Estimating the Independent Short-Term Effect of Temperature and Humidity on Mortality: Assessing Various Modelling Strategies

Abstract

ISEE-859 Objective: Within the PHEWE (Project on Health Effects of Weather conditions in Europe) project, an extensive discussion on how to model temperature and humidity in the investigation of acute effects of meteorological conditions on mortality took place. The common use of relative humidity (RH) was criticized and it was proposed to use alternatively either absolute humidity (AH) or the difference between dry bulk and dew point temperature (Dry-Dew). We will present the application of different models and compare their results. Material and Methods: The data base consists of data on meteorological variables and total natural, cardiovascular and respiratory mortality from 10 cities during the cold period (September–March) spanning across Europe. GEE models were used for city-specific analysis and the estimated effects combined using second stage models. Results: The high correlation coefficient between temperature and AH (>0.80) precluded the simultaneous use of these 2 variables in 1 model, due to high collinearity. When they were introduced to models alternatively, both were highly significant, each reflecting also the effect of the other and it was impossible to separate the effects of the 2. When temperature (either mean or max) was introduced in a model together with Dry-Dew or RH, the temperature effect was more important and highly statistically significant, whilst the humidity effect was much smaller. Conclusions: When the objective of an investigation is to estimate the separate (independent) acute effects of temperature and humidity on health, AH and temperature cannot be used simultaneously in a model because of collinearity. Therefore their independent effects cannot be assessed. However when using Dry-dew or RH, it is shown that the temperature effect is more important compared to humidity in the cold period.