1. Introduction

Excess mortality, the difference between the number of deaths observed during a crisis and the number expected had the crisis not occurred, has become the reference measure of the population impact of the COVID-19 pandemic [1-3]. Early multi-country analyses showed that the first wave alone produced large and heterogeneous mortality increases [4]. Excess mortality captures deaths that were not certified as COVID-19 because of limited testing or variable certification practice, deaths caused indirectly by disrupted health services, and deaths averted by reduced mobility or injury [5,6]. Global analyses estimated that excess deaths in 2020 and 2021 were roughly three times the number of COVID-19 deaths reported to the World Health Organization (WHO), with wide variation between countries and regions [2,3,7].

Countries of the Gulf Cooperation Council (GCC) occupy an unusual position in this literature. Their populations are young, with large expatriate workforces that often live in shared accommodation and work in essential occupations; nationals and migrants differ in age structure and access to care, and population size changed abruptly during the pandemic as workers left and returned [8-10]. National studies documented excess mortality in Oman during the first months of 2020 [11], a disproportionate excess among migrant workers in Kuwait in 2020 [9], and low non-COVID-19 all-cause mortality but elevated COVID-19 mortality among craft and manual workers in Qatar [10]. Comparable work in neighbouring Arab countries reported a 21% rise in standardised mortality in Jordan in April to December 2020 [12] and surveillance signals from civil registration in Egypt [13]. Other regional work has described COVID-19 deaths from hospital and surveillance records, their predictors and their economic cost [14-18], or modelled transmission [19]. Most of these studies end in 2020 or 2021, use different baselines and do not compare countries.

Two methodological questions complicate cross-country comparison. First, excess mortality depends on the model used to project expected deaths: the choice of reference years, trend specification and model family can change estimates materially, sometimes reversing conclusions [20-23]. Second, extending estimates beyond 2021 requires extrapolating pre-pandemic trends further, which raises the question of whether any later excess reflects lingering pandemic effects, population change or baseline drift [24-26]. Machine-learning forecasters are increasingly proposed for baseline estimation, yet their performance relative to simple statistical models on short annual series remains uncertain [27,28].

Saudi Arabia and Bahrain do not publish monthly all-cause death counts in the openly available international compilations, so a GCC-wide analysis must rely on the member states that do. We therefore aimed (i) to estimate cumulative and annual excess all-cause mortality from 2020 to 2023 in Kuwait, Oman, Qatar and the United Arab Emirates, with Egypt, Jordan and Tunisia as Arab comparators; (ii) to describe the timing of excess across countries; (iii) to compare excess deaths with officially reported COVID-19 deaths; and (iv) to evaluate, by out-of-sample back-testing on pre-pandemic years, how statistical and machine-learning baselines perform on these series. We also report 2024 as a check on whether mortality returned to the projected baseline.

2. Materials and Methods

2.1. Study Design

This was a multi-country population time-series study with the country-month as the unit of analysis. The design is ecological: exposure (the pandemic period) and outcome (all-cause deaths) are defined for whole national populations, and the estimand is a population-level contrast between observed and expected deaths, so inferences apply to countries rather than individuals [29]. The statistical analysis plan, including the primary model, secondary outcomes and sensitivity analyses, was written and frozen before excess deaths were computed; one implementation correction to a comparator model was made before drafting and is described in Section 2.6.

2.2. Data Sources

Monthly all-cause death counts were taken from the World Mortality Dataset (WMD), an openly maintained compilation of national vital statistics described by Karlinsky and Kobak [1], in the version available on 6 November 2026. The WMD records the source of each national series: Kuwait from the Central Statistical Bureau, Oman from the National Centre for Statistics and Information, Qatar from UN data (2015 to 2019) and the Planning and Statistics Authority (from 2020), the United Arab Emirates from UN data, Egypt from UN data (2015 to 2019) and the Central Agency for Public Mobilization and Statistics (from 2020), Jordan from the Civil Status and Passports Department, and Tunisia from the National Institute of Statistics. Officially reported COVID-19 deaths were obtained from the Our World in Data compilation of WHO dashboard data [30,31] and summed by calendar month. Mid-year population totals came from the World Bank World Development Indicators (indicator SP.POP.TOTL). All files were downloaded on 6 November 2026, hashed and archived unchanged.

2.3. Countries, Period and Eligibility

Countries were eligible if they belonged to the GCC or were Arab League members in the Middle East or North Africa, and had complete monthly series from no later than 2018 to December 2024 in the WMD. Kuwait, Oman, Qatar and the United Arab Emirates met these criteria among GCC states; Saudi Arabia and Bahrain lacked sub-annual data. Egypt, Jordan and Tunisia were the eligible comparators; Lebanon (series ending in 2021) and Iran (non-Arab, weekly series ending in 2022) were excluded a priori. The reference period was January 2015 to December 2019 (January 2018 to December 2019 for the United Arab Emirates; Jordan had no 2016 data), the pandemic period was January 2020 to December 2023, and 2024 was analysed separately. In total, 792 country-months were analysed, 372 of them in the reference period.

