The Evolution of Excess Deaths in the United States During the First 2 Years of the COVID-19 Pandemic

Abstract

Understanding consequences of the COVID-19 pandemic requires information on the excess mortality resulting from it. Multiple studies have examined excess deaths during the pandemic’s initial stages, but how these have changed over time is unclear. National- and state-level death counts and population data from 2009 to 2022 were used in this analysis to evaluate excess fatalities from March 2020 to February 2021 and March 2021 to February 2022, with deaths from earlier years used to project baseline counts. The outcomes were total, group-specific, cause-specific, and age-by-cause excess fatalities, and numbers and percentages directly involving COVID-19. Excess deaths declined from 655,735 (95% confidence interval: 619,028, 691,980) during the first pandemic year to 586,505 (95% confidence interval: 532,823, 639,205) in the second. The reductions were particularly large for Hispanics, Blacks, Asians, seniors, and residents of states with high vaccination rates. Excess deaths increased from the first to second year for persons younger than 65 years and in low-vaccination states. Excess mortality from some diseases declined, but those from alcohol, drug, vehicle, and homicide causes likely increased between the first and second pandemic year, especially for prime-age and younger individuals. The share of excess fatalities involving COVID-19 decreased modestly over time, with little change in its role as an underlying versus contributing cause of death.

Abbreviations

CI

confidence interval

COVID-19

coronavirus disease 2019

Patterns of deaths may have varied over the course of the coronavirus disease 2019 (COVID-19) pandemic. In its beginning stages, vaccines had not been developed and there were few effective treatments. The United States authorized emergency use of the Pfizer–BioNTech COVID-19 vaccine in December 2020 (1), 10 months after the first significant COVID-19 deaths in the United States, and by May 2021, vaccines were widely available to US adults. More recently, several medications have been approved for the treatment of COVID-19 (2), and medical providers have become more effective in managing the disease (3).

There have also been changes in numerous other factors that may differentially affect the vulnerability of population subgroups. Early in the crisis, the economy went into a tailspin, many businesses closed, and the most vulnerable were the elderly, particularly those in nursing homes, as well as persons in jobs not allowing social isolation, or those with preexisting structural risk factors (4–6). Over time, the economy largely recovered, many individuals returned to in-person employment, and there were substantial regional and local variations in mask wearing, social distancing, and vaccination rates. Similarly, stress on the medical system varied across time and space (7), shifting the relative risk of death from non–COVID-19 causes. There may also have been changes in risk factors for a variety of external sources of death, and these may have accumulated over time and had heterogeneous effects across groups (8, 9).

The standard method of calculating the death toll of the COVID-19 pandemic involves estimating the counterfactual number of fatalities expected in its absence and taking the difference between this number and observed counts. Numerous estimates of these “excess deaths” have been obtained for the United States and internationally (10–20), but less is known about how patterns of excess mortality have varied over the course of the pandemic.

These issues are addressed here by examining how excess deaths changed overall and across groups, causes, and locations between the first year of the substantial number of COVID-19 deaths in the United States, from March 2020 through February 2021 (year 1), and the second year, covering March 2021 to February 2022 (year 2). In addition, the relative contributions of COVID-19 versus other causes of mortality to excess deaths are investigated, as is the distinction between COVID-19 as an underlying versus a contributory cause of death. This investigation is particularly salient given the exceptionally large reductions in life expectancy observed in the United States since the COVID-19 pandemic began (21, 22).

METHODS

Data

Data on the universe of deaths for 2021 and earlier years came from the Centers for Disease Control and Prevention (CDC) Multiple Cause of Death data, downloaded from CDC Wonder (23). Final death counts for 2022 were not available at the time of this study, so data for January and February of that year were obtained from the CDC’s data set Provisional Mortality Statistics, 2018 Through Last Month, also available on CDC Wonder (24). These data sets provided information from death certificates on demographic characteristics and underlying and contributory causes of death, categorized using 4-digit International Classification of Diseases, Tenth Revision (ICD-10) codes. Population data were obtained from the CDC’s Bridge-Race and Single-Race Population Estimates data sets (25, 26).

