Excess mortality
Excess mortality refers to the additional deaths from all causes during a specific time period or in a defined population that exceed the number expected under normal conditions, calculated as the difference between observed deaths and a baseline of anticipated deaths derived from historical data adjusted for factors such as population size, age structure, and long-term trends.[1][2][3] This approach provides a comprehensive gauge of a crisis's mortality burden, encompassing both direct fatalities and indirect effects like disruptions to healthcare or socioeconomic stressors, independent of potentially unreliable or delayed cause-specific attributions.[3][4] During the COVID-19 pandemic, excess mortality metrics highlighted underreporting of the virus's toll in numerous locations, with peer-reviewed estimates indicating around 14.9 million global excess deaths by late 2021, far surpassing confirmed COVID-19 fatalities.[5][6] In the years following the initial waves, sustained excess mortality has been documented across Western countries and beyond, totaling over 3 million additional deaths from 2020 to 2022, with notable rises in non-respiratory conditions including ischemic heart disease, cerebrovascular events, and diabetes—attributed in analyses to pandemic-induced healthcare deferrals, behavioral shifts, and possible lingering physiological impacts rather than ongoing acute infections.[7][8][9]Definition and Measurement
Core Definition
Excess mortality, also known as excess deaths, quantifies the difference between the observed number of deaths in a population during a specific period and the number of deaths expected under baseline conditions absent any unusual events.[1] This baseline is typically derived from historical mortality data, such as averages from preceding years, adjusted for factors including population size, age structure, and seasonal trends.[2] The metric captures mortality from all causes, making it a robust indicator for assessing the total impact of crises like epidemics, natural disasters, or policy interventions, as it avoids underreporting biases inherent in cause-specific death registrations.[10] Expected deaths are estimated using statistical models that account for long-term trends in mortality rates, ensuring comparability across periods; for instance, the World Health Organization defines excess mortality as mortality above what would be anticipated based on non-crisis rates in the relevant population.[10] Negative excess mortality, where observed deaths fall below expectations, can also occur due to factors like reduced exposure to other illnesses during lockdowns or behavioral changes.[4] This all-cause approach distinguishes excess mortality from metrics focused solely on confirmed cases or direct attributions, providing a more complete picture of indirect effects such as disruptions to healthcare systems.[2] Excess mortality is often expressed in absolute terms (e.g., total excess deaths) or relative terms (e.g., percentage deviation from baseline), facilitating cross-country and temporal comparisons.[11] For example, during the COVID-19 pandemic, global estimates highlighted substantial discrepancies between reported COVID-19 deaths and total excess mortality, underscoring potential undercounting in official tallies.[1] Its utility lies in its empirical foundation, relying on verifiable death registrations rather than diagnostic accuracy, though interpretations require caution regarding baseline assumptions and data quality.[2]Calculation Methods and Baselines
Excess mortality is calculated as the difference between observed all-cause deaths and expected deaths over a defined period, with the latter derived from baseline mortality patterns in prior years to isolate deviations attributable to specific events.[3] Baselines exclude the period of interest to prevent circular incorporation of anomalous mortality, typically spanning multiple years to average out yearly fluctuations while capturing seasonality through weekly or monthly alignments.[12] Simple baseline methods rely on arithmetic averages of historical deaths, adjusted pro-rata for the reference unit (e.g., weeks or months) and sometimes scaled by population size or demographic shifts. Eurostat, for instance, computes excess as a percentage by comparing observed monthly deaths—derived from weekly data transmitted by member states—to the unadjusted average monthly deaths across 2016–2019, applying completeness corrections for recent data lags but not explicit age- or trend-based modifications.[11] This approach yields straightforward comparability across European countries but may overlook long-term declines in age-standardized rates. More sophisticated techniques employ regression models to forecast expected deaths, incorporating covariates for trends and seasonality. The World Health Organization models baselines using negative binomial regression on pre-2020 data: for monthly-reporting countries, 2015–2019 records inform linear annual trends and cyclic cubic splines for intra-year patterns; annual-data countries draw from extended histories (e.g., 2000–2019) apportioned monthly via multinomial logistic models proxying temperature-driven seasonality.[10] Uncertainty is quantified through gamma-distributed sampling from model posteriors, with further adjustments for subnational data via proportionality assumptions or covariates like Human Development Index in integrated nested Laplace approximation frameworks. The U.S. Centers for Disease Control and Prevention estimates expected deaths via historical averages benchmarked against prediction intervals, often Poisson-distributed to flag significant excesses when observed counts exceed upper bounds (e.g., 95th percentile equivalents).[2] These baselines draw from multi-year pre-pandemic aggregates, weighted for reporting lags and demographic factors, enabling provisional surveillance. Comparative analyses of such methods—contrasting basic averages, WHO regressions, and advanced forecasting like Acosta-Irizarry's trend-extrapolative approach—reveal variances in sensitivity to disruptions, with parametric models outperforming static averages in capturing pre-event trajectories but risking overfitting to noise.[12] [13] Proposals for refined baselines include retrospective minima from low-mortality weeks or within-year comparisons to mitigate trend extrapolation errors during volatile periods, though standard practice prioritizes multi-year stability for causal attribution.[14] Overall, method selection balances parsimony with predictive accuracy, influencing excess estimates by up to several percentage points in tested scenarios.[12]Data Sources and Limitations
Primary data sources for excess mortality derive from national vital registration systems, which compile death certificates and demographic information to produce all-cause mortality counts. In the United States, the Centers for Disease Control and Prevention (CDC) utilizes the National Vital Statistics System (NVSS), drawing on provisional and final data from state registries to estimate excess deaths by subtracting observed deaths from expected baselines.[2] Similarly, in the European Union, Eurostat aggregates weekly and monthly mortality data from member states' statistical offices, enabling cross-country comparisons of excess mortality ratios relative to pre-pandemic periods.[11] Globally, the Human Mortality Database (HMD) provides high-quality historical data for select high-income countries, while organizations like Our World in Data integrate these with national agency reports for broader coverage.[3] The World Health Organization (WHO) supplements official statistics with modelled excess mortality estimates, incorporating vital registration where available and statistical adjustments for underreporting in low-data regions, yielding global figures such as 14.83 million excess deaths from 2020 to 2021.[15] These models account for baseline trends using methods like Poisson regression on historical data, but rely on assumptions about reporting completeness.[10] Independent trackers, such as The Economist's, cross-verify official reports from over 200 locations, highlighting discrepancies between reported COVID-19 deaths and total excess.[16] Limitations arise from inconsistencies in data timeliness and completeness; provisional figures, as used by the CDC, often undercount due to reporting lags of weeks to months, with weighting adjustments that may incompletely capture surges.[2] Coverage gaps persist in low- and middle-income countries lacking robust vital registration, restricting global analyses to primarily high-income nations with reliable historical series.[3] Baseline estimation poses methodological challenges: simple averages of prior years ignore trends like aging populations or seasonality, while advanced techniques such as spline regressions or Bayesian models can yield varying results depending on parameters, potentially overstating or understating excess by 10-20% in sensitivity tests.[12][17] Cross-country comparisons are further complicated by differing categorization practices, population adjustments, and external factors like migration, which affect denominator accuracy.[8]Historical Context
Ancient and Medieval Epidemics
The Plague of Athens, occurring in 430 BC amid the Peloponnesian War, afflicted the overcrowded city-state, where refugees swelled the population to an estimated 250,000–400,000 within its walls. Contemporary historian Thucydides reported daily deaths reaching 600–1,100 at peak, with the epidemic persisting in waves through 426 BC and claiming approximately 25–30% of inhabitants, or 60,000–80,000 lives, far exceeding baseline annual mortality rates of 1–2% derived from pre-war demographic stability.[18][19] This excess mortality, inferred from skeletal evidence and population reconstructions, disrupted Athenian society, contributing to military defeats and long-term demographic decline without modern vital records for precise baselines.[20] The Antonine Plague of 165–180 AD, likely caused by smallpox or measles introduced via trade routes, ravaged the Roman Empire's estimated 50–60 million population, resulting in 5–10 million excess deaths or roughly 7–10% overall mortality, with urban centers like Rome experiencing up to 2,000 daily fatalities in peak years.[21][22] Historical demographers calculate this surplus against expected peacetime mortality of under 2% annually, drawing from fiscal records showing depopulation in provinces and army recruitment shortfalls, though literary sources like Galen may inflate urban impacts due to eyewitness bias.[23] The plague's recurrent waves amplified cumulative excess, weakening imperial frontiers and economy without offsetting factors like migration fully restoring numbers.