The Application of Machine Learning Algorithms to Predict HIV Testing Using Evidence from the 2002-2017 South African...

Area of Science:

• Public Health
• Epidemiology
• Machine Learning

Background:

• A significant portion of South Africa's population has an unknown HIV status, hindering epidemic control efforts.
• Existing predictive models are insufficient for guiding targeted HIV testing interventions in South Africa.
• Machine learning (ML) offers potential for identifying individuals at higher risk for HIV infection, informing testing recommendations.

Purpose of the Study:

• To identify consistent predictors of HIV testing among adults in South Africa using supervised machine learning (SML) algorithms.
• To evaluate the performance of four SML algorithms (decision trees, random forest, SVM, logistic regression) on population-based survey data.
• To inform data-driven policy initiatives for enhancing HIV testing efficacy and achieving the UNAIDS 2030 goal.

Main Methods:

• Utilized the South African National HIV Prevalence, Incidence, and Behavior and Communication Survey (SABSSM) datasets from five cross-sectional cycles.
• Applied four SML algorithms: decision trees, random forest, support vector machines (SVM), and logistic regression.
• Employed an 80% training and 20% testing split with 5-fold cross-validation for each dataset.

Main Results:

• Random forest demonstrated superior performance across all datasets, achieving the highest accuracy, precision, F1-score, and AUC.
• Key predictors of HIV testing included knowledge of testing sites, being female, younger age, high socioeconomic status, and digital information access.
• SVM showed high recall but lower precision, while logistic regression and decision trees had moderate performance, with decision trees prone to overfitting.

Conclusions:

• Supervised machine learning, particularly random forest, effectively identifies predictors of HIV testing in South Africa.
• Targeted interventions should focus on increasing awareness of testing sites, leveraging digital platforms, and addressing socioeconomic factors.
• Improving HIV testing efficacy through data-driven strategies is crucial for South Africa's progress towards the UNAIDS 2030 goal.