Background: We investigated the impact of environmental temperature on mortality in São Paulo, Brazil, and examined differences in the temperature-mortality relationship with respect to cause, age group, and socioeconomic position (SEP).
Methods: Generalized additive Poisson regression models adjusted for non-temperature related seasonal factors (including air pollution) were used to analyse daily mortality counts for selected causes from 1991 to 1994. Individuals were classified by the aggregate SEP of their area of residency. These were analysed as potential modifiers of the temperature-mortality relationship.
Results: Among the elderly we observed a 2.6% increase in all-cause mortality per degree increase in temperature above 20 degrees C, and a 5.5% increase per degree drop in temperature below 20 degrees C, after adjustment for confounding. Relationships were similar in children, but somewhat weaker in adults. Cold effects were present for deaths due to cardiovascular disease (CVD), respiratory, and other causes, with effects being greatest in the respiratory group. Heat effects were not found for CVD deaths in adults, but otherwise varied little by cause of mortality. There was little evidence for a modification of the mortality effects of cold or heat by SEP.
Conclusions: These findings show that the U-shaped pattern of the temperature-mortality relationship found in cooler northern countries occurs also in a sub-tropical city. In addition, the relative effects of temperature were similar in each socioeconomic grouping.