The short-term harvesting effects of ambient particulate matter on mortality in Taiyuan elderly residents: A time-series analysis with a generalized additive distributed lag model.

The evaluation on mortality displacement and distributed lag effects of airborne particulate matter (PM) on death risks is important to understand the positive association of short-term pollution from both ambient PM10 and PM2.5 with daily mortality. Herein, short-term influences of urban PM10 and PM2.5 exposure on the mortality of respiratory diseases (RD) and cardiovascular diseases (CVD) were studied at Taiyuan, China, a typical inland city suffering from heavy ambient PM loading and having high morbidity of RD and CVD. Using a time-series analysis with generalized additive distributed lag model (DLM), the potential mortality displacement was determined and the single-day and cumulative lag-day effects of PM on mortality were estimated after the daily mass concentrations of urban PM2.5 and PM10 from January 2013 to October 2015 and the daily number of non-accidental death (NAD) and cause-specific mortality in the residents aged more than 65 years old were obtained. Results showed there were significant associations of PM2.5 and PM10 with daily mortality on the current day and within one week. And a statistically significant increase (P < 0.05) in the cumulative effect estimates of PM2.5 and PM10 on CVD, ischemic heart disease (IHD), and myocardial infarction (MI) mortality (as well as PM2.5 on NAD) was observed, while the associations of PM2.5 with RD and pneumonia mortality, PM10 with NAD and RD mortality were not statistically significant, when the exposure window was extended to lag 0-30 days. It was concluded that there were harvesting effects and cumulative effects of ambient PM2.5 and PM10 on the elderly residents' mortality due to RD and CVD at Taiyuan and they could be estimated quantitatively when the broader time window was used, suggesting that the underestimation on the association of ambient PM with non-accidental death can be avoided using the present method in our study.