Total mortality counts were computed, as were those by sex, race/ethnicity, age, and the following causes: COVID-19; heart disease; malignant neoplasms (cancer); cerebrovascular disease (stroke); Alzheimer disease; diabetes; chronic lower respiratory disease; influenza; pneumonia; kidney disease; and alcohol-related causes (alcohol), drug poisonings (drugs), motor vehicle accidents (vehicle), intentional self-harm (suicide), and assault (homicide). Web Table 1 (available at https://doi.org/10.1093/aje/kwad127) indicates the ICD-10 codes corresponding to each of these causes. For COVID-19, the primary measure indicates deaths for which it was either an underlying or a contributory cause; cases for which COVID-19 was the underlying cause only were also studied. The race/ethnicity groups examined included Hispanics of any race and non–Hispanic Whites, Blacks, Asians (including Pacific Islanders), and Indigenous peoples (namely, American Indians and Alaska Natives). Age-specific deaths were additionally computed for COVID-19 and the 4 external causes.

Results were also separately analyzed for states with “low,” “medium,” and “high” vaccination rates, defined as those where less than 60%, 60% to less than 70%, and 70% or greater, respectively, of the population had received the second dose of a 2-dose COVID-19 vaccine or a single dose of the J&J/Janssen vaccine as of April 6, 2022, using data from the CDC as reported by the New York Times (27). Details on the states included in each category are contained in Web Table 2. Web Appendix 1 supplies additional details.

Methods

Excess mortality was calculated as the difference between actual deaths and the counterfactual “baseline” that would have been expected without the COVID-19 pandemic. These total, group-specific, cause-specific, and age-by-cause baselines were determined using a variation of recently developed procedures designed to use specifications providing the best goodness of fit for the particular outcome (19, 20). Monthly mortality counts from March 2009 through February 2022 were first computed. When using provisional data, deaths reported as of February 5, 2023, and January 2, 2023, were compared to examine whether there was undercounting due to delayed recording of deaths. There was little or no indication of this, as expected, because the provisional data were updated more than 11 months after February 2022, the latest month analyzed (Web Table 3).

The monthly counts were then aggregated into 12-month March through February periods, hereafter denoted by the year containing the majority of months (e.g., 2020 refers to the March 2020 through February 2021 period). Yearly deaths were next adjusted for excess idiosyncratic volatility in January of each year, and these adjusted counts were converted into mortality rates. Poisson regression was then used to regress the adjusted mortality rates on linear and quadratic time trends, with total or group population as the exposure variable and with robust standard errors used to account for potential heteroscedasticity. Population was estimated at the beginning of the calendar year over which deaths were measured, with adjustments for race-specific populations incorporated as needed to adjust for changes in reporting categories (Web Tables 4 and 5). Mortality rates were not age adjusted, because the inclusion of linear or quadratic trends accounted well for changes in the age distribution of the population over time (Web Figure 1).

The analysis period started between 2009 and 2013, and ended in 2019. A root-mean-square error goodness-of-fit criterion was used to determine the starting analysis year and the choice of linear versus quadratic trend specification (Web Table 6). Baseline death counts were estimated using the preferred Poisson regression specification, with the trend variables set to their 2020 and 2021 values for years 1 and 2 of the pandemic; 95% confidence intervals (CIs) were computed on the basis of the estimated robust standard errors.

Excess deaths were calculated as the difference between observed year 1 or year 2 fatality and baseline counts, with CIs obtained from the baseline estimates and actual fatalities treated as known values. Excess deaths as a percentage of baseline were also computed, as were deaths with COVID-19 as an underlying or contributory cause. The latter were also expressed as a percentage of all excess deaths. Finally, fatalities with COVID-19 as the underlying cause were calculated as a proportion of deaths for which COVID-19 was listed as either an underlying or contributing cause. Further details are provided in Web Appendices 2 and 3.

RESULTS

Evolution of excess deaths

Excess deaths fell 10% between the first and second year of the COVID-19 pandemic, from 655,735 (95% CI: 619,028, 691,980) to 585,605 (95% CI: 532,823, 639,205), largely because of a 62,597 reduction in actual counts. This represented a decrease in excess mortality from 22.8% (95% CI: 21.3, 24.4) to 20.4% (95% CI: 18.2, 22.6) of baseline. These results, and corresponding findings for population subgroups, are summarized in Table 1 and Figure 1, with additional details in Web Tables 7 and 8.

Table 1.

