The number of publicly reported deaths from coronavirus disease 2019 (COVID-19) may underestimate the pandemic’s death toll. Such estimates rely on provisional data that are often incomplete and may omit undocumented deaths from COVID-19. Moreover, restrictions imposed by the pandemic (eg, stay-at-home orders) could claim lives indirectly through delayed care for acute emergencies, exacerbations of chronic diseases, and psychological distress (eg, drug overdoses). This study estimated excess deaths in the early weeks of the pandemic and the relative contribution of COVID-19 and other causes.
Weekly death data for the 50 US states and the District of Columbia were obtained from the National Center for Health Statistics for January through April 2020 and the preceding 6 years (2014-2019).1,2 US totals excluded Connecticut and North Carolina because of missing data. The analysis included total deaths and deaths from COVID-19, influenza/pneumonia, heart disease, diabetes, and 10 other grouped causes (Supplement). Mortality rates for causes other than COVID-19 were available only for underlying causes. Death data with any mention of COVID-19 on the death certificate (as an underlying or contributing cause) were used to capture all deaths attributed to the virus. Population counts for calculating mortality rates were obtained from the US Census Bureau.3,4
Observed deaths for the 8 weeks between March 1, 2020, and April 25, 2020, were taken from provisional data released on June 10, 2020.2 Expected deaths (and 95% CIs) for these same weeks were estimated by fitting a hierarchical Poisson regression model to the weekly death counts for the period of December 29, 2013, through February 29, 2020 (assembled from final data for 2014-20181 and provisional data for January 1, 2019, through February 29, 20202). The model with the optimal fit (Supplement) used a combination of harmonic functions to capture seasonality and adjusted for annual trends with a categorical year effect. The model allowed season and time trends to vary by state.
Excess deaths equaled the difference between observed and expected deaths and were summed across the 8 weeks to estimate total excess deaths. To explore increases in cause-specific mortality in jurisdictions overwhelmed by COVID-19, mortality trends for 14 grouped causes (4 reported here) were examined in the 5 states with the most COVID-19 deaths from March through April 2020 (Massachusetts, Michigan, New Jersey, New York, and Pennsylvania). Deaths in these states peaked in the week ending on April 11, 2020, and the proportional increase above baseline (weighted mean of weekly deaths over 9 weeks in January to February 2020) was measured. All calculations were performed using SAS, version 9.4 (SAS Institute Inc).
Between March 1, 2020, and April 25, 2020, a total of 505 059 deaths were reported in the US; 87 001 (95% CI, 86 578-87 423) were excess deaths, of which 56 246 (65%) were attributed to COVID-19. In 14 states, more than 50% of excess deaths were attributed to underlying causes other than COVID-19; these included California (55% of excess deaths) and Texas (64% of excess deaths) (Table). The 5 states with the most COVID-19 deaths experienced large proportional increases in deaths due to nonrespiratory underlying causes, including diabetes (96%), heart diseases (89%), Alzheimer disease (64%), and cerebrovascular diseases (35%) (Figure). New York City experienced the largest increases in nonrespiratory deaths, notably those due to heart disease (398%) and diabetes (356%).
These estimates suggest that the number of COVID-19 deaths reported in the first weeks of the pandemic captured only two-thirds of excess deaths in the US. Potential explanations include delayed reporting of COVID-19 deaths and misattribution of COVID-19 deaths to other respiratory illnesses (eg, pneumonia) or to nonrespiratory causes reflecting complications of COVID-19 (eg, coagulopathy, myocarditis). Few excess deaths involved pneumonia or influenza as underlying causes.
This study has limitations, including the reliance on provisional data, potentially inaccurate death certificates, and modeling assumptions. For example, modeling epidemiologic years instead of calendar years would reduce the excess deaths estimate to 73 524.
Large increases in mortality from heart disease, diabetes, and other diseases were observed. Further investigation is required to determine the extent to which these trends represent nonrespiratory manifestations of COVID-19 or secondary pandemic mortality caused by disruptions in society that diminished or delayed access to health care and the social determinants of health (eg, jobs, income, food security).
Corresponding Author: Steven H. Woolf, MD, MPH, Center on Society and Health, Virginia Commonwealth University School of Medicine, 830 E Main St, Ste 5035, Richmond, VA 23298-0212 (steven.woolf@vcuhealth.org).
Accepted for Publication: June 16, 2020.
Published Online: July 1, 2020. doi:10.1001/jama.2020.11787
Author Contributions: Drs Woolf and Chapman had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Woolf, Chapman, Sabo, Weinberger.
Acquisition, analysis, or interpretation of data: Chapman, Sabo, Hill.
Drafting of the manuscript: Woolf, Chapman, Sabo, Weinberger.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Chapman, Sabo, Weinberger, Hill.
Administrative, technical, or material support: Woolf, Chapman.
Supervision: Woolf.
Conflict of Interest Disclosures: Dr Weinberger reported receiving grants from Pfizer and the National Institute of Allergy and Infectious Diseases (R01AI137093) and personal fees from Pfizer, Merck, GlaxoSmithKline, and Affinivax outside the submitted work. No other disclosures were reported.
Funding/Support: This study was partially funded by the National Center for Advancing Translational Sciences (grant UL1TR002649).
Role of the Funder/Sponsor: The funder had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.