A point-by-point rebuttal of the Washington Post's double down on Covid deaths being "overcounted."

Last week, a Washington Post columnist with whom I am friendly (but have had important disagreements with) wrote that we are overcounting Covid-19 deaths in Massachusetts. The opinion relied upon anecdotal evidence and unrelated hospitalization data. Much of what it said was incorrect, unsupported, or both. Predictably, the column has been co-opted by conspiracy theorists as proof that they’ve been right all along. They could not be more wrong, and I’m saddened to see them emboldened by a column that was so factually inaccurate.

I know the column is factually incorrect because I study “excess mortality,” the most relevant field to the question at hand. I study excess mortality for the US as a whole, but I study it especially closely in Massachusetts, my home state. So, this is kind of my lane. I have data from as recently as last week showing that the opposite of what the Post column purported is actually true; In fact, we are still undercounting Covid-19 deaths. (There have been periods when we were slightly overcounting Covid-19 deaths in this state, which I have pointed out before, but we certainly are not overcounting now).

Yes, we are in a better place than we were a year ago. But you simply cannot use that fact to make a claim that the approximately 450-600 Covid-19 deaths that are occurring nationwide each day are overcounts.

Yesterday evening, The Mehdi Hasan Show on MSNBC covered my response to the column in a terrific 18-minute segment that summarized and highlighted what I wrote right here in Inside Medicine.

The first 9+ minutes are a deep dive on the issue, laid out expertly by Mehdi, followed by an interview with me and Stanford infectious diseases physician Dr. Abraar Karan (which we recorded earlier in the day).

In fact, after we recorded the segment, and while I was seeing and treating patients in the emergency room, the Post published “The Check Up,” a weekly newsletter by the same author as the piece we were debunking. I had hoped that the author would use the opportunity to walk back the overcounting claims, based on my work and that of others in this field. That did not happen. The follow-up column doubles down.

The follow-up piece requires a response. So, even though I just spent 9 hours seeing patients, I’m going to write a point-by-point rebuttal below.

Now, everything you need to know to correct the factual errors in the Post is already available for free. Again, here’s Inside Medicine from Monday, and here’s the MSNBC segment from last night.

But—I’ll admit, I’m uncharacteristically incensed right now—in large part because the thoughtful, careful, and meticulous data analyses that my research colleagues and I have carried out are being cast aside in favor of unsupported opinions.

So, I again want to set the record straight and, yes, blow off some steam. Accordingly, I’m putting the rest of this piece behind a paywall. (Update: I am removing the paywall as of January, 24th, due to broader interest than I anticipated. If you’d like to support this work, please consider clicking below to become a paid supporter of Inside Medicine.)

While the information below is helpful for refuting the factual inaccuracies advanced in the Post, is not strictly essential.

What follows is a line-by-line rebuttal of the Post piece, zero punches pulled.

So…buckle up…

“Opinion. The Checkup With Dr. Wen: We need an honest accounting of covid’s toll.” (The Washington Post).

Well, let’s just start with the headline. In fairness, this headline is an improvement from the one introducing the previous post, which now infamously reads “We are overcounting covid deaths and hospitalizations. That’s a problem.”

Yes, we do need an honest accounting of Covid’s toll. That’s exactly why rigorous research by my colleagues and I requires extreme care and dedication. It’s why we spent countless hours (months and years) making sure everything is right. We want our communities to accurately understand where we are in the pandemic, and, also, so that future historians look back on this time, and say “Wow, these people really had a handle on the epidemiology, at least.”

Let’s get into the prose:

“My column last week on how we are overcounting covid hospitalizations and deaths drew outrage from both sides. Right-wing commentators claimed that the admission came “two and a half years late.” Some on the left insisted that, based on excess mortality data, covid-19 deaths are being undercounted.

The author does not refute the misinformation coming from the right, but, rather, lets it stand, uncorrected. That’s dereliction of duty when your writing is literally being used by extremists to spread lies about the pandemic having been blown out of proportion.