Excess Deaths by Population Subgroup, United States, March 2020 to February 2022

	No. of Excess Deaths				in Excess Deaths as % of Baseline: Year 2 vs. Year 1a
Group	Year 1 (March 2020–February 2021)		Year 2 (March 2021–February 2022)
	Estimate	95% CI	Estimate	95% CI
All	655,735	619,028, 691,980	586,505	532,823, 639,205	–2.5
Sex
Male	363,614	349,187, 377,904	328,603	306,794, 350,102	–2.6
Female	292,116	269,493, 314,375	257,274	224,572, 289,216	–2.4
Race/ethnicityb
White	379,395	350,703, 407,716	388,627	346,484, 429,974	0.5
Black	117,534	113,471, 121,552	88,757	81,655, 95,724	–8.5
Hispanic	116,289	113,095, 119,438	74,553	69,726, 79,280	–20.0
Asian	27,049	26,576, 27,519	15,485	14,533, 16,425	–14.8
Indigenous	9,130	8,970, 9,288	9,938	9,685, 10,188	4.9
Age, years
<25	8,078	5,686, 10,373	14,159	10,570, 17,524	12.0
25–44	41,733	36,646, 46,647	66,975	57,824, 75,573	18.0
45–64	131,297	126,131, 136,414	167,317	159,462, 175,058	6.9
≥65	469,026	444,085, 493,680	352,433	314,558, 389,651	–5.4
State vaccination ratec
Low	170,400	160,885, 179,802	178,306	163,576, 192,766	0.9
Medium	259,472	240,198, 278,439	264,631	236,756, 291,870	0.4
High	225,750	217,197, 234,222	143,830	131,143, 156,340	–9.2

Abbreviations: , change; CI, confidence interval; COVID-19, coronavirus disease 2019.

^a Point estimate of change in excess deaths as a percentage of baseline between year 1 and year 2.

^b White, Black, Asian, and Indigenous refer to non-Hispanic people.

^c Indicates percentage of persons in state who are fully vaccinated against COVID-19, meaning they had received a second dose in a 2-dose COVID-19 vaccine series or 1 dose of the single-shot J&J/Janssen vaccine. “Low,” “medium,” and “high” refer to rates <60%, 60%–69%, and ≥70%, respectively, as of April 6, 2022.

Figure 1.

Excess deaths in the United States as a percentage of baseline deaths from March 2020 to February 2022, by pandemic year and group. Years 1 and 2 cover March 2020 to February 2021 and March 2021 to February 2022, respectively. Error bars show 95% confidence intervals. VR, vaccination rate.

Declines in excess deaths between years 1 and 2 were observed overall and for males, females, Hispanics, Asians, and Blacks. Although the reductions were fairly similar as a percentage of baseline for both sexes, those for Hispanics, Asians, and Blacks contrasted with no change or increases for White people and Indigenous peoples. This partially reflected the greater percentage growth in excess deaths of minorities and Hispanics than of White people in the first year of the pandemic. As a result, Hispanics continued to have significantly more excess deaths as a percentage of baseline in year 2 (32.5%; 95% CI: 29.8, 35.3) than White people (17.8%; 95% CI: 15.6, 20.1), and there also were modestly higher rates for Black people (24.4%; 95% CI: 22.0, 26.9). However, by year 2, the excess deaths of Asians, relative to baseline, (18.6%; 95% CI: 17.3, 20.0) were similar to those of White people. By contrast, the increase in the excess death rates of Indigenous peoples was relatively large in year 1 and then increased more in year 2, leaving this group with the largest excess-to-baseline death share in the later period (51.5%; 95% CI: 49.5, 53.5).

Patterns of excess mortality differed dramatically with age. Seniors (age 65 years or older) had the most excess deaths in both years, but, as a percentage of baseline, this declined from 22.8% (95% CI: 20.4, 23.2) in year 1 to 16.5% (95% CI: 14.4, 18.5) in year 2, a decrease of almost 147,000 excess deaths. Conversely, excess mortality, relative to baseline, increased by 12, 18, and 7 percentage points for those younger than 25 years, those aged 25–44 years, and those aged 45–64 years, respectively, albeit with large CIs. These changes imply that excess mortality was 26%, 47%, and 32%, respectively, above baseline in year 2 for these 3 age groups.

The results also varied sharply with state vaccination rates. In year 1, excess deaths, as a percentage of baseline, were greater in high- than low-vaccination states, but the pattern reversed in the year 2: excess deaths were 22%, 22%, and 25% above baseline in low-, medium-, and high-vaccination states, respectively, in the first year, compared with 23%, 22%, and 16%, respectively, in the second year. There were 55,000 more excess deaths in high- than low-vaccination states in year 1 but 35,000 fewer in year 2.

Table 2, Figure 2, and Web Tables 7 and 8 provide corresponding information for causes of death. Figure 2 separates these into “major causes” (Figure 2A), with at least 50,000 deaths in the year before the pandemic (March 2019–February 2020), and “other causes” that had fewer fatalities in that year (Figure 2B). This is useful because the estimated percentage changes and their CIs were frequently so much larger for the latter group. (Note the different x-axes in panels A and B.)

Table 2.

Excess Deaths by Cause of Death, United States, March 2020 to February 2022

	No. of Excess Deaths
Cause	Year 1 (March 2020–February 2021)		Year 2 (March 2021–February 2022)
	Estimate	95% CI	Estimate	95% CI	in Excess Deaths as % of Baseline: Year 2 vs. Year 1a
COVID-19
Anyb	538,309		442,146
Underlying cause	489,808		391,318
Heart	37,441	30,678, 44,136	36,491	26,490, 46,343	–0.2
Cancer	−772	−3,628, 2,070	7,922	3,198, 12,609	1.5
Stroke	11,191	7,914, 14,398	14,673	9,837, 19,358	2.3
Alzheimer disease	21,226	14,613, 27,477	15,166	6,066, 23,538	–4.1
Diabetes	15,215	14,440, 15,984	12,053	11,122, 12,975	–3.9
Lower respiratory diseases	–8,189	–10,604, –5,812	–6,200	–9,827, –2,658	1.2
Influenza	–5,823	–9,515, –3,264	–7,202	–14,209, –3,425	–18.1
Pneumonia	3,251	1,797, 4,656	1,151	–583, 2,815	–4.8
Kidney	–1,280	–3,756, 1,088	–415	–4,619, 3,494	1.6
Alcohol	11,142	10,163, 12,098	13,507	11,894, 15,059	5.3
Drugs	23,754	17,775, 29,272	37,820	26,746, 47,374	21.2
Vehicle	5,030	3,536, 6,468	11,506	8,970, 13,875	18.7
Suicide	–3,682	–5,719, –1,726	–1,722	–4,495, 906	4.0
Homicide	7,074	5,814, 8,252	8,564	6,457, 10,442	10.8

Abbreviations: , change; CI, confidence interval; COVID-19, coronavirus disease 2019.

^a Point estimate of change in excess deaths as a percentage of baseline between year 1 and year 2.

^b Includes deaths for which COVID-19 was either an underlying or a contributory cause.

Figure 2.

Excess deaths in the United States as a percentage of baseline deaths from March 2020 to February 2022, by pandemic year and cause. Years 1 and 2 cover March 2020 to February 2021 and March 2021 to February 2022, respectively. A) Results for major causes of death (with at least 50,000 deaths in the year before the pandemic). B) Results for other causes of death. Error bars show 95% confidence intervals.

The largest number of non–COVID-19 excess deaths was due to heart disease, but these were a relatively modest 6% above baseline in both years 1 and 2. By comparison, diabetes mortality was 17% over baseline in year 1 and 13% above it in year 2, and Alzheimer disease deaths were 19% and 15% higher than baseline, respectively. Deaths resulting from lower respiratory disease were 4%–5% below baseline in both years, and influenza deaths almost disappeared, particularly by year 2, in which they were almost 90% under baseline.

Some of the largest increases, in percentage terms, were for deaths from alcohol and external causes. Alcohol-related mortality was 28% above baseline in year 1 and 33% higher in year 2. Point estimates for drug, vehicle, and homicide fatalities were large and increasing over time: from 33% to 54%, 13% to 32%, and 39% to 49% over baseline; however, wide CIs imply that although we can be confident that the COVID-19 pandemic substantially increased the number of these deaths, their evolution over time is less certain. Finally, suicide rates were 7% and 3% below baseline in the 2 years.

COVID-19 deaths

COVID-19 directly accounted for approximately four-fifths of excess mortality in year 1 and three-quarters in year 2 (see Figure 3A and Web Table 9). There was little change in the fraction of these deaths for which COVID-19 was an underlying rather than a contributory cause, with the former accounting for approximately 90% of all COVID-19 fatalities in both years (Figure 3B). The point estimates suggest that the COVID-19 share of excess deaths declined between year 1 and year 2 for most groups, but only modestly, and the measures were imprecise.

