Using Annual Data to Estimate the Public Health Impact of Extreme Temperatures

Abstract

Short-term associations between both hot and cold ambient temperatures and higher mortality have been found worldwide. Few studies have examined these associations on longer time scales. Age-standardized mortality rates (ASMRs) were calculated for 1976-2012 for Hong Kong SAR, People's Republic of China, defining "annual" time periods in 2 ways: from May through April of the following year and from November through October. Annual frequency and severity of extreme temperatures were summarized by using a degree-days approach with extreme heat expressed as annual degree-days >29.3°C and cold as annual degree-days <27.5°C. For example, a day with a mean temperature of 25.0°C contributes 2.5 cold degree-days to the annual total. Generalized additive models were used to estimate the association between annual hot and cold degree-days and the ASMR, with adjustment for long-term trends. Increases of 10 hot or 200 cold degree-days in an annual period, the approximate interquartile ranges for these variables, were significantly (all P's ≤ 0.011) associated with 1.9% or 3.1% increases, respectively, in the annual ASMR for the May-April analyses and with 2.2% or 2.8% increases, respectively, in the November-October analyses. Associations were stronger for noncancer and elderly mortality. Mortality increases associated with extreme temperature are not simply due to short-term forward displacement of deaths that would have occurred anyway within a few weeks.

Keywords: ambient temperature; biometeorology; climate change; mortality; time series.