Autopsies performed on COVID-19 patients have provided critical information about SARS-CoV-2’s tropism, mechanisms of tissue injury, and spectrum of disease.
To provide an updated database of postmortem disease in COVID-19 patients, assess relationships among clinical and pathologic variables, evaluate the accuracy of death certification, and correlate disease variables to causes of death.
The 272 postmortem examinations reported in this paper were submitted by 14 pathologists from 9 medical or forensic institutions across the United States. The study spans the eras of the 3 principal COVID-19 strains and incorporates surveyed demographic, clinical, and postmortem data from decedents infected with SARS-CoV-2, including primary and contributing causes of death. It is the largest database of its kind to date.
Demographics of the decedents reported here correspond well to national statistics. Primary causes of death as determined by autopsy and official death certificates were significantly correlated. When specifically cited disease conditions found at autopsy were correlated with COVID-19 versus non–COVID-19 deaths, only lung findings characteristic of SARS-CoV-2 infection or the absence of lung findings were significantly associated.
Changes in hospitalization and disease likely stem from longer lifespans after COVID-19 diagnosis and alteration in treatment approaches. Although Omicron variants preferentially replicate in the upper airways, autopsied patients who died of COVID-19 in that time period showed the same lung damage as earlier decedents. Most importantly, findings suggest that there are still unelucidated risk factors for death from COVID-19 including possibly genetic susceptibility.
The COVID-19 pandemic caused by worldwide infection with SARS-CoV-2 has established itself as one of the most disruptive infectious disease events in recorded history. The impact of this pandemic has extended well beyond healthcare, with the full scope of its consequent social, economic, cultural, and political upheaval yet to be determined. The pandemic has also been characterized by a succession of pathologic viral variants,1 changes in masking and social distancing policy and practice,2 improved clinical management of COVID-19,3 and the rapid deployment of effective vaccines against SARS-CoV-2.4
As of March 10, 2023, 676 million cases of COVID-19 have been reported around the world, with more than 6.8 million associated deaths. In the United States, nearly 104 million confirmed cases and more than 1.1 million associated deaths have been reported.5 However, achieving consensus on the death toll from COVID-19 has proven surprisingly difficult.6 Clinical determination of the cause of a patient’s death can be inconsistent, even under the best of circumstances, and statistical estimates of excess deaths can significantly vary depending on the chosen models and underlying assumptions in such determinations.7
From the earliest days of the pandemic, autopsies performed on those dying of or with COVID-19 in both hospital and forensic settings have provided critical information about SARS-CoV-2’s tropism, mechanisms of tissue injury, and the spectrum of disease,8–15 and this information has helped advance the treatment of COVID-19. Moreover, through the examination of postmortem pathology in decedents with SARS-CoV-2 infection, autopsies have provided valuable perspective in interpreting COVID-19 death statistics.16 Our present study incorporates comprehensive demographic, clinical, postmortem, and cause of death data from 272 decedents infected with SARS-CoV-2. This represents the largest and most comprehensive database of its kind to date. As a follow-up to a previous multi-institutional autopsy paper by many of the same authors,16 this study provides an updated overview of postmortem pathology in COVID-19, assesses relationships among various clinical and pathologic variables in the disease, including an infected patient’s vaccination status, and evaluates the accuracy of clinical and autopsy-determined death certification in COVID-19.
This multi-institutional autopsy study helps identify critical risk factors for death from COVID-19 and informs both individual patient treatment and public policy guidance relevant to the disease.
METHODS
Study Design and Participants
A written invitation to participate in this multi-institutional survey was disseminated to all participants from a previous autopsy study.16 The 272 postmortem examinations reported in this study were submitted by 14 resident and/or attending pathologists from 9 medical or forensic institutions across the United States. The autopsies were performed between June 1, 2020, and January 31, 2022, and data were collected, compiled, and analyzed in March and April 2022. All contributing pathologists confirmed complying with applicable policies and procedures governing human subjects’ research at their respective institutions, as well as obtaining appropriate consents for the autopsy procedure.
Data from each autopsy were collected through online submission of responses to a 26-question survey using the REDCAP database, which was also used for compilation and preliminary analysis. The first 6 survey questions elicited submitter and patient demographics and the time frame of autopsy performance. Questions 7 through 14 evaluated clinical variables of the decedents, including vaccination status, preexisting conditions, and hospital course. The next 9 questions documented postmortem disease findings by organ system. Finally, the last 3 questions defined primary and contributing causes of death determined by the decedent’s clinician or other reporter on the death certificate, and by the pathologist in the autopsy report.
Meaningful relationships among variables highlighted in the survey were then quantified by χ2 correlations. Notably, the vaccination status of each decedent was correlated with the primary cause of death both as recorded in the autopsy report and on the death certificate. Percentages of preexisting diseases and acquired clinical conditions during hospitalization were compared to the national Centers for Disease Control and Prevention (CDC) estimate of the population average, using a 1-sample proportions test with continuity correction. Major diseases documented in autopsy reports were analyzed in relation to COVID-19 versus non–COVID-19 death as determined from autopsy and death certificates using univariable logistic regression, in order to identify possible risk factors associated with death from COVID-19. All raw P values reported from univariable logistic regression models were adjusted for multiple comparisons using the false-discovery rate (FDR) correction by Benjamini and Hochberg.17 We also performed sensitivity analyses using multivariable logistic regression to confirm the significant findings.
Role of the Funding Source
This study was supported by resources internal to the Stanford University School of Medicine and used no external funding sources in its creation.
RESULTS
Database Demographics
The mean (60 years) and median (62 years) ages of decedents in the present study are extremely close to those reported in the previous data set (61 and 63 years, respectively) (Table 1). The eldest decedent in each survey was nearly 100 years old, and while the youngest decedent in the previous cohort was an infant (0.6 years), an adolescent (15 years) was the youngest decedent in the present study. The data set of the present study included a nearly equal proportion of males and females, with a slightly greater percentage of autopsied females (49%, 132 of 272 cases) compared to that in the previous survey (41% females, 55 of 135 cases). The most notable demographic change in the present cohort compared to the prior study was a greater percentage of Black or African American decedents (46%, 124 of 272 compared to 30%, 40 of 135 cases) and a corresponding decrease in the percentage of White decedents, from 50% (68 of 135) to 39% (106 of 272). The number of Hispanic decedents evaluated in the present data set decreased compared to the previous survey (17%, 23 of 135 cases compared to 13%, 35 of 272), most likely because several Brazilian pathologists participated in the previous study but not the current study.
Demographic Comparison of Initial Study of 135 COVID-19–Positive Autopsies (February 1—May 30, 2020) and Current Study of 272 COVID-19–Positive Autopsies (June 1, 2020–January 31, 2022)

The majority of COVID-19 autopsies in the present data set were performed in the first time period (June 1, 2020–April 30, 2021) when the Alpha strain of COVID-19 was most prevalent (177 of 272 autopsies, 65% of study cases) (Table 2). From the time period when the Delta strain was most common (May 1–November 30, 2021), 73 of 272 autopsies (27% of study cases) were performed. Finally, in the third time period of the Omicron strain (December 1, 2021–January 31, 2022), 22 of 272 autopsies (8% of study cases) were reported.
General Environmental Characteristics of Patients in 272 COVID-19–Positive Autopsies (June 1, 2020–January 31, 2022)

The COVID-19 vaccination status of approximately one third of autopsied decedents was unknown (90 of 272 or 33% of study cases) in the present survey. However, 156 of 272 decedents (57% of study cases) were reported to be unvaccinated, 13 decedents (5% of study cases) were partially vaccinated, and 13 decedents (5% of study cases) were fully vaccinated (ie, they had received 1 or 2 primary vaccinations and at least 1 booster). In January 2022 when the present cohort’s data collection period ended, 76% of the US population (251 518 114 of 332 403 650 people) was at least partially vaccinated, with 64% (212 738 336 of 332 403 650) having received 1 or 2 shots.17 As expected, it appears that more unvaccinated patients died (and were autopsied) than vaccinated patients. However, among decedents with known vaccination status, the presence of any vaccination exposure was not significantly correlated to death from COVID-19 versus non–COVID-19 death as diagnosed both by death certificate or autopsy. While intriguing, this finding must be interpreted with caution, as the vaccination status of so many decedents in this database is unknown.
Decedents in the present study were hospitalized from 0 (ie, death outside of the hospital or in the emergency department) to 151 days, with a mean and median hospital stay duration of 17 and 11 days, respectively. Hospitalization (and overall life span from diagnosis to death) was shorter earlier in the pandemic measurement period (mean and median hospitalization of 10 and 8 days, respectively, in the previous study). Decedents in the current data set spent from 0 to 517 days on a ventilator, with a mean and median ventilator dependency of 11 and 3 days, respectively. The decedent who spent 517 days on a ventilator was not hospitalized for most of his terminal illness and was likely in a rehabilitation facility. The mean and median ventilator durations in the earlier study were 6.5 and 4 days, respectively.
Reported Preexisting Diseases and Acquired Clinical Conditions in SARS-CoV-2–Infected Decedents
Each decedent in the present cohort had an average of 3.04 preexisting conditions, compared to 2.88 such conditions among autopsied decedents in the previous COVID-19 study. Systemic hypertension represented the most frequently cited preexisting condition for decedents in both the present (177 of 272, 65% of study cases) and previous (86 of 135, 64% of study cases) data sets. These numbers are both significantly higher than the national CDC estimate of 47% of the US adult population (both P values <.001).18 Diabetes mellitus was also frequently cited in both study populations. Although its prevalence decreased from the previous to the present survey (52%, 70 of 135, to 38%, 102 of 272 study cases), it remained much more prevalent in the study cohort than the CDC estimate of the disease’s prevalence among US adults of 11.3% (both P values <.001).19 Obesity was more frequently seen among decedents in the present study (112 of 272, 41% of study cases) compared to the previous data set (46 of 135, 34% of study cases) (P = .16). Cardiovascular disease (including coronary artery disease, vascular disease, and hyperlipidemia), as well as malignancy or rheumatologic disease, each demonstrated an increase in citations among decedents between the previous and present surveys (Figure 1). Only the present cohort included substance use (eg, smoking, alcohol, drugs) as a preexisting condition. Of note, the prevalence of preexisting liver disease remained about the same between the previous and present studies. The percentages of preexisting neuromuscular disease, lung disease, kidney disease, and heart disease (other than hypertension or coronary artery/vascular disease) all remained within 2 percentage points of the previous data set’s findings.
Preexisting conditions in initial survey of 135 COVID-19–positive autopsies (February 1—May 30, 2020) and current study of 272 COVID-19–positive autopsies (June 1, 2020–January 31, 2022).
Preexisting conditions in initial survey of 135 COVID-19–positive autopsies (February 1—May 30, 2020) and current study of 272 COVID-19–positive autopsies (June 1, 2020–January 31, 2022).
As in the previous survey, clinical conditions acquired by decedents during their terminal hospitalizations were categorized by organ system in the present cohort (Figure 2). Decedents in the present study developed an average of 2.42 conditions, compared to an average of 2.85 acquired conditions among autopsied decedents in the previous data set. Acute respiratory disease remained the most frequently cited acquired condition and was slightly more frequent in the current cohort. Acute kidney or genitourinary disease remained the second most commonly cited condition during terminal hospitalization (44%, 122 of 272 versus 54%, 72 of 135 cases). There was a striking decrease in the incidence of acute coagulopathy/disseminated intravascular coagulation among decedents in the present compared to the previous cohort (16%, 43 of 272 compared to 34%, 46 of 135; P = .001). Furthermore, in the present compared to the previous study, fewer decedents presented with newly diagnosed acute myocardial disease, neurologic disease, or gastrointestinal disease during their terminal hospitalizations (all P > .05). Acute liver disease remained largely unchanged between data sets. Finally, the incidence of sepsis acquired during terminal hospitalization slightly increased from the previous survey.
Clinical hospitalization conditions in initial survey of 135 COVID-19–positive autopsies (February 1—May 30, 2020) and current study of 272 COVID-19–positive autopsies (June 1, 2020–January 31, 2022). Abbreviations: DIC, disseminated intravascular coagulation; GI, gastrointestinal.
Clinical hospitalization conditions in initial survey of 135 COVID-19–positive autopsies (February 1—May 30, 2020) and current study of 272 COVID-19–positive autopsies (June 1, 2020–January 31, 2022). Abbreviations: DIC, disseminated intravascular coagulation; GI, gastrointestinal.
Autopsy Disease Findings
In the present cohort, disease pathologies found at autopsy of individuals with COVID-19 most frequently involved the lungs, heart, and kidneys, similar to what was demonstrated in the previous study (Table 3). Acute and/or organizing phases of diffuse alveolar damage (DAD) remained the most common pathologies, accounting for nearly half of all the present data set’s citations in the pulmonary system. However, the overall percentage of decedents in whom DAD was cited decreased from 75% (101 of 135) to 54% (148 of 272) for acute-phase DAD (P < .001) and from 47% (63 of 135) to 37% (101 of 272) for organizing-phase DAD (P = .08). Approximately half of decedents in the current study had macrothrombi or microthrombi in the lungs, though (where evaluated), deep leg venous thrombi were infrequent. There were also very few thrombi in other sites, including the kidneys.
Organ System and Disease Most Cited in 272 COVID-19–Positive Autopsies (June 1, 2020–January 31, 2022)

The prevalence of some chronic pathologies—which we interpret to reflect preexisting conditions—remained similar across the present and previous surveys, including cardiac myocyte hypertrophy, as well as renal arteriosclerosis/arteriolosclerosis and focal glomerulosclerosis. However, the prevalence of coronary atherosclerosis was greater in the present compared to the previous cohort (41%, 111/272, compared to 25%, 34 of 135; P = .002), while interestingly, the prevalence of reported myocardial fibrosis was slightly lower in the present compared to the previous study (36%, 99 of 272, compared to 42%, 57 of 135; P = .27). Taken together, these chronic pathologies accounted for 70% (313 of 447) and 46% (230 of 501) of cited cardiovascular and renal system pathologies, respectively, in the present data set. Regarding liver disease, the present compared to the previous survey demonstrated a slight increase in incidence of cirrhosis among decedents (P = .63), whereas incidence of hepatitis slightly decreased (P = .08). Also, histologic findings compatible with low perfusional states (ie, hypoxia or ischemia) were again frequently cited in the liver and kidney.
Primary and Contributing Cause of Death Determinations
Overall, in the present cohort, there was strong agreement (κ = 0.63, 95% CI: 0.53–0.73) between cause-of-death determinations in death certificates completed by clinicians (based on antemortem clinical information) and autopsy reports completed by pathologists (based on postmortem findings), as borne out by unweighted Cohen κ (Table 4). Respiratory disease related to COVID-19 represented the leading cause of death determined by both clinicians and autopsy pathologists in the present study (Table 5). DAD or pneumonia related to COVID-19 together accounted for 60% (162 of 272) of death certificate and 61% (166 of 272) of autopsy cause-of-death determinations. In comparison, acute respiratory disease, most often cited as “COVID-19 pneumonia” or “acute lung injury,” accounted for 75% (101 of 135) of decedents’ cause of death determinations in the previous data set.
Primary Cause of Death From Postmortem Examination and Primary Cause of Death From Death Certificate in the Current Study of 272 COVID-19–Positive Autopsies (June 1, 2020–January 31, 2022)

Primary Cause of Death and Most Important Other Cause of Death From Postmortem Examination (PME) in the Current Study of 272 COVID-19–Positive Autopsies (June 1, 2020–January 31, 2022)

Of the 272 decedents included in the current survey, 190 individuals (approximately 70%) had permutations of “COVID-19” listed as the cause of death, while the remaining 82 had some other unrelated cause of death listed. There was no significant difference in the number of comorbid conditions between these 2 groups of decedents.
