with an increase in lower respiratory symptoms. These authors discuss the importance of misclassification of exposure in diluting the findings of the effects of nitrogen dioxide exposure.
In a large-scale analysis of ambient daily air pollution and mortality, Schwartz and Marcus (1990) found a strong relation between air-pollution levels and daily mortality over the period 1958-1972 in London, England. Inclusion of terms for changes in temperature and humidity increased the strength of the relation between particles and mortality. The inclusion of weather data in the analysis illustrates how indirect measures of exposure can strengthen existing associations.
Information derived from daily health diaries can be particularly important. For example, diaries can capture variability in the occurrence of symptoms, which can be related to variation in air-pollutant concentrations. Schwartz et al. (1991) described the advantages of linking ambient-air measurements and individual indicators of household exposure to daily diaries of respiratory symptoms. Diary studies record the health status of each study participant repeatedly over time and can define the impact of short-term changes in the environment on human health. In this analysis of data from the Six Cities Study of Air Pollution and Health and from a nurses' diary study in Los Angeles, exposure information was obtained from air-monitoring stations as well as from data on the type of cooking stove and history of parental and roommate smoking. A statistically significant relation between air pollution and reported symptoms remained after the effects of autocorrelation and heterogeneity were addressed. The reader is referred to volume 1 of the present report (NRC, 1991a, pp. 166-167) and the report Human Exposure Assessment for Airborne Pollutants (NRC, 1991b, pp. 157-159) for a more detailed discussion of the use of diaries.
The Health Interview Survey (HIS) of a sample of thousands of US residents was used by Ostro and Rothschild (1989) to assess the relation between acute respiratory morbidity and air pollution. The large number of subjects in this study aided in the identification of statistically significant associations. The exposure information was derived from EPA's SAROAD monitoring network of ambient-air pollution. Portnoy and Mullahy (1986) have also used the HIS data to advantage. In each study, statistically significant findings were documented despite the limited exposure assessment from ambient monitoring.
Some investigators (Bates and Sizto, 1987; Bates et al., 1990; Samet et al., 1981) have reported a relation between visits to hospital emergency departments for respiratory ailments and various types of air pollution. The statistically significant relation between ozone concentration and asthma visits was identified by a multiple regression analysis that controlled for temperature. These authors conclude that, among regions with