Data-driven identification of post-acute SARS-CoV-2 infection subphenotypes, 2022, Zhang et al

Data-driven identification of post-acute SARS-CoV-2 infection subphenotypes
Hao Zhang, Chengxi Zang, Zhenxing Xu, Yongkang Zhang, Jie Xu, Jiang Bian, Dmitry Morozyuk, Dhruv Khullar, Yiