Mortality risk attributable to high and low ambient temperature: a multicountry observational study

Summary Background Although studies have provided estimates of premature deaths attributable to either heat or cold in selected countries, none has so far offered a systematic assessment across the whole temperature range in populations exposed to different climates. We aimed to quantify the total mortality burden attributable to non-optimum ambient temperature, and the relative contributions from heat and cold and from moderate and extreme temperatures. Methods We collected data for 384 locations in Australia, Brazil, Canada, China, Italy, Japan, South Korea, Spain, Sweden, Taiwan, Thailand, UK, and USA. We fitted a standard time-series Poisson model for each location, controlling for trends and day of the week. We estimated temperature–mortality associations with a distributed lag non-linear model with 21 days of lag, and then pooled them in a multivariate metaregression that included country indicators and temperature average and range. We calculated attributable deaths for heat and cold, defined as temperatures above and below the optimum temperature, which corresponded to the point of minimum mortality, and for moderate and extreme temperatures, defined using cutoffs at the 2·5th and 97·5th temperature percentiles. Findings We analysed 74 225 200 deaths in various periods between 1985 and 2012. In total, 7·71% (95% empirical CI 7·43–7·91) of mortality was attributable to non-optimum temperature in the selected countries within the study period, with substantial differences between countries, ranging from 3·37% (3·06 to 3·63) in Thailand to 11·00% (9·29 to 12·47) in China. The temperature percentile of minimum mortality varied from roughly the 60th percentile in tropical areas to about the 80–90th percentile in temperate regions. More temperature-attributable deaths were caused by cold (7·29%, 7·02–7·49) than by heat (0·42%, 0·39–0·44). Extreme cold and hot temperatures were responsible for 0·86% (0·84–0·87) of total mortality. Interpretation Most of the temperature-related mortality burden was attributable to the contribution of cold. The effect of days of extreme temperature was substantially less than that attributable to milder but non-optimum weather. This evidence has important implications for the planning of public-health interventions to minimise the health consequences of adverse temperatures, and for predictions of future effect in climate-change scenarios. Funding UK Medical Research Council.

Welfare, is represented by counts of deaths for non-external causes only (ICD-9: 0-799; ICD-10: A00-R99). Mean daily temperature (in ˚C) and relative humidity (in %), computed as the 24-hour average based on hourly measurements, were obtained from the Environment and Health Administration. A single weather station, located at Torkel Knutssongatan in Central Stockholm, was selected. Measures ozone (O3, in ppb) and nitrogen oxides (NOx, in ppb) were available in the same period. Daily level of pollutants were computed as the 24-hour mean based on hourly measurements. In total, missing data amount for 0.00% and 6.59% of the mortality and temperature series, respectively. These data were used and described in previous publications. 8,9 Taiwan We collected data in Kaohsiung, Taipei and Taichung between 1 st of January 1994 and 31 st of December 2007. Daily mortality is represented by counts of deaths for all causes and for non-external causes only (ICD-9: 0-799; ICD-10: A00-R99). Mean daily temperature (in ˚C) and relative humidity (in %) were computed as the 24-hour average based on hourly measurements. Measures of carbon monoxide (CO, in ppb), ozone (O3, in ppb), nitrogen dioxide (NO2, in ppb), particles (PM10, in ppb) and sulphur dioxide (SO2, in ppb) were available for the same period. Fine particles measures (PM2.5, in ppb) were available only in [2005][2006][2007]. Daily level of pollutants were computed as the 24-hour mean based on hourly measurements. Data were pooled from 1 meteorological station and 11 air quality monitoring stations in Kaohsiung, 2 meteorological station and 5 air quality monitoring stations in Taichung, and 3 meteorological station and 15 air quality monitoring stations in Taipei, respectively. In total, missing data amount for 0.03% and 0.00% of the mortality and temperature series, respectively.

