Contribute Try STAT+ Today

Tens of thousands of Covid-19 deaths are going unreported in the U.S., with far more missed in counties that strongly supported former President Trump, according to new research.

The figures suggest that political leanings have helped suppress the true scale of deaths. In cases where the deceased didn’t have a Covid-19 test, a coroner or medical examiner has the freedom to interpret symptoms.

“There’s potentially latitude to make a judgement call conditional on a set of beliefs about Covid and whether it represents a serious problem or a hoax,” said Andrew Stokes, a professor of global health at Boston University School of Public Health who performed the analysis for STAT.


More than 413,000 people have died in the U.S. with Covid-19 attributed somewhere on their death certificate. The real death toll, as reflected in the number of excess deaths in 2020 compared with annual deaths from 2013 to 2018, is even higher. A separate study, led by Stokes, of 787 counties with more than 20 Covid-19 deaths from Feb. 1 to Oct. 17, 2020, found that while there were 199,124 official Covid-19 deaths in that time period, an additional excess 88,142 deaths weren’t attributed to the virus.

Some of these excess deaths are likely due to factors that were exacerbated by the pandemic, such as overdoses and suicide in response to isolation and economic hardship, or subpar health care in an overrun system. But researchers believe many are Covid-19 deaths that go uncounted. Overall, the true Covid-19 death toll is 44% higher than official figures, according to the study, which has been submitted to PLOS Medicine.


The researchers found that unattributed Covid-19 deaths were significantly higher in rural areas than urban; in the South compared to other regions; and in areas with lower levels of education. All these factors tend to correlate with support for Trump.

The gap between excess deaths and deaths officially attributed to Covid-19 was greatest in the most rural areas and the least in the most urban areas. Patrick Skerrett / STAT Source: Andrew Stokes

“We see a lot of variation in excess deaths, and it’s kind of hard to explain,” said Katherine Hempstead, another researcher on the study and a senior policy adviser at the Robert Wood Johnson Foundation. There seems to be correlation between areas that were less inclined to take Covid-19 seriously and adopt preventive measures, such as mask-wearing and social distancing, and those with higher levels of unrecognized Covid-19 deaths, she said.

Not all the variations in excess deaths point to a Trump effect: Lower-income counties, which tend to vote Democrat, also have far higher rates of excess deaths. But preliminary research by Stokes, which isn’t part of the broader study, explicitly examined the political correlations and found that excess deaths unattributed to Covid-19 are far more common in the most Trump-supporting counties, based on both 2020 and 2016 presidential election results, than in the least.

In total across the country, there were 44 excess deaths that weren’t officially recognized as Covid-19 for every 100 official Covid-19 deaths. But the uncounted deaths are far higher in the 25% of counties with the most Trump voters in 2020, with 163 excess deaths for every 100 Covid-19 deaths. In comparison, there were just 18 excess deaths per 100 Covid-19 deaths in the bottom quarter of these counties.

Throughout the pandemic, the response to Covid-19 has been intensely partisan, and the politicized reactions to the virus could easily shape how death itself is recorded. Former President Trump spoke out against testing, falsely claiming that it created more cases. As patients are less likely to be registered as a Covid-19 death if they haven’t been tested, this would contribute to cases being missed.

Testing uptake is likely to be lower in areas that are more conservative, said Stokes. “People who are undiagnosed are less likely to get Covid-19 assigned on their death certificate than people who are diagnosed.”

There were higher rates of excess deaths not officially attributed to Covid-19 in counties with the highest percentage of Trump voters than in those with the least Trump voters. Patrick Skerrett / STAT Source: Andrew Stokes

In areas where the pandemic is downplayed or seen as a myth, the death reporting process leaves room for those who don’t take Covid-19 seriously to leave the virus off death certificates. Covid-19 is typically the underlying cause of death, leading to conditions such as pneumonia or a heart attack that is the more direct cause. In cases where someone hasn’t been tested, coroners are free to attribute a death to the more direct condition, and leave Covid-19 out entirely.

If someone is hospitalized with breathing problems, is tested and found to have Covid-19, and dies from those breathing problems, then the virus is clearly an underlying cause. “Those deaths are relatively straightforward to keep track of,” said Greg Hess, chief medical examiner in Pima County, Ariz. “On a death certificate, you would make mention of the fact that you diagnosed the person with Covid, and they died from viral pneumonia due to Covid.”

