, and Department for International Development.