Spatio-temporal risk prediction of leptospirosis: A machine-learning-based approach

Area of Science:

• Environmental Science
• Epidemiology
• Public Health

Background:

• Leptospirosis is a neglected zoonotic disease with a global prevalence, especially in tropical areas.
• Climate change, rainfall, and cyclones intensify the spread of Leptospira bacteria, transmitted via contaminated water and soil.
• Pacific islands face a high burden, necessitating identification of factors influencing disease distribution for policy development.

Purpose of the Study:

• To create a detailed spatio-temporal risk map for leptospirosis nationwide.
• To identify key meteorological, topographic, and socio-demographic factors influencing Leptospira transmission.
• To enhance disease surveillance and risk assessment for an effective early warning system.

Main Methods:

• Utilized machine learning models trained on binarized incidence rates for prediction.
• Conducted spatial analysis at a finer resolution than city level and monthly temporal analysis (2011-2022).
• Integrated meteorological, topographic, and socio-demographic variables into the analysis.

Main Results:

• Achieved 83.29% concordance and 83.93% sensitivity in predicting contamination risk.
• Identified seasonal patterns, including the influence of the El Niño Southern Oscillation.
• Rainfall and humidity (one-month lag) significantly contribute to Leptospira contamination; high organic matter soil may reduce bacterial presence.

Conclusions:

• Environmental factors significantly drive the seasonal spread of Leptospira in tropical and subtropical regions.
• Machine learning models offer a robust approach for disease surveillance and risk assessment.
• Findings support targeted public health policies and the development of an early warning system for leptospirosis.