Skip to Main Content

While early studies of who was dying of Covid-19 identified risks such as obesity and having diabetes, there is a growing realization that those initial conclusions might have been misleading, obscuring a more significant explanation.

As researchers pull back their lens from individuals to population-level risk factors, they’re finding that, in the U.S., race may be as important as age in gauging a person’s likelihood of dying from the disease.

The higher the percentage of Black residents in a county, the higher its death rate from Covid-19 — even after accounting for income, health insurance coverage, rates of diabetes and obesity, and public transit use, finds a new study by researchers at the MIT Sloan School of Management. With those plausible explanations ruled out, “the causal mechanism has to be something else,” said applied economist Chris Knittel, the study’s senior author. “If I were a public official, I’d be looking at differences in the quality of insurance, conditions such as chronic stress, and systemic discrimination.”


The county-by-county analysis of Covid-19 death rates in the U.S. comes as more and more studies shift from the initial focus on individual-level factors that seem to increase people’s risk of dying to population-level ones, too, said experts in public health, demographics, and infectious disease.

One reason for the new focus is that “we’re not learning much more beyond what’s been observed from the start of the pandemic: that sex, age, and preexisting conditions put you at greater risk of dying” from Covid-19, said Aaron Glatt, an infectious disease physician at the Icahn School of Medicine at Mount Sinai. In almost every country, for instance, more men than women are dying of Covid-19, an imbalance that likely reflects both biology — women have stronger immune systems — and socialization: They seem to be following social distancing guidelines more than men, which could decrease the viral load they’re exposed to.


Also driving the shift in focus is that society-level influences are potentially more “actionable” than individual risk factors; you can’t change your blood type. A study posted to the preprint site medRxiv this month reported that type O is associated with lower risk of respiratory failure from Covid-19 and type A with a higher risk, but the paper hasn’t been peer-reviewed and it’s not clear how much difference blood type might make. “I wouldn’t tell one patient, thank God you have type O, but another, start preparing your will because you’re type A,” Glatt said.

To investigate population-level factors, MIT’s Knittel and graduate student Bora Ozaltun analyzed county-by-county mortality rates — the number of deaths from Covid-19 as a percentage of population, from April 4 to May 27. The mortality rate is more precise than the infection rate or the case fatality rate (the percent of diagnosed cases who die), which are imprecise because of inadequate testing. In contrast, although some Covid-19 deaths were incorrectly attributed to other causes, especially early in the U.S. outbreak, “deaths, sadly, are an absolute,” Knittel said.

They then used standard statistical tools to tease out which factors are most strongly correlated with mortality rates. Race stood out. Nationwide, the average county-level death rate from Covid-19 is 12 per 100,000 people. Counties with a Black population above 85% had a death rate up to 10 times higher. For every 10 percentage point increase in a county’s Black population, its Covid-19 death rate roughly doubles, Knittel said.

That meshes with other research. A study last month of 1,052 Covid-19 patients treated at Sutter Health hospitals in California, for instance, found that Black patients had 2.7 times the odds of hospitalization as non-Hispanic white patients, indicating more severe disease. And an analysis by scientists at the Harvard T.H. Chan School of Public Health found that the death rate in predominantly non-white areas is six times that in non-Hispanic white areas.

“Black people are dying of Covid-19 at a rate more than twice our share of the population,” said Malika Fair, an emergency medicine physician in Washington, D.C., and senior director of health equity programs at the Association of American Medical Colleges.

The MIT researchers’ key finding is that the underlying reasons for the link between race and death rate are not the usual suspects.

“Policymakers’ natural instinct is to think this correlation is because of income disparities, or having health insurance, or diabetes, obesity rates, smoking rates, or even use of public transit,” Knittel said. “It’s not. We controlled for all of those. The reason why [Black people] face higher death rates is not because they have higher rates of uninsured, poverty, diabetes, or these other factors.”

The Sutter study, too, adjusted for age, sex, comorbidities, and income; the higher hospitalization rate for Black patients wasn’t explained by any of those.

That leaves other factors. “If I were a policymaker,” Knittel said, “I’d be looking at things like the systemic racism that affects the quality of insurance African Americans have and the quality of the health care they receive.”

People with Medicaid or high-deductible plans, for instance, are less likely to have a primary care physician; 34% of Black people and 15% of white individuals are covered by Medicaid and therefore less likely than people with employer-sponsored insurance to have a regular physician.

“People who didn’t have a relationship with a primary care provider were much less likely to get tested,” said Georges Benjamin, a physician and executive director of the American Public Health Association. “Testing sites were put in affluent communities, or required a car, and testing kits were in short supply. Any time there is a shortage of something, minorities are less likely to get it.”

