Background:
Late HIV diagnosis and suboptimal linkage to antiretroviral therapy (ART) remain major challenges in Cameroon, contributing to ongoing transmission and preventable morbidity. Traditional testing strategies often fail to efficiently identify individuals at the highest risk due to limited resources and heterogeneous population dynamics. Artificial intelligence (AI), specifically machine learning (ML), offers a novel approach to predict undiagnosed HIV and optimise targeted testing interventions.
Methods:
We developed and validated a predictive ML model using routine health system data, demographic information, behavioural surveys, geographic indicators, and clinical encounter histories. The model generates individual and community-level risk scores to identify populations with a high probability of undiagnosed HIV. These risk scores were integrated into an interactive dashboard to support district health planning, prioritising outreach and testing campaigns. Model performance was evaluated using area under the receiver operating characteristic curve (AUC), sensitivity, and specificity metrics, alongside pilot implementation to assess feasibility, uptake of testing, and linkage to ART.
Results:
Preliminary analysis indicates the AI model accurately identifies high-risk individuals and communities, achieving an AUC of 0.89. Targeted interventions guided by model predictions increased HIV testing uptake by 27% and improved same-day ART initiation rates by 19% compared to standard outreach approaches. Geographic visualisation of risk hotspots enabled more efficient resource allocation and enhanced engagement with previously underserved populations.
Conclusions:
AI-driven predictive modelling can significantly improve efficiency in HIV testing and linkage to care in Cameroon, reducing late diagnoses and supporting decentralised, data-informed interventions. Integration of such models into routine health planning demonstrates strong potential for scalability and impact on national HIV strategies. Future work will focus on continuous model refinement, operational integration, and expansion to additional regions.