Figure 3.

US COVID-19 deaths as a percentage of excess deaths and US deaths with COVID-19 as underlying cause as percentage of all COVID-19 deaths, by pandemic year and group. Years 1 and 2 cover March 2020 to February 2021 and March 2021 to February 2022, respectively. A) COVID-19 deaths as a percentage of all excess deaths. Error bars show 95% confidence intervals. B) Deaths with COVID-19 as an underlying cause as a percentage of all deaths involving COVID-19 (either as an underlying or contributing cause). Results for the first pandemic year (March 2020–February 2021) are shown in the blue-shaded bars; results for the second pandemic year (March 2021–February 2022) are shown in the gray-shaded bars. VR, vaccination rate.

There were 2 exceptions to these general patterns. First, the proportion of excess deaths from COVID-19 was lower for Black people (69% in both years 1 and 2) and Indigenous peoples (72% and 51% in years 1 and 2, respectively) than in years 1 and 2, respectively, for White people (86% and 78%), Hispanics (85% and 80%), or Asians (83% and 77%). Second, the COVID-19 share of excess deaths increased monotonically with age: from 13% and 16% for those younger than 25 years, to 31% and 39%, 69% and 68%, and 92% and 90% for ages 25–44 years, 45–64 years, and 65 years or older. For seniors, the CIs include the possibility that all excess mortality was due to COVID-19. Among persons younger than 25 years, a relatively high proportion (slightly more than one-fifth) of COVID-19 deaths involved the virus as a contributory, rather than underlying, cause.

Age-specific mortality

Total mortality and the number of deaths from most diseases increase sharply with age. In the year prior to COVID-19, persons aged 65 years or older accounted for 74.2% of all deaths, versus just 2.1% and 5.0%, respectively, for those younger than 25 years and those 25–44 years old. COVID-19 mortality was even more skewed, with those younger than 25 years old, 25–44 years old, 45–64 years old, and 65 years or older accounting for 0.2%, 2.4%, 16.8%, and 80.6%, respectively, of such deaths during the first pandemic year. However, because a much larger share of external causes of death occurs among younger adults, it is informative to examine age-specific changes in these over the course of the pandemic. This is done in Table 3, which displays estimates of age-specific excess deaths from all sources, COVID-19, and the 4 external causes. Additional documentation is provided in Web Table 10.

Table 3.

Excess Deaths by Age Group and Selected Causes, United States, March 2020 to February 2022

	No. of Excess Deaths				% of Baseline a
Cause by Age, years	Year 1 (March 2020–February 2021)		Year 2 (March 2021–February 2022)
	Estimate	95% CI	Estimate	95% CI	Year 1 (March 2020–February 2021)	Year 2 (March 2021–February 2022)
All causes
<25	8,078	5,686, 10,373	14,159	10,570, 17,524	14.2	26.2
25–44	41,733	36,646, 46,647	66,975	57,824, 75,573	29.1	47.1
45–64	131,297	126,131, 136,414	167,317	159,462, 175,058	24.8	31.7
≥65	469,026	444,085, 493,680	352,433	314,558, 389,651	21.8	16.5
COVID-19b
<25	1,026		2,399
25–44	12,828		25,771
45–64	90,594		119,729
≥65	433,856		294,240
Drugs
<25	3,202	2,344, 3,923	3,630	2,294, 4,630	71.1	91.2
25–44	13,227	10,115, 16,071	20,335	14,742, 25,074	40.0	65.6
45–64	5,362	3,797, 6,851	8,559	6,681, 10,337	17.5	26.0
≥65	446	227, 655	1,621	1,344, 1,884	9.1	30.3
Vehicle
<25	1,904	1,471, 2,309	2,932	2,320, 3,486	29.1	50.0
25–44	2,675	2,119, 3,207	4,833	3,826, 5,758	22.4	42.1
45–64	1,148	819, 1,468	2,544	1,922, 3,131	10.6	24.2
≥65	–784	–1,093, –485	915	416, 1,385	–9.2	10.9
Suicide
<25	–389	–842, 37	–431	–964, 66	–5.5	–5.8
25–44	–272	–737, 180	791	–18, 1,561	–1.7	4.9
45–64	–2,113	–2,790, –1,462	–1,472	–2,389, –604	–13.0	–9.0
≥65	–308	–505, –115	353	57, 641	–3.2	3.7
Homicide
<25	2,259	1,663, 2,795	2,678	1,745, 3,467	42.3	52.7
25–44	3,731	3,076, 4,339	4,490	3,361, 5,477	44.7	57.3
45–64	1,017	833, 1,191	1,222	898, 1,517	29.5	37.1
≥65	97	24, 165	131	48, 209	8.5	11.4

