The Coronavirus’ Hidden Death Toll in America’s Counties

Our examination of 366 counties across the United States finds that COVID-19 deaths were two to 30 times higher than reported COVID-19 death rates.

Coronavirus has taken a major toll on the US, and we likely don’t yet know the full extent of its damage. At Surgo Foundation, we’ve been concerned that the severity of the epidemic is being obscured by a massive underreporting of “excess deaths” — or the number of deaths above the average recorded for the same period in previous years.

Measuring “excess deaths” provides a window into the true toll of this pandemic. Excess death counts reflect not only coronavirus deaths, but increases and decreases in all causes of death — such as motor vehicle accidents, illnesses unrelated to COVID-19, suicides, and homicides. In the absence of widespread coronavirus testing, these counts help us better understand, and plan for, the true impact of the disease.

Researchers analyzing excess deaths in Italy concluded that the COVID-19 death toll may be twice as high as the official government count.

And in the US, a state-level analysis of excess deaths has reached similar conclusions: in the hardest-hit states of New York and New Jersey, researchers estimated that COVID-19 deaths may be 1.5–3 times higher than officially reported.

But we miss something important if we restrict ourselves to a state-by-state picture of the coronavirus death toll. If we really want to understand where the virus is having the greatest impact, we need to drill down to a more precise, hyper-local level.

At Surgo Foundation, we’ve analyzed county-level mortality data from the Centers for Disease Control and Prevention to paint a more granular picture, highlighting disparities in death reporting that were previously unavailable. Now, for the first time since the pandemic began, we’re able to study excess death counts at the county level in the United States.

And what we’ve learned is disconcerting. Our data suggest that we may be especially undercounting COVID-19 deaths in Southern parts of the US. In more than 200 of the counties we examined, excess death rates were between two times to 30 times higher than reported COVID-19 death rates.

Why is this important? Knowing that some communities are being hit much harder by coronavirus than their official COVID-19 death counts suggest can help state leaders direct resources more effectively to fight the pandemic. It can also help local officials give more accurate and confident messaging about the importance of preventive measures. And it helps policymakers and scientists look for patterns of vulnerability that can inform our response — not just to this pandemic, but to a future one.

Drilling down a bit deeper, here’s what we found out about COVID-19’s toll at the county level:

1. Death rates this year are higher than normal across many US counties.

Across the 366 counties we examined, there were on average 51 more deaths per 100,000 people than the average for the same period in 2014–2018. If this rate held against the entire US population of 328 million, this would amount to over 167,000 excess deaths this year to date.

The graph below shows the distribution of historic death rates (in blue) compared with current death rates (in red). In past years, 62 of the counties surveyed had death rates exceeding 300 per 100,000 people. This year, 176 of the counties do. (Note that this sample represents about 12% of US counties.)

The CDC only reports excess death data for counties with 10 or more COVID-19 deaths. 45 states had at least one county represented in this database.

2. The rate of excess deaths is even greater than the rate of reported COVID-19 deaths.

As the graph below shows, counties with more COVID-attributed deaths also had a greater number of overall excess deaths — that is, deaths above the average number for the same period in 2014–2018.

We expect to see this: it’s not surprising that much of the excess death rate in 2020 is attributed to COVID-19. But COVID-related deaths don’t account for all the excess mortality. On average, excess death rates are 30% higher than reported COVID-19 death rates.

In some counties, this number is far higher. In Lee County, Mississippi, for example, the COVID-19 death rate is 31 per 100,000 people, but the excess death rate is 398 per 100,000 — about 13 times higher.

3. The discrepancy between reported COVID-19 deaths and overall excess deaths isn’t the same across the country.

We also observed that a number of counties in the Southern US (in yellow in the graph above) have very low reported COVID-19 death rates, but high excess death rates, even if they’re not as extreme as they are in Lee County. This raises the possibility that deaths resulting directly from COVID-19 are being underreported.

For example, this may be because people weren’t tested, and their death was ascribed to pneumonia or influenza instead. Another possible reason is that more people than usual are dying from other medical conditions, such as heart attacks or strokes, because they fail to seek help at clinics or hospitals from fear of being exposed to the coronavirus. Thus, they are indirect victims of the pandemic, and contribute to the excess deaths. Finally, COVID-19 has been linked to other potentially lethal conditions, such as kidney failure, strokes, and blood clots. Because these conditions can occur after typical COVID symptoms have passed, these deaths may not be attributed to COVID-19.

4. We see more COVID-19 attributed deaths where we have done more testing.

The data suggest we could be missing large numbers of COVID-19 deaths in places where testing is low.

Using data from the Johns Hopkins University coronavirus database, we compared the rate of coronavirus testing with the reported rate of coronavirus deaths. Our graph (below) shows that states with higher rates of testing have also seen higher rates of deaths attributed to COVID-19.

Why might this be? Testing has ramped up in areas hardest hit by coronavirus, so the COVID-19 death rate may be higher in these states simply because their outbreaks are worse and there are, in fact, more deaths per head among the population.

But another possible explanation is that states with lower testing rates are not catching, and therefore not counting, COVID-related deaths, which could make their reported death rates lower than in high-testing states — and lower than they are in reality.

The takeaway: In the absence of adequate testing, county-level excess death rates reveal a more accurate picture of the toll of COVID-19.

What does all of this scrutiny of deaths at the county level tell us? In short, more hyperlocal and precise data means a more impactful response. The US has still not achieved the level of coronavirus testing needed to capture the full extent of the pandemic. That makes an accurate count of the COVID-19 death toll critical for informing policymakers and public health officials about the severity of the pandemic. Undercounted death tolls give policymakers a false sense of security about the safety of reopening the country, and put residents at risk.

Looking at excess death rates provides another important way to understand how COVID-19 is impacting communities. And drilling down to the county level can give us a far more precise picture than statewide data alone. As more data becomes available, it can help target resources better, improve local messaging, and understand the relationship between the hardest-hit areas and measures of community vulnerability — to prepare better for future outbreaks.

This work was made possible by everyone at the Surgo Foundation, including but not limited to (in alphabetical order): James Baer, Yael Caplan, Grace Charles, Bethany Hardy, and Sema Sgaier.

Technical notes:

• We retrieved provisional county-level mortality data released by the CDC for 366 US counties across 45 states for the period of February 1 — May 9. For each county in this limited sample, the CDC reports the number of COVID-attributed deaths, as well as the total number of deaths from all causes.
• We calculated excess death rates in each county during this three-month period by comparing current death rates with average rates for the same period in 2014–2018.
• Because of low mortality reporting, we excluded Connecticut counties from our analysis.
• CDC death data for 2020 are provisional. It takes anywhere from 2 to 8 weeks (sometimes even longer) for a reported death to be processed and added to the data — meaning that the final numbers will almost certainly be higher still. CDC methodology is available on their website.
• State-level testing data and COVID death rate data were retrieved on May 13th from the Johns Hopkins University COVID-19 Github repository.
• Questions, concerns, comments? Get in touch at covid19@surgofoundation.org.