The burden of heat in arid regions of the Middle East: an analysis from Jordan and Kuwait

The Middle East, with its vast arid landscape, is facing escalating health risks due to intensifying heat under climate change. Kuwait and Jordan, two representative countries from the region, have no heat action plan in place. This study aims to (1) quantify the mortality burden of extreme heat in these countries, and hence (2) identify critical temperature thresholds. We collected 17 years of daily mortality records from 2000 to 2016 in Amman, Jordan, and the entire state of Kuwait. We fitted a time series design restricted to the summer months (June–August) for each location. We used distributed lag non-linear models to estimate non-linear associations and lagged effects of temperature on mortality. We then calculated attributable mortality for a range of temperature percentiles. We analyzed a total of 56 654 (39 996 all-cause deaths in Amman and 16 658 non-accidental deaths in Kuwait). Kuwait’s average summer temperature (38.7 °C) was higher than Amman (26.5 °C). In Kuwait, 202.1 (95% eCI: 17.7, 344.8) attributable heat deaths occurred over 79 d at temperatures above 41.9 °C (>95th percentile), averaging 2.6 heat deaths per day. Amman experienced 500.7 (95% eCI: 17.7, 344.8) attributable heat deaths over 77 d at temperatures above 30.75 °C, with an average of 6.5 deaths per day. This study equips Kuwait and Jordan with critical data to develop and implement targeted heat action plans. The two Middle Eastern countries face extreme heat challenges and are undergoing serious demographic changes with an influx of migrant workers and refugees. We provide an essential quantification of the mortality burden attributable to extreme heat in Kuwait and Jordan, two countries in the Middle East facing unprecedented heat and yet without existing heat action plans. Policymakers face the question of which thresholds represent negligible risk and which cannot be ignored. By analyzing 56 654 deaths, we reveal the number of deaths per day attributable to specific temperatures. This study presents the first step towards establishing targeted heat action plans to mitigate the health risks posed by intensifying heat. The methods used for Kuwait and Amman can be replicated in similar Middle Eastern countries.


Introduction
The Middle East, encompassing approximately 7 million km 2 and home to over 490 million people [1], is characterized by its arid and hyper-arid regions.These conditions are increasingly critical in the context of global climate change, with the region bracing for more frequent and severe heat, posing significant threats to human health and survivability [2][3][4].Despite these challenges, the region faces a chronic shortage of population health data.
We collected multi-decadal daily mortality data in two, what we believe are, representative areas with arid climates in the Middle East.Kuwait and Jordan, with their harsh climatic conditions, can serve as key case studies for understanding the broader impacts of extreme heat in arid regions.Not just that, but the two countries are undergoing serious demographic changes.
Kuwait, known for its scorching summers and status as a major petroleum producer, economic prosperity has attracted millions of migrant workers, significantly altering its demographic landscape.Simultaneously, Jordan, traditionally a land of stability amidst regional turmoil, has experienced a demographic shift due to the influx of refugees from conflict zones like Palestine, Syria, and Iraq.Although these changes in population dynamics in Kuwait and Jordan arise from different circumstances, they collectively increase the vulnerability of these populations to the health risks that can be posed by extreme heat [5].
In this paper, we (1) quantified the burden of hot summers on mortality, and (2) identified possible and locally relevant thresholds for extreme summer heat using 17 years of mortality records in Kuwait and Amman, Jordan.

Data sources
We collected daily all-cause mortality records for the city of Amman, Jordan, spanning from 1 January 2000, to 31 December 2016, from the Jordanian Civil Status and Passport Department.Cause specific mortality data was not available for Amman, and the dataset contained two columns: date and all-cause mortality count from all ages and demographics.Concurrently, daily maximum and minimum temperature data for the Amman station were obtained from the Jordanian Ministry of Water and Irrigation for the same period.Average temperature (in • C) was calculated as the mean of the maximum and minimum for any given day.Station humidity data was not available for Amman.
Kuwait can be described as a city country.For the same study period, we collected daily non accidental mortality records from the National Center for Health Information, Department of Vital Statistics, Ministry of Health, Kuwait.The data we obtained included all ages and demographics, but excluded all external causes and was limited to non-accidental mortality A00 to R99 according to the International Classification of Diseases tenth version.While Kuwait's mortality data was cause specific and had individual characteristics information, they were not used to keep the analysis consistent with Amman's data.Daily ambient temperatures (24 h average in • C) and relative humidity (24 h average in %) data were obtained from the meteorological services in Kuwait Airport for the same study period.
Basic data checks were performed on the temperature dataset, including logic checks and tests to ensure data homogeneity, specifically verifying that the daily maximum temperature was greater than the daily minimum temperature.These steps were taken to identify and address any apparent outliers or irregularities in the temperature records [6].No extensive data homogenization techniques were applied to the temperature dataset.
For context, according to the World Bank in 2022, Jordan has a population of approximately 11.3 million and a GDP per capita of $4311, compared to Kuwait, which has a smaller population of around 4.3 million but a significantly higher GDP per capita of $41 079.

