EMERGENCY ROOM ADMISSIONS, METEOROLOGIC VARIABLES, AND AIR POLLUTANTS: A PATH ANALYSIS

Abstract

Goldsmith, J. R. (Ben-Gurlon University of the Negev, Beer Sheva 84 120, Israel), H. L. Griffith, R. Detels, S. Beeser and L. Neumann. Emergency room admissions, meteorologic variables, and air pollutants. Am J Epidemiol 1983; 118: 759–78. Daily hospital emergency room admissions at hospitals located within 8 km of Los Angeles Basin monitoring stations at Long Beach, Lennox, Azusa, and Riverside, California, were examined for correlations with pollutant and meteorologic variables for 1974–1975. By conventional correlation and regression with lagged and temporospatlal analysis, the authors could not distinguish effects of pollution by particulate sulfate from those due to meteorologic variables and oxidant. The authors use a variety of structural models and path analysis to estimate “direct” effects on emergency room admissions of maximum temperature, humidity, wind velocity, barometric pressure, carbon monoxide, sulfur dioxide, nitric oxide, high-volume suspended particulates, coefficient of haze, nitrogen dioxide, oxidant, and sulfate. Criteria for choice of models included plausibility of pairwise dependence relationships, magnitude of the correlations with emergency room admissions, and examination of the partial correlation matrix. Their results show that a variety of models gave similarly large path coefficients for a given location for the following variables: maximum temperature at each site; sulfate at Long Beach and Lennox but not at Riverside which nevertheless had the highest sulfate means; oxidant at Azusa which had the highest oxidant levels. At other locations, despite substantial and significantly elevated correlation coefficients, oxidant had a small or negative path coefficient After considering other possible factors, the authors conclude that sulfate pollution at Lennox and Long Beach had an important and possibly causal association with demand for emergency room admissions. This demonstrates the usefulness of using a variety of structural models in the analysis of ecologlc data.