Comparison of Temperature–Mortality Associations across the Middle East Using Different Exposure Estimation Approaches

BACKGROUND: Traditional temperature–health
studies have predominantly relied on temperature measurements from
stations or modeled spatial averages from gridded temperature data
sets. It has been suggested that population-weighted spatial averages
would perform better in remote regions with large temperature and
population variability. This would be particularly true in regions
other than North America and Europe where outcome data are often only
available on a crude spatial scale, but no studies have examined this
in such regions, where temperatures can be particularly hot. OBJECTIVE: Using the Middle East as a climate hotspot, our
objective was to illustrate the utility of population weighting temperature
exposures in understudied regions with large health data aggregation
areas. METHODS: We used a daily 1 km × 1 km temperature
data set for 152 administrative regions in 12 Middle Eastern countries.
From 2003 to 2020, for each administrative region, we computed daily
minimum and maximum population-weighted and unweighted spatial average
temperatures. To illustrate, we examined temperature–mortality
associations in two countries: Kuwait and Jordan. We used distributed
lag nonlinear models to estimate the daily time series temperature–mortality
associations in using three temperature exposure measurement approaches:
station temperatures, unweighted spatial averages, and population-weighted
temperatures. For each scenario, we fitted country-specific optimized
parameters and compared them using three metrics: 1) exposure–response
relationships, 2) minimum mortality temperatures, and 3) attributable
mortality estimates. RESULTS: The study region had geographically
sporadic yet densely populated areas within each country. In both
Kuwait and Jordan, population-weighted and unweighted spatial average
temperatures resulted in fairly similar exposure–response curves,
whereas both were notably different from station temperatures. Minimum
mortality temperatures were 30.2, 28.6, and 28.3 °C in Kuwait
for station, unweighted spatial average, and population-weighted temperatures,
respectively. In Jordan, the corresponding temperatures were 20.6,
20.9, and 20 °C. Heat attributable mortality calculated using
population-weighted temperatures increased by 15% compared to the
traditionally used station temperatures in Kuwait and Jordan, respectively,
and −0.4% and 5% compared to unweighted spatial average temperatures. DISCUSSION: Spatial averaging, whether weighted or unweighted,
is a valuable tool for estimating heat-attributable mortality. This
is especially true in regions like the Middle East, where granular
temperature data are often unavailable and health studies are urgently
needed. Population-weighted temperatures may better capture localized
exposures in areas with significant population clustering, though
their exact added effect on top of unweighted spatial averages remains
a tentative conclusion.