USING TIME SERIES OF ANNUAL DEATH COUNTS TO OBTAIN HARVESTING-RESISTENT ESTIMATES OF TEMPERATURE- MORTALITY ASSOCIATIONS

摘要

Background and Aims: Daily time series regression analyses have proved powerful ways to identify acute effects of temperature extremes and of air pollution, but associations may reflect just short term displacement of deaths ("harvesting"). To avoid this we apply time series methods to associate annual deaths counts with annual summaries of temperature extremes. We illustrate the method by application to impacts of low and high temperature in London 1949-2006. Methods: Analysis units were Sept-Aug years (n=57), for which total deaths (Y) and summaries of cold and heat (X: means degree-days less and more than 18oC daily mean respectively) were obtained. The 18 degree threshold was based on the minimum mortality temperatures in previous daily time series studies. The coefficient of the cold and heat variables was estimated using an overdispersed Poisson GLM. Secular trends were modeled using splines with 5 degrees of freedom and influenza by including count of pneumonia and influenza deaths (linear term). Results: The estimated increase in total annual mortality per degree was 2.3% (95%CI 0.6,4.1) for cold and 1.6% (-4.1%,8.1%) for heat. Sensitivity analyses showed similar results for a range of alternative model choices including more flexible secular trends, negative binomial regression, indirect adjustment for black smoke, and different thresholds (15 and 21 degrees). Conclusions: In this London annual series we saw an association of cold with mortality which was broadly similar in magnitude to that found in daily studies, suggesting that most deaths due to cold were displaced by at least six months. The estimated association with heat was imprecise, with confidence interval including magnitudes found in daily studies but also including zero. The imprecision reflects the smaller number of heat-related deaths in temperate London which makes this approach ineffective here to identify harvesting in deaths due to heat.