Relative risk for HIV in India - An estimate using conditional auto-regressive models with Bayesian approach

Area of Science:

• Epidemiology
• Biostatistics
• Spatial Analysis

Background:

• Current HIV risk classification for Indian states relies on prevalence counts and clinic data.
• Existing methods lack spatial dependence analysis and statistical uncertainty measures.
• A novel model-based approach is needed to address these limitations.

Purpose of the Study:

• To develop and apply a model-based approach for classifying Indian states into HIV risk categories.
• To incorporate spatial dependence and covariate information for more accurate risk assessment.
• To identify key factors influencing HIV risk for targeted intervention strategies.

Main Methods:

• Utilized Poisson log-normal models with conditional autoregressive structures.
• Employed neighborhood-based and distance-based weight matrices for spatial analysis.
• Incorporated covariate data including female sex workers, literacy rate, and intravenous drug users.
• Used R and WinBugs software for model fitting and analysis of 2011 HIV data.

Main Results:

• The convolution model with a distance-based weight matrix and specific covariates demonstrated the best fit (Deviance Information Criterion).
• Estimated relative HIV risk for various Indian states using the selected model.
• Classified states into distinct HIV risk categories based on estimated relative risks.
• Generated an HIV risk map of India visualizing state-level risks.

Conclusions:

• The model-based approach provides a more robust method for HIV risk classification in Indian states.
• Focusing HIV control strategies on female sex workers, intravenous drug users, and literacy rates is recommended for effectiveness.
• The study highlights the importance of spatial factors and specific covariates in understanding and managing HIV epidemics.