AI-Driven Tuberculosis Hotspot Mapping to Optimize Active Case-Finding: Implementing the Epi-Control Platform in...

Area of Science:

• Public Health
• Epidemiology
• Artificial Intelligence

Background:

• Tuberculosis (TB) is a significant public health issue in Uganda, a high-burden country.
• Challenges in TB control include asymptomatic cases, diagnostic limitations, and unequal healthcare access.
• Effective strategies are needed to improve active case-finding (ACF) and reduce transmission.

Purpose of the Study:

• To implement and evaluate the Epi-control platform, an AI tool, for predicting community-level TB hotspots.
• To support data-driven ACF by identifying high-risk areas for targeted screening.
• To assess the model's ability to prioritize areas with higher TB case detection.

Main Methods:

• Utilized retrospective chest X-ray screening data from Uganda.
• Integrated demographic, environmental, and human development indicators into a predictive model.
• Employed a proprietary Bayesian modelling framework to predict TB risk at the sub-parish level.
• Validated the model by comparing TB yields in predicted hotspots versus non-hotspot areas.

Main Results:

• The AI model identified significantly higher TB yields in predicted hotspot areas (Risk Ratio = 1.69).
• The Central and Western regions exhibited the highest concentration of predicted hotspots.
• The model successfully prioritized areas with higher observed ACF yield in the retrospective data.

Conclusions:

• AI-based predictive modelling, like the Epi-control platform, can enhance the efficiency of ACF in Uganda.
• Targeting high-risk areas identified by the model can optimize screening efforts and resource allocation.
• Integration into national TB programs could improve planning and prioritization, though prospective validation is recommended.