Background: Kawasaki disease exhibits a distinct seasonality, and short-term changes in weather may affect its occurrence.
Methods: To investigate the effects of weather variability on the occurrence of this syndrome, we conducted a time-between-events analysis of consecutive admissions for Kawasaki disease to a large pediatric hospital in Chicago. We used gamma regression to model the times between admissions. This is a novel application of gamma regression to model the time between admissions as a function of subject-specific covariates.
Results: We recorded 723 admissions in the 18-year (1986–2003) study period, of which 700 had complete data for analysis. Admissions for Kawasaki disease in Chicago were seasonal: The mean time between admissions was 34% shorter (relative time = 0.66, 95% confidence interval 0.54–0.81) from January–March than from July–September. In 1998, we recorded a larger number of admissions for Kawasaki disease (n = 65) than in other years (mean n = 37). January–March months of 1998 were warmer by a mean of 3°C (1.5°C–4.4°C) and the mean time between admissions was 48% shorter (relative time = 0.52, 0.36–0.75) than in equivalent periods of other study years.
Conclusions: Our findings show that atypical changes in weather affect the occurrence of Kawasaki disease and are compatible with a link to an infectious trigger. The analysis of interevent times using gamma regression is an alternative to Poisson regression in modeling a time series of sparse daily counts.