What is the REAL U.S. COVID-19 Fatality Rate? 5.8%? 16.8%? Somewhere In-Between?
How are you behaving given the knowledge that one in seventeen people who get COVID-19 DIE? Will you change your behavior if you know that one in six people who get COVID-19 DIE?
How is the fatality rate for COVID-19 actually calculated? Most websites that provide statistical information regarding COVID-19 include, at a minimum, Total Cases and Total Deaths, and many of those also provide a calculated Case Fatality Rate that is simply Total Deaths divided by Total Cases. In the United States, today, June 1, U.S. Total Deaths are 106,323 and U.S. Total Cases are 1,842,873, so the simply calculated U.S. Case Fatality Rate is (106,323/1,842,873)*100 = 5.8%. That means that one in seventeen cases ends in death.
Is the COVID-19 fatality rate really just 5.8%?
After someone is tested and diagnosed with COVID-19, it can be a fairly long time before we know if that person will actually recover. A person can be in the hospital for more than a month before we know if they will survive. Likewise, death can be sudden, or a person can linger on life support for many, many weeks.
Pending cases should not be included as “recovered” cases in fatality rate calculations
The thing is, calculating a fatality rate based on the number of total cases and the number of total deaths on a given date is patently false, primarily because it makes the assumption that every case that is not already a death is automatically counted as a recovery. Well, we know for a fact that, on any given date, that assumption is wildly inaccurate. On any given day, there are hundreds of thousands of open cases—people who have been diagnosed with COVID-19 and who may or may not recover. Likewise, they may, or may not, die. A more accurate fatality rate calculation would look at deaths relative to cases that have had a resolution, not just total cases. Total cases refers to all the people who have tested positive or been diagnosed as having COVID-19. But an accurate fatality rate should be based only on cases where we have a resolution, meaning, of the people who have had a actual resolution—who have either recovered or died from COVID-19—how many died?
This has been bothering me for a while. The main website I was using to check on COVID-19 statistics on a daily basis was www.ncov2019.live by Avi Schiffmann. I still check it every day. Avi Schiffmann, a 17-year old kid in Oregon (or maybe it’s Washington) was the first person to develop a COVID-19 tracking website because he realized IN DECEMBER that this was going global and going global fast. I heard about Avi and his website on Democracy Now! with Amy Goodman in January, I think.
Where my COVID-19 tracking journey began
Back when I first started recording data daily from Avi’s site in mid-March, the virus had not yet spread to all countries. That was one of the things I found interesting: that only 144 out of 195 countries had suffered an outbreak by mid-March. Sure, it was in only a handful of countries in January, but I was sure coronavirus was going to cover the planet, so I tracked the country count in addition to the COVID-19 stats that are most relevant to me personally, which are cases/deaths/recoveries for the globe, the USA, California (where I live), and South Dakota (where my parents live). Back then, on March 18th (Day 12 of my own self-quarantine), none of these stats were terrifying in and of themselves. The terror for me was the knowledge that we were not taking coronavirus seriously in the United States and it was definitely going to get a LOT worse. I had no doubt. I didn’t care what politicians were saying. I watched the daily increase in virus spread and I KNEW it was going to get bad in the United States because it was obviously already here and spreading and we were doing absolutely nothing to contain or even identify the spread. With data from Avi Schiffman’s site, here’s where we were then, and where we are now (as of May 31th, Day 86 of my self-quarantine):
| 3/18 Cases | 3/18 Deceased | 3/18 Recovered | |
| Global | 214,797 | 8,779 | 83,574 |
| USA | 7,539 | 118 | 8 |
| California | 636 | 11 | 1 |
| South Dakota | 11 | 1 | 0 |
| 5/31 Cases | 5/31 Deceased | 5/31 Recovered | |
| Global | 6,087,409 | 368,494 | 2,687,332 |
| USA | 1,802,180 | 104,166 | 384,821 |
| California | 107,059 | 4,137 | unknown |
| South Dakota | 4,960 | 62 | 4,336 |
So, back in the beginning—back in the beginning of my own tracking in March, that is—there were less than 8,000 cases in the country, 118 people had died in America, 11 people had died in California, and South Dakota suffered its first reported death (which happened to be in Rapid City, on the opposite end of the state from where my parents live). Avi Schiffmann also tracked how many people had recovered, though I found out later that very few states in America were reporting on recoveries so the US recovery numbers were vastly underreported. To be fair to the states, no one was paying attention to recoveries back then. (I’m not sure why California bothered to report a single recovery back in March, given that no recovery number is available for California on Avi’s site today (hence the “Unknown” in my table above). Heck, given that it can take about three months to completely get through COVID-19, it turns out it was rather ridiculous to report on recoveries at all back in March. I’m not sure it’s altogether reasonable to trust state numbers regarding recoveries now, to be honest.
