Air pollution effects on hospital admission rates: A random effects modeling approach

Statistical methods are proposed to analyze parallel time series of hospital-based health data and measurements of ambient air pollution. Specifically, associations between the number of daily health events (hospital admissions or emergency-room visits for respiratory illnesses) and daily levels of ambient air pollutants in the vicinity of several hospitals are examined. A relative-risk regression model is proposed in which the regression parameters are assumed to vary at random among hospitals. Adjustment for seasonal trends in admissions are also considered. Simple computational methods based on generalized estimating equations are explored as the basis for statistical inference. The proposed methods are illustrated on data obtained from 164 acute-care hospitals in Ontario over the May-to-August period for 1983 to 1988. These admission rates are related to ozone levels obtained from 22 monitoring stations maintained by the Ontario Ministry of the Environment.