Statistical analysis
We used a time series design, restricting to the hottest summer months from June to August each year.For each country, we fitted separate negative binomial regression models with mortality counts as the outcome [7].Conventionally, the relationship between mean temperature and mortality was modeled using distributed lag non-linear models (DLNMs).The DLNM function is a flexible modeling framework that estimates non-linear associations with delayed effects [8].Based on previous multi-country evidence of summer only DLNM-time series analyses choices of multi-country, we used a natural spline with two degrees of freedom for the exposure-response dimension placed at the 50th and 90th percentile, with an extended lag of 10 d modeled using a natural spline with two degrees of freedom spaced equally in the log scale [9].To account for the time trend and seasonality, we applied natural splines with 3 degrees of freedom per summer, and controlled for day of the week (categorical).The minimum mortality temperature (MMT) was defined as the temperature that is associated with the least mortality in the summer period.Relative risks of death with 95% confidence intervals were reported for given temperatures compared to the MMT.
The attributable mortality from a range of percentiles were calculated using the following equation [10]: where for temperature x t in a given day summer day t, the attributable number AN x,t of deaths are those experienced in the next l lag days and L as the maximum lag period, and ∑ β xt,l is the overall cumulative log-relative risk for temperature x t and n t is the number of deaths in day t [10].We calculated empirical confidence intervals (eCI) using Monte Carlo simulations assuming a multivariate normal distribution of the coefficients.This method allows for a more flexible and data-driven approach by repeatedly sampling from the estimated distribution of simulated model coefficients.
As a sensitivity analysis, we adjusted for relative humidity in the temperature-mortality models in Kuwait.All statistical analyses were performed using R statistical software version 4.1.2.The dlnm package [11] was used for non-linear delayed ambient temperature effects.

Results
A total of 56 654 mortalities were analyzed, consisting of 39 996 all-cause deaths in Amman and 16 658 all-cause non-accidental deaths in Kuwait.The mortality data was examined during the summer seasons, from June 1st to August 31st, for 17 years from 2000 to 2016.This analysis included a total of 1561 summer days in both countries.
Distribution of summer temperatures in Amman and Kuwait is shown in figure 1.The mean average summer temperature in Kuwait was notably higher at 38.7 • C, whereas Amman recorded a lower mean average summer temperature of 26.5 • C. The 99th percentile average summer temperature in Kuwait reached 42.9 • C, surpassing Amman's corresponding value of 33.3 • C.
The overall exposure-lagged relationships between temperature and mortality are shown in figure 2. In Amman, the MMT was 22.9 • C, while in Kuwait, the MMT was 40 • C. Very hot days in Amman and Kuwait were associated with increased relative risks of death.
In table 1, we present attributable mortalities from summer temperature in Amman and Kuwait from 2000 to 2016.Notably, different temperature percentiles corresponded to different temperature ranges in each location.Over 17 summers, 79 and 77 d exceeded the 95th percentile of summer temperatures in Kuwait and Amman, respectively.Other moderately hot summer temperatures (between 60th and 70th percentile) corresponded to 157 in Kuwait and 150 d in Amman.
Similarly, the relative risk of death increased with higher temperature percentiles (table 1).At the 65th percentile (mid-range of the 60th-70th percentile), the relative risk of death was 1.00 (95% CI: 0.97, 1.04) and 1.03 (95% CI: 0.97, 1.09) in Kuwait and Jordan, respectively, compared to the summer MMT in the respective locations.At extreme percentiles (mid-range of >95th), the relative risk increased to 1.25 in both countries.Adjusting for relative humidity in Kuwait did not change the results.