[Be right back. Seoul is apparently starving and keeps pulling at my hands with her paws so I can’t type. I love my cat, but being a cat-slave can be tedious when you have things you want to get done for yourself, ya know?]
Reported fatality rates by global region
From the beginning, one thing that I found interesting on Avi’s site (www.ncov2019.live) was the way he grouped nations by geography and provided aggregate stats by global region. The first time I actually recorded those regional fatality numbers in my journal (April 3rd), I felt the fatality rates were interesting in their variation:
| Global Region | 4/2 Fatality Rate | 4/13 Fatality Rate | Change |
| World | 6% | 7% | 1% increase |
| Asia | 4% | 5% | no change |
| Europe | 8% | 9% | 1% increase |
| USA* | 3% | 5% | 2% increase |
I wondered to myself: why was the fatality rate so high in Europe compared to Asia? Further research suggested it had to do with a lack of hospital beds/ICUs/ventilators, but there’s also the point that Asia, China in particular, had already spent massive efforts in strict quarantines and contact-tracing by the time April rolled around, whereas other global regions were just beginning to grasp the scope of coronavirus at that point. I’d be interested in analysis that compares countries based on how many days into coronavirus they are, and some such data is out there, but gathering data for every country based on a equal status starting point is beyond me, even if I do think it would be interesting to know.
Getting at more complete data
I never found out precisely why Avi changed his website, but though he continues to report stats by global region today he stopped showing fatality rates in mid-April. Maybe he came to the realization that the simple total deaths to total cases was an inappropriate calculation, I don’t know. When Avi stopped reporting fatality rates, I started looking for additional sources of COVID-19 data. I was pleasantly surprised to find a good site for U.S. state-level data, www.infection2020.com, which includes a map of county-level data that is updated daily. So I started keeping track of COVID-19 not just by infections and deaths in a few states that I was particularly interested in personally, but also for the county my parents live in (Yankton) and my county here in California (El Dorado). That website, which is U.S. centric, also provides a Case Fatality Rate calculated as I mentioned in the beginning of this piece: “Case Fatality Rate (CFR) is the percentage of people diagnosed with COVID-19 that die from the disease. CFR = (Number of confirmed deaths / Number of confirmed cases) * 100.” Note there is no mention of the fact that an unknown number of the “Number of confirmed cases” will actually result in death, not recovery. That website, like most, doesn’t track Recoveries. Well, it has a little block that says more than 500,000 people have recovered, but it also has the caveat that: “Data on recovered cases is limited. The actual number of recovered is likely substantially higher.”
Well, let’s be honest. The actual number of cases and the actual number of deaths are likely substantially higher as well. Thanks to a federal government that refuses to establish or agree to any standards of defining or reporting COVID-19 infections, deaths, or recoveries, the only thing I know for sure is that the state-level numbers are NOT all recorded or reported by the same standards. Given additional flubbing being reported on anecdotally in states like Florida or by the Veterans Administration, any state-level data is suspect at best. It took me hours of searching over multiple days to finally find a website that even attempts to collect data on COVID-19 Recoveries for every state. I eventually got there by searching out various websites referred to on the CDC (U.S. Center for Disease Control and Prevention) website.
The motherlode of state-level coronavirus data
I finally ended up at a very interesting website, www.covidusa.net, that provides a whole lot of very interesting data analysis both nationally and at the state level.
They also use multi-day rolling averages to calculate simple predictions based on those rates. For example, to cite www.covidusa.net:
“Yesterday (May 31st), there were 20,007 new confirmed* cases, 28,297 recoveries, and 605 deaths. The current 7-day average of 20,990 new cases/day has declined –8.2% from 14 days ago. At that rate. . .”
| in 7 days | in 14 days | in 30 days | |
| Total Cases | 1,935,614 | 2,068,909 | 2,344,229 |
| Estimated New Cases | 141,834 | 275,129 | 550,449 |
| Total Deaths | 112,266 | 119,997 | 135,965 |
| Estimated New Deaths | 7,816 | 15,547 | 31,515 |
They also have testing stats, hospitalization/ICU stats, and a flu comparison that shows that if we had as many COVID-19 cases as we do flu cases in a year (COVID-19 being 45 times more deadly than the flu) then COVID-19 deaths THIS YEAR would total 1.69 MILLION. That’s DEATHS, people, not cases. So, to all you idiots still quoting He-Who-Shall-Not-Be-Named and saying “It’s like the flu,” well, grow up and stop with the willful ignorance. Please.
