A new study utilized a novel machine learning method to identify unrecognized COVID-19 deaths during the first two years of the pandemic and found that 19 percent more COVID-19 deaths occurred than official records indicate. These deaths disproportionately affected certain populations, suggesting that the US death investigation system hid the true extent of pandemic mortality and inequities.
For decades, health experts have urged reform of the death investigation system in the United States, calling it a fragmented and underfunded network of investigators, many of whom lack the scientific training and resources necessary to determine causes of death accurately and equitably.
A new analysis of undercounted COVID-19 deaths led by researchers at Boston University School of Public Health (BUSPH) and Stanford University signals that this reform is urgently needed.
The study found that more than 155,000 US deaths between March 2020 and December 2021 were not officially recorded as COVID-19 deaths, which equates to 19 percent more deaths due to COVID-19 than federal records indicate. These unrecognized COVID-19 deaths disproportionately burdened certain populations more than others, including racial and ethnic minorities, as well as people who were low-income, had preexisting health conditions, lived in the South, or did not have a high school education. The findings were published in the journal Science Advances .
Previous studies have estimated unrecognized COVID-19 deaths using excess mortality models, which compare actual deaths of all causes to expected deaths as a result of the pandemic. While this method has provided a window into the inequitable effects of the pandemic, it does not differentiate between deaths specifically caused by SARS-CoV-2 (the virus that causes COVID-19 infections) and deaths that resulted from indirect factors, such as delayed healthcare, mental health crises, or other social and economic challenges. The new analysis utilizes a novel machine learning method to predict these unrecognized COVID-19 deaths, which offers a more specific estimate of deaths directly caused by SARS-CoV-2 infection.
"Without an accurate count of who is dying and where, public health resources can't reach the communities that need them most," says study senior and corresponding author Dr. Andrew Stokes , associate professor of global health at BUSPH. "The COVID-19 pandemic exposed the long-standing gaps within the American death investigation system, and the changes that are urgently needed to improve the quality of cause-of-death data—for all deaths, not just those caused by COVID-19."
Since the early days of the pandemic, Dr. Stokes has led a team of researchers to develop an extensive body of research on excess mortality, aiming to capture the pandemic's hidden death toll, the populations and communities most burdened, and how these trends have shifted over time. The machine learning algorithms in the latest study are more reliable than the excess mortality models in the team's prior work because the algorithms were informed by federal health datasets of hospital-verified, inpatient COVID-19 death information, as opposed to estimates derived from broader all-cause mortality data.
Based on evidence from the team's prior work that indicated that non-COVID deaths that occur out of the hospital during the pandemic typically mirrored inpatient COVID mortality patterns, the researchers then used their inpatient death certificate data to predict COVID-19 deaths in out-of-hospital settings. They focused on adults 25 years or older dying from natural causes between March 2020 and December 2021.
Previously, few studies had applied this type of machine learning modeling to track mortality trends.
"Because COVID-19 testing was near-universal in hospitals, deaths in those settings were more likely to be accurately classified," says Dr. Stokes. "That gave us a strong foundation to train our model, which we then applied to out-of-hospital settings, where testing was far less consistent and COVID-19 deaths were more likely to go unrecognized."
The number of COVID-19 deaths that occurred at home was 160 percent higher than official records reflected, suggesting that there could have been more than 111,000 uncounted COVID-19 deaths that occurred in homes.
"It's especially striking that these results are so similar to the excess mortality estimates produced by the CDC, since those CDC estimates reflect the consequences of the pandemic itself, as well as the pandemic response," says study coauthor Dr. Elizabeth Wrigley-Field , associate professor of sociology at the University of Minnesota, Twin Cities. "Right now there is a big public debate about whether the response did more harm than good. Our results strongly suggest that the immense harm in 2020-2021 stemmed from the virus, not from attempts to mitigate the virus."
Deaths from COVID-19 were also more likely to be overlooked among people in the South. Alabama, for example, had 67 percent more total predicted COVID-19 deaths than officially reported.
"Our findings underscore that well-documented disparities in the impact of the pandemic were likely even worse than previously described," says study lead author Dr. Mathew Kiang , assistant professor of epidemiology and population health at Stanford. "The variation of underreporting across counties not only suggests there are places where death reporting can be improved, but also that there are places that did really well, and we can learn from them. Machine-learning methods may have a role in helping with death reporting more broadly in this resource-constrained environment."
COVID-19 deaths were also more likely to be uncounted for people between the ages of 65 to 84, males, those who did not have a high school education; people who identified as Hispanic, American Indian and Alaska Native, Asian, and Black, and those who lived in counties with lower socioeconomic status and more pre-pandemic health issues. Many of the predicted uncounted COVID-19 deaths were attributed to underlying causes such as Alzheimer's disease and related dementia, cardiovascular disease, and diabetes.
"Under-counting inequities in COVID-19 deaths can be viewed as both a manifestation of structural racism, ableism, and classism and as a mechanism preventing responsive policy action," says study coauthor Dielle Lundberg , research fellow in the Department of Global Health at BUSPH. "Undercounts functioned to absolve health policymakers of their failures to enact pragmatic health policies during the pandemic. Policies such as increased unemployment benefits, expanded paid sick and medical leave, rent moratoriums, funding for services that would have allowed disabled people to remain at home rather than enter or stay in institutional settings, and universal access to healthcare would have disproportionately benefited marginalized communities."
The study did not directly examine the specific reasons why many death certifications list inaccurate causes of death, but other sources point to COVID-era challenges such as inadequate staffing to conduct postmortem COVID-19 testing, a lack of standardized training and protocol for death investigators, and partisan beliefs that may cloud investigators' judgment, particularly for county coroners, who are politically reported and are not required to have medical backgrounds. The researchers recommend increased funding and training for death investigators, as well as increased hiring of medical examiners. They also note that machine learning methods could potentially be adapted to other settings where cause-of-death data may be incomplete, delayed, or suspected to be biased, including drug overdoses, deaths in police custody, and extreme-heat-related deaths – though these tools should complement, not replace, broader systemic reform.
**
About Boston University School of Public Health
Founded in 1976, Boston University School of Public Health is one of the top ten ranked schools of public health in the world. It offers master's- and doctoral-level education in public health. The faculty in six departments conduct policy-changing public health research around the world, with the mission of improving the health of populations—especially the disadvantaged, underserved, and vulnerable—locally and globally.
