The study in question seems heavily biased toward people who were hospitalized with COVID:
> Findings: From a total of 2100 studies identified, 57 studies with 250 351 survivors of COVID-19 met inclusion criteria. The mean (SD) age of survivors was 54.4 (8.9) years, 140 196 (56%) were male, and 197 777 (79%) were hospitalized during acute COVID-19.
79% hospitalization rate is obviously not correct for the average COVID infection.
However, the study results are still significant: A large portion of the hospitalized COVID-19 patients are experiencing long COVID symptoms.
A separate question (that isn't really answered by this study) is how many non-hospitalized people are experiencing long COVID. That's a difficult question to answer because "long COVID" has various vague definitions. Unfortunately, I think we need to introduce multiple levels of long COVID to describe the symptoms, as the current definition is so broad that it covers everything from people experiencing mild nagging symptoms to people so debilitated that they struggle to be upright for any significant amount of time. Lumping the entire spectrum into a single definition seems like a quick way to downplay the severity of severe long COVID.
I'd first questioned your characterisation. If anything it's an understatement. This is both deceptive characteristation by Science Daily and poorly a communicated study by Groff, et al, and JAMA Open Networks, a peer-reviewed publication. How the paper itself was accepted and published as-is is concerning. This is despite my own concerns over long-Covid --- they're not helped by poor research and misleading media.
Note that that quote comes from the paper itself, not the submitted summary, for anyone trying to find it:
And I'd agree, this suggests a significant issue with data selection and inclusion which would greatly weaken if not contradict the headline and summary from Science Daily.
For the record, US CDC (as of July, 2020) was reporting an overall hospitalisation rate of 3.4% and a peak observed rate of 7.4% for the age 65+ population:
I've contacted several of the paper's authors, Science Daily's editor, and the editors of JAMA Open Networks. Both the study and communication of it are greatly flawed.
Note: Edited to add CDC reference above, initial paragraph, and note that I've contacted editors and authors.
I've heard back from one of the contributing authors who states:
As stated in the supplement we conducted sensitivity analysis which showed no difference in the rate of PASC between hospitalized and non-hospitalized patients. Therefore expecting 50% of all survivors to develop PASC is not unreasonable.
(Via email.)
The supplement (it's a PDF attachement to the JAMA Open Networks article) ... consists of two figures, one identifying data flow process in the study, the second showing response of PASC to income, proportion hospitalised, and study quality. The 2nd of these, a simple box-and-whiskers plot, is apparently meant to address the sampling concerns raised here.
I'm not satisfied that this is either sufficient explanation or effective communication of the point. Though I'll note that the information is provided, if well-buried and exceedingly scant.
I got tinnitus from a bad illness I caught from a friend’s sick kid several years ago, and have since noticed the effects of other illnesses more carefully. I feel that all these long COVID symptoms are no different than lingering symptoms that perhaps all illnesses have. It’s just that we are so focused on COVID that we are only now broadly recognizing this phenomenon.
6 comments
[ 2.2 ms ] story [ 28.3 ms ] threadThe study in question seems heavily biased toward people who were hospitalized with COVID:
> Findings: From a total of 2100 studies identified, 57 studies with 250 351 survivors of COVID-19 met inclusion criteria. The mean (SD) age of survivors was 54.4 (8.9) years, 140 196 (56%) were male, and 197 777 (79%) were hospitalized during acute COVID-19.
79% hospitalization rate is obviously not correct for the average COVID infection.
However, the study results are still significant: A large portion of the hospitalized COVID-19 patients are experiencing long COVID symptoms.
A separate question (that isn't really answered by this study) is how many non-hospitalized people are experiencing long COVID. That's a difficult question to answer because "long COVID" has various vague definitions. Unfortunately, I think we need to introduce multiple levels of long COVID to describe the symptoms, as the current definition is so broad that it covers everything from people experiencing mild nagging symptoms to people so debilitated that they struggle to be upright for any significant amount of time. Lumping the entire spectrum into a single definition seems like a quick way to downplay the severity of severe long COVID.
Note that that quote comes from the paper itself, not the submitted summary, for anyone trying to find it:
https://jamanetwork.com/journals/jamanetworkopen/fullarticle...
And I'd agree, this suggests a significant issue with data selection and inclusion which would greatly weaken if not contradict the headline and summary from Science Daily.
For the record, US CDC (as of July, 2020) was reporting an overall hospitalisation rate of 3.4% and a peak observed rate of 7.4% for the age 65+ population:
https://web.archive.org/web/20200709001525/https://www.cdc.g...
I've contacted several of the paper's authors, Science Daily's editor, and the editors of JAMA Open Networks. Both the study and communication of it are greatly flawed.
Note: Edited to add CDC reference above, initial paragraph, and note that I've contacted editors and authors.
As stated in the supplement we conducted sensitivity analysis which showed no difference in the rate of PASC between hospitalized and non-hospitalized patients. Therefore expecting 50% of all survivors to develop PASC is not unreasonable.
(Via email.)
The supplement (it's a PDF attachement to the JAMA Open Networks article) ... consists of two figures, one identifying data flow process in the study, the second showing response of PASC to income, proportion hospitalised, and study quality. The 2nd of these, a simple box-and-whiskers plot, is apparently meant to address the sampling concerns raised here.
I'm not satisfied that this is either sufficient explanation or effective communication of the point. Though I'll note that the information is provided, if well-buried and exceedingly scant.