Autopsy findings in the pulmonary, cardiac, renal, and neurologic systems were analyzed against death from COVID-19 versus death from other causes (ie, non–COVID-19 death) as documented by autopsy and death certificate. After correcting for the FDR, surprisingly only some pulmonary pathologic findings were significantly correlated with death from COVID-19. Specifically, DAD (both acute and organizing, both FDR-adjusted [FDR-adj] P < .001), thrombi in the lungs (FDR-adj P < .001), and pulmonary fibrosis (FDR-adj P = .01) were significantly and positively associated with death from COVID-19 versus death from other causes, and the absence of any pulmonary pathologic findings (FDR-adj P < .001) was associated with death from other causes (non–COVID-19 death). Other reported categories of pulmonary pathology (tracheitis/bronchitis/bronchiolitis, pulmonary hemorrhage, pulmonary malignancy) were not significantly associated with death from COVID-19. Similarly, no significant differences were recorded between the 2 groups with respect to prevalence of reported cardiac, renal, or cerebral pathologic findings.
For 198 of the 272 (73%) autopsies included in the present cohort, submitting pathologists indicated the next most important (significant) disease process at autopsy that contributed to death. Combined chronic diseases, including systemic hypertension, coronary artery atherosclerosis, and (to a lesser extent), diabetes mellitus and obesity, were frequently cited as contributing causes of death. When primary and contributing causes of death at autopsy were combined, the percentage of deaths from COVID-19 substantially decreased (Table 6). In patients with a primary cause of death other than COVID-19, COVID-19 was only rarely listed as the most important contributing cause.
Limitations
This data set incorporated only autopsied decedents who were COVID-19–test positive at the time of death. Autopsy findings were not centrally reviewed, although submitters were or were supervised by experienced autopsy pathologists. Some differences between the 2 study cohorts might be accounted for by the absence of Brazilian engagement in the present cohort (only US institutions are represented).
DISCUSSION
We report findings from the largest COVID-19 autopsy cohort to date, which included data from 272 autopsies performed at 9 institutions from across the United States. All cases were performed or supervised by experienced pathologists, and all institutions had participated in our previous COVID-19 autopsy study. Evaluation in this study spans the eras of the 3 principal COVID-19 strains to date: Alpha, Delta, and Omicron. The demographics of decedents included in this data set are comparable in age, sex, and ethnicity to those in the United States believed to have died from COVID-19 overall. Of note, a change in the racial composition of decedents included in the prior survey from earlier in the pandemic and the current survey performed later in the pandemic corresponds to the overall increased proportion of Black or African American decedents during the pandemic.20 The respective proportions of autopsies included in this cohort’s 3 time periods correspond well to US mortality rates caused by each dominant COVID-19 strain during those same time periods.20 There does not appear to be significant selection bias for patients who were autopsied and reported in this study.
Major comorbid diseases cited remained mostly consistent across our 2 data sets; the increase in reported obesity might simply reflect more frequent clinical documentation of this comorbidity in clinical charts because of increasing acknowledgment that obesity represents a risk factor for lethal COVID-19. According to the CDC, 42% of the US population is obese,21 and thus the frequency of obesity among decedents in our survey was similar to that seen in the general population. The increased incidence of cirrhosis might correlate with the inclusion of alcohol and drug use in the present cohort.
Primary causes of death as determined by autopsy significantly correlated with clinical causes of morbidity listed on death certificates, indicating that previous concerns about death certificate accuracy or bias are not supported by this study’s data, at the very least for deaths that occurred in the middle to later time periods of the pandemic.
