Better in their words…

Eric Topol recently tweeted out a Jama Health Forum article, but he has made several errors in its interpretation. Shall we examine them?

Here is Topol’s tweet

Eric Topol @EricTopol

Heart attacks, strokes and other major adverse outcomes doubled in people post-Covid at 1 year compared with matched uninfected controls #LongCovid, replicating previous reports jamanetwork.com/journals/jama-…@JAMAHealthForum

4:01 PM ∙ Mar 3, 2023

---

3,275Likes1,722Retweets

Topol tweet contains error #1. He claims that the paper compares people post-covid to uninfected controls. See here

But that is 100% false.

All he needed to do was read the abstract.

The paper compares people who meet a PCC (post covid condition) definition, using a CDC definition to uninfected controls. Many people who had COVID do not meet this definition and are fine. It compares a tiny set of people who had covid— those with a post covid condition— to the uninfected.

Topol expands his analysis on his blog, where naturally he repeats the error, “It compared 13,345 people with Covid matched with 26,870 control without Covid” Ouch.

It is in the abstract, man.

Later in the paper it explains the PCC cohort used for analysis is ~6% of the ‘I had COVID’ cohort.

Let’s examine the paper.

The paper explains it uses billing codes placed 5-12 weeks after a COVID diagnosis to identify the group with PCC. (Uhh, but that will also identify people who hadn’t been to a doctor who have pre-existing, undiagnosed diseases).

It then uses propensity score matching on baseline variables to create the control cohort. The authors don’t have home testing results, so it cannot assure you the control arm didn’t have people with COVID in there. It compares people whose covid was so bad it drove them to the doctor to people (actually a subgroup with some ICD10 codes) to people who didn’t get covid that bad (but may have still gotten covid). It then tracks ICD-10 codes in the subsequent 12 months, and mortality. Here is mortality.

Hmmm, that is interesting, death rate starts to separate in weeks 5-12, but remember this is the soonest the PCC diagnosis can be made. PCC kills instantly! And look how bad PCC is!

Omg, it causes COPD!

Hmm, wait, how do you define PCC again?

Do you see it?

Having a dypnea diagnostic code is linked to a COPD diagnostic code.

And the authors would have you believe COVID causes COPD? haha

There you are, an innocent person who gets COVID, and it destroys your alveoli, and by week 5 you have dypnea, which is later found to be COPD.

Alternatively: people so sick with COVID19 who seek medical care at hospitals & clinics have underlying medical problems that are undiagnosed, and in the weeks after it is revealed they have COPD, when the dyspnea continues. They have probably had COPD all along, and that is why covid hit so hard they went to the doctor. Now, plugged into the system, they get the diagnosis.

It is basically saying: people who have more diagnostic codes placed in the 5 to 12 weeks after a covid diagnosis have more diagnostic codes placed in the first 12 months after a covid diagnosis. That should be in the journal of DUH!

Tachycardia can give you PCC, which then causes cardiac arrhythmia. Hahaha, I actually think the method allows you to have both diagnoses on the same visit!

The paper is saying people who get medical diagnoses placed in EHRs after COVID diagnosed in EHRs get more medical diagnoses in EHR than healthy people who may or may not have had covid.

Absolutely, 100% useless analysis, and yet it is broadcast widely & uncritically by Eric Topol.

I will reiterate my point. It is concerning to me that the media so frequently quotes someone who is not reading papers in a detailed way. The real reason the public is misinformed is because we have a crisis where many doctors, even those with long titles, do not read papers at a high level.