But it’s also nonsense. Am I on the left because I track data closely and find that we still have excess mortality that tracks with Covid’s rises and falls? Was I “on the right” when I wrote in Lancet Infectious Diseases that for a period last spring we did in fact over count Covid deaths in Massachusetts? No. I am not on the left nor the right on issues like this: I’m on the data. I go where the data take me. When we someday are overcounting Covid deaths—which likely will occur again—I will not have suddenly switched political affiliations. I will have stayed on the same team—the one that is data-driven.

“I agree with both sets of critics in one crucial respect: We must have honest accounting and transparent reporting. That requires acknowledging that data changed over time and that deaths due to the pandemic are not necessarily the same as deaths due to covid.”

An honest accounting is precisely what all-cause mortality is about. It takes out the subjectivity. The fact that Covid deaths rise and fall in lockstep with all-cause excess mortality and that for the most part, there have been fewer Covid deaths than excess deaths, argues strongly that Covid itself is driving these deaths. But the author could be correct. Semantically speaking, these may not be the same deaths. But with data like ours, the burden of proof is on the author. What is responsible for these deaths and what evidence is offered to support those explanations? The author offers nothing.

“At the beginning of the pandemic, many people — including a lot of young and otherwise healthy individuals — became very ill from covid pneumonia. Testing was limited, and it’s likely many deaths weren’t attributed to covid that should have been.”

Correct. In journalism, this kind of paragraph is called “B matter.” It’s a set of irrefutable facts, meant to provide context. In this case, they’re unrelated to the main argument. A cynical read would be that it’s a smokescreen because the correct place to deploy such a passage would have been higher up: to refute the lies advanced by purveyors of misinformation who are now saying that the column is a de facto confession that we’ve been exaggerating Covid all along.

“This is almost certainly the case in other parts the world, such as Eastern Europe, where the official death counts from covid were probably much lower than the actual mortality from the disease. In the United States, there were probably quite a few missed covid cases early on, so even if some deaths were improperly attributed to covid, they were still likely undercounted in 2020 and much of 2021.”

The author may be right about Europe. I don’t actually know. As for the United States, it’s clearly known that we missed Covid cases early on. But the authors also says some deaths were “improperly” attributed to Covid. How many? 0.1%? 1%? 5%? 10%? 25%? 50%? 75%? 85%? 90%? All of them? What’s the source on this? And how do we know what part of 2021 is being referred to here? I am getting this granular precisely to point out how sloppy this all is, even when the micro-point being advanced is one I happen to agree with.

“Things changed by 2022 with the spread of the milder omicron variant and as the vast majority of Americans acquired excellent protection against severe disease through vaccination or infection. My colleagues and I noticed a dramatic shift among patients with covid-19. Far fewer were being hospitalized with covid pneumonia and other direct impacts of the virus.”

Actually this started happening much earlier than 2022 in many places. It really started in July and August, of 2021—at least here in Massachusetts. Now, you might think that means that the author and I agree on this. And in a sense, we half do. Yes, most of the hospitalized patients in highly vaccinated places like Massachusetts were no longer being hospitalized with Covid pneumonia. But they were being hospitalized because of SARS-CoV-2 infections that were tipping the vulnerable over the edge—sometimes to the hospital, and sometimes to the morgue. I’ve been saying this for longer than almost anyone. The whole point of excess mortality is to measure whether there are more deaths than usual. Since Covid cases (or wastewater levels) clearly kept tracking with excess deaths, even in 2021, 2022, and now, it’s not credible to say Covid is not responsible, without an incredibly brilliant explanation to the contrary. One has not been offered.

“Now, most covid patients fall into two categories: The first is those who are admitted to a hospital for a non-covid health issue but who have to take a test because of hospital protocols. Many of these patients test positive incidentally, especially in communities with higher rates of covid.”

Yup. And if these people do not die at a higher rate than they normally would (all cause), then Covid would not be a factor. But we know that all-cause deaths are up. So these patients are either not contributing to all-cause mortality, or they are because they’re dying more often than they normally would. And why would that be? Hmm!

By the way, these patients often spread Covid in hospitals. So even if they themselves were not high-risk Covid patients, they may have unwittingly killed a few others. Hospital-acquired Covid rates remain too high. (The number ought to be zero.)

