Will The Real Coronavirus Death Toll Please Stand Up?
The actual number of people who have died due to coronavirus is a critical number that determines both the direction of public health policy and tens of trillions of dollars of excess spending. Hence, it’s likely the single most important number we’re tracking, perhaps even more important than cases. However, there are all sorts of reasons to believe that the numbers we’re all tracking in the media every day are seriously inaccurate. Let’s dig in.
The Colorado Shocker
I have reported before that I have been very proud of the way that my home state, Colorado has handled the coronavirus. Our Governor has done a great job of keeping politics out of the state response and has threaded the needle between keeping the maximum number of people safe and not destroying the local economy by enforcing shutdowns beyond what the data warrants. That’s why I was first very concerned and then pleasantly surprised this week about changes in the Colorado coronavirus death reporting.
First, came local news reports over the past month that showed several examples of where attending physicians or local coroners had stated that the cause of death was not coronavirus, but the state had added these patients to the coronavirus death toll because they happened to have a positive test (1). Three of these highlighted cases were in nursing homes and one was a guy who died with a blood alcohol level of 0.55, which is twice the lethal limit.
However, on Friday, the state decided to adjust it’s reported coronavirus deaths. It now reports numbers two ways:
- Died with coronavirus
- Died because of coronavirus
That took the state coronavirus death toll down from 1,150 to 878. Meaning, that the prior number dropped by 24%!
I have to say that I remain proud of Colorado’s response. They clearly had an issue here and unlike other states who are still reporting the “died with” number, they decided to do the right thing and trust the attending physician’s assessment. Hence, I applaud the change as this will help drive public policy more accurately.
Probable versus Confirmed
The dispute between New York City and the State of New York is also emblematic of how confusing this death toll data can get (4). There, the issue is different and similar, whether the doctor placed COVID-19 on the death certificate with or without a positive test result. If there was no positive test, then the death is “probable”. If there is then it’s definite. The NY State guidance document online refers to the CDC’s document (11):
In cases where a definite diagnosis of COVID–19 cannot be made, but it is suspected or likely (e.g., the circumstances
are compelling within a reasonable degree of certainty), it is acceptable to report COVID–19 on a death certificate as “probable” or “presumed.”
NYC has been counting both in its numbers and the State of New York only includes the confirmed cases. The difference between the two is again about 25%. Hence, in New York, they haven’t really addressed the “died with” versus “died of” dilemma. They’re still arguing over if a test was done or not.
Academics have been arguing that NY state is underestimating the death counts. However, there’s a confounder that few academics would have the experience to understand. It’s called insurance reimbursement.
The average academic scientist has no earthly idea of how doctors and hospitals get reimbursed. They also have never directly experienced how reimbursement drives physician and hospital behavior. Hence, they wouldn’t know that their data likely needs to be adjusted for the effects of the financial incentive of a 20% Medicare kicker payment on the diagnosis of COVID-19 (8). Yes, you read that right, the CARES act added a 20% bonus to those hospital cases with a COVID diagnosis.
The concept of reimbursement impacting physician behavior is well-vetted (5). From that referenced article: “Under fee-for-service reimbursement, in which physician compensation is primarily a function of the supply of services and procedures, physicians have an implicit financial incentive to increase the quantity of medical services…”. Hence, if you incentivize physicians to add COVID to the hospital problem list by throwing in a 20% payment kicker, based on the peer-reviewed literature, you’re more likely to get more COVID diagnoses which will end up on the death certificate.
So we now have several ways where the diagnosis of COVID-19 can erroneously end up on a death certificate:
- The patient didn’t die because of COVID but tested positive
- The patient didn’t have a COVID test, but the attending physician thought that it was more likely than not that the patient died of COVID, but was wrong
- The hospital had a financial incentive to place COVID-19 on the problem list which ended up being transferred to the death certificate
Excess Deaths Early in the Pandemic or Due to the Pandemic?
Early Pandemic Deaths
One of the consistent cries of many in the media is excess deaths in the early pandemic. The claim is that there were likely deaths that occurred early in the pandemic before testing was common and that these were likely due to coronavirus and not counted in the death toll (9). There have been reports of possibly tens of thousands of uncounted deaths.
Since most of this media reporting is based on the study I referenced above, let’s critically review that research. First, none of the study authors who make this claim actually reviewed a single hospital record. The claim is simply based on the idea that more respiratory disease deaths occurred in x month in 2020 than in 2019 or other prior years. Hence, the reported number of excess deaths may be significantly inaccurate. Why? Any study looking at excess deaths due to COVID-19 would need to have reviewed a sample of hospital charts to ensure that the symptoms, lab tests, and other findings were consistent with COVID-19.
Deaths Due to the Pandemic
The discussion on excess deaths rarely takes into account what I see on the ground every day as a physician, that sick people without coronavirus are terrified of going to a hospital or an emergency room. Hence, how many of these excess deaths could easily be due to the fact that more people with existing respiratory problems (who were told through the media that they were high risk to die from COVID if they went to the hospital) were choosing to neglect medical care?
