Effect of seasonal and temperature variation on hospitalizations for stroke over a 10-year period in Brazil.

Abstract

Seasonal variation in stroke incidence remains controversial. We aimed to describe the pattern of seasonality in hospitalizations for stroke in Brazil. We evaluated age-adjusted hospitalization rates for stroke per month using data from the Brazilian Unified Healthcare System and median monthly temperature data obtained from the National Institute of Meteorology. To detect a seasonality pattern in time series, we used seasonal-trend decomposition using LOESS. We calculated a seasonal strength statistic and used Kruskal-Wallis test to evaluate the presence of seasonality in Brazil and its five regions. We also assessed the association of temperature and stroke hospitalization rates using Spearman's rho correlation. We identified 1,422,496 stroke-related hospitalizations between 2009 and 2018. Mean age was 67 years, 51% were male and 77.5% of stroke diagnoses were not specified as ischemic or hemorrhagic. Median temperature was 23.8℃ (IQR 22.3-24.4). Age-adjusted hospitalizations demonstrated significant seasonal variation during all the years analyzed, with increased rates during the winter. When regional differences were analyzed, seasonal behavior was present in the south, southeast and northeast regions of the country. These were also the regions with lower median temperatures during the winter months and greater amplitude of average temperatures between warmer and colder months. In this large national cohort of stroke patients in Brazil, we demonstrated the presence of seasonal variation in the age-adjusted hospitalization rate, with peak rates during the winter months. The regional gradient of incidence of stroke was directly associated with colder winters and greater amplitude of temperature.