2.4. Variables and Outcomes

The primary outcome was cumulative excess deaths from January 2020 to December 2023 in each country, defined as the sum over months of observed minus expected deaths. It was expressed as a count with a 95% prediction interval (PI), as a P-score (excess as a percentage of expected deaths) and as excess deaths per 100,000 population, using the mean World Bank population over 2020 to 2023 as denominator. Secondary outcomes were annual excess and P-scores for 2020 to 2024; monthly P-scores, the month of peak P-score and the number of months in which observed deaths exceeded the upper limit of the 95% PI; the ratio of cumulative excess to reported COVID-19 deaths; and pooled estimates for the four GCC states and the three comparators.

2.5. Statistical Analysis

Time-series regression of aggregate counts is the standard approach for such questions [32,33]. The primary baseline followed the specification used for the WMD [1]: an ordinary least squares regression of monthly deaths on a linear term for calendar year and twelve month-of-year indicators, fitted separately for each country to the reference period and projected forward. For the United Arab Emirates, with only two reference years, the year term was dropped and expected deaths equalled the month-specific means, as prespecified. Uncertainty was quantified by simulation with 10,000 draws: regression coefficients were drawn from their sampling distribution (multivariate t with residual degrees of freedom) and independent normal residual noise with the estimated variance was added to each projected month. The 95% PI of any sum (annual or cumulative excess, pooled groups) was taken from the 2.5th and 97.5th percentiles of the simulated sums; pooled intervals combined independent country draws.

To evaluate baseline choice we performed a rolling-origin out-of-sample test on pre-pandemic data only [34,35]: each model was trained on 2015 to 2017 and used to forecast 2018, then trained on 2015 to 2018 to forecast 2019. The United Arab Emirates was excluded from back-testing because its series begins in 2018, leaving 12 country-origin pairs. Six baselines were compared: (M1) the primary linear trend and month model; (M2) a negative binomial generalised linear model with log link, linear trend and month indicators; (M3) a seasonal autoregressive integrated moving average (SARIMA) model with orders selected by the Akaike information criterion over a restricted grid (non-seasonal p and q from 0 to 2, d of 0 or 1, and seasonal terms of none, one autoregressive term or one seasonal difference, period 12), with a linear trend or drift; (M4) Holt-Winters exponential smoothing with damped additive trend and additive seasonality; (M5) gradient-boosted regression trees (XGBoost [36]) using a time index and calendar month as features; and (M6) the equal-weight mean of M1 to M5. Accuracy was summarised by the mean absolute percentage error (MAPE) of monthly forecasts and the absolute percentage error of the annual total. Back-testing informed interpretation and did not change the primary model.

Sensitivity analyses re-estimated the cumulative 2020 to 2023 excess (i) with each alternative baseline refitted to the full reference period; (ii) with a population-rate baseline, a quasi-Poisson model with an offset for the logarithm of annual population, trend and month terms; (iii) without a trend term for all countries; (iv) with a reference period restricted to 2017 to 2019; and (v) for Jordan, excluding September to December 2019, months that were unusually high relative to earlier years. Robustness was judged by the stability of the sign of cumulative excess and of the ranking of P-scores (Spearman correlation with the primary ranking). The association between excess and reported COVID-19 deaths per 100,000 across the seven countries was described with Spearman's rank correlation; with seven units this was descriptive. No formal hypothesis tests were planned and no multiplicity adjustment was applied.

Analyses used Python 3.13 with statsmodels [37], scikit-learn and XGBoost. The primary point estimates and prediction intervals were recomputed independently with a different implementation (formula-based regression with an analytic t-based prediction variance for the sum of months). Point estimates agreed exactly; interval bounds agreed to within 2.1% of the interval width, the small difference reflecting the distributional approximation used by each method.

2.6. Protocol Deviation

In the first implementation of the SARIMA comparator, differenced specifications included a time trend that, after differencing, implies a quadratic trend in levels and produced implausible projections. This was corrected to a drift term (a linear trend in levels), as the plan intended, before the manuscript was drafted. The primary model was unaffected.

2.7. Reporting Guideline

Reporting follows the STROBE statement adapted for ecological studies [38] and, for the estimates, the Guidelines for Accurate and Transparent Health Estimates Reporting (GATHER) [39]. A completed checklist is provided as supplementary material.

2.8. Ethics

The study used publicly available, aggregated national statistics containing no individual-level or identifiable information. Ethical approval and informed consent were therefore not required.

3. Results

3.1. Data and Expected Mortality

Table 1 summarises the seven series. Over 2020 to 2023 the four GCC states registered 135,359 deaths in a combined mean population of 21.8 million, and the three comparators registered 3,079,440 deaths in 135.1 million. The primary model reproduced the seasonal pattern of the reference period in each country, with winter peaks most pronounced in Egypt, Jordan and Tunisia and flatter seasonality in the GCC states (Figure 1).

Table 1. Characteristics of the national series and deaths observed, expected and reported as COVID-19, January 2020 to December 2023.