[24] The Plague of Justinian, erupting in 541 AD and recurring until circa 750 AD, marked the first documented bubonic plague pandemic, originating in Egypt and decimating the Byzantine Empire and Mediterranean basin. In Constantinople alone, peak mortality hit 5,000–10,000 daily, equating to 40–50% of its 500,000 residents over initial waves, while empire-wide estimates suggest 25–50 million excess deaths against a baseline population of 50–100 million, representing 25–50% depopulation in affected regions per maximalist reconstructions from tax rolls and chronicles.[25][26] Procopius's accounts, corroborated by genomic evidence of Yersinia pestis, indicate surplus mortality dwarfing normal rates, though minimalist views from archaeological data argue for lower impacts outside urban cores due to rural underreporting and adaptive quarantines.[27] Long-term waves sustained elevated death rates, hindering Justinian's reconquests and accelerating antiquity's transition. Medieval Europe's Black Death of 1347–1351, another Y. pestis outbreak, inflicted 30–50% excess mortality across the continent's 75–100 million population, totaling 25–50 million deaths beyond expected annual baselines of 1.5–2%, as evidenced by parish records, manorial rolls, and mass graves showing synchronized spikes uncorrelated with famines or wars alone.[28][29] Regional variations included 40–60% losses in England and Italy from 1348–1350, per skeletal isotope analysis and poll tax data indicating abrupt population halving without recovery until the 16th century.[30] This demographic shock, calculated via retrospective cohort methods on surviving ledgers, stemmed from high case-fatality ratios (60–90%) in untreated pneumonic and septicemic forms, outpacing any natural mortality and reshaping labor markets.[31] Subsequent waves, like the 1361 pestis secunda, added 10–20% further excess, per urban burial registries.[32]Modern Pandemics Prior to 2020
The 1918 influenza pandemic, caused by the H1N1 virus, resulted in an estimated 50 to 100 million global deaths, representing one of the highest excess mortality events in modern history, with approximately 675,000 deaths in the United States alone.[33][34] This equated to excess all-cause mortality rates far exceeding baseline expectations, particularly among young adults aged 15-34, where influenza and pneumonia mortality surged over 20 times pre-pandemic levels.[35] Peaks occurred in waves, with a notable early excess of 4,600 all-cause deaths in certain regions during the 1917-1918 season, escalating dramatically by October 1918.[36][37] The 1957-1959 Asian influenza pandemic, driven by the H2N2 virus, led to approximately 1.1 million excess global deaths (95% confidence interval: 0.7-1.5 million), with an average pandemic-associated excess respiratory mortality rate of 1.9 per 10,000 population.[38][39] In the United States, it caused around 80,000 deaths, concentrated from September 1957 to March 1958, primarily affecting older populations but with notable excess across age groups.[40][41] The event's mortality burden varied regionally, underscoring differences in baseline health and response capabilities.[42] The 1968-1970 Hong Kong influenza pandemic, associated with the H3N2 virus, produced 1 to 4 million excess deaths worldwide, including about 100,000 in the United States, with most fatalities among those aged 65 and older.[43][44] Excess mortality rates reached peaks such as 64 per 100,000 in Australia by 1970, reflecting sustained circulation rather than a single acute wave.[45] The pandemic affected 30-57% of the global population, yet its case fatality rate remained low at 0.02-0.03%, highlighting how demographic vulnerabilities amplified total excess deaths.[46] The 2009 H1N1 swine flu pandemic resulted in an estimated 150,000 to 575,000 global excess deaths, far exceeding the World Health Organization's initial tally of 18,500 laboratory-confirmed cases, with respiratory mortality approximately 10 times higher than reported.[47][48] A modeling study pegged the figure at around 284,000 excess deaths, driven by disproportionate impacts in certain regions and age groups, including higher burdens in those over 50 in some areas.[49] Unlike prior influenza events, vaccination and antiviral availability mitigated some excess, though underreporting persisted due to diagnostic limitations.[50]| Pandemic | Virus | Global Excess Deaths | Key Notes on Excess Mortality |
|---|---|---|---|
| 1918 Spanish Flu | H1N1 | 50-100 million | Highest in young adults; multiple waves with sharp all-cause spikes.[33][34] |
| 1957 Asian Flu | H2N2 | ~1.1 million | Respiratory excess rate 1.9/10,000; peaked 1957-1958.[38] |
| 1968 Hong Kong Flu | H3N2 | 1-4 million | Primarily elderly; sustained over 2 years.[43] |
| 2009 H1N1 | H1N1 | 150,000-575,000 | Underreported; regional variations, higher in adults.[47] |
Non-Pandemic Events like Wars
