In several studies, investigators have reported associations among air poll
ution, weather, and daily deaths, usually from all causes. In the current s
tudy, we focused on the difference in lag time between exposure to total su
spended particulates or extreme weather and cause-specific mortality in an
effort to understand the potential underlying mechanism. We used a robust P
oisson regression in a generalized additive model to investigate the associ
ation between air pollution and daily mortality. We used a loess smooth fun
ction to model season, weather, and humidity; indicator variables for hot d
ays were also used. To examine the relationship in a currently meaningful r
ange, we excluded all days with a total suspended particulate concentration
higher than 200 mu g/m(3). We found a significant association on the concu
rrent day, both for respiratory infection deaths (11% increase/100 mu g/m(3
) increase in total suspended particulate; 95% confidence interval = 5, 17)
and for heart-failure deaths (7% increase; 95% confidence interval = 3, 11
). The associations with myocardial infarction (i.e., 10% increase; 95% con
fidence interval = 3, 18) and chronic obstructive pulmonary disease (12% in
crease, 95% confidence interval = 6, 17) were found for the means of 3 and
4 d prior to death. We observed an effect of cold weather at lag 1 for resp
iratory infections and an effect of hot weather at lag 0 for heart failure
and myocardial infarctions. The association for all causes and cause-specif
ic deaths was almost identical to that noted previously in Philadelphia, Pe
nnsylvania. Smoothed functions of total suspended particulates suggested a
higher slope at lower concentrations, and this finding may account for diff
erences noted between European and U.S. studies. Given that both the depend
ence between weather and daily mortality and the lag between exposure and d
eath varies by cause of death, analyses by specific causes of death would b
e very useful in the future.