Decedents in the present data set were hospitalized for longer periods of time than in the previous study, which could be due to longer lifespans resulting from more effective COVID-19 treatments, greater prehospitalization awareness of COVID-19 as a serious illness, or a combination of these and other factors. The reduced median ventilation days between the cohorts during the course of the pandemic may reflect shifts in treatment paradigms, including greater use of steroids and attempted avoidance of ventilation in patients with different respiratory failure phenotypes, including the use of continuous positive airway pressure devices as well as non-invasive ventilation.22
The number of preexisting diseases in the decedents increased slightly from the previous study, while the types of diseases cited were highly similar between the 2 data sets. The autopsies showed many end-organ signs resulting from the cited chronic conditions, including frequently cited cardiac fibrosis (likely reflecting hypertensive cardiomyopathy and/or remote infarction[s] due to coronary artery atherosclerosis), kidney glomerulosclerosis and arteriosclerosis (likely related to systemic hypertension and/or diabetes mellitus), and liver fibrosis and inflammation (likely due to hepatitis and/or substance use). Remarkably, only 9 decedents in the current cohort had no preexisting conditions indicated. Preexisting disease has been widely accepted as a crucial risk factor for severe COVID-19 manifestations and for death from SARS-CoV-2 infection. However, the sheer number of preexisting conditions did not significantly correlate with death from COVID-19 versus death from other causes as determined by both autopsy and death certificate. This analysis included up to 6 individual conditions and conditions grouped into sets of 1–2 and 3+, and 1–3 and 4+. There may have been some concern that extant disease that did relate to death may have been underreported by autopsy pathologists. However, since the numbers of preexisting diseases cited per decedent actually increased from the previous data set, the data suggest that the assumed importance of preexisting conditions to COVID-19 lethality may be more complex than previously suspected and should be the subject of further study.
Categorization of conditions acquired during terminal hospitalization showed little change from the previous survey conducted earlier in the pandemic, except for an observed reduction in the incidence of coagulopathy/disseminated intravascular coagulation. This reduced incidence of coagulopathy among individuals dying with COVID-19 may reflect improved clinical recognition and early treatment of coagulopathy in COVID-19.22 Additionally, the slight increase in clinical citations for sepsis and the relatively consistent demonstration of autopsy findings consistent with shock in the previous and present studies suggest that sepsis may have evolved from superinfections in those who survived longer due to better clinical management of COVID-19.
Changes in disease findings at autopsy over time could also be explained by changes in treatment approaches altering pathologic manifestations of disease. Though the decrease in acute DAD could be attributed to overall increasing length of survival among patients with COVID-19, this would not explain the decrease in organizing DAD. However, the decrease in organizing changes could be attributed to the more judicious use of ventilation in COVID-19 patients. Similarly, the incidence of some other disease pathologies also decreased between the previous and present data sets, including acute tubular injury (from 52% to 19% of study cases) and pulmonary hemorrhage (from 43% to 21% study cases). These findings are of significant interest, since the percentages of patients reported as having clinical pulmonary disease during hospitalization did not decrease substantially between the datasets. Although the SARS-CoV-2 virus has evolved and adapted during the course of the pandemic and its associated mortality rates have declined, the mechanism of lethality, namely acute respiratory distress syndrome/DAD and the subsequent scarring, appears remarkably unchanged. Although Omicron variants in particular will preferentially replicate in the upper airways, autopsied patients who died of COVID-19 in that time period show the same lower airway damage as decedents examined when the Alpha and Delta strains were dominant.23
Further, when all specifically cited disease conditions found at autopsy were correlated with death due to COVID-19 versus death due to other causes, only lung findings very characteristic of SARS-CoV-2 infection (eg, DAD and thrombi) versus the complete absence of any lung findings were significantly associated. It is remarkable that no other pathologic findings, including those caused by preexisting conditions or acute hospitalization conditions, correlated with death caused directly by SARS-CoV-2 infection. This was uniformly true in both death certificate and autopsy-determined causes of death.
CONCLUSIONS
The findings from this large and multi-institutional data set may suggest that there are as yet unelucidated risk factors that correlate with death from COVID-19. Patient genetic susceptibility may be a key component of this relationship. Genetic sequencing of fresh, frozen, or formalin-fixed tissue from autopsied COVID-19–positive decedents, correlating that information with genetic strains of SARS-CoV-2 causing infection, and linking these data to causes of death would likely be valuable in improving understanding of the clinical implications of specific strains of SARS-CoV-2. Furthermore, major autopsy studies spanning different geographic areas and care settings may be a crucial source of population morbidity and mortality data that can signpost areas for future in-depth investigation. These data invite consideration that current common assumptions about risk factors for COVID-19 severity may not be all that the pandemic has to teach medical science.
The authors would like to gratefully acknowledge the conceptual contributions and insights of Anthony Rosen, MD, PhD, Johns Hopkins University, Baltimore, Maryland.
References
Competing Interests
The authors have no relevant financial interest in the products or companies described in this article.