Predicting malaria epidemics in Burkina Faso with machine learning

Area of Science:

• Epidemiology
• Public Health
• Machine Learning

Background:

• Accurate malaria forecasting is crucial for timely public health interventions.
• Traditional malaria prediction methods rely on complex simulations.
• A need exists for data-driven early warning systems.

Purpose of the Study:

• To develop and validate the first data-driven malaria epidemic early warning system.
• To predict the 13-week malaria case rate in Burkina Faso.
• To leverage high-fidelity health data for improved forecasting.

Main Methods:

• Utilized the Integrated e-Diagnostic Approach (IeDA) infant consultation data.
• Employed a combination of Gaussian Processes and Random Forest Regressors.
• Trained and tested the algorithm on historical malaria epidemic data.

Main Results:

• The system predicts weekly malaria cases over a 13-week period.
• Achieved >99% recall at a 30% precision for low-threshold alerts.
• >99% precision with 5% recall for high-threshold alerts.
• Provided precise 1σ (5 cases) and 2σ (30 cases) predictions.

Conclusions:

• The developed system offers a significant advancement in malaria forecasting.
• Data-driven early warning systems can enhance public health preparedness.
• This approach has the potential to save lives by enabling proactive interventions.