Without a primary care provider, Black people who thought they were infected were also likely to be turned away from hospitals, Benjamin said. “Someone without a primary care doctor doesn’t get into the ER as fast as someone whose doctor calls ahead,” he said. “At what point were your symptoms severe enough that you got into the health care system?” For people of color, it was likely later, he suggests.

“Black patients presenting with fever and cough were less likely to receive a referral for a Covid-19 test,” Fair said. That delayed appropriate care.

And once they do get into the system, research has found, the quality of care Black people receive for a variety of conditions, such as cardiovascular disease, is likely to be lower. Racism “is apparent in how we treat patients,” Fair said. “We still see differences in the care [for many conditions] given to Blacks and whites.”

Another possible factor in the high death rate among Black Americans is the well-documented health effects, including on the immune system, of chronic stress such as that caused by a lifetime of discrimination.

In the MIT study, the correlation between a county’s Covid-19 death rate and its proportion of Black residents was stronger within any given state than between states. “That tells me that there are important state-level differences that drive these deaths,” Knittel said. “African Americans are more likely to live in states with poor health care systems.”

Commuting via public transportation, relative to telecommuting, is also linked to a higher death rate. When public transit use is 20.6 percentage points higher in one county than another, its death rate is about tenfold higher.

Driving to work was also linked to a higher death rate. Many people working at their place of employment rather than home were in public-facing and therefore risky occupations such as health care, grocery stores, and public safety. “The correlation between death rate and driving to work suggests that just being at work, no matter how you get there, increases your risk of dying,” Knittel said.

Those positions are also filled disproportionately with people of color, likely contributing to the correlation between race and death rates. “Clearly, people who have jobs where they come into contact with others, from health care workers to bus drivers, are more likely to become infected,” the APHA’s Benjamin said. Although much about the pathology of the new coronavirus remains a mystery, the chance of becoming infected is partly a function of how much virus one is exposed to; the more infected people someone encounters, the higher that viral load can be.

The new population-level approach to understanding risk has prompted a rethinking of the role of conditions such as obesity and type 2 diabetes.

“Obesity is a marker of poverty and therefore of access to high-quality health care,” said Nina Schwalbe of the Mailman School of Public Health at Columbia University. Although the physiological consequences of obesity, notably high rates of inflammation, might contribute to Covid-19 severity, “obesity is a signal for so many of the social determinants of health, and we have to ask what this signal is telling us about vulnerabilities,” Schwalbe said.

That holds for type 2 diabetes as well. An analysis of 13 separate studies found that the disease is associated with 3.7 times the risk of having severe Covid-19 or dying from it compared to not having any underlying illness. This, too, is a disease of poverty, which means those who have it are more likely to live in crowded homes where “social distancing” is impossible, more likely not to have a primary care physician, and more likely to have jobs that increase their exposure to infected people.

The Sutter and MIT studies cast doubt on whether individual risk factors are as important as social determinants of health in affecting someone’s chances of contracting severe and even fatal Covid-19. “It should cause us to ask a different set of questions about what puts you at risk of hospitalization or death,” Schwalbe said.

More and more evidence is pointing to social determinants of risk, which puts the role of underlying health conditions in a new light. “Comorbidities are still used to blame people for how hard they are hit by Covid-19,” said Philip Alberti, senior director for health equity research at the AAMC. To reduce the U.S. death toll now that many states are seeing a new surge in cases, he said, “our response to this disease” must look beyond the strictly medical.

  • “There needs no ghost …to tell us that”. Common sense and personal observation inform us of the devastating effects that low self-esteem, continual discrimination and the hopeless impossibility of self-improvement will have on health and psychology. One underground line in London shows a drop in life expectancy of a year for every stop it takes towards the poorer districts. Life expectancy in The Gorbals , Glasgow , is 53.9 years, less than even war-torn Iraq. To be hated and despised because of where you live and how you look will obviously affect your physical health. It needs no expensive investigation to tell us what we surely already know.

  • It’s one thing to shed light on the causes of poor lifestyle, but it’s another to deflect too much attention from the personal responsibility factor in lifestyle. I know lots (and I mean lots) of people who have plenty of education and access and still live risky lifestyles. That must be conceded in every discussion about covid and lifestyle, or people get the idea that they have no control over the outcome of an actual infection. That helps no one.

    • Risky lifestyles are often a matter of people’s personal philosophies and personalities as best I can tell – i saw a rant one time by a guy who claimed society wanted to deny the reality anyone can die any time, and his evidence was guard rails on highways – he said they were a form of denial – seemed to completely reject the idea they were to protect people – or to make money for contractors, which I think is at least half of it.

      Many of the “Open Up Now” people are probably the high spirited inpatient types who do risky stuff all the time – like sports where you can bust yourself up pretty easily – and/or they believe they are young and healthy enough to go through the infection with a couple weeks in bed, no lasting damage, and if Grandma gets it and passes away, well, she was getting pretty old anyway.
      But I am not sure anyone can do anything to fight an infection once they get it, if that is what you meant. Or very easily stave a severe case off by losing weight and such, that is a long term project.

