Influência de variáveis climáticas defasadas na incidência de dengue em Uberlândia-mg: uma análise de séries temporais

Abstract

Dengue is an endemic disease in Brazil, and its transmission is associated with climatic condi tions, particularly temperature and precipitation. This study aimed to analyze the historical behavior of dengue cases in Uberlândia, Minas Gerais, Brazil, from January 2014 to April 2026, through SARIMA and SARIMAX time series models, evaluating the influence of time lagged climatic variables. Case data were obtained from the Notifiable Diseases Information System (SINAN), and climatic data on temperature and precipitation were obtained from the NASA POWER platform. The methodology followed the Box-Jenkins approach, including a trend test (Cox-Stuart) and a seasonality test (Fisher), logarithmic transformation to stabilize variance, identification and estimation of SARIMA models, and subsequent incorporation of lagged climatic exogenous variables, selected through cross-correlation function (CCF) analy sis. The SARIMA(1,1,0)(1,1,1)12 model was selected as the reference model, presenting the lowest information criterion values among the models with statistically significant coefficients. The inclusion of temperature lagged by five months as an exogenous variable resulted in the final SARIMAX model, which showed a lower AIC and a statistically significant coefficient, evidencing an inverse association between temperature and the subsequent incidence of the disease. Residual diagnostics indicated the absence of autocorrelation at the first lags, but sig nificant residual autocorrelation at the longer seasonal lags, attributed to the atypical behavior of the 2025 outbreak. Model validation was performed through rolling (one-step-ahead) fore casting, given the atypical nature of that outbreak. The models proved to be complementary, as SARIMA showed slightly higher predictive performance, while SARIMAX made it possible to quantify the influence of temperature on dengue incidence. The results reinforce the poten tial of climatic variables as a tool to support epidemiological surveillance and vector control planning in the municipality.