But other cases are less clear-cut, said Hess. A patient could contract Covid-19 while in the hospital for a hip fracture, and this could exacerbate their underlying heart disease and diabetes. They could then recover from Covid-19, but eventually die from their deteriorated health overall. “So now what happens on a death certificate? Well, it really depends on who’s writing it,” said Hess.

The system for recording death in the United States is fragmented. Though most deaths are declared by a doctor, when a patient dies at home and isn’t under the care of a physician, either a medical examiner or coroner will typically be called on to declare the death.

The study led by Stokes found that counties that used coroners, who are elected and can be laypeople, had higher rates of excess Covid-19 deaths than counties with medical examiners, who are appointed medical officials.

Counties with elected coroners had higher rates of excess deaths not attributed to Covid-19 than those with medical examiners. Patrick Skerrett / STAT Source: Andrew Stokes

“They’re going to be voting for you. I’m sure there’s pressure, that’s one of the difficulties of being elected,” said Sally Aiken, medical examiner in Spokane County, Wash., and chair of the board of directors of the National Association of Medical Examiners. “In these smaller areas, there are fewer voters, so it doesn’t take as much to get rid of you.”

Death investigators are used to facing pressure from family members. “There are certain expectations that families have or certain narratives that they have in their mind about why their loved one died that might be contrary to or supported by what is available in the medical record,” said Hess. “Sometimes someone might tailor the way a death certificate is written to try to get some relief from a family member who might be bothering or badgering them.” There’s no evidence of that happening locally, he added.

Aiken said she’s investigated several deaths where the family insisted that Covid-19 couldn’t have played a role. “There’s an element of pressure from certain individuals who just, frankly, don’t believe Covid exists. So they don’t want that on the death certificate,” she said.

Most of them relented once the deceased was tested, but she said one family insisted, even when a test came back positive, that their relatives’ other underlying conditions were the true cause, rather than Covid-19.

“Families don’t get to decide what goes on the death certificate, at least in my jurisdiction, but in some jurisdictions that probably influences what goes on the death certificate, particularly in borderline cases where someone has underlying significant natural disease, then they get Covid,” said Aiken.

The South had higher rates of excess deaths not attributed to Covid-19 than other regions. Patrick Skerrett / STAT Source: Andrew Stokes

Covid-19 deaths are more likely to be missed in areas of the country where there’s less testing. If a patient dies without a Covid-19 test, then it’s up to the death investigator to choose to order a post-mortem test. Though both Hess and Aiken have done this for all cases where Covid-19 is suspected, Hess said it’s not possible for all coroners and medical examiners to do so systematically across the country. “There’s just no bandwidth in the system to do that,” he said. “There’s going to be deaths that have occurred during the course of this pandemic, where Covid was a component of the death, and it is not recorded.”

The cost of ordering a test is also a deterrent, especially in areas where coroners have to pay for post-mortem testing out of their budgets, said Aiken. “That’s probably not happening in some not-well-funded, very rural death investigation systems,” she said. “Even if maybe they’ve had an exposure, or you know, maybe the coroner doesn’t believe in it, maybe the family doesn’t believe in it, they just think they had the flu or got sick or they died suddenly.”

Political beliefs and finances aren’t the only reasons why Covid-19 might be left off a death certificate. In New Jersey, the state health department declared that Covid-19 should be listed as the underlying, not the primary cause of death; the specific cause, such as pneumonia or heart disease, must be included on the death certificate.

Sherry Bensimon, a New Jersey and New York funeral director on the board of the Metropolitan Funeral Directors Association, said she’s received death certificates that simply state Covid-19 and, when she calls to explain that they’ll be rejected, the physicians simply switch out one primary cause for another, without adding Covid-19 as an underlying cause. “I’ve seen it many times where the doctors will just leave it off,” she said. “They’re exhausted, they’re busy, and they have a funeral director hounding them.”

Unfortunately, said Aiken, even the most rigorous system will miss some Covid-19 deaths. Most Covid-19 death certificates are signed by hospital or community physicians, who get little training on how to sign death certificates. And there are plenty of false negatives on Covid-19 tests, which haven’t been studied for their accuracy post-mortem.

The true Covid-19 death toll may never be definitively determined, but it’s certain that deaths from the virus are undercounted, said Aiken. “This is a real disease. People are dying. It shouldn’t be anything political, it should just be the truth.”

Correction: A previous version of this story misstated how much higher the true Covid-19 death toll is than official figures. It’s 44% higher, according to the new study.

