The mortality burden attributed to regional indoor temperatures in China

Publication Information

Title:The mortality burden attributed to regional indoor temperatures in China

Author(s):Zhongguo Huang, Jianxiong Hu, Jinghua Gao, Min Yu, Mengen Guo, Ruilin Meng, Chunliang Zhou, Yize Xiao, Biao Huang, Jiangmei Liu, Maigeng Zhou, Ryan J. Gainor, Ramune Reliene, Guanhao He, Tao Liu, Wenjun Ma

‌Journal Name, Year, Volume(Issue): Page range: 

Environment International, 2025, 204: 109822

DOI:10.1016/j.envint.2025.109822

Abstract:

Background: Despite predominant indoor occupancy patterns, mortality risks and burdens associated with regional (county/district level) indoor temperature remain underexplored in epidemiological research.

Objective: To construct a reliable regional indoor temperature prediction model and estimate the disease burden attributed to non-optimal regional indoor temperature.

Design: The outdoor meteorological parameters were from the Fifth Generation European Reanalysis dataset, while regional determinants were from national statistical yearbooks. Indoor temperature and building characteristics were collected from 99 buildings across 33 cities. Employing a random forest (RF) algorithm, we developed a prediction model of regional indoor temperatures based on outdoor meteorological parameters and regional determinants for building characteristics. Subsequently, we estimated the regional temperature-mortality associations for both indoor and outdoor temperatures using a distributed lag non-linear model (DLNM) based on cause-specific mortality data collected from 364 counties/districts in China during 2006-2017. Finally, we compared the temperature-related mortality burdens associated with both indoor and outdoor temperature.

Results: The RF algorithm identified outdoor meteorological parameters (temperature, relative humidity, wind speed, and precipitation) and regional determinants (green space, latitude, longitude, education attainment, penetration rate of air conditioner, and seasonal variation) as primary determinants of regional average indoor temperature, whereas building characteristics exhibited limited influence. The developed prediction model demonstrated superior predictive accuracy with performance metrics including a root mean square error (RMSE) of 1.473°C, mean absolute error (MAE) of 1.034°C, and R² value of 0.938. Analysis of 6.5 million non-accidental death records revealed consistent inverse J-shaped associations for both regional indoor and outdoor temperature-mortality relationships, with indoor temperature demonstrating greater mortality risks. Comparative assessment showed higher temperature-attributable fractions for indoor exposure (18.09%, 95%CI: 17.87-18.31%) versus outdoor exposure (14.46%, 95%CI: 14.41-14.52%), particularly notable for heat-related mortality burden (indoor: 8.38% vs outdoor: 3.66%).

Conclusions: Meteorological parameters and regional determinants emerged as primary predictors of indoor temperature. Regional indoor temperature exposure exhibited greater mortality risks and burden compared to regional outdoor temperature, particularly during heat condition.

First Page of Article: