HIV risks are heterogeneous in nature even in generalized epidemics. However, data are often missing for those at highest risk of HIV, including female sex workers. Statistical models may be used to address data gaps where direct, empiric estimates do not exist.
We proposed a new size estimation method that combines multiple data sources (the Malawi Biological and Behavioral Surveillance Survey, the Priorities for Local AIDS Control Efforts study, and Malawi Demographic Household Survey). We employed factor analysis to extract information from auxiliary variables, and constructed a linear mixed effects model for predicting population size for all districts of Malawi.
On average, the predicted proportion of FSW among women of reproductive age across all districts was about 0.58%. The estimated proportions seemed reasonable in comparing with a recent study PLACE II. Compared to using a single data source, we observed increased precision and better geographic coverage.
We illustrate how size estimates from different data sources may be combined for prediction. Applying this approach to other sub-populations in Malawi and to countries where size estimate data are lacking can ultimately inform national modeling processes and estimate the distribution of risks and priorities for HIV prevention and treatment programs.

Copyright © 2020. Published by Elsevier Inc.

Author