Expected deaths from the primary linear trend and month model fitted to the reference years (month means without trend for the United Arab Emirates). Reported COVID-19 deaths are WHO-reported counts compiled by Our World in Data. GCC, Gulf Cooperation Council; WHO, World Health Organization.

Figure 1. Observed monthly all-cause deaths and expected deaths with 95% prediction intervals, 2015 to 2024 (dotted line, January 2020). Observed deaths repeatedly exceeded the expected range in 2020 and 2021 in every country.

3.2. Cumulative Excess Mortality, 2020 to 2023

All seven countries recorded more deaths than expected over 2020 to 2023 (Table 2, Figure 2). Pooled across the four GCC states, cumulative excess was 21,692 (95% PI 18,760 to 24,660) deaths, a P-score of 19.1% (95% PI 16.5 to 21.7) and 99.6 excess deaths per 100,000 (95% PI 86.1 to 113.2). Among GCC states the P-score was highest in the United Arab Emirates (27.9%; 95% PI 26.1 to 29.8), followed by Qatar (20.9%; 95% PI 14.9 to 27.0), Kuwait (20.5%; 95% PI 15.9 to 25.1) and Oman (9.4%; 95% PI 3.1 to 15.7). Pooled excess in the comparators was 400,840 (95% PI 264,980 to 538,942), a P-score of 15.0% (95% PI 9.9 to 20.1), with Egypt at 16.1% (95% PI 10.0 to 22.2), Jordan at 13.9% (95% PI 6.7 to 21.2) and Tunisia at 7.6% (95% PI 2.6 to 12.6).

Because GCC populations are young, the excess per 100,000 population ranked countries differently from the P-score. Excess per 100,000 was highest in Egypt (322), Tunisia (202) and Jordan (145), intermediate in Kuwait (138) and the United Arab Emirates (101), and lowest in Oman (77) and Qatar (70). The pooled GCC rate (100 per 100,000) was about one third of the comparator rate (297 per 100,000).

Table 2. Cumulative excess all-cause mortality, January 2020 to December 2023, primary model.

P-score, excess deaths as a percentage of expected deaths; per 100,000 uses the mean World Bank population for 2020 to 2023. Pooled intervals combine independent country simulations. PI, prediction interval; GCC, Gulf Cooperation Council.

Figure 2. Cumulative excess mortality, 2020 to 2023: (A) P-score and (B) excess deaths per 100,000, with 95% prediction intervals. GCC states had higher relative excess but lower excess per head than the Arab comparators.

3.3. Timing and Annual Pattern

Excess was concentrated in 2020 and 2021 (Table 3, Figure 3). Observed deaths first exceeded the upper prediction limit in March or April 2020 in Qatar, Kuwait and the United Arab Emirates, in May and June 2020 in Egypt and Oman, and in October 2020 in Jordan and Tunisia. The largest monthly P-score occurred in May 2020 in the United Arab Emirates (125%) and June 2020 in Egypt (83%), whereas in Jordan (March 2021, 122%), Qatar (April 2021, 116%), Oman (June 2021, 133%), Kuwait (July 2021, 143%) and Tunisia (July 2021, 140%) it occurred during the 2021 waves. Kuwait was the only country with an annual P-score above 40% in both 2020 (43.6%) and 2021 (44.5%); in Tunisia almost all excess fell in 2021 (34.6%, against 3.0% in 2020).

Patterns diverged after 2021. In Kuwait, Oman and Tunisia, mortality returned to or fell below the projected baseline in 2022, and in 2023 observed deaths were lower than expected (P-scores -7.0%, -7.8% and -6.4%); this deficit persisted in 2024. Jordan was close to baseline in 2022 and 2023. Egypt retained a smaller positive excess in 2022 (8.3%) and 2023 (5.0%). In contrast, Qatar and the United Arab Emirates remained at least 20% above baseline in every year from 2022 to 2024 (Qatar 21.6%, 21.0% and 22.9%; United Arab Emirates 32.2%, 29.4% and 43.2%). Over 2020 to 2023, observed deaths exceeded the upper prediction limit in 36 of 48 months in the United Arab Emirates and 32 in Qatar, compared with 12 to 19 months elsewhere.

Table 3. Annual P-scores (%) and excess deaths, primary model, 2020 to 2024.

Cells show the P-score in percent with excess deaths in parentheses; negative values indicate fewer deaths than expected. Values for 2024 extend the 2015 to 2019 trend five years and are shown as a return-to-baseline check; they should be interpreted cautiously. P-score, excess deaths as a percentage of expected deaths.

Figure 3. Monthly P-scores, January 2020 to December 2024 (red, above expected; blue, below; line separates GCC states from comparators). Peaks clustered in mid-2020 and mid-2021; Qatar and the United Arab Emirates stayed above baseline later.

