Air pollution and health effects in Sao Paulo, Brazil : a time series analysis.

A time series study was conducted to investigate the association between variations in daily levels of air pollution and health effects in the city of sao Paulo, Brazil. This study was prompted by positive associations reported in other time series studies, principally in North AmeriE:a and Europe, and preliminary results from some limited analyses reported for sao Paulo. Its aims were to examine specific causes of mortality and hospital admissions, to identify more vulnerable subgroups defined in terms of age, to assess the role of socio-economic conditions in modifying the association and to detail the impact of other potential risk factors, especially meteorological. Daily measurements of air pollutants (PM10, S02, N02, 0 3 and CO) for 12 monitoring stations across the city and several meteorological variables, along with daily counts of mortality for all ages during 3 years and hospitalisations for children during 23 months were available. The time series models used Poisson regression analysis and were adjusted for effects of trend, cyclical patterns (including season), weekday, holidays, meteorological factors, and autocorrelation. Increases in PM10 and S02 were associated with a 3-4% increase in daily deaths for all causes in the elderly (results are presented for an increase from the 10th to the 90th centile of pollution measurements). Cardiovascular deaths were additionally associated with CO (4% increase). Respiratory deaths in the elderly showed higher increases (6%) associated with PM10• No significant effects for children's mortality were observed. Nevertheless, respiratory or pneumonia hospital admissions for children showed significant increases associated with 0 3 (5-8%), N02 (9%), and to a lesser extent with PM10 (9%). There was a significant trend of increasing risk of death according to age with effects only evident for older subjects. However, this age effect was more evident for all cause mortality. There was a weak suggestion of larger effects on mortality for areas economically more affluent. Some indication was found of a harvesting effect occurring in the mortality and hospital admission series in sao Paulo. Results are broadly consistent with those previously reported but somewhat smaller in magnitude. In contrast with an earlier preliminary analysis in Sao Paulo, there were no effects on mortality for children. However, new analyses for hospital admissions indicated that children are at an increased risk of non-fatal illness in relation to air pollution.