Effects of climate variables on dengue risk in Natal-RN by generalized models

Background: In recent years, investment in research has increased to explain the epidemiological behavior of dengue. Several authors have investigated the relationship between climatic variables and dengue, often using mathematical models to analyze time series to describe temporal trends, identify patterns, and even for control, planning and prediction. This work aimed to identify climatic influences on dengue in Natal, city from Rio Grande do Norte State - Brazil, from Jan 2001 to Dec 2012, and to evaluate the ability of prediction of models to describe their associations and their occurrence mean parameters. Methods & Materials: We assessed the potential contribution of climate-driven (rainfall and maximum temperature) on Dengue Fever (DF) cases based in the following models: Generalized Linear (GLM), Generalized Additive (GAM) and Nonlinear Lags Distributed (DLNM), all by Poisson distribution, and provide suggestions to improve their performance. These models use all the same families and link functions; but in DLNM flexibly describe associations that show potentially lagged and nonlinear effects in the time series, with a cross-based basis, a two-dimensional functional space specifying relationships in the prediction and their lags. Results: GLM showed that DF cases are highly correlated with rainfall at low rainfall (<70 mm). Increases in DF cases are observed at pluviometric rates around 100-200 mm and 350-500 mm, when there is a marked decrease in the number of cases. In relation to the maximum temperature (tested by GAM), the optimal limit varies around 28.0 °C to 30.2 °C, with dengue peaks at a maximum average of 29.5 °C. At 30.6 °C, there is a negative correlation, with a marked decrease in the number of cases. With DLNM, the optimal limit for maximum temperature varies around 28.0 °C and 32 °C, with risk of dengue, respectively 2x and 4x the average number of cases. Conclusion: DLNM proved to be more efficient in the analyzes, since it made it possible to find the average parameters of maximum temperature for the occurrence of dengue and time lag, including viewing in 3D, respectively, 28-29.5 °C, delayed in 2 months in relation to the occurrence of the cases.