    • A risky lifestyle is knowing that you have diabetes type II but simply refuse to change your habits at all. The same holds true for high blood pressure, high cholesterol and heart disease. While being overweight isn’t an indicator of an underlying condition per se, the correlation between weight and such conditions is well established.

      Sports can cause some serious injuries, but serious such injuries are rare and death is even rarer.

      Age isn’t an underlying condition. We tend to think of immunosenescence as inevitable, but there is plenty of research to show that diet and exercise can greatly curtail immunosenescence. In other words, did granny do what she should have during her lifetime to extend it?

      Lifestyle is certainly more involved than just diet and exercise, but those are pretty core to longevity, but we continue to make endless excuses for avoiding the topic.

  • I wonder if they looked at the VitaminD levels in this study, as we do see a problem in the Scandinavian countries with africans inability to get enough Vitamin D from the northern sun.It is dark here and they are supposed to live in a different environment. women from the african continent to a large extent suffer from osteoporosis and there are many somali men that are badly affected by Covid 19. I do think we are supposed to live where we were designed to live in order to stay healthy.
    I would appreciate a reply

    • Hi Anna, Can you tell us how much more Covid 19 Somali men have? I have not read this and did not know if the Vitamin D is really a problem for Covid 19 or not. Do you have some statistics? Thanks.

  • Can anyone give some insight into this?
    They say if a county was 85% black, the death rate from coronavirus would be ten times the national average.
    But, elsewhere I have seen the claim the black case fatality is about twice the white rate.
    So, does that mean the infection rate in those counties was five times, roughly, the national average? (Two times the case fatality rate x 5 times the national average infection rate = 10 times the national average mortality rate?)
    If so, it is really hard to see how the social factors they hypothesize could cause that much higher infection rate. I can not see any big difference in the way black and white people live that would make the disease spread that much more in one group rather than another.
    If the numbers are real, and black people have no special genetic vulnerability to infection, so they are getting infected that much more due to things which can in theory be eliminated by just changing practices, then clearly we need to do something different in those counties to slow this down, but the authors are not offering us any suggestions.

    • You said, “If so, it is really hard to see how the social factors they hypothesize could cause that much higher infection rate.” I think a better (though less politically interesting) explanation is that this is a disease to which poorer individuals are more susceptible, due to working conditions, living conditions, general health, use of public transportation, etc. This is showing up more in the Black community because African-Americans are disproportionately poor. That’s where the policy questions need to begin—economic inequality, not medical. I hypothesize that if you were able to abstract out poor white people, you’d see similar, or even worse, numbers. But tracking class is much harder with the medical data. In this case, race is serving as a proxy for class but being pushed as the cause because of the political capital involved.

    • Doing some more research on this claim of very high mortality rate in counties which 85% black – in a Wikipedia article – for what that is worth – there are no counties which are 85% black, but there are 4 which are close – two in Mississippi and two in Alabama – all but one have populations below 10,000 people. Two have high coronavirus death rates, one average, one had only one death which put it below the national average.
      I am not a statistician, nor do I know if the Wikipedia list of the most black counties is right or complete – but I am suspicious this statistic from the “blackest” county in the US is not representative. As I say, it does not seem to be borne out by the numbers available online, but even it if was, would that have any statistical significance – depending on when the paper made the claim, the death rate in the US was 1 in 20,000 not long ago. that would mean on average .3 people in a county of around 7,000. So, if you got 3 people dying you are at “10 time the national mortality rate” but it is a statistical blip, isn’t it? Another 80+% black county had an extremely low death rate – which could have been changed to a very high one with a couple more fatal infections – so?????

    • In case anyone cares:
      Jefferson County Mississippi
      Claiborne County, Mississippi
      Macon County, Alabama
      Greene County, Alabama –
      One of these contains Tuskegee Institute – and has a population over 20,000, but the rest are, no offense to anyone please – but really backwaters with mostly long term declines in population, so while they are fairly large, they have very few people and a few cases can really skew the numbers – I said two have above average coronavirus death rates but it may have been one.
      At least one was one of the poorest counties in the entire US – maybe two.
      I do not doubt one bit there are a lot of medical treatments not available at all in these places – but what percentage of the population lives in similar areas? One medium size city has the population of 50 or more of these tiny rural counties.

    • Not to beat it to death, but I should also point out, none of the statistics I could find online bear out the claim of “10 times the national mortality rate in counties which are 85% black” claim.
      Either the numbers online are very inaccurate, or, very early on in the epidemic, when the national mortality rate was much lower, those counties could have had ten times the national rate due to a very few fatal cases- a statistical blip, for as I say, some of these tiny counties still have lower than the national mortality rate.
      I respectfully ask the authors, of the study and the article, to give some more details on this question.

