Identifying spatio-temporal patterns in the association between PM10 and mortality

Abstract

BACKGROUND AND AIM: It is well known that the short-term effects of PM10 on mortality are not homogeneous across regions. In recent years, a few studies have revealed that the mortality effects of PM10 could also vary over time with specific patterns. Since the effects of PM10 could be heterogeneous over regions, its temporal pattern might also be different according to geographical locations. In this study, we tried to investigate whether there are specific spatio-temporal patterns in the association between PM10 and mortality in Seoul, Korea. By scrutinizing the spatio-temporal variations, we expect to gain more insight than just investigating spatial or temporal variations. METHODS: The study area is Seoul, Korea, for 2004-2018. For the 25 districts in Seoul, we estimated district and time-window specific effect of PM10 on the non-accidental mortality using the time series analysis. We used 5-year moving time window that shifts every month for the study period (window 1: Jan 2004 - Dec 2008, …, window 120: Jan 2014, - Dec 2018). The estimated 120 coefficients in each district are then smoothed using B-spline. After that, we clustered the coefficients of the B-spline using the K-means. RESULTS:We discovered that the mortality effects of PM10 have been changed with specific temporal patterns and that patterns are different over districts. By using K-means clustering, we could identify four different patterns. In one cluster, the effects of PM decreased in the early years but increased in recent years. Another cluster showed the opposite patterns. There was also a linearly decreasing pattern. CONCLUSIONS:Identifying the spatio-temporal pattern is important since it could improve our understanding of the PM. Further studies would be needed to explain the observed spatio-temporal variations. KEYWORDS: Air pollution, Particulate matter, Short-term exposure, Mortality