Misleading Covid Death Numbers

This article (and future ones) are available at my Substack – “Check Your Work”

For a variety of reasons, the number of Covid deaths being reported in the media or by “experts” are misleading and not accurate. I’ve had a few Twitter threads about this, but there are two simple and obvious things that many people continue to get wrong, and why numbers you see cited by various sources are often misleading or sometimes just patently false:

  • Daily numbers are not representative of trends (versus 7-day averages or weekly totals)
  • Deaths by report date include deaths from weeks to months ago (they didn’t happen the day before)

In addition to these inconsistent and delayed reporting issues, surveillance deaths are deaths among cases, including deaths where Covid isn’t on the death certificate, and even when it is on the death certificate, it’s often not listed as the underlying cause of death. This is the with versus from debate about incidental Covid deaths that often comes up, and it is more of a problem in 2022 than ever before. I’ll address those issues in a future post.

Single Day Numbers are Misinformation

Recently, I discovered some big accounts on Covid Twitter reporting single day death numbers from BNO News. Two and a half years into the pandemic, there is simply no excuse for this. It’s completely wrong if a single day’s number is presented as if they represents the number of people who are dying “daily” or who died “yesterday.” Most states are no longer reporting daily – many report only once per week. This means many states report 7 days worth of death reports at once. If someone has a regular job and gets paid $1000 every Monday, it’s false to say they get paid “$1000 per day” or that their $1000 pay check is for one day’s worth of work. These “experts” inevitably hype when there are a large number of deaths reported in one day, but ignore days where only single digits of deaths are reported. As usual, I’m left wondering if the “experts” who share these numbers on Twitter are incompetent or dishonest, and neither are good.

There is literally NO reason to look at the daily number of reported deaths, because it’s a meaningless number – way too high some days and way too low others. This is the reason sites calculate 7-day averages of Covid metrics. The 7-day averages have always been better than single day numbers, but have become more important in 2022, now that many states don’t report daily.

In the past week, many people fear mongered about over 1200 reported deaths in one day. One of the people who tweeted this is Alex Meshkin, CEO of Covid testing company Flow Health and well-known Covid fear monger. Most people ignored that Kansas reported almost 500 deaths as part of a “death reconciliation” process, and that only 7 of the deaths they reported were from the previous week. (By the way, death reconciliation often means matching deaths to known Covid cases – incidental Covid surveillance deaths. More on this in a future post.)

Then on Saturday, many people reported over 600 Covid deaths, even though the 7-day average was much lower. Megan Ranney tweeted that there were 616 Covid deaths “per day” – complete misinformation. It’s absolutely the same as if I were to tweet that there are zero Covid deaths per day – the number reported by BNO on Monday (10/3). Jerome Adams also QTd BNO News and called out the 600 deaths with a snarky comment about Covid not being over. Note how different the new cases and new deaths numbers are in the two tweets below, just a few days apart. That’s because a lot of states report multiple days’ worth of data on Fridays, and only a few states report at all on Sundays. The lines directly below New cases and New deaths list the 7-day averages for cases and deaths. You’ll see those are fairly similar, because their whole purpose is to smooth out the day-to-day reporting inconsistencies.

A Specific Issue with BNO News Tweets…

Another misleading aspect of the BNO News tweets is that they list how many states reported each day, so when the number is inevitably less than 50/50, people get the impression that the numbers are incomplete. They don’t understand that many of the states are submitting several days’ worth of reports at once. Georgia, for example, only reports deaths on Wednesday afternoons, so our weekly report is equivalent to 7 daily reports. (And we’re one of the 10 largest states by population, so it’s a big impact.)

All states are reporting deaths – they just don’t all report daily. Just because states aren’t reporting daily, doesn’t mean deaths are going unreported.

On a day when it says 28/50 states reported, some people tweet that you have to almost double the number because that means it’s only about half the states reported. But if 10 of those 28 states reported a full week’s worth of deaths at once, that’s 70 days’ worth of reports, plus the other states reporting, so it’s equal to much MORE than a single day’s worth of death reports. Many people don’t understand this, and it’s a crucial point.

I think BNO News should remote the number of states reporting , as well as the daily case and death numbers, from their daily tweets. They could just provide the 7-day averages which smooth out the issues of differing state reporting frequencies. I reached out to BNO News about how misleading their tweets were, but they stand by them and have no interest in changing their format to be more accurate, so I told them I will start calling out their tweets for misinformation, and exposing any of the “experts” who retweet them or share their numbers. I encourage others to do the same. Feel free to link to this post.

For what it’s worth, BNO News does include the 7-day averages for cases and deaths in their tweets (though it’s not labeled well so you may not realize it). I recommend unfollowing BNO News and ignoring their tweets entirely, but if you do see one of their tweets and want to glean what you can from it, ignore the parts that I crossed out and focus on the highlighted lines for case and death data. The 7-day averages of cases and deaths are much more relevant than any single day’s number of cases or deaths.

Personally, I don’t think it’s a good idea to follow the numbers daily. It’s an unhealthy obsession at this point, and even the 7-day average numbers have a lot of issues. But if you want to periodically check the 7-day average of cases and/or deaths or see how things are trending, here are a few sites that you can check instead of following the BNO News Twitter account:

Reported Deaths are Severely Lagged

Another issue to remember that has been true since the beginning is that reported deaths are NOT deaths from the previous day. They are mainly deaths from weeks or even months prior. So reported deaths remain high well after deaths are falling in real time. This is because when someone dies, it takes time for information about their death to get reported to the state health department, and then it can take additional time for the state to determine/indicate that death is associated with a Covid case. Different states use various mechanisms for doing this (like the reconciliation process in Kansas that I mentioned earlier). After cases and hospitalizations peak during a wave, deaths start falling shortly after in reality. But due to these reporting lags, the death report numbers take a longer time to fall. Deaths also sometimes spike erratically due to a backlog of deaths reported all at once, like the Kansas report from last week. These backlogs make it appear that numbers are going up, but then the numbers drop back down when the backlog falls out of the 7-day average. (Some sources like The New York Times Covid tracker attempt to adjust for these backlogs when they are announced by states, so their data is further smoothed to avoid misleading trends.)

This is one reason why even the 7-day average of deaths is not a very good representation of the number of people dying with Covid at the current time. And yes, I said “with Covid” because these surveillance counts aren’t looking at whether Covid was the underlying cause of death or not. More on that in my follow up post.