Finding data on recoveries is harder than you think
These days, I’m doing my own daily tracking on a lot more than case and death totals for just a couple of counties. The hardest data to find has been the bit of data most important to determining an accurate fatality rate: Recovery . . . not just Total Recoveries for the entire United States or other countries, but Recoveries reported at the state level that can be used to calculate the Total Recoveries in the United States.
The question that irritates me most
What I want to know is: why is the Fatality Rate based on Total Cases, when it should actually be based on Cases with Outcomes? To me, that is much more relevant. The only way we can calculate an Outcome Fatality Rate is by disregarding cases that are still pending. We only want to include cases that have had an outcome. And the only “outcomes” of a COVID-19 case are recovery and death. Pending cases, which are people still in treatment or simply waiting out the virus at home, shouldn’t count in calculating fatality rates.
Calculating an Outcome-Based Fatality Rate
Total Cases is much too inclusive a number. What matters here is Total Cases MINUS Pending Cases. Remember, there are only two outcomes to COVID-19: Recovery or Death. Anyone still pending (not yet recovered and not yet dead) doesn’t count when calculating the outcome fatality rate of COVID-19. The actual math is slightly more confusing, since no one does reporting on “pending” cases. You can figure out pending cases by subtracting Cases with Outcomes from the Total Cases figure. Calculation for “Cases with Outcomes” is thus Cases Resulting in Recovery plus Cases Resulting in Death. Briefly, the Outcome Fatality Rate can be calculated as:
Outcome Fatality Rate % = (Total Deaths / Cases with Outcomes) * 100
where Cases with Outcomes is Total Deaths plus Total Recoveries. Here are the U.S. stats for yesterday (5/31) using data from www.covidusa.net
| Total Cases | 1,774,680 |
| Total Deaths | 103,873 |
| Total Recoveries | 514,842 |
| Cases with Outcomes | 618,715 |
| Pending Cases | 1,155,965 |
The Outcome Fatality Rate Calculation
This makes our calculation:
| Outcome Fatality Rate = (103,873/618,705)*100 = 16.8% |
An Outcome Fatality Rate of 16.8% means that one in every SIX people who get COVID-19 DIE.
Case Fatality Rate 5.8% vs. Outcome Fatality Rate 16.8%
For the United States as a whole, the commonly reported Case Fatality Rate may be around 5.8%, but that’s based on Total Cases and vastly undercounts deaths because it assumes all open cases result in recovery. That’s a huge false assumption in my book.
I believe the Outcome Fatality Rate is much higher than the Case Fatality Rate, possibly as high as my calculated Outcome Case Fatality Rate 16.8%.
I was going to paste my entire spreadsheet of state-level data to support my calculation of the fatality rate of COVID-19 based on resolved cases only (disregarding pending cases), but you can more simply access all the state-level base data and more from the source at www.covidusa.net. The big potential for error in my calculation is in the underlying data, because I’m sure the recovery numbers are actually greater than what has been reported. Heck, Washington D.C. actually reported 1,075 recoveries one day and then changed that number to zero the next day. Either every single one of those people actually died, or, more likely, the District of Columbia decided their data on recoveries was so inaccurate that it’s preferable to simply report nothing at all on the subject of COVID-19 recoveries.
Just like the Case Fatality Rate number of around 5.8% reported on most websites is much too low because it’s based on Total Cases which counts pending cases as recovered, you can say that my Outcome Fatality Rate number of 16.8% is too high because the data on recoveries is underreported. Suppose my recovery data is underreported by half (so, assume recoveries are double what have been reported). That would put the COVID-19 Fatality Rate at 11.4%, which is a less frightening and possibly more reasonable statistic. But who really knows? Given that most reporters agree that the number of total cases is vastly underreported, you could make an argument that pretty much all analytics are suspect because of the vagaries of the underlying data.
5.8% Case Fatality Rate is Definitely Low
What we do know with some certainty is that anyone claiming that the fatality rate for COVID-19 is anywhere below 6% is flat-out wrong because they make the false assumption that anyone not already dead is definitely going to recover.
How are you going to behave now?
If you are fine with the reported one-in-seventeen chance of COVID-19 death, then go ahead with your daily business without a mask, without washing your hands, without physical distancing. But will you think twice about all that, if you know the fatality rate is more like one-in-five? Think about it. For every five people you know, one dies. Are you sure you still want to act like coronavirus doesn’t matter?