Impact of control for air pollution and respiratory epidemics on the estimated associations of temperature and daily mortality
- First Online:
- Cite this article as:
- O’Neill, M.S., Hajat, S., Zanobetti, A. et al. Int J Biometeorol (2005) 50: 121. doi:10.1007/s00484-005-0269-z
We assessed the influence of control for air pollution and respiratory epidemics on associations between apparent temperature (AT) and daily mortality in Mexico City and Monterrey. Poisson regressions were fit to mortality among all ages, children (ages 0–14 years) and the elderly (ages ≥65 years). Predictors included mean daily AT, season, day of week and public holidays for the base model. Respiratory epidemics and air pollution (particulate matter <10 μm in aerodynamic diameter and O3) were added singly and then jointly for a fully adjusted model. Percent changes in mortality were calculated for days of relatively extreme temperatures [cold (10–11°C) for both cities and heat (35–36°C) for Monterrey], compared to days at the overall mean temperature in each city (15°C in Mexico City, 25°C in Monterrey). In Mexico City, total mortality increased 12.4% [95% confidence interval (CI) 10.5%, 14.5%] on cold days (fully adjusted). Among children, the adjusted association was similar [10.9% (95% CI: 5.4%, 16.7%)], but without control for pollution and epidemics, was nearly twice as large [19.7% (95% CI: 13.9%, 25.9)]. In Monterrey, the fully adjusted heat effect for all deaths was 18.7% (95% CI: 11.7%, 26.1%), a third lower than the unadjusted estimate; the heat effect was lower among children [5.5% (95% CI: −10.1%, 23.8%)]. Cold had a similar effect on all-age mortality as in Mexico City [11.7% (95% CI: 3.7%, 20.3%)]. Responses of the elderly differed little from all-ages responses in both cities. Associations between weather and health persisted even with control for air pollution and respiratory epidemics in two Mexican cities, but risk assessments and climate change adaptation programs are best informed by analyses that account for these potential confounders.