Data-driven analysis to understand long COVID using electronic health records from the RECOVER initiative 2023 Zang et al

Even with that important proviso, I thought this study was well done and made for very interesting reading.

The differences between the two data sets, Florida and New York, are really remarkable. There was a lower incidence of PASC associated conditions in Florida and a much smaller range. The Florida data set had more disadvantaged people who will have had less access to care, and possibly less focussed care, on average. The Florida data set was also younger, and most infections happened later than in New York, so knowledge of appropriate treatment of Covid-19 was better.

They analysed the large datasets a number of ways, including splitting the data out just for people with no co-morbidities prior to infection. Even in that healthy group, there were still patterns consistent with ME/CFS.

It's notable that anxiety is not one of the more common post-covid conditions in the New York dataset, and there was actually less anxiety in the post-Covid Florida data set than in the uninfected controls. Depression, oddly, did not even get a mention in both datasets.

 

Here's just an example of the data presentation - the charts are very clear throughout, and well worth a look through. These excerpts are for Adjusted excess burden (numbers per 1000), just for the New York sample:





This is a massive sample, so it carries some weight. Diagnoses were added to patient records between 30 and 180 days after infection. Look at the incidence of Malaise and fatigue, 23 people per 1000 infected (over the prevailing rate in uninfected people). It's not different between males and females. The difference in rates for white and black people probably reflects a difference in age distributions as well as differences in access to quality care, rather than a biological difference. People with prior mental illness were not more likely to get 'Malaise and fatigue' on their records (22 people per 1000). People with no prior comorbidity had a much lower incidence (12.4 per 1000).

There is no mention of CFS.

I assume the big health insurance companies have done hundreds of these types of analyses; they probably know a lot more about the effects of Covid-19 than anyone.

 

I'll just add this, to explain those two codes on the last line of the chart excerpt above.
U099 - post-COVID-19 condition, unspecified, and
B948 - sequelae of other infection and parasitic diseases

These probably overlap with ME/CFS.

Edit to add - supplementary Table 2 shows which codes they analysed and which they didn't. 'Chronic fatigue- unspecified' is probably included in the Malaise and fatigue group. 'Post-viral fatigue syndrome' was a code included in the analysis but it was either grouped in Malaise and fatigue, or wasn't a common enough diagnosis to rate.

 

That chart shows increased drug prescription rates in people infected with
covid-19 in the 30 to 180 days after the acute infection.

The top blue section is for neurological drugs - so that is where the anti-depressants would show up if there was a big spike in them. There is some increase (moderate relative to the increase in drugs for lung function, for example, and, really weirdly, witch hazel. What's with that? Is it a New York thing? Maybe the witch hazel rep was offering doctors a big bonus for prescribing it?)

Figures are adjusted hazard ratios
quetiapine (1.59) - antipsychotic for schizophrenia, bipolar disorder and major depressive disorder
melatonin (1.55) - sleep medication
mirtazapine (1.45) - antidepressant, also used for OCD and anxiety
lorazepam (1.34) - Ativan - a benzodiazepine, for anxiety, sleep problems, seizures

witch hazel (2.28) - used topically to soothe skin and eye irritation and itching

ibuprofen (1.32) - painkiller