In addition, the American Academy of Family Physicians recently warned that the COVID economic recession may kill between 60-150,000 people through deaths of despair (10). Why? Lost jobs, social isolation, and increased stress will push some over the proverbial edge.
Neither of these concepts of deaths due to the Pandemic is commonly accounted for in the discussion of excess deaths.
More Confusing Statistics
As Mark Twain once said, there are three kinds of lies: “Lies, damn lies, and statistics”. A paper came out this week (really more of an opinion piece with a model by an academic) that appears to have pegged the infection fatality rate (IFR) of COVID-19 at an astounding 1.3% (3). The IFR is your risk of dying if you get the coronavirus, hence it’s really the rubber meets the road risk number.
Reading between the lines in the paper shows that the 1.3% number is the number of people out there with symptoms who perished, ignoring the huge number of people who get coronavirus and are asymptomatic. Hence, this 1.3% number is an inflated IFR. The real infection fatality rate is the number of people who died versus the total number who are infected with and without symptoms.
What Happens to the IFR if We Adjust NY Deaths Like Colorado?
New York’s state’s IFR had come in around 0.5% which was higher than Germany and that seen in Santa Clara California (6,7). However, what happens if we take 24% off the NY state death toll because there are patients who didn’t die because of the coronavirus, they died with the coronavirus. Interestingly, that takes if from 0.5% to 0.38%. That’s remarkably similar to the rate calculated by German scientists of 0.37%.
What’s the Real Total Coronavirus Death Toll?
How can we compare the two most common media narratives for accuracy? One argues that we need more aggressive measures to shutdown society and keep it shut because we have under-reported COVID-19 deaths. The opposite narrative argues that we have over-reported COVID-19 deaths and the reaction to that data has caused more people to die and not fewer. I created a diagram above.
Ultimately, you need to decide for yourself which narrative to believe. However, as a physician, it’s scary that we have two narratives at all. This is all supposed to be hard data on which we’re basing public health policy and not politics.
The upshot? We need to understand the actual COVID-19 death numbers as our ability to have a functioning society depends on accurate pandemic math. However, as you can see, the reported COVID-19 death numbers are all over the map right now. Hopefully, we’ll take politics out of this calculus and proceed, as always, with a heavy heart and a clear mind.
(1) CBS 4 Denver. New COVID-19 Death Dispute: Colorado Coroner Says State Mischaracterized Death. https://denver.cbslocal.com/2020/05/14/coronavirus-montezuma-county-coroner-alcohol-poisoning-covid-death/ Accessed 5/16/20.
(2) Colorado Department of Public Health. Case Data. https://covid19.colorado.gov/data/case-data Accessed 5/16/20
(3) Basu A. Estimating The Infection Fatality Rate Among Symptomatic COVID-19 Cases In The United States. Health Affairs. https://doi.org/10.1377/hlthaff.2020.00455
(4) Politico. Cuomo, de Blasio can’t agree on how many New Yorkers have died from coronavirus. https://www.politico.com/states/new-york/city-hall/story/2020/05/13/cuomo-de-blasio-cant-agree-on-how-many-new-yorkers-have-died-from-coronavirus-1283901 Accessed 5/16/10.
(5) Armour BS, Pitts MM, Maclean R, et al. The Effect of Explicit Financial Incentives on Physician Behavior. Arch Intern Med. 2001;161(10):1261–1266. doi:10.1001/archinte.161.10.1261
(6) Bendavid E, et al. COVID-19 Antibody Seroprevalence in Santa Clara County, California. medRxiv 2020.04.14.20062463; doi: https://doi.org/10.1101/2020.04.14.20062463
(7) Streeck, et al. The University of Bonn. Infection fatality rate of SARS-CoV-2 infection in a German community with a super-spreading event. https://www.ukbonn.de/C12582D3002FD21D/vwLookupDownloads/Streeck_et_al_Infection_fatality_rate_of_SARS_CoV_2_infection2.pdf/%24FILE/Streeck_et_al_Infection_fatality_rate_of_SARS_CoV_2_infection2.pdf Accessed 5/5/20.
(8) US Department of Health and Human Services. CARES Act Provider Relief Fund. https://www.hhs.gov/coronavirus/cares-act-provider-relief-fund/index.html Accessed 5/5/20.
(9) Weinberger D, et al. Estimating the early death toll of COVID-19 in the United States.
medRxiv 2020.04.15.20066431; doi: https://doi.org/10.1101/2020.04.15.20066431
(10) The Well Being Trust and the Robert Grahm Center (American Academy of Family Physicians). PROJECTED
DEATHS OF DESPAIR from COVID-19. https://wellbeingtrust.org/wp-content/uploads/2020/05/WBT_Deaths-of-Despair_COVID-19-FINAL-FINAL.pdf Accessed 5/8/20.
(11) Centers for Disease Control and Prevention. Vital Statistics Reporting Guidance. Report No. 3 ▪ April 2020. https://www.cdc.gov/nchs/data/nvss/vsrg/vsrg03-508.pdf Accessed 5/17/20.