Discussion
We analyzed tens of thousands of death cases in two arid Middle Eastern countries and provided new and critical insights into the burden of hot summer days on mortality.To date, Kuwait and Jordan have no heat warning systems nor action plans to protect population health despite extremely harsh conditions and an influx of vulnerable migrant workers and refugee communities in the two countries.In this analysis, we showed attributable heat deaths per day and across 17 summers with potential cutoffs that provide evidence-based actionable information for policymaking in the two countries.In Kuwait, when average summer temperatures exceeded 41.9 • C, an average of 2.6 deaths per day were attributed to heat.Similarly, in Jordan, 6.5 heat deaths per day were attributed to days with summer temperatures above 30.8• C. The warming of our planet is not evenly distributed [12].In certain regions, temperatures are rising at a faster pace and setting record-high levels [2,13,14].Recent evidence is now suggesting that areas in the Middle East could be facing significant risks to maintain human survivability due to climate change [3,4,15].One projection for the Middle East and North Africa region estimated an eight to 20 times increase in mortality rates in the last 30 years of the 21st century with respect to the historical period (1951-2005) [16].Kuwait, commonly known as one the hottest country in the world, registered one of the highest temperatures on record; a 54.0 • C in the summer of 2016 [17].In recent years, Kuwait summer days are commonly exceeding 40 • C and 50 • C [18].Jordan, on the other hand, known for its historically moderate temperatures, has been experiencing increasingly intense hot days in recent years.Jordan recorded temperatures in excess of 44.8 • C in the summer of 2023, marking one of the highest temperatures ever documented in the country [19].The rising frequency of days exceeding 35 • C and occasionally reaching 40 • C or more in Jordan's summers is an indication of the changing climate patterns affecting the region.All in all, extreme temperatures are particularly concerning for Middle Eastern countries who also grapple with lack of data-driven evidence of heat mortality.
In previous analyses within the region, particularly in Kuwait, there was evidence indicating a significant increase in mortality risks-up to double or triple-at the most extreme temperature ranges (99th percentile) [5,20,21].However, while these relative risk figures are valuable from an epidemiological standpoint, they offer limited practical utility for burden analysis.Such coefficients do not sufficiently address the need to characterize risk in a context where the costs and benefits of potential interventions must be carefully balanced.In contrast, our paper advances this regional knowledge by providing two pieces of information: (1) defining specific summer temperature thresholds, and (2) quantifying the actual burden of heat mortality.
Quantifying the burden of heat deaths in terms of deaths per day for different temperature ranges, as demonstrated in our study, is central for economic valuation.This approach allows for the identification of summer days where public health interventions could yield the most significant benefits.It can also be used to inform tools to assess how much society is willing to pay to reduce the risk of death.For instance, if a notable-potentially actionable-increase in mortality is observed at temperatures above the 90th percentile, resources can be concentrated on days when such temperatures are forecasted.This targeted allocation of resources ensures that interventions, such as opening cooling centers or issuing heat warnings, are activated precisely when most needed, optimizing their effectiveness and cost-efficiency.
The different impact observed in Kuwait and Amman underscores the influence of contextual and local climatic conditions on health outcomes.While both cities are located in arid regions, the distinct temperature profiles and associated health risks highlight the need for country-specific strategies to mitigate the impact of heat on public health.The higher summer threshold for increased mortality risk in Kuwait (40 • C) compared to Amman (22.9 • C) suggests differing adaptation levels to local climates, possibly influenced by socio-economic and infrastructural factors such as air conditioning prevalence [22] and urban characteristics [23].In this context comes the importance of establishing localized heat action plans [24].These include heat warning systems, public awareness campaigns, and targeted strategies to protect vulnerable populations, such as outdoor workers and communities in rural areas and refugee camps.This analysis, while offering valuable insights, is constrained by several limitations.First, the comparison between Amman and Kuwait is not like-for-like.Amman represents a single city, albeit Jordan's largest, whereas the mortality data from Kuwait comes from the entire country, albeit is relatively small.Additionally, Kuwait's mortality data excludes accidental deaths (e.g.external causes), whereas Jordan's mortality data includes all deaths.Also, we limited our analysis to aggregated death counts and were not able to separate or explore effect measure modification by cause of death, age, sex, and other demographics, since these strata were not available in Amman.Secondly, our approach to modeling the data from these two locations was guided by predetermined statistical choices, as carried out by similar multi-city analyses [9].This means the model we selected a priori may not necessarily be the best fitting.Alternative models incorporating different lag times or exposure knots might yield more accurate results.Thirdly, our models were not adjusted for other environmental exposures such as relative humidity and air pollution, as these variables were not available in Jordan.Our study relied on central site temperature data.The study areas covered by the central site temperature data vary, with an approximate radius reaching tens of kilometers.This approach may not capture localized temperature variations within each study area, potentially affecting the accuracy of our population exposure assessments.Finally, the calculation of attributable mortality comes with its own set of uncertainties, including some values that fell below zero.While these negative values warrant consideration, we believe they should not completely deter decision-making.It is rather prudent to interpret these findings within the context of their limitations.

Conclusion
Middle Eastern countries are grappling with the escalating challenge of extreme heat.In Kuwait and Jordan, two arid nations characterized by harsh climates and heightened demographic vulnerabilities, the toll of summer temperatures on mortality rates is too high for inaction.This new data equips Kuwait and Jordan with the critical information needed to develop and implement their own targeted heat action plans, addressing this urgent public health concern.

Figure 1 .
Figure 1.Summer temperature distribution in Amman and Jordan in the study period.

Figure 2 .
Figure 2. Two dimensional plots of relative risk (with 95% confidence intervals-shaded area) of mortality risk and summer temperatures in Amman and Kuwait (2000-2016).The curves are centered at the minimum mortality temperature (MMT) which was 22.9 • C and 40 • C in Amman and Kuwait, respectively.

Table 1 .
Heat thresholds and attributable heat deaths across 17 summers (June-August) in Amman and Kuwait from 2000 to 2016.