A reproducible ensemble machine learning approach to forecast dengue outbreaks

Area of Science:

• Epidemiology
• Machine Learning
• Public Health

Background:

• Dengue fever is a significant global health and economic burden, particularly in tropical and subtropical regions.
• Predicting dengue outbreaks is challenging due to complex spatiotemporal variations in disease incidence.

Purpose of the Study:

• To develop and validate a machine learning ensemble model for forecasting the dengue incidence rate (DIR) in Brazil, focusing on individuals under 19.
• To assess the model's transferability to Peru and its potential for aiding public health interventions.

Main Methods:

• An ensemble machine learning model integrating spatial and temporal data was developed for one-month-ahead DIR forecasting at the state level.
• The model's efficacy was evaluated through comparative analyses against a dummy model and ablation studies.
• The approach was tested for transferability to Peru, considering its distinct epidemiological profile.

Main Results:

• The ensemble model demonstrated significant qualitative and quantitative efficacy across Brazil's 27 Federal Units.
• Successful transferability and consistent performance were observed in Peru and during Brazil's 2019 dengue outbreak.
• The model identified key factors triggering dengue outbreaks in both countries.

Conclusions:

• The developed machine learning model offers a scalable and practical solution for dengue incidence forecasting.
• This approach advances climate services for health and demonstrates the value of integrating advanced analytics into public health frameworks.
• The study highlights the importance of interdisciplinary collaboration for addressing global health challenges like arboviral diseases.