ENDCast: A framework for El Niño driven disease forecasting in Latin America and the Caribbean

FLETCHER, CLOE; LOTTO BATISTA, MARTÍN; LÜHRSEN, DANIELA; CARVALHO, BRUNO; LLABRÉS BRUSTENGA, ALBA; MÜLLER, GABRIELA VIVIANA; GÓMEZ, ANDREA; LÓPEZ, MARÍA SOLEDAD; CÁRCAMO GARCÍA, PALOMA; CARRASCO ESCOBAR, GABRIEL; UMAÑA, JUAN; SANTOS VEGA, MAURICIO; GRACIE, RENATA; BARCELLOS, CHRISTOVAM; ROLLOCK, LESLIE; BORBOR CORDOVA, MERCY; LOWE, RACHEL

Bruselas

Conferencia; Research Perspectives on the Health Impacts of Climate Change; 2024

El Niño influences extreme climatic events around the globe, such as floods, droughts and storms, which can impact the timing and intensity of climate-sensitive infectious disease outbreaks. Early warning systems and decision-support tools that integrate seasonal climate forecasts can predict the risk of outbreaks 1-6 months in advance, enabling sufficient time to implement actions for epidemic preparation. However, there is currently a lack of accessible and reproducible prediction tools that can be implemented in response to emerging climatic events. To address this gap, we present a decision-support framework, ENDCast (El Niño Driven Disease Forecasting), that produces probabilistic predictions of infectious disease outbreaks in response to El Niño or La Niña events and associated seasonal climatic anomalies. We have opted to focus our initial prototype in the Latin American and Caribbean (LAC) region, an area experiencing multiple concurrent outbreaks during the current 2023/24 El Niño. An operational platform that issues monthly disease forecasts with a 1-6 month outlook will be hosted via an R Shiny web application throughout the event. At present, we produce forecasts for two vector-borne diseases (dengue and malaria) and one water-borne zoonosis (leptospirosis) in LAC hotspots that are sensitive to El Niño, including Northeastern Argentina, Barbados, the Brazilian Amazon, Northeastern and Southern Brazil, Colombia,Ecuador and Peru. However, the ENDCast framework represents a simple, reproducible methodology that could be rapidly and flexibly deployed to any location or climate-sensitive disease during a future El Niño or La Niña. Using climate data obtained from open global products and locally provided epidemiological data, we undertake a comprehensive model fitting, selection and verification process to establish bespoke climate-integrated disease prediction models in each site. We employ a hierarchical Bayesian mixed modelling framework, where models contain up to three climatic covariates including lagged El Niño-, temperature- and precipitation-based indicators. By integrating climate forecasts that have been individually calibrated through tailored post-processing techniques, we produce probabilistic 1-6 month forecasts for dengue, malaria and leptospirosis each month. These forecasts are hosted on a web application that has been co-produced with local collaborators. The platform includes visualisations, alerts and risk levels for predicted outbreaks at varying spatial and temporal scales, providing decision-makers with early warning of potential outbreaks to trigger early action. The ENDCast approach could be easily adapted to other endemic settings, including Asia or Africa, to predict the probability of outbreaks for a host of different climate-sensitive diseases, contributing to the Wellcome Trust funded projects (HARMONIZE and IDExtemes) and EU-funded projects (E4Warningand IDAlert).