A groundbreaking study published in the journal Science Advances has revealed that the early death toll from the Covid-19 pandemic in the United States was significantly higher than official counts indicated. While approximately 840,000 Covid-19 deaths were recorded on death certificates during 2020 and 2021, researchers estimate an additional 155,000 unrecognized deaths likely occurred outside hospital settings. This suggests that about 16% of Covid-19 fatalities went uncounted during those critical years.
Disparities in Undercounted Deaths
The research team employed sophisticated artificial intelligence techniques to analyze mortality data, uncovering stark disparities in which deaths were missing from official tallies. The undiagnosed fatalities were disproportionately concentrated among Hispanic individuals and other people of color, particularly during the initial months of the pandemic. Geographically, these uncounted deaths clustered in southern and southwestern states including Alabama, Oklahoma, and South Carolina.
Systemic Barriers to Accurate Counting
Elizabeth Wrigley-Field, a University of Minnesota researcher and co-author of the study, explained that while hospital patients underwent routine Covid-19 testing, many who became ill and died outside medical facilities were never tested. This testing gap was especially pronounced early in the pandemic when at-home testing options were not readily available.
The study further identified structural problems within America's death investigation system. In many regions, elected coroners without specialized medical training handle death investigations, creating inconsistencies in how causes of death are determined. Some research indicates that partisan attitudes may have influenced whether individuals sought Covid-19 testing or whether coroners pursued postmortem coronavirus examinations.
"Our antiquated death investigation system is one key reason why we fell short of accurate counts, particularly outside of big metropolitan areas," said Andrew Stokes of Boston University, the paper's senior author.Broader Implications and Expert Perspectives
Steven Woolf, a Virginia Commonwealth University researcher not involved in the study, noted that six years after the coronavirus first swept through the United States, many of the same vulnerable populations continue to face barriers to healthcare access. "People on the margins continue to die at disproportionate rates because they can't access care," he emphasized in written comments.
The Centers for Disease Control and Prevention currently reports more than 1.2 million Covid-19 deaths since the pandemic began in early 2020, with over two-thirds occurring during 2020 and 2021. These numbers have been subject to ongoing debate, with some social media misinformation falsely claiming inflated death counts and former President Donald Trump amplifying such claims during his administration.
Methodological Approach and Limitations
The research team focused specifically on deaths caused by coronavirus infection, distinguishing them from other pandemic-related fatalities such as those resulting from overwhelmed healthcare systems or increased drug overdoses. Using machine learning algorithms, scientists analyzed death certificates of infected patients who died in hospitals, then applied observed patterns to evaluate certificates of people who died outside hospitals with causes listed as pneumonia or diabetes.
While scientific understanding of machine learning applications in medical research continues to evolve, Woolf described this team's methodology as "intriguing" and noted that their overall findings align closely with estimates from other pandemic mortality studies.
The research underscores persistent challenges in public health data collection and highlights how systemic inequities can manifest even in how deaths are counted during national emergencies. As the scientific community continues to refine pandemic mortality estimates, this study provides crucial insights into the true human cost of Covid-19 and the populations most affected by undercounting.