Abbreviations: CI, confidence interval; COVID-19, coronavirus disease 2019.

^a Point estimate of excess deaths as a percentage of baseline.

^b Includes deaths for which COVID-19 was either an underlying or a contributory cause.

The share of excess deaths resulting from COVID-19 increased monotonically with age, but this pattern was reversed for drug, vehicle, and homicide fatalities. For example, vehicle-related deaths were responsible for 24% and 21% of the excess mortality in years 1 and 2, respectively, for those younger than 25 years and approximately 7% for those aged 25–44 years, but less than 2% of excess deaths for those aged 45–64 years and below 1% for those aged 65 years or older. Similarly, homicides were responsible for 19%–28% and 7%–9% of excess mortality for the 2 youngest age groups, respectively, but less than 1% for those aged 45–64 years, and virtually none among seniors.

The greater contributions of drug, vehicle, and homicide deaths to the excess mortality of younger people rather than older individuals reflected larger pandemic effects and, often, higher relative growth between years 1 and 2. The pandemic increased drug-associated deaths for all 4 age groups, with the largest initial increase in the age groups of 25–45 and 45–64 years—approximately 19,000 excess deaths combined in year 1—and the largest increase between the first and second year, to more than 29,000 excess deaths in year 2.

Excess vehicle- and homicide-related deaths were concentrated among persons younger than 45 years, who experienced approximately 4,600 excess vehicle-associated deaths and 6,000 excess homicides in year 1, increasing to approximately 7,800 and 7,200, respectively, in year 2, although with wide CIs. However, excess vehicle-related and homicide fatalities increased between the first and second pandemic year for all age groups, usually by statistically significant amounts. Deaths with COVID-19 as an underlying or contributory cause also approximately doubled from year 1 to year 2 for those younger than 45 years, from low absolute levels, and increased 13% for the 45–64 year age group, while falling by one-third for those 65 years and older.

The patterns for suicides were quite different, with the overall reduction in excess deaths dominated by decreases among those aged 45–64 years, but with less favorable outcomes in year 2 than year 1 for all age groups except the youngest. By contrast, suicide numbers were below baseline levels for persons younger than 25 years and with no evidence of a change between the 2 pandemic years for them.

DISCUSSION

Hopes that the COVID-19 pandemic would end quickly did not come to pass. In its second year, there were 586,505 excess deaths (95% CI: 532,823, 639,205) in the United States, approximately three-fourths of these directly involving COVID-19. This was approximately 90% as large as the 655,735 (95% CI: 619,028, 691,980) excess fatalities occurring during the pandemic’s first year. However, the patterns of excess mortality changed substantially over time. The entire reduction occurred in states with relatively high vaccination rates, with excess deaths increasing in less-vaccinated states. This probably reflected a direct protective effect of the vaccines and possibly reductions in delayed or foregone medical care among vaccinated individuals (28).

The decreases between years 1 and 2 of the pandemic were heavily concentrated among seniors, with an estimated decline of more than 143,000 excess deaths in this group. By contrast, excess mortality increased for younger individuals, particularly those younger than 45 years. Reductions in excess deaths were much larger, in relative terms, for Black people and Hispanics than for White people, although, because minorities were initially so much more negatively affected (13, 15, 29), these groups continued to have higher excess death rates than those of White people in year 2. By contrast, Asians had essentially the same ratio of excess to baseline deaths in the second year as White people, and Indigenous peoples had the most unfavorable experiences, with large relative growth in excess deaths in year 1 of the pandemic and further increases in year 2.

These patterns are closely connected. The decline in excess deaths for seniors was almost certainly related to their rapid uptake of vaccinations (30, 31), with particularly sharp fatality reductions among nursing-home residents when vaccines became available (32). This also probably explained some of the sizeable declines in excess diabetes and Alzheimer disease deaths, given linkages between COVID-19 and both of these diseases (33–35). Alzheimer disease mortality may also have decreased because vaccinations allowed for reductions in social isolation among nursing-home residents (36–38).