3.4. Excess Compared with Reported COVID-19 Deaths

In total, 78,612 COVID-19 deaths were reported by the seven countries for 2020 to 2023, against 422,532 excess deaths, a ratio of 5.37 (Table 2, Figure 4). The ratio was 2.12 for the GCC states combined. Excess exceeded reported deaths by more than twofold in the United Arab Emirates (4.23), Qatar (2.69) and Kuwait (2.44), a pattern compatible with under-ascertainment even in health systems that deployed testing and vaccination early [40], and by a factor of 14.5 in Egypt. In Jordan the two measures were similar (1.15), and in Oman (0.78) and Tunisia (0.83) reported COVID-19 deaths exceeded cumulative excess, because deficits in 2022 and 2023 offset part of the earlier excess. Across countries, excess deaths per 100,000 showed no monotonic relation with reported COVID-19 deaths per 100,000 (Spearman's rho 0.07, n = 7).

Figure 4. Cumulative excess versus reported COVID-19 deaths per 100,000, 2020 to 2023 (log axes; ratio in parentheses; dashed line, equality). Reported deaths were a poor guide to the ranking of countries by excess mortality.

3.5. Baseline Back-Testing and Sensitivity Analyses

In the pre-pandemic back-test, all six baselines forecast annual totals for 2018 and 2019 with mean absolute errors between 4.0% and 5.4% (Table 4, Figure 5A). The primary model had a mean annual error of 5.3% with negligible mean bias (0.06%); gradient-boosted trees had the lowest annual error (4.0%) and Holt-Winters the next lowest (4.6%), while SARIMA had the highest monthly error (7.4%). The differences between models were small relative to the pandemic excess, and the back-test errors were of a similar order to year-to-year variation in the reference series.

Table 4. Out-of-sample accuracy of baseline models forecasting 2018 and 2019 from earlier years.

Rolling-origin test: trained on 2015 to 2017 to forecast 2018 and on 2015 to 2018 to forecast 2019, in Kuwait, Oman, Qatar, Egypt, Jordan and Tunisia. Positive bias indicates over-forecasting. MAPE, mean absolute percentage error; APE, absolute percentage error; SARIMA, seasonal autoregressive integrated moving average.

The sign of cumulative excess was positive under every sensitivity analysis in every country (Table 5, Figure 5B). Magnitudes varied: across all baselines, cumulative P-scores ranged from 15.2% to 37.5% in Kuwait, 8.0% to 36.6% in Oman and 4.6% to 23.4% in Jordan, whereas Egypt (8.4% to 17.3%) and Qatar (16.3% to 29.7%) were more stable. Oman's primary estimate was at the lower end of the range because its reference period included a step increase in deaths between 2015 and 2016 that steepened the projected trend. The ranking of countries was preserved under the negative binomial (Spearman correlation with the primary ranking 1.00), ensemble (0.89) and population-rate (0.86) baselines, but not when the trend term was dropped (0.14). Excluding the late-2019 months in Jordan raised its P-score to 23.4%. With the population-rate baseline, Qatar and the United Arab Emirates still showed P-scores of 32.2% and 23.3% in 2022 and 31.2% and 22.4% in 2024, whereas Kuwait, Oman and Tunisia remained at or below baseline in 2023 and 2024.

Table 5. Cumulative P-scores (%) for January 2020 to December 2023 under the primary and selected alternative baselines.

Population-rate, quasi-Poisson model with offset for log population; no trend, month-specific means of the reference years; range covers the primary model, five alternative model families and the sensitivity analyses (including exclusion of September to December 2019 for Jordan). P-score, excess deaths as a percentage of expected deaths.

Figure 5. (A) Out-of-sample error of six baselines forecasting 2018 and 2019; (B) cumulative P-scores, 2020 to 2023, under the primary model (diamonds) and alternative baselines (grey). Baseline choice changed the size but not the direction of excess.

4. Discussion

4.1. Principal Findings

In four GCC states with openly available monthly vital statistics, the COVID-19 period was associated with about 21,700 more deaths than pre-pandemic patterns predicted over 2020 to 2023, an increase of nearly one fifth. Relative excess was larger in the GCC states than in the Arab comparators, but excess per head was about three times lower, reflecting the younger GCC populations. Most excess occurred in two waves, in mid-2020 and mid-2021, and in Kuwait, Oman and Tunisia mortality fell below the projected baseline from 2023. Excess deaths exceeded officially reported COVID-19 deaths by approximately two- to fourfold in Kuwait, Qatar and the United Arab Emirates and by more than an order of magnitude in Egypt. Simple and machine-learning baselines forecast pre-pandemic years with similar accuracy, and none changed the direction of the findings.

4.2. Comparison with Other Studies

Our 2020 estimate for Kuwait (3,211 excess deaths) is close to the 3,346 reported by Alahmad and colleagues, who projected expected deaths from historical trends and showed that migrant workers bore a disproportionate share [9]. The Omani Ministry of Health reported a 15% increase in all-cause mortality between March and August 2020 using a Farrington model [11]; our annual 2020 P-score for Oman (14.3%) is consistent once months before the first wave are included. For Jordan, Khader and Al Nsour estimated 4,583 excess deaths in April to December 2020 [12], similar to our calendar-year estimate of 4,569, and found that most excess deaths had been attributed to COVID-19, in line with the near-unity ratio we observed over the whole period. In Qatar, a national cohort found low non-COVID-19 mortality among craft and manual workers, consistent with a healthy worker effect, but higher COVID-19 mortality than among Qatari nationals [10], which may help explain why relative excess in Qatar was high although excess per head was the lowest in our panel.

