Résumé
Leptospirosis outbreaks are frequent in Pacific islands where tropical climate and lifestyle provide suitable conditions for human infection (1). With challenging clinical and laboratory diagnoses and without effective vaccines, preparedness and prevention remain key to mitigate outbreaks (2). Such strategies imply that epidemics are anticipated months in advance, underlying the need to design operational early warning systems. These systems could focus on the strong association between leptospirosis and climate to produce operational models and inform public health surveillance (3,4).We adjusted climate-driven machine learning models of leptospirosis dynamics on 7 tropical islands: Fiji, New Caledonia, Tahiti, Futuna, Mayotte, Guadeloupe, and Reunion island. A global seasonal model was first fitted and could be used to estimate leptospirosis seasonal dynamics where the disease is not routinely monitored. Island-specific inter-annual models provided epidemic risk predictions based on climate for islands with known historical dynamics. We tested several explanatory variables based on temperature and precipitation data, including averages, the number of extreme events, and their dispersion over the previous three months. We selected the best models based on mean average error scores obtained in cross-validation.Our global model was mainly driven by precipitation. The inter-annual dynamics was best estimated with island-dependent models driven by both precipitation and temperature variables. Including variables that describe extreme climate events improved the predictions. The inter-annual dynamic remained difficult to capture in Tahiti. The developed models have been implemented in an experimental dashboard connected to climate data, designed to inform decision-makers on leptospirosis dynamics. While climate data can be routinely collected, regular and timely access to health data remains a challenge to build operational early warning systems. The model will be further improved with the addition of environment-based risk determinants and through the incremental addition of more leptospirosis surveillance data. Références(1)Costa F, Hagan JE, Calcagno J, Kane M, Torgerson P, Martinez-Silveira MS, et al. Global Morbidity and Mortality of Leptospirosis: A Systematic Review. Plos Neglect Trop D. 2015 Sep;9(9).(2)Goarant C. Leptospirosis: risk factors and management challenges in developing countries. Research and Reports in Tropical Medicine [Internet]. 2016 Sep 28 [cited 2024 Jan 25];7:49–62. Available from: https://www.tandfonline.com/doi/abs/10.2147/RRTM.S102543(3)Rees, E. M., Lotto Batista, M., Kama, M., Kucharski, A. J., Lau, C. L., & Lowe, R. (2023). Quantifying the relationship between climatic indicators and leptospirosis incidence in Fiji: A modelling study. PLOS Global Public Health, 3(10), e0002400.(4)Lotto Batista, M., Rees, E. M., Gómez, A., López, S., Castell, S., Kucharski, A. J., ... & Lowe, R. (2023). Towards a leptospirosis early warning system in northeastern Argentina. Journal of The Royal Society Interface, 20(202), 20230069.