Non-White people had elevated excess death rates compared with that of White people in the pandemic’s first year, which has been attributed to increased exposure risk in work, transportation, and living situations; lower vaccination rates; and a variety of institutional factors (39–41). Many of these influences became less pronounced over time as more individuals returned to in-person jobs, disparities in vaccination rates diminished, and, for Asians and Hispanics, came to exceed those of White people (42). By contrast, the high excess death rates of Indigenous peoples occurred despite relative high vaccination rates (43) and likely reflected the effects of preexisting health problems and large increases in external sources of death such as those related to drugs, overdoses, and traffic accidents (21, 44).

The findings for younger people largely reflected heterogeneity in pandemic effects across causes of death. Vehicle, drug, and homicide fatalities are leading causes of death among adults younger than 45 years, whereas cancer, heart disease, and (recently) COVID-19 dominate for those aged 55 years or older (45). Overall deaths from these latter causes increased in the first year of the pandemic but either declined (COVID-19), were little changed (heart disease), or only modestly increased (cancer) between year 1 and year 2. By contrast, vehicle, drug, and homicide deaths all increased at the start of the pandemic and then often increased further in its second year, particularly for prime-age and younger adults. This resulted in higher excess deaths for these groups as well as relatively large contributions of non–COVID-19 causes to this excess mortality. These results are consistent with prior research emphasizing the role of non–COVID-19 deaths, particularly those from external causes, in explaining excess mortality among working-age adults (46, 47). However, deaths attributed to COVID-19 also increased between the first and second pandemic years for individuals younger than 65 years, while falling substantially for seniors.

Fatalities from drugs, suicide, and chronic liver disease have been called “deaths of despair” (48), and there was a widespread belief that the numbers of these would increase during the pandemic (49–52). But the evidence suggests that despair was not the major factor. Suicide rates were below baseline levels in both years 1 and 2, and other research indicates only modest changes in mental distress during the epidemic (53). Instead, what probably increased was a variety of behavioral and environmental risks. The number of drug deaths may have increased because of social isolation, heightened stress, and decreased access to substance use treatment and emergency services (54, 55), as well as the rapid penetration of fentanyl, which increased the probability of death at given levels of use (56). Higher numbers of traffic fatalities reflected increases in speeding, driving while impaired, and reductions in seatbelt use (9), with further growth in year 2 as miles driven recovered from sharp but transitory pandemic-related decreases in vehicle use (57). The number of alcohol-related deaths likely increased because of higher overall consumption and binge drinking (58, 59). The growth in homicide numbers has been associated with elevated stress, increased proliferation of guns, and diminished trust in the police (60, 61).

The findings for excess deaths from some external causes should be interpreted with caution because changes observed immediately prior to the pandemic present difficulties for predicting baseline mortality counts. For example, drug deaths rose monotonically over time through 2017 but then declined by 4% from 2017 to 2018, before increasing again (to roughly 2017 levels) in 2019 (62). Similarly, the number of suicides declined from 2018 to 2019 (63), making it hard to determine whether the subsequent reductions represented pandemic effects or a continuation of recent trends. Given these complexities, it is not surprising that estimates of excess deaths from these sources were sometimes sensitive to the specific analysis period used. More generally, although allowing the starting year of analysis to vary over groups or causes of death improved goodness of fit of the prediction models, it may complicate comparisons of the calculated excess deaths across them.

This study has additional limitations. First, the provisional mortality data may be incomplete, reported with delays, or misclassify causes of death (64), although, as discussed, this is unlikely to have created major errors here, given the amount of time elapsed since the last month of data analyzed. Second, even when using final death data, cause of death may sometimes have been misclassified, particularly early in the pandemic when coroners and medical examiners had less experience identifying COVID-19. Third, multiple factors potentially affecting excess deaths—such as the availability of hospital beds and staff, the ability of individuals employed in different occupations to engage in social distancing, and the specific effects of a variety of COVID-19–related restrictions—were not considered in this analysis. Therefore, the results should be viewed as descriptive rather than necessarily causal. Finally, this study only considers excess deaths and not other adverse consequences of the COVID-19 pandemic.