Influence of climate and heatwaves on dengue transmission in Sao Paulo and Natal, Brazil

PLOS One | https://doi.org/10.1371/journal.pone.0334838 October 24, 2025 4 / 13 epidemiological week”. We also investigated the effects of maximum temperature on dengue infections. We defined a heatwave as one when there were two or more days where the daily maximum temperature exceeded the 90th percentile of its historical distribution. We obtained weekly reports of confirmed dengue infections aggregated by onset date from 2014 and 2023 from the Brazilian Ministry of Health [25]. Statistical analysis Distributed lag non-linear model. This approach builds upon the methodology of Seah et al. [24], who examined the connection between heatwaves and dengue fever in Singapore. Taking into account the typical mosquito lifespan (10–14 days), the human infectivity period (7 days) [26], the extrinsic incubation period in Aedes mosquitoes (8–12 days), and the intrinsic incubation period in humans (3–14 days) [27], the complete dengue transmission cycle lasts approximately 28–47 days, or about seven weeks. Based on this, we considered both the immediate effects at lag week 0 and the delayed effects of meteorological variations for lag weeks 1–6, covering a total lag period of seven weeks (weeks 0–6). This approach aligns with the lag duration reported by Sang et al. in a previous study [28]. Building on previous studies that identified non-linear relationships between temperature and dengue incidence [29,30], we employed the Distributed Lag Non-Linear Model (DLNM) package in R (version 2.3.9) [31] to examine the temporal

effects of heatwaves on weekly reported dengue cases over the maximum lag period. The DLNM framework captures both non-linear and delayed associations between meteorological variables and dengue incidence by combining basis matrices for the exposure–response and lag–response relationships into a bi-dimensional cross-basis function, allowing these effects to be modeled simultaneously [31]. Negative Binomial Regression. Our primary focus was on dengue case counts in São Paulo/SP and Natal/RN from 2014 to 2023. To analyze count data (positive integers), especially in cases of overdispersion – where the variance exceeds the mean – we employed negative binomial regression models. We first examined univariate associations between each variable and the outcome, then included all significant predictors in subsequent multivariable model iterations. Model significance was assessed using the omnibus test, while the significance of individual parameters was evaluated with the Wald Chi-Square test. Variables with p-values above 0.05 in the Wald Chi-Square test were excluded from the multivariable model. For models meeting all significance criteria, we compared Akaike Information Criterion (AIC) Table 1. Climatic and demographic characteristics of the two cities analyzed: São Paulo/SP and Natal/RN. Data obtained from IBGE [22] and INMET [23]. Data São Paulo/SP Natal/RN Population (inhabitants) 11,895,578 751,300 Area (km²) 1,580 167 Human density (hab/km²) 7,528 4,498 Climate Humid subtropical Humid coastal tropical

Seasons Rainy summer – dry winter Dry summer – mild/rainy winter Average annual temperature (ºC) 20.0 26.0 Absolute Maximum temperature (ºC) 37.8 34.4 Absolute Minimum temperature (ºC) 3.5 18.2 Average Thermal amplitude (ºC)* 15.1 7.4 Average Total yearly rainfall (cm)* 218.8 392.5 Tree planting on public roads (%) 74.8 44.7 Ideal sewage system (%) 92.6 61.8 * Considering the period from 2014 to 2023. https://doi.org/10.1371/journal.pone.0334838.t001