“Importantly, this kind of across-the-board testing isn’t being done for flu, RSV, adenoviruses or other coronaviruses. If it were, we would see a much higher rate of hospitalizations due to those viral infections, too.”

We are doing a ton more testing for those viruses than we used to, RSV in particular. It will be interesting to see what we learn. Every winter, all-cause mortality goes up. The author is right that in the past, other viruses have driven spikes in all-cause deaths (albeit at lower rates). But our failure to measure those—and our improvement in measuring them now—is not evidence that we are overcounting Covid. It’s inexplicable to say that since we didn’t test for flu and RSV enough in the past, that this means we are now overcounting Covid. But again, if we were, a lack of excess mortality (or more Covid deaths than all-cause excess deaths) would be the surest sign of that.

“The second category includes medically frail patients for whom covid might exacerbate an underlying condition. People with severe chronic lung disease, congestive heart failure or other serious underlying conditions are at risk of being “tipped over” with any respiratory infection. If they catch covid, they could become hospitalized. Unfortunately, some die.”

Correct. And if more people die than statistically should have during a given period—and that increase tightly tracks with Covid waves—that’s Covid killing people until proven otherwise, regardless of what’s on a death certificate (in either direction).

“Because there is incredible variability in how hospitals document covid, this translates directly into variability on death certificates, where it’s common practice to enter multiple causes. If covid-19 appears on a death certificate along with several other diagnoses, it’s unclear whether covid was the primary reason for death, a contributing cause or incidental.”

This is a non-sequitur. This is why we study all-cause mortality and then look to see if that tracks with Covid waves. Some types of deaths (like accidental overdoses) do not track with Covid waves. So we can’t say that when Covid spikes, people’s behaviors change, leading to more overdose deaths. Instead, accidental overdose deaths are consistently up since the summer of 2020, and they do not correlate with Covid waves. That said, unintentional overdose deaths are a tiny fraction of excess deaths overall (but a far higher one in young adults, and a major problem). Most excess deaths—even non-Covid ones—track with, you guessed it, Covid waves. It’s almost as if Covid is bad for people with things like fragile heart conditions. (Sarcasm font.)

“This is why we need more rigorous research like what Shira Doron from Tufts Medical Center is spearheading. Hospitals and health departments should use a uniform set of criteria to classify covid hospitalizations and deaths. The accounting should also be retrospective so that we can put to bed the criticism that severe illness was overcounted all along.”

Again, hospitalization data is a non-sequitur to this conversation. Imagine saying that because most cars in the shop need minor repairs that nobody is dying on the road. Connecting these issues is a bizarre conceit in the face of the mortality data. But the kind of research that is being proposed here is both lower quality than the comprehensive mortality data we have and also, it has not even been done. Sorry, but I’m not convinced by research that does not yet exist—and which, when it does, will have almost zero bearing on this question.

“Such analysis is more precise than the often-cited excess mortality data, which measures the number of deaths that surpass the expected number in a given period. It’s tempting to compare the current level of deaths to pre-pandemic mortality and attribute all additional deaths to covid. But this confuses correlation with causation.”

The first half of the first sentence (in italics) is wrong. It’s simply unsupported. Next, yes, correlation and causation are different, we agree. But at least my data have correlations! And strong ones, I might add. The author does not have or even refer to data that one could correlate to anything, let alone establish any causation.

“That’s because the pandemic has caused vast disruptions to the health-care system. Hospitals have been so overwhelmed that patients with heart attacks and strokes couldn’t receive timely care. Waiting times continue to be astronomical, exacerbated by the shortage of nurses and other health-care workers.”

Well, this is just an argument for better control of Covid, via masking and testing. That said, outside of extreme circumstances (which unfortunately have occurred, such as last year when hospitals were overrun by supposedly “mild” Omicron cases), for the most part, patients continue to receive the timely care that they need. If the author’s experience in the emergency room is different than that, or can show data that to support that this is the main reason why people are dying, it has not been offered.

In case you have not noticed, a troubling theme has emerged here. The author shrugs off data that undermines the apparently pre-determined conclusion, and offers pure conjecture in its place. I honestly expect more from my respected colleagues.