  • Perhaps the so-called “under reporting” is the correct figure while the rest are inflated to suit a Covid narrative? Seems just as reasonable assumption to me!

  • Your bar chart shows absolutely no difference in how many excess deaths (the number of deaths in 2020 in excess of the usual number of deaths) were ascribed to Covid between counties in the 1/4 (25%) with the highest number of Trump votes (high Trump quarter) and the lowest number of Trump votes (low Trump quarter). Your data proves the contrary – that they were equal and the number of Trump votes had no effect at all on the determination of the cause of death.

    The high Trump quarter had .9 of its 1.6 [excess] deaths per 1000 people ascribed to Covid and the low Trump quarter had 1.3 of its 2.4 deaths per 1000 people ascribed to Covid. And guess what? If we divide the Covid deaths by the deaths per 1000 for both quarters for which data is provided, both the high Trump quarter and the low Trump quarter have .5 of their deaths per 1000, a whopping 50%, ascribed to Covid.

    The graphs that determined the quarters based on Trump votes in 2016 and those in 2020 are nearly identical, which is suspicious in itself, and no data is provided for the two intermediate quartiles of counties, the 25% with the second highest numbers of Trump votes and the 25% with the second lowest numbers. Without that data, one cannot determine whether there is a dose response effect, which is an increase in effect with increase in “dose” – here the number of Trump votes.

    Had there been a dose response, you would have described it. Had there been any statistical significance, you would have described that too. Any difference by an additional decimal, .54 vs .56, is irrelevant because the numbers used in the calculation (.9, 1.3, 1.6, and 2.4) had only one decimal place. The numbers used here were obtained by looking at the bar graph and dividing the space between 1 and 1.5 into 5 spaces, each denoting .1 death.

    Your claim of a difference between counties with high Trump vote totals and low ones is false and misleading. It was clearly intentional because, for one example, half of the counties were omitted, and that was not by accident. The graph was expanded to make numbers between .5 and 2.4 per 1000, all less than a quarter of a percent (10 per 1000 is 1%), appear meaningful and the lack of statistical significance was not even addressed.

    Shame on you, Olivia Goldhill, and shame on STAT for publishing this. It has changed my heretofore positive impression for good.

    I like my name spelled plain, with no added letters, but I do hold doctorates in both medicine (M.D.) and jurisprudence (J.D.) making me well qualified to render my opinions on this type of data.



    Nancy Lord

  • You article is sick and divisive. Please remove politics from this pandemic. A lot of countries had undercounted dead. US are already doing better that other countries.

  • Your article is out of date. WHO just yesterday stated that too many Covid-19 deaths are false and that only pcr tests will be accepted as positive proof of infection.

    • That’s odd. I was intrigued by your comment and went to look for this directive from the WHO and could not find anything like what you are suggesting. Can you please specify the name of the document?

  • The headline could be stated differently. Perhaps it should read, “Covid death numbers exaggerated in anti Trump areas”

  • Facts are facts. Excess deaths are a well understood way of evaluating pandemic outcomes. INTERPRETING those results is less well understood. Once the actual study has been published, epidemiologists will be able to evaluate the results. Hopefully there will be many more studies aimed at determining ACTUAL DEATHS, as well as (more importantly) why particular groups or areas varied. We all know that response in the US WAS fragmented and politicized, its vitally important that we learn how and why.

  • The way to evaluate reports is to look at the methods, not diss the conclusions. You can debate what the broader implications are, but that starts to get pretty far away from The methods reported in BMJ seem pretty straight forward. Peer review. Science.

    You can say that the numbers reported by states are bogus, but unless there are truckloads of corpses being secreted away in the night…

  • Come on, STAT and investigators. How hokey is this?! Whether it is “undercounting” or overcounting depends on the bias of the investigators. And if the idea of a counting explanation were wrong, as it might very well be, one could conclude that Trump-assigned communities were experiencing fewer COVID effects. One bogus conjecture just as uncertain as the other. In addition, there is also almost certainly selective reporting of the numerical analyses with the same bias (e.g., providing quartile number comparisons instead of some other -tile). And as far as the statistical confidence of the numerical comparisons, the word “statistical” does not even appear in the article. Geesh! Is this what you are feeding to readers? Slop to pigs soon to go to slaughter!

  • Data obviously is important to be accurate in any reporting and based upon news media’s both ignorance and bias in reporting data over the last few decades I do not believe any data is accurately being reported especially by government

Comments are closed.