    • Remember that US fatality statistics are pretty much irrelevant due to the number of fatalities of people in elder care facilities. Over 90% of the fatalities in my city came from one elder care facility.

      An outbreak in an elder care facility will skew any county level numbers.

  • The article states, “The mortality rate is more precise than the infection rate or the case fatality rate (the percent of diagnosed cases who die), which are imprecise because of inadequate testing.” Is there more justification for this? This is important because CFR for African-Americans is essentially the same as for whites in every location I’ve analyzed. This seems to indicate that the issue is more about why are more African-Americans getting infected, and not the care they receive after the fact. That’s a makes a huge difference in policy matters.

    • I haven’t liked the reporting about deaths associated with (Dead of COVID? With COVID? From COVID? Because of COVID?) the pandemic virus for a couple of months already. It has all been way, way too loose jointed, too lacking in definition, to be called science in my view. And now STAT offers the kind of cherry picked click bait represented by this article. It is more I don’t like and smells a lot like abuse of statistics. So I dug a little. The editors at STAT might want dig a little before they insult their readership again.

      This in particular caught my eye: “Nationwide, the average county-level death rate from Covid-19 is 12 per 100,000 people. Counties with a Black population above 85% had a death rate up to 10 times higher.” Wow.

      As far as I can tell, as of the 2010 census, there was exactly ONE COUNTY in the US with a black population above 85%. Jefferson County MS, estimated 2018 population 7,100. At 2010, there were exactly FOUR MORE COUNTIES in the US with black population above 80%: Claiborne and Holmes in MS and Greene and Macon in AL. Total estimated population for all five is about 60,000; they are among the poorest and least populous counties in the country. To compare the NATIONWIDE county level death rates to those from counties which very likely have no health care facilities or personnel of any kind, not even pharmacies, and to then suspect “social factors” like race is nonsensical. At ten times 12 deaths per 100,000, in a county where 7,000 people live, it comes to 9 deaths. Of COVID or from COVID or with COVID, or maybe not. This is not trend spotting, it is dishonesty wearing a lab coat.

  • I think we shouldn’t be so quick to make claims without assessing all of the data.

    Case in point: My best friend was positive for covid. He is young (late 20s) morbidly obese, his weight has fluctuated between 350-500lbs over the last 5 years. He has kidney failure from type 2 diabetes. He has been on dialysis for about 5 years. He also has asthma bad enough that he needs regular nebulizer treatments (at least monthly) He was positive for about 6 weeks. Remained largely asymptomatic throughout.

    I was miserably ill with what I believe to be covid in late January. I am otherwise healthy with slight asthma that only flares up with illness. It ravaged me. Fever wouldn’t go down with meds, I slept for a week straight, couldn’t move, lungs felt full and were heavy. Somehow, after a week I began improving and fever went away completely.

  • Are these Vitamin D claims valid? If we could really fortify ourselves against Coronavirus by taking some Vitamin D, that would be the best news since this thing started.
    All I have seen in mainstream science publishing has been one paper, maybe just a preprint -which tied the course of the pandemic to alleged reduced sun exposure by country – and theorized the fairly low rates in the north of Europe were due to vitamin D supplementation in those places – but did not directly tie disease severity to individual vitamin D levels. The authors of the study, IIRC, made a point of saying correlation does not prove causation, they just asked for more study. But despite it becoming to some extent a topic of theorizing, I have not seen anything which confirmed it.
    One would think, in these Nordic countries where, we are told, the government has medical records on everyone, it would be fairly easy to find out if people who got very sick had low levels of vitamin D beforehand.

    This is kind of interesting -I recall several years back, someone had found that immigrants to one of the scandinavian countries, immigrants from Africa I think – had high rates of autism in the kids born in Scandinavia and low Vitamin D.

    Also, IIRC, there is a theory that depleted Vitamin D has been suspected of being the reason for the observed fact, if a mother get’s pregnant right after having a baby, the second baby has a much higher than normal risk of being autistic. The theory is, the mother has inadequate Vitamin D (I do not know much about this but it’s fat soluble and maybe after it is taken out of her tissues she can not get it back in where needed very rapidly) and vitamin D is somehow involved in brain development.

    Vitamin D is one of those molecules which does many important things in the body and is being very actively investigated.

    Having said all that – there should be a study that actually looks at the patients with severe disease to see if they have a Vitamin D deficiency. Without that, the emotion over it seems premature.

    • Hi, I just posted something on Vitamin D. I am swedish and we have a serious problem with our immigrants lack of vitamin D as they have lesser ability of taking it up via their skin. we are living in darkness for long periods up here. Somali men have been badly affected by Covid 19. There is no openness about this and it is a shame.

Comments are closed.