Bayesian spatial semi-parametric modeling of HIV variation in Kenya

Area of Science:

• Epidemiology and Public Health
• Spatial Statistics
• Biostatistics

Background:

• Traditional spatial statistics models often underutilize geographical data and assume linear covariate effects.
• Previous research suggests a nonlinear relationship between Human Immunodeficiency Virus (HIV) infection and age.

Purpose of the Study:

• To develop and apply a Bayesian semi-parametric regression model for analyzing HIV prevalence data.
• To investigate the influence of various factors, including spatial effects and nonlinear age associations, on HIV infection risk in Kenyan men.

Main Methods:

• A Bayesian semi-parametric regression model was developed using Markov Chain Monte Carlo (McMC) for estimation and inference.
• Penalized regression splines were employed to model the nonlinear association between HIV infection and age.
• Structured and unstructured spatial effects were incorporated to account for geographical variations in HIV prevalence.

Main Results:

• Circumcision was found to reduce the risk of HIV infection.
• Men in urban areas and those with a history of Sexually Transmitted Infections (STIs) showed higher HIV prevalence.
• Higher education levels were associated with a lower risk of HIV infection, and a nonlinear age-HIV risk relationship was confirmed, peaking around age 40.

Conclusions:

• The Bayesian semi-parametric model offers a practical and flexible approach for analyzing complex epidemiological data, effectively capturing nonlinear effects and spatial variations.
• Significant spatial variations in HIV prevalence across Kenyan counties were identified.
• Key modifiable risk factors and protective measures for HIV infection in men were elucidated, providing valuable insights for public health interventions.