Wars have consistently produced excess mortality through mechanisms such as direct combat fatalities, civilian targeting, forced displacement, and induced famines, elevating all-cause death rates far beyond peacetime baselines in affected populations.[51] These indirect effects, including breakdowns in food supply and sanitation, often amplify mortality independently of infectious disease outbreaks, though historical conflicts frequently exacerbated vulnerabilities leading to secondary health crises.[52] During the American Civil War (1861–1865), linkage of full U.S. census records reveals approximately 497,000 excess deaths among military-age non-Black white males, attributable to battlefield losses, wounds, and war-induced hardships like malnutrition and inadequate medical care.[53] This figure surpasses traditional estimates of 360,000 Union and Confederate combat deaths, highlighting undercounted indirect impacts on civilian-adjacent demographics.[53] In World War I (1914–1918), excess mortality among young adult males in belligerent nations was stark; for German men aged 20–25, death rates increased by 500% over normal peacetime levels, driven primarily by conscription, trench warfare, and exposure to combat hazards.[54] Similar spikes occurred across Europe, where mobilization disrupted economies and heightened famine risks in rural areas, contributing to broader demographic shortfalls without relying on pandemic-scale epidemics.[54] World War II (1939–1945) provides further evidence of war-induced surges; in the Netherlands, total excess civilian mortality from conflict reached an estimated 160,000, with about 65,000 occurring in 1945 amid "Hunger Winter" starvation and infrastructure collapse.[55] In Finland's Karelian Isthmus, crude death rates climbed to 11.8 per 1,000 inhabitants in 1944—nearly 40% above the pre-war 1939 baseline of 8.6 per 1,000—due to evacuation, bombing, and food shortages.[52] Such patterns underscore how wartime blockades and scorched-earth tactics generated sustained excess deaths through caloric deficits and exposure, distinct from pathogen-driven events.[55]Excess Mortality During the COVID-19 Pandemic
Global Scale and Timeline
Excess mortality during the COVID-19 pandemic manifested globally from early 2020, with the first significant deviations from baseline death rates occurring in January 2020 in regions affected by initial outbreaks, such as Wuhan, China, and escalating worldwide by March 2020 as SARS-CoV-2 spread via international travel.[3] The World Health Organization's modeled estimates indicate that cumulative excess deaths reached approximately 14.9 million from January 2020 to December 2021, encompassing both direct COVID-19 fatalities and indirect effects, far exceeding the 5.4 million confirmed COVID-19 deaths reported globally in official tallies during this period.[56] [57] This excess was calculated by comparing observed all-cause mortality to expected levels derived from 2015-2019 trends, adjusted for demographic changes.[15] Temporal patterns revealed multiple waves of elevated mortality. The initial global peak aligned with the first wave in spring 2020, particularly intense in Europe and North America from March to May, followed by surges in Latin America and India during mid-2020.[3] A second major wave occurred in late 2020 to early 2021 across hemispheres, driven by variant emergence and seasonal factors, with excess mortality p-scores—measuring percentage deviation from baseline—exceeding 20% in affected areas.[58] By mid-2021, the Delta variant contributed to renewed peaks, notably in South Asia and parts of Europe, while Omicron-driven waves in late 2021 and 2022 showed varying excess depending on vaccination coverage and healthcare capacity.[59] Alternative modeling, such as that referenced in The Economist database, corroborates these trends, estimating around 18.2 million excess deaths through December 2021, highlighting underreporting in low-data regions like Africa and parts of Asia.[60] [59] Excess mortality persisted beyond the initial pandemic peaks into 2022 and 2023, albeit at lower intensities in many high-income countries, with global estimates suggesting ongoing deviations attributable to lingering pandemic effects.[8] For instance, analyses of 47 Western countries reported over 3 million cumulative excess deaths from 2020 to 2022, with non-COVID causes comprising a substantial portion in later years.[7] Projections indicate gradual normalization, though full recovery to pre-2020 baselines remains uncertain as of 2023 data.[17] These patterns underscore the pandemic's prolonged demographic impact, varying by region due to differences in virus transmission, public health responses, and reporting infrastructure.[61]Country-Level Patterns
Excess mortality during the COVID-19 pandemic exhibited stark country-level variations, with cumulative rates per 100,000 population from 2020 to 2021 reaching approximately 6,994 in Peru and 4,897 in Bulgaria, while remaining below 100 in New Zealand (33) and Australia (50).[3] These disparities reflected differences in baseline health vulnerabilities, healthcare capacity, demographic structures, and reporting completeness, though data limitations persist in underreporting nations.[62] In 34 high-income countries analyzed from 2020 to 2023, total excess deaths totaled 2,097,101, with the United States accounting for 58% despite comprising only 25% of the population.[63]| Country | Cumulative Excess Deaths per 100,000 | Period | Source |
|---|---|---|---|
| Peru | ~6,994 | 2020–2021 | WHO via OWID[3] |
| Bulgaria | ~4,897 | 2020–2021 | WHO via OWID[3] |
| New Zealand | ~33 | 2020–2021 | WHO via OWID[3] |
| Australia | ~50 | 2020–2021 | WHO via OWID[3] |
| United States | ~1,700 (estimated total excess aligned with p%=17.9% nonelderly) | 2020–2023 | PNAS/HMD[63] |