Survival analysis to estimate association between short-term mortality and air pollution.

BACKGROUND: Ecologic studies are commonly used to report associations between short-term air pollution and mortality. In such studies, the unit of observation is the day rather than the individual. Moreover, individual data on the subjects are rarely available, which limits the assessment of individual risk factors. These associations can also be investigated using case-crossover studies. However, by definition, individual risk factors are not studied, and such studies analyze only dead subjects, which limits the statistical power. OBJECTIVE: We suggest that the survival analysis is more suitable when cohorts are examined with a time-dependent ecologic exposure. To our knowledge, to date this type of analysis has never been proposed. DESIGN, PARTICIPANTS, MEASUREMENTS: In the present study we used a Cox proportional hazards model to investigate the distribution over time of the short-term effect of black smoke and sulfur dioxide in 439 nonaccidental and 158 cardiorespiratory deaths among the 1,469 subjects of the Personnes Ag? QUID (PAQUID) cohort in Bordeaux, France. The model has a delayed entry and a polynomial distributed lag from 0 to 5 days. Results are adjusted for individual risk factors, temperature, relative humidity, weekday, season, influenza epidemics, and a time function to control temporal trends. RESULTS: We identified a positive and significant association between cardiorespiratory mortality and black smoke, with a 24% increase in deaths 3 days after a 10-microg/m3 increase in black smoke (95% confidence interval, 4-47%). CONCLUSIONS: We conclude that the Cox proportional hazards model with time-dependent covariates is very suitable to investigate simultaneously the short-term effect of air pollution on health and the effect of individual risk factors on a cohort study