Modeling HIV and Aids Transmission Dynamics in Kenya Using an Adaptive Spatial Hierarchical Bayesian Approach

HIV and AIDS remain a significant public health challenge in Kenya, which has the third-largest HIV epidemic globally with 1.3 million people living with HIV. Despite progress in prevention and treatment, the complex spatiotemporal dynamics of HIV transmission continue to challenge intervention strategies, necessitating advanced modeling approaches that capture spatial heterogeneity and temporal evolution. This study developed and validated an adaptive spatial hierarchical Bayesian SEIR model for HIV transmission dynamics in Kenya. The research addressed four objectives, that is, developed an adaptive spatial hierarchical Bayesian SEIR model integrating transmission patterns and geographic distribution, established the model’s theoretical properties including equilibria, stability, and convergence, validated the model through empirical testing using historical HIV data; and applied the model for policy scenario analysis and intervention optimization. The study employed a quantitative design integrating spatial-temporal SEIR modeling, hierarchical Bayesian inference, and spatial statistics. The enhanced SEIR model incorporated HIV-specific compartments, adaptive spatial weights, and time-varying parameters to capture Kenya’s heterogeneous transmission landscape. Analysis utilized comprehensive HIV surveillance data from 15 strategically selected Kenyan counties (2015-2024), including data from Kenya AIDS Indicator Survey, National AIDS Control Council, Kenya Health Information System, and Kenya National Bureau of Statistics. The dataset encompassed 1,620 county-month observations with variables spanning demographic, epidemiological, and intervention domains. Computational implementation used R Statistical Software (v4.2.0) for data preprocessing, Stan for Bayesian inference via Hamiltonian Monte Carlo, INLA for spatial component approximation, and Python for machine learning integration. The model demonstrated superior predictive accuracy with 48% improvement in prediction errors (RMSE = 0.035) compared to non-spatial approaches and 22% improvement over static spatial models. Key findings included: identification of substantial spatial heterogeneity in basic reproduction numbers (R₀ ranging from 0.3 in Wajir to 3.2 in Homabay); successful capture of both spatial clustering (Moran’s I residuals = 0.03, p = 0.45) and temporal dynamics (𝑅² ≥ 0.87); reliable uncertainty quantification (coverage probabilities ≥ 85%); and optimal intervention strategies showing combined approaches could avert 2,341 new infections at cost-effectiveness of $920 per infection averted. The validated model provides Kenya’s HIV control program with enhanced capabilities for evidence-based resource allocation, early warning systems for emerging transmission hotspots, policy scenario evaluation, and adaptive program management. The framework offers valuable insights for spatial epidemiology researchers and public health policymakers in similar resource-limited settings, contributing to more effective HIV prevention and control strategies globally.