Relative Humidity Modifies The Effect Of Temperature On Influenza Incidence: A Time Series Study In A Subtropical City

Abstract

Introduction Influenza peaks in winter months in most temperate regions, yet it could have multiple peaks in one epidemic year in tropical and subtropical regions. Relative humidity (RH) is often considered as a confounder when assess association between temperature and influenza, and few studies have considered RH might be an effect modifier. This study aims to assess the modification role of RH on effect of temperature on influenza incidence in a subtropical city. Methods Time series generalized additive Poisson models were used to assess the effect of temperature on laboratory confirmed influenza incidence from 2007–2011. In model I, weekly temperature and RH were simultaneously added into the model using thin plate spline adjusting for long term trend, seasonal trend, rainfall and the pandemic period. In model II, we modified model I to include a bivariate term of temperature and RH using thin plate spline. In model III, RH was categorized into three levels, and we modified model I to include an interactive term between temperature and RH levels. Sensitivity analyses were conducted by lagging the environmental factors for one and two weeks. Results Model I suggested non-linear effects of temperature on influenza incidence. Influenza incidence began to decrease when weekly mean temperature reached 10℃, but turned upward when it reached 22℃. The surface plot from model II showed that at low temperature, influenza incidences were high regardless of RH. But at high temperature, influenza incidence was positively associated with RH. In model III, the interactive term was statistically significant (P