The United States recorded an estimated 37,100 excess deaths as the novel coronavirus spread across the country in March and the first two weeks of April, nearly 13,500 more than are now attributed to covid-19 for that same period, according to an analysis of federal data conducted for The Washington Post by a research team led by the Yale School of Public Health.
The Yale team's analysis suggests that the number of excess deaths accelerated as the pandemic took hold. There were 16,600 estimated excess deaths just in the week of April 5 to April 11, compared with 20,500 over the prior five weeks.
Though the team's estimate of the impact early in the outbreak already paints a picture of unusually high mortality, the number is certain to grow as more deaths are reported to the federal government on a rolling basis.
"I think people need to be aware that the data they're seeing on deaths is very incomplete," said Dan Weinberger, a Yale professor of epidemiology who led the analysis for The Post.
Those excess deaths - the number beyond what would normally be expected for that time of year - are not necessarily attributable directly to covid-19, the disease caused by the coronavirus. They could include people with unrelated maladies who avoided hospitals for fear of being exposed or who couldn't get the care they needed from overwhelmed health systems, as well as some number of deaths that are part of the ordinary variation in the death rate. The number is affected by increases or decreases in other categories of deaths, such as traffic fatalities and homicides.
But excess deaths are a starting point for scientists to assess the overall impact of the pandemic.
The nation surpassed 64,000 coronavirus deaths on Friday, according to The Post's compilation of state health department data. Weinberger said his team's estimates of excess deaths indicate that the pandemic's true toll to date is likely substantially higher.
"It's hard to say how much higher, but our best guess might be it's in the range of one and a half times higher," he said.
The analysis is based on death certificate data that states compile and send to the National Center for Health Statistics, a part of the Centers for Disease Control and Prevention.
It often takes weeks for a death to be counted in these federal numbers, so tallies for the most recent weeks are provisional and inevitably missing many deaths. As time passes and more information becomes available, NCHS backfills data for those weeks - and the death totals grow.
For example, earlier this week The Post reported - based on the Yale team's model - that the nation had recorded an estimated 15,400 excess deaths between March 1 and April 4. But more up-to-date NCHS data released Friday revealed that more than 5,000 additional people had died overall during that span - bringing the total estimated excess deaths to 20,500.
The fact that the picture of overall death during that period changed so substantially this week underlines how real-time counts of covid-19 deaths are only a partial snapshot.
On April 4, state health departments were reporting that about 8,100 people had died of covid-19, according to The Post's tally at the time. Those figures excluded deaths of people who probably died of covid-19, based on symptoms and exposure, but were never tested. They also excluded some deaths that had occurred but had not yet been publicly reported.
NCHS corrects for those exclusions. On Sunday, it was reporting that as of April 4, about 10,500 people had died of covid-19. On Friday, after more backfilling with updated reports, that number stood at about 11,361 - more than 9,000 short of the estimated 20,500 excess deaths.
Examining the number of excess deaths is a vital exercise for understanding the extent of the pandemic, said Eleanor Murray, a professor of epidemiology at Boston University School of Public Health, who was not involved in the Yale team's analysis.
"We know because we have not been doing enough testing that we are absolutely missing covid cases," she said. "So whatever death counts we have are certainly going to be underestimates of the people dying from covid."
State and local leaders should be studying the estimates of excess deaths in their communities and basing consequential decisions about reopening businesses and social activities on those figures, rather than strictly using reported coronavirus infections and covid-19 fatalities, figures that are often incomplete and misleading, Murray said.
"If you're basing your decision to open on your reported case numbers or your reported covid deaths, you're making the wrong decision because those numbers aren't reflective of what's actually happening out in your community," she said.
Estimating the death count from any disaster is a difficult task. After Hurricane Maria struck Puerto Rico in 2017, an accurate assessment of the dead eluded researchers and the public for months, Murray said.
For nearly a year, the territory's leaders refused to acknowledge that the count was higher than the official tally of 64 - even after a Harvard study estimated that more than 4,600 had died. In August 2018, a government-requested report from George Washington University put the number of excess deaths at 2,975, a figure then-Gov. Ricardo Rosselló finally embraced, saying he had "made mistakes."
And Hurricane Maria was a discrete event in a single place. Calculating mortality during the coronavirus pandemic - a global disaster that unfolded over months - will be significantly more challenging, Murray said.
"Identifying exactly how many people died from covid and from pandemic-related factors, that's going to be an open scientific question for years," she said. "And we're probably never going to have an exact number."
- - -
The Washington Post's Jacqueline Dupree and Monica Ulmanu contributed to this report.
- - -
About this story: A research team led by the Yale School of Public Health used historical data on all deaths between 2015 and early 2020, published by the National Center for Health Statistics (NCHS), to model the number of deaths that would normally be expected each week from March 1 to April 11. The estimate takes into account seasonal variations, intensity of flu epidemics, as well as year-to-year variations in mortality levels. Details on the team's statistical approach estimating seasonal baseline deaths can be found in an article posted online at the preprint server medRxiv. The method used for this analysis differs in that researchers did not attempt to correct for data reporting delays, as they did for their previous article. Instead, the analysis for The Post relied only on reported deaths, a more conservative approach to estimating excess deaths.
The number of overall deaths in the United States and for each state was obtained from provisional death data published weekly by the NCHS, part of the Centers for Disease Control and Prevention. Figures for North Carolina were not up-to-date, and that state is not included in this analysis.
NCHS data are collected from state health departments, which report deaths at different rates. It usually takes about three weeks for death data to stabilize, but even then, they are still not complete. As a result, it is expected that the numbers of total deaths as of April 11 will continue to increase as states continue reporting additional data to NCHS.
The number of excess deaths was calculated by subtracting the expected seasonal baseline from the number of all deaths. Because the seasonal baseline is an estimate, there is some uncertainty associated with the excess-death figure of 37,100. Based only on the deaths reported so far, there is a 90 percent chance that the actual number of excess deaths is greater than 33,200, and a 70 percent chance that it is greater than 35,500. (There is a 2.5 percent chance that the actual number of excess deaths is lower than 31,400, and an equal chance that it is higher than 43,500).
The covid-19 death toll as of April 11 comes from figures reported by NCHS.