Abstract

Background: Human immunodeficiency virus (HIV) remains a public health priority in Latin America. While the
burden of HIV is historically concentrated in urban areas and high-risk groups, subnational estimates that cover
multiple countries and years are missing. This paucity is partially due to incomplete vital registration (VR) systems
and statistical challenges related to estimating mortality rates in areas with low numbers of HIV deaths. In this
analysis, we address this gap and provide novel estimates of the HIV mortality rate and the number of HIV deaths
by age group, sex, and municipality in Brazil, Colombia, Costa Rica, Ecuador, Guatemala, and Mexico.
Methods: We performed an ecological study using VR data ranging from 2000 to 2017, dependent on individual
country data availability. We modeled HIV mortality using a Bayesian spatially explicit mixed-effects regression
model that incorporates prior information on VR completeness. We calibrated our results to the Global Burden of
Disease Study 2017.
Results: All countries displayed over a 40-fold difference in HIV mortality between municipalities with the highest
and lowest age-standardized HIV mortality rate in the last year of study for men, and over a 20-fold difference for
women. Despite decreases in national HIV mortality in all countries—apart from Ecuador—across the period of
study, we found broad variation in relative changes in HIV mortality at the municipality level and increasing relative
inequality over time in all countries. In all six countries included in this analysis, 50% or more HIV deaths were
concentrated in fewer than 10% of municipalities in the latest year of study. In addition, national age patterns
reflected shifts in mortality to older age groups—the median age group among decedents ranged from 30 to 45
years of age at the municipality level in Brazil, Colombia, and Mexico in 2017.
Conclusions: Our subnational estimates of HIV mortality revealed significant spatial variation and diverging local
trends in HIV mortality over time and by age. This analysis provides a framework for incorporating data and
uncertainty from incomplete VR systems and can help guide more geographically precise public health intervention
to support HIV-related care and reduce HIV-related deaths.