Urban heat island intensity and chikungunya incidence: a multi-city time-series analysis in Brazil.
TL;DR
UHII-CHIKV associations were heterogeneous; city-specific thermal context should inform integrating satellite-derived heat metrics into arboviral surveillance, as several associations attenuated or lost statistical significance under overdispersion-robust models and were sensitive to individual epidemic years.
Abstract
Objectives
Urban heat islands (UHIs) produce 2°C-8°C in urban cores excesses, accelerating Aedes aegypti development and shortening chikungunya virus (CHIKV) extrinsic incubation, yet UHI intensity's (UHII) independent contribution to CHIKV transmission remanins unquantified.
Methods
We conducted an ecological time-series analysis of five Brazilian cities (2015-24; n = 563 city-months) . Monthly UHII was derived from MODIS land surface temperature (urban-core minus peri-urban) via Google Earth Engine; CHIKV incidence came from InfoDengue. Lag-specific (0-3 months) Poisson generalized linear models were adjusted for temperature, humidity, and seasonality. Given severe overdispersion, quasi-Poisson and negative-binomial models were the preferred sensitivity analyses.
Results
Across 261 353 cases, each 1°C UHII increase was associated with higher lag-1 incidence in subtropical Rio de Janeiro (RR = 2.857, 95% CI 2.799-2.915) and Salvador (RR = 1.168, 1.156-1.179; both P < .001). Brasília showed inverse associations (lags 0-3), Fortaleza (highest burden, 133 371 cases) a mixed lag pattern, and São Paulo minimal effects despite the highest UHII. Incidence peaked in April-May.
Conclusions
UHII-CHIKV associations were heterogeneous; city-specific thermal context should inform integrating satellite-derived heat metrics into arboviral surveillance, as several associations attenuated or lost statistical significance under overdispersion-robust models and were sensitive to individual epidemic years.