Thailand
We collected data from 62 provinces (see full list in Table S2 below) between 1 st of January 1999 and 31 st of December 2008. Daily mortality, obtained from the Ministry of Public Health, Thailand, is represented by counts of deaths for non-external causes only (ICD-9: 0-799; ICD-10: A00-R99). Mean daily temperature (in ˚C) and relative humidity (in %), computed as the average between daily minimum and maximum, were obtained from the Meteorological Department, Ministry of Information and Communication Technology, Thailand. A total of 117 weather stations in 62 provinces, with at least one weather monitoring station in each province. In total, missing data amount for 0.00% and 6.10% of the mortality and temperature series, respectively. Humidity measurements were missing in at least 10% of days in 12 provinces.

UK
We collected data in 9 regions of England and in Wales (see full list in Table S2 below) between 1 st of January 1993 and 31 st of December 2006. Daily mortality, obtained from the Office of National Statistics, is represented by counts of deaths for all causes and for non-external causes only (ICD-9: 0-799; ICD-10: A00-R99). Mean daily temperature (in ˚C) and relative humidity (in %), computed from the 24-h average of hourly measurements) were obtained from the British Atmospheric Data Centre. An average of 29 stations contributed data to each regional series, from a minimum of 7 in London to a maximum of 44 in Wales. In total, missing data amount for 0.00% and 0.00% of the mortality and temperature series, respectively. These data were used and described in previous publications. 10,11 USA We collected data from 135 cities (see full list in Table S2 below) between 1 st of January 1985 and 31 st of December 2006. Daily mortality, obtained from the National Center for Health Statistics (NCHS), is represented by counts of deaths for non-external causes only (ICD-9: 0-799; ICD-10: A00-R99). Mean daily temperature (in ˚C, computed as the 24-hour average based on hourly measurements) and relative humidity (in %, computed from the 24-h average of hourly measurements of dew point temperature) were obtained from the National Climatic Data Center (NCDC) of the National Oceanic and Atmospheric Administration (NOAA). A single weather station was selected for each city in the land-based station data or NCDC, based on the proximity to the city's population centre. In 6 cities where multiple observations were missing from all the nearby monitors, hourly data from the Integrated Surface Database Lite of NCDC were converted in daily values. For 25 stations missing dew point data, dew point data were obtained from the nearest station with dew point data. In total, missing data amount for 0.32% and 1.89% of the mortality and temperature series, respectively. These data were used and described in previous publications. 12,13 Additional information on the statistical methods

Details on the computation of the attributable risk
The mortality risk attributable to a temperature x t for a given day t in the series is defined as the number AN x,t and fraction AF x,t of deaths experienced in the next L days, with L as the maximum lag period, defined by: and AN x,t = AF x,t • ∑ n t+l L + 1 L l=0 with ∑ β x t ,l as the overall cumulative log-relative risk for temperature x t in day t, and n t as the number of deaths in day t. To be noted how the number of attributable deaths AN x,t is computed as the fraction of the average mortality in the future L days. The risk estimate ∑ β x t ,l is obtained by the BLUP of the overall cumulative exposure-response association, re-centered on the temperature of minimum mortality (MMT). The MMT is therefore the counterfactual condition for the definition of the attributable risk. Therefore, the attributable risk can be interpreted as the excess deaths due to non-optimal temperature, if compared to a hypothetical situation in which temperature is constantly equal to the MMT.
The total attributable number of deaths AN tot due to temperature is given by the sum of AN x,t for all the days t of the series, and its ratio with the total number of deaths provides the total attributable fraction AF tot . The components attributable to cold and heat are computed by summing the subsets of AN x,t lower or higher than the temperature corresponding to the location-specific MMT. These components are further separated in moderate and extreme contributions by selecting low and high temperature cut-offs. Here extreme cold and heat are defined as the temperatures lower than the 2.5 th and higher than the 97.5 th city-specific percentiles, respectively. Moderate temperatures are instead defined as the ranges between the MMT and these cut-offs. Other cut-offs are used for a further stratification for different temperature ranges in Table S3.
The method is described in details in a previous publication. 14