At the global level, WHO and the Institute for Health Metrics and Evaluation estimated that excess deaths in 2020 and 2021 were roughly three times reported COVID-19 deaths [2,3], and a 67-country analysis found that the gap was widest where testing and certification were weakest [7]. The ratios we observed, below one in Oman and Tunisia over four years but above four in the United Arab Emirates and above fourteen in Egypt, illustrate how strongly this gap varies even within one region. Egypt's ratio is in keeping with the large gaps between excess and reported deaths that global models estimated for many lower-middle-income settings with incomplete testing and certification [2,6,41]. Our finding that excess mortality largely receded by 2023 in most countries matches multi-country analyses showing a peak in 2020 and 2021 followed by decline [42], and European evidence that sustained post-pandemic excess is sensitive to the assumed trend [25].

The persistent excess in Qatar and the United Arab Emirates needs careful interpretation. Both countries changed rapidly in population size and composition during and after the pandemic, and the United Arab Emirates' population grew by 16% between 2019 and 2024. For the United Arab Emirates the reference period covered only 2018 and 2019, so the baseline carried no trend and could not anticipate this growth; the population-rate baseline reduced but did not remove the post-2021 excess. For Qatar, the national source of the monthly series changed in 2020, from compilations held by the United Nations to the Planning and Statistics Authority, so a discontinuity in coverage or registration practice cannot be excluded. Post-pandemic excess has been reported elsewhere, for example in Hong Kong [26], but commentators have stressed that whether later excess reflects real mortality change or baseline drift depends on how the baseline is defined [24]. We therefore regard the 2022 to 2024 excess in these two countries as unexplained rather than as evidence of ongoing pandemic mortality.

Methodologically, our back-test agrees with forecasting research showing that simple statistical models are hard to beat on short, seasonal series and that combinations perform reliably [27,28]. Gradient-boosted trees, which cannot extrapolate a trend, performed well over a one-year horizon because the reference series were nearly flat, but their projections over four years diverge from trend-based models by construction. Previous sensitivity analyses showed that method and reference period are the main sources of variation in excess estimates [20,21,43,44]; our results add that this variation is large relative to the excess itself in small populations such as Oman, where the cumulative P-score varied more than fourfold across defensible baselines.

4.3. Relevance to Saudi Arabia and the Region

Saudi Arabia shares the demographic features that shaped these results: a young population, a large migrant workforce and a rapidly changing population size. Its pandemic response built on experience with Middle East respiratory syndrome [45], and published Saudi analyses of COVID-19 mortality rely on hospital records rather than population vital statistics [14]. The pattern seen in its neighbours, a large relative excess concentrated in 2020 and 2021, a gap between excess and reported deaths, and difficulty separating late excess from population change, is likely to be relevant to Saudi Arabia as well, but it cannot be confirmed because monthly all-cause death counts are not available in open international compilations. Publishing monthly deaths by age, sex and nationality, together with matching population estimates, would allow Saudi Arabia and Bahrain to be included in regional surveillance and would let excess mortality be decomposed between nationals and migrants, as was possible in Kuwait [9] and Qatar [10]. Such data would also support preparedness for future respiratory epidemics, for which routine mortality surveillance has been proposed as an early warning system in the region [13].

4.4. Strengths and Limitations

The study covers the full four-year period from 2020 to 2023, applies one prespecified model to all countries, quantifies uncertainty for every estimate, compares six baseline methods in a pre-pandemic back-test, and recomputed the primary estimates independently. All inputs are public and the analysis can be reproduced from archived files.

Several limitations apply. First, the design is ecological: the estimates describe national populations and do not identify which deaths were caused by infection, by disrupted care or by other factors, nor which groups were affected. Second, the data contain no age or sex breakdown, so P-scores and rates could not be standardised; crude comparisons between young GCC populations and older comparator populations are affected by age structure [46,47]. Third, the series are compiled from national registration systems whose completeness, timeliness and treatment of non-resident deaths may differ between countries and over time; Qatar and Egypt changed source in 2020, the 2023 and 2024 Egyptian figures are rounded and provisional, and deaths of migrant workers who left the country before death are not captured. Fourth, the reference period was short, particularly for the United Arab Emirates (two years) and Jordan (four years with a gap), which increases both statistical uncertainty and sensitivity to anomalous years. Fifth, population denominators for GCC states fluctuated sharply in 2020 and 2021 and may not match the population at risk of death registration. Sixth, extrapolating a 2015 to 2019 trend to 2023 and 2024 becomes progressively less reliable, as the data compilers themselves note [1]. Finally, with seven countries the cross-country correlation analyses are descriptive only.

4.5. Implications

For public health agencies in the region, the results indicate that reported COVID-19 deaths understated the mortality impact of the pandemic in most GCC states and should not be used alone to compare countries or evaluate responses. Excess mortality estimates should be reported with the baseline clearly stated and with sensitivity analyses, particularly in small populations. Further research should use individual-level death registration by nationality, age and cause to establish whether the post-2021 excess in Qatar and the United Arab Emirates reflects changes in mortality, in population composition or in recording.