“In addition, primary care offices canceled in-person visits, and surgeons postponed procedures. Patients deferred cancer screenings and fell behind on blood pressure and diabetes management. In fact, a recent BMJ paper estimates that more than 30,000 excess deaths in Britain are due to worsened cardiovascular disease, which, as the top cause of death, was already exerting a heavy toll.”

I just think it’s amazing that the paper linked is a news item summarizing two studies. The first study says that, in Britain, cardiovascular deaths have gone up since 2020. No clear (or compelling) explanation is given as to why, and we are not given the timing of those deaths. If Britain is anything like the United States, then cardiovascular deaths have not gone up steadily during the pandemic; rather, they have gone up and down with Covid waves.

But the second study is from the US and it purports to show that among patients who recovered from Covid, major cardiovascular events (including death) were far more likely in the following year. Now, I’m not sure that paper has the methods to back that conclusion up (in fact, I have my reservations about the study), but it literally goes against the columnist’s point.

“We’ve also had a terrible RSV and flu season, probably because of the immunity gap from covid mitigation measures. Other infections, such as sexually transmitted diseases, are on the rise, in part because community health resources have been diverted to address the coronavirus.”

We suppressed RSV and flu so much in 2020-2021 that, yes, there’s catch up now. STIs may be up, but I don’t see data that STI deaths are up—and we are talking about deaths here. HIV deaths appear to have gone down during the pandemic. Next you’ll tell me that this is because the pandemic has made everyone so lonely and that they’re killing themselves instead of having Chemsex and seroconverting. Nice try but we looked at that too. Suicides are down since the pandemic began.

“And let’s not forget that “diseases of despair” are escalating, with skyrocketing deaths from opioid overdose and alcoholism and increasing depression and mental health distress, including among youths.”

As per above, deaths from mental health are down. Suicides went up slightly for young adults in 2021 (after being down during the “shelter-in-place" period and 2020 overall), but went down for other demographics. And, good news, suicide mortality is back down for 2022, for the months that reliable data exist (through June 30, 2022).

Yes, accidental overdoses are up. That doesn’t negate the Covid deaths. And I haven’t checked alcohol deaths. Ok, fine, I’ll go check. Stand by, I’ll be right back. And…I’m back. See? It’s good to check these things. Alcohol deaths were up some during 2020-2021, possibly around 5,000-7,000 per year in adults ages 18-50. But in 2021, 42,146 adults ages 18-50 died of Covid—and that’s with Covid as the underlying cause on the death certificate. The number is higher if we include deaths where Covid was listed as a contributing cause.

“We could consider these deaths pandemic-related deaths, since many of them would not have happened if it were not for the last three years of covid. But these deaths need to be separated from those directly caused by the SARS- CoV-2 virus.”

We absolutely do separate these deaths. In fact, my team is among the first that I’m aware of to do this properly. We’ll show the results in a forthcoming paper. Spoiler: it’s called the Covid-19 pandemic for a reason. (That said, for younger adults, overdoses, although not suicides, are a problem that has gone from bad to worse).

“Reader Chris from Minnesota summarizes this well: “There are excess deaths occurring since covid started, but if we assign all of them to covid, then we are missing other ways that our health-care system is failing.””

True. But, we are not assigning all of excess deaths to Covid. We are, however, noticing that most excess deaths track with Covid and making a responsible inference. There remains a gap between Covid deaths and all-cause deaths (indicative of undercounting Covid deaths). That’s actually the bigger epidemiologic question: Are we undercounting Covid deaths? Probably. If so, why? While I think there are strong data implying that Covid also fills that remaining gap, even if am wrong about that, Covid would be still responsible for a gigantic number of deaths.

Again, all relevant evidence is that we are undercounting Covid deaths in Massachusetts currently, and other places as well.

When we no longer have excess mortality and Covid deaths do not tightly track with known waves, we’ll be in a much better place.

And yes, we’ve made progress. But sadly, a lot of that is because so many died of Omicron, mistakenly thinking it was mild.

Let’s celebrate when the pandemic is really over, rather than declaring it over just because we are no longer seeing a 9/11 worth of Covid deaths every single day. A 9/11 worth of Covid deaths every week is also not good.