Forecasting dengue fever in Brazil: An assessment of climate conditions

Abstract

Local climate conditions play a major role in the biology of the Aedes aegypti mosquito, the main vector responsible for transmitting dengue, zika, chikungunya and yellow fever in urban centers. For this reason, a detailed assessment of periods in which changes in climate conditions affect the number of human cases may improve the timing of vector-control efforts. In this work, we develop new machine-learning algorithms to analyze climate time series and their connection to the occurrence of dengue epidemic years for seven Brazilian state capitals. Our method explores the impact of two key variables—frequency of precipitation and average temperature—during a wide range of time windows in the annual cycle. Our results indicate that each Brazilian state capital considered has its own climate signatures that correlate with the overall number of human dengue-cases. However, for most of the studied cities, the winter preceding an epidemic year shows a strong predictive power. Understanding such climate contributions to the vector’s biology could lead to more accurate prediction models and early warning systems.

Evaluation of the Models for Forecasting Dengue in Brazil from 2000 to 2017: An Ecological Time-Series Study

“…Baqueiro et al [ 47 ] comprehensively compared deterministic (ARIMA) and stochastic models (neural networks) for predicting dengue with a forecasting horizon of 1 month in the city of São Paulo. Storlerman et al [ 48 ] employed machine learning algorithms to detect climatic signatures that correlated with the total number of dengue cases in some Brazilian capitals. Guo et al [ 49 ] developed dengue prediction models using artificial intelligence algorithms for Guangdong, China.…”

Probabilistic seasonal dengue forecasting in Vietnam: A modelling study using superensembles

“…Dengue early warning systems driven by Earth observations and seasonal climate forecasts have been proposed using a range of modelling approaches [ 38 ], including autoregressive integrative moving average (ARIMA) [ 39 ], point forecasts [ 32 , 40 ], spatiotemporal Bayesian hierarchical models [ 18 , 19 ], least absolute shrinkage and selection operator (LASSO) regression [ 41 , 42 ], and machine learning [ 43 ]. Often, models are validated using block cross-validation to select the model specification with the lowest out-of-sample predictive error [ 18 , 32 ].…”

Unlocking the Predictive Power of Heterogeneous Data to Build an Operational Dengue Forecasting System

“…Broadly speaking, dengue prediction studies have focused on the Americas 19 and Southeast Asia 20 . Time scales range from predicting categorical risk for an entire transmission season 21 to weekly case counts 19 . Spatial scales are most commonly limited to cities 22 , small countries such as Singapore 20 , or specific regions (e.g., county, state) within a country with particularly good human case counts and/or mosquito time series data 23 .…”

“…Spatial scales are most commonly limited to cities 22 , small countries such as Singapore 20 , or specific regions (e.g., county, state) within a country with particularly good human case counts and/or mosquito time series data 23 . Windows of prediction include nowcasting (necessary since reported case counts often have a lag of 1-4 weeks) 19 , forecasting weekly cases 4-12 weeks into the future 20 , and several-months ahead categorical risk forecasting 21 . Many of these studies have exploited multiple data streams including socioeconomic drivers 9 , weather data 6 , satellite imagery 11 , topography (e.g., altitude), entomological factors 24 , Internet data (e.g., social media) 13,25 , and clinical surveillance data 19 .…”