5. Conclusions

Between 2020 and 2023, deaths in Kuwait, Oman, Qatar and the United Arab Emirates exceeded pre-pandemic expectations by close to one fifth, mostly during the 2020 and 2021 waves, and in most countries the excess was considerably larger than reported COVID-19 deaths. Relative excess in the GCC states was comparable to or greater than in Egypt, Jordan and Tunisia, while excess per head was lower. Mortality returned below baseline in several countries by 2023, but persistent excess in Qatar and the United Arab Emirates could not be separated from population change and data discontinuities. Timely monthly vital statistics by age, sex and nationality, including from Saudi Arabia, are needed for reliable regional mortality surveillance.

Declarations

Author Contributions: NA

Funding: NA

Conflicts of Interest: NA

Institutional Review Board Statement: Not applicable: secondary analysis of publicly available, de-identified aggregate data.

Informed Consent Statement: Not applicable.

Data Availability Statement: Monthly deaths: World Mortality Dataset, https://github.com/akarlinsky/world_mortality. Reported COVID-19 deaths: Our World in Data, https://github.com/owid/covid-19-data (public/data/cases_deaths/full_data.csv). Population: World Bank World Development Indicators, SP.POP.TOTL, https://data.worldbank.org/indicator/SP.POP.TOTL. All files were accessed on 6 November 2026. Analysis code is available from the authors.

References

[1] Karlinsky A, Kobak D. Tracking excess mortality across countries during the COVID-19 pandemic with the World Mortality Dataset. eLife. 2021;10:e69336. doi:10.7554/elife.69336. PMID: 34190045.

[2] COVID-19 Excess Mortality Collaborators. Estimating excess mortality due to the COVID-19 pandemic: a systematic analysis of COVID-19-related mortality, 2020–21. The Lancet. 2022;399(10334):1513-1536. doi:10.1016/s0140-6736(21)02796-3. PMID: 35279232.

[3] Msemburi W, Karlinsky A, Knutson V, Aleshin-Guendel S, Chatterji S, Wakefield J. The WHO estimates of excess mortality associated with the COVID-19 pandemic. Nature. 2022;613(7942):130-137. doi:10.1038/s41586-022-05522-2. PMID: 36517599.

[4] Kontis V, Bennett JE, Rashid T, Parks RM, Pearson-Stuttard J, Guillot M, et al. Magnitude, demographics and dynamics of the effect of the first wave of the COVID-19 pandemic on all-cause mortality in 21 industrialized countries. Nat Med. 2020;26(12):1919-1928. doi:10.1038/s41591-020-1112-0. PMID: 33057181.

[5] Islam N, Shkolnikov VM, Acosta RJ, Klimkin I, Kawachi I, Irizarry RA, et al. Excess deaths associated with covid-19 pandemic in 2020: age and sex disaggregated time series analysis in 29 high income countries. BMJ. 2021;373:n1137. doi:10.1136/bmj.n1137. PMID: 34011491.

[6] Whittaker C, Walker PGT, Alhaffar M, Hamlet A, Djaafara BA, Ghani A, et al. Under-reporting of deaths limits our understanding of true burden of covid-19. BMJ. 2021;375:n2239. doi:10.1136/bmj.n2239.

[7] Sanmarchi F, Golinelli D, Lenzi J, Esposito F, Capodici A, Reno C, et al. Exploring the Gap Between Excess Mortality and COVID-19 Deaths in 67 Countries. JAMA Netw Open. 2021;4(7):e2117359. doi:10.1001/jamanetworkopen.2021.17359. PMID: 34269809.

[8] Nahari TH, Alkhidir MA, Ibrahim HM, Al Mamun M. Migrant workers with COVID-19: a major challenge for Gulf Cooperation Council (GCC) countries to curb the spread of infection. Glob Health Promot. 2024;31(3):141-146. doi:10.1177/17579759231216108.

[9] Alahmad B, AlMekhled D, Odeh A, Albloushi D, Gasana J. Disparities in excess deaths from the COVID-19 pandemic among migrant workers in Kuwait. BMC Public Health. 2021;21(1):1668. doi:10.1186/s12889-021-11693-w. PMID: 34521360.

[10] AlNuaimi AA, Chemaitelly H, Semaan S, AlMukdad S, Al-Kanaani Z, Kaleeckal AH, et al. All-cause and COVID-19 mortality in Qatar during the COVID-19 pandemic. BMJ Glob Health. 2023;8(5):e012291. doi:10.1136/bmjgh-2023-012291. PMID: 37142299.

[11] Al Wahaibi A, Al-Maani A, Alyaquobi F, Al Harthy K, Al-Jardani A, Al Rawahi B, et al. Effects of COVID-19 on mortality: A 5-year population-based study in Oman. International Journal of Infectious Diseases. 2021;104:102-107. doi:10.1016/j.ijid.2020.12.054. PMID: 33359442.

