Attributable mortality risk of temperature: a multi-country study

Gasparrini, Antonio

Attributable mortality risk of temperature: a multi-country study

Thursday, 21 August 2014: 9:00 AM

Ballroom C (Dena'ina Center)

Antonio Gasparrini, PhD , London School of Hygiene and Tropical Medicine, London, United Kingdom

Michela Leone , Lazio Regional Health Service, Rome, Italy

Yuming Guo , University of Queensland, Herston, Australia

Eric Lavigne , University of Ottawa, Ottawa, ON, Canada

Antonella Zanobetti , Harvard School of Public Health, Boston, MD

Joel Schwartz , Harvard School of Public Health, Boston, MD

Aurelio Tobias , Spanish Council for Scientific Research (CSIC), Barcelona, Spain

Shilu Tong , Queensland University of Technology, Brisbane, Australia

Michelle L Bell , Yale University, New Haven, CT

Yue-Liang L Guo , National Taiwan University, Taipei, Taiwan

Chang-fu Wu , National Taiwan University, Taipei, Taiwan

Haidong Kan , Fudan University, Shanghai, China

Seung-Muk Yi , Seoul National University, Seoul, South Korea

Masahiro Hashizume , Nagasaki University, Nagasaki, Japan

Yasushi Honda , University of Tsukuba, Tsukuba, Japan

Ho Kim , Seoul National University, Seoul, South Korea

Ben Armstrong , London School of Hygiene and Tropical Medicine, London, United Kingdom

INTRODUCTION: while few studies investigated the attributable mortality risk for either heat or cold in selected countries, none so far has provided estimates for the whole temperature range in different climates.

METHODS: we collected data for 326 cities in Australia (1988-2009), Canada (1986-2009), China (1996-2008), Italy (1987-2010), Japan (1972-2009), Korea (1992-2010), Spain (1990-2010), Taiwan (1994-2007), Thailand (1999-2008), UK (1993-2006), and USA (1985-2009), totalling over 48 million deaths. A standard time series Poisson model was fit in each city controlling for trend and day of the week. The temperature-mortality relationship was estimated with a distributed lag non-linear model through a bi-dimensional spline, then reduced to the overall risk cumulated over lag 0-21. City-specific best linear unbiased predictions were computed from a multivariate meta-analytical model. Attributable risk were calculated for heat and cold, defined as temperatures above and below the point of minimum mortality.

RESULTS: temperature is attributed in total 6.31% (95%CI:6.05-6.50%) of mortality, with substantial inter-country variation, from 3.3% in Thailand to 11.3% in China. The temperature percentile of minimum mortality varies from around 60^th in (sub)tropical countries (Thailand and Taiwan) to around 80^th-90^th in the other countries. Most of the attributable deaths are due to cold, with a fraction of 5.90% (5.65-6.09%), if compared to 0.41% (0.37-0.44%) due to heat, a ratio relatively stable across countries. Sensitivity analyses show that the estimates are robust to modelling choices and confounding control.

CONCLUSIONS: this represents by far the largest study on temperature-mortality associations. While most of previous research has focused on effects of heat, a large part of the mortality burden appears to be attributable to the contribution of cold. This evidence has important implications to the planning of public health interventions to prevent the health consequences of temperature, and to predict future impact under climate change scenarios.