Uncovering the impact of randomness in HIV hotspot formation: A mathematical modeling study

Abstract

Background HIV hotspots, regions with higher prevalence than surrounding areas, are observed across Africa, yet their formation and persistence mechanisms remain poorly understood. We hypothesized that random fluctuations during the early stages of the HIV epidemic (1978-1982), amplified by positive feedback between HIV incidence and prevalence, play a critical role in hotspot formation and persistence. To explore this, we applied a network-based HIV transmission model, focusing on randomness in the spatial structure of the epidemic. Methods We adapted a previously validated agent-based network HIV transmission model, EMOD-HIV, to simulate HIV spread in western Kenya communities. The model includes demographics, age-structured social networks, and HIV transmission, prevention, and treatment. We simulated 250 identical communities, introducing stochastic fluctuations in network structure and case importation. Outliers were defined as communities with prevalence > 1.5x the median, and persistence as meeting these criteria for >70% of 1980-2050. We systematically varied community size (1,000-10,000), importation timing (1978-1982), and importation patterns (spread over 1, 3, or 5 years), and calculated the proportion of outliers and persistent outliers. Results HIV prevalence outliers were more common in smaller communities: in 1990, 25.3% (uncertainty interval: 22.3%-28.2%) of 1,000-person communities vs. 9.1% (uncertainty interval: 6.9%-11.4%) of 10,000-person communities. By 2050, 21.6% of 1,000-person communities were persistent outliers, compared to none in larger communities. Autocorrelation of HIV prevalence was high (Pearson's correlation coefficient 0.801 [95% CI: 0.796-0.806] for 1,000-person communities), reflecting feedback that amplified early fluctuations. Conclusions Early random fluctuations contribute to the emergence and persistence of prevalence outliers, especially in smaller communities. Recognizing the role of randomness in prevalence outlier formation in these settings is crucial for refining HIV control strategies, as traditional methods may overlook these areas. Adaptive surveillance systems can enhance detection and intervention efforts for HIV and future pandemics.