[12] Khader Y, Al Nsour M. Excess Mortality During the COVID-19 Pandemic in Jordan: Secondary Data Analysis. JMIR Public Health Surveill. 2021;7(10):e32559. doi:10.2196/32559. PMID: 34617910.

[13] Meckawy R, El Weshahi HMT, Foda E, Sultan EA. Mortality surveillance as an early warning system for respiratory infection outbreaks: lessons from the COVID-19 pandemic in Alexandria, Egypt. J. Egypt. Public. Health. Assoc. 2026;101(1):4. doi:10.1186/s42506-025-00205-y. PMID: 41729422.

[14] Mahmoud MA, Altaluoni AA, Alshargi AA, Al-Zalabani AH. Clinical and epidemiological characteristics of COVID-19 mortality in Saudi Arabia: A retrospective multi-center study. Journal of Family Medicine and Primary Care. 2024;13(10):4270-4275. doi:10.4103/jfmpc.jfmpc_128_24. PMID: 39629401.

[15] BuHamra SS, Al-Kandari NM, Hussam E, Almetwally EM, Gemeay AM. A case study for Kuwait mortality during the consequent waves of COVID-19. Heliyon. 2024;10(5):e26790. doi:10.1016/j.heliyon.2024.e26790. PMID: 39669480.

[16] Almari M, O’Neill S, Sadique Z. The economic burden of COVID-19 premature mortality in Kuwait. BMC Public Health. 2025;26(1):405. doi:10.1186/s12889-025-25940-x. PMID: 41466402.

[17] Ben Fredj M, Gara A, Kacem M, Dhouib W, Ben Hassine D, Bennasrallah C, et al. Longitudinal analysis of Covid-19 infection trends and in-hospital mortality across six pandemic waves in Tunisia. Arch Public Health. 2025;83(1):86. doi:10.1186/s13690-024-01485-0. PMID: 40170097.

[18] Dhaouadi S, Hechaichi A, Letaief H, Safer M, Mziou E, Talmoudi K, et al. Caractéristiques cliniques et épidémiologiques des décès COVID-19 en Tunisie avant l´émergence des VOCs: mars 2020-février 2021. Pan Afr Med J. 2022;43. doi:10.11604/pamj.2022.43.172.35544. PMID: 36879635.

[19] Ayoub HH, Chemaitelly H, Seedat S, Makhoul M, Al Kanaani Z, Al Khal A, et al. Mathematical modeling of the SARS-CoV-2 epidemic in Qatar and its impact on the national response to COVID-19. J Glob Health. 2021;11:05005. doi:10.7189/jogh.11.05005. PMID: 33643638.

[20] Nepomuceno MR, Klimkin I, Jdanov DA, Alustiza‐Galarza A, Shkolnikov VM. Sensitivity Analysis of Excess Mortality due to the COVID‐19 Pandemic. Popul Dev Rev. 2022;48(2):279-302. doi:10.1111/padr.12475. PMID: 35600716.

[21] Levitt M, Zonta F, Ioannidis JP. Comparison of pandemic excess mortality in 2020–2021 across different empirical calculations. Environmental Research. 2022;213:113754. doi:10.1016/j.envres.2022.113754. PMID: 35753371.

[22] Acosta RJ, Irizarry RA. A Flexible Statistical Framework for Estimating Excess Mortality. Epidemiology. 2022;33(3):346-353. doi:10.1097/ede.0000000000001445. PMID: 35383642.

[23] Wakefield J, Knutson V. Excess Mortality Estimation. Annual Review of Statistics and Its Application. 2025;12(1):45-68. doi:10.1146/annurev-statistics-112723-034236. PMID: 42100554.

[24] Klein J. Excess mortality is not self-interpreting: Baseline choice as a governance issue in public health analytics. Scand J Public Health. 2026;54(4):429-431. doi:10.1177/14034948261424605. PMID: 41766197.

[25] Alicandro G, Gerli AG, La Vecchia C. Excess mortality at ages 20–64 from 2020 to 2024: a comparative analysis of 16 European countries. bmjph. 2025;3(2):e003328. doi:10.1136/bmjph-2025-003328. PMID: 41321369.

[26] Li K, Wei Y, Hung CT, Wong CKH, Xiong X, Chan PKS, et al. Post-pandemic excess mortality of COVID-19 in Hong Kong: a retrospective study. The Lancet Regional Health - Western Pacific. 2025;58:101554. doi:10.1016/j.lanwpc.2025.101554. PMID: 40336577.

[27] Makridakis S, Spiliotis E, Assimakopoulos V. Statistical and Machine Learning forecasting methods: Concerns and ways forward. PLoS ONE. 2018;13(3):e0194889. doi:10.1371/journal.pone.0194889. PMID: 29584784.

[28] Makridakis S, Spiliotis E, Assimakopoulos V. The M4 Competition: 100,000 time series and 61 forecasting methods. International Journal of Forecasting. 2020;36(1):54-74. doi:10.1016/j.ijforecast.2019.04.014.

[29] Morgenstern H. Ecologic Studies in Epidemiology: Concepts, Principles, and Methods. Annu. Rev. Public Health. 1995;16(1):61-81. doi:10.1146/annurev.pu.16.050195.000425.

[30] Dong E, Du H, Gardner L. An interactive web-based dashboard to track COVID-19 in real time. The Lancet Infectious Diseases. 2020;20(5):533-534. doi:10.1016/s1473-3099(20)30120-1. PMID: 32087114.

[31] Mathieu E, Ritchie H, Ortiz-Ospina E, Roser M, Hasell J, Appel C, et al. A global database of COVID-19 vaccinations. Nat Hum Behav. 2021;5(7):947-953. doi:10.1038/s41562-021-01122-8.

[32] Bhaskaran K, Gasparrini A, Hajat S, Smeeth L, Armstrong B. Time series regression studies in environmental epidemiology. International Journal of Epidemiology. 2013;42(4):1187-1195. doi:10.1093/ije/dyt092. PMID: 23760528.

[33] Scortichini M, Schneider dos Santos R, De’ Donato F, De Sario M, Michelozzi P, Davoli M, et al. Excess mortality during the COVID-19 outbreak in Italy: a two-stage interrupted time-series analysis. International Journal of Epidemiology. 2020;49(6):1909-1917. doi:10.1093/ije/dyaa169. PMID: 33053172.

[34] Tashman LJ. Out-of-sample tests of forecasting accuracy: an analysis and review. International Journal of Forecasting. 2000;16(4):437-450. doi:10.1016/s0169-2070(00)00065-0.

[35] Hyndman RJ, Koehler AB. Another look at measures of forecast accuracy. International Journal of Forecasting. 2006;22(4):679-688. doi:10.1016/j.ijforecast.2006.03.001.

[36] Chen T, Guestrin C. XGBoost. In: Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. 2016. p. 785-794. doi:10.1145/2939672.2939785.

[37] Seabold S, Perktold J. Statsmodels: Econometric and Statistical Modeling with Python. In: Proceedings of the Python in Science Conference. 2010. p. 92-96. doi:10.25080/majora-92bf1922-011.

[38] von Elm E, Altman DG, Egger M, Pocock SJ, Gøtzsche PC, Vandenbroucke JP. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. The Lancet. 2007;370(9596):1453-1457. doi:10.1016/s0140-6736(07)61602-x.

[39] Stevens GA, Alkema L, Black RE, Boerma JT, Collins GS, Ezzati M, et al. Guidelines for Accurate and Transparent Health Estimates Reporting: the GATHER statement. The Lancet. 2016;388(10062):e19-e23. doi:10.1016/s0140-6736(16)30388-9.

[40] Al Kaabi N, Oulhaj A, Ganesan S, Al Hosani FI, Najim O, Ibrahim H, et al. Effectiveness of BBIBP-CorV vaccine against severe outcomes of COVID-19 in Abu Dhabi, United Arab Emirates. Nat Commun. 2022;13(1):3215. doi:10.1038/s41467-022-30835-1. PMID: 35680857.

[41] Knutson V, Aleshin-Guendel S, Karlinsky A, Msemburi W, Wakefield J. Estimating global and country-specific excess mortality during the Covid-19 pandemic. Ann. Appl. Stat. 2023;17(2). doi:10.1214/22-aoas1673.

[42] Jang D, Choi S, Choi B, Kim S. Structural gradients in excess mortality during and after the COVID-19 pandemic: economic development, preparedness, and cross-national differences, 2020–2024. J Glob Health. 2026;16:04246. doi:10.7189/jogh.16.04246. PMID: 42533632.

[43] Barnard S, Chiavenna C, Fox S, Charlett A, Waller Z, Andrews N, et al. Methods for modelling excess mortality across England during the COVID-19 pandemic. Stat Methods Med Res. 2021;31(9):1790-1802. doi:10.1177/09622802211046384. PMID: 34693801.

[44] Niewiadomski P, Wojtasik J, Łyszczarz B. Beyond death counts: how COVID-19 affected excess mortality and productive life-years lost across 28 European states, 2020–2023. Front. Public Health. 2026;13:1720864. doi:10.3389/fpubh.2025.1720864. PMID: 41641044.

[45] Algaissi AA, Alharbi NK, Hassanain M, Hashem AM. Preparedness and response to COVID-19 in Saudi Arabia: Building on MERS experience. Journal of Infection and Public Health. 2020;13(6):834-838. doi:10.1016/j.jiph.2020.04.016. PMID: 32451260.

[46] Aburto JM, Schöley J, Kashnitsky I, Zhang L, Rahal C, Missov TI, et al. Quantifying impacts of the COVID-19 pandemic through life-expectancy losses: a population-level study of 29 countries. International Journal of Epidemiology. 2021;51(1):63-74. doi:10.1093/ije/dyab207. PMID: 34564730.

[47] Schöley J, Aburto JM, Kashnitsky I, Kniffka MS, Zhang L, Jaadla H, et al. Life expectancy changes since COVID-19. Nat Hum Behav. 2022;6(12):1649-1659. doi:10.1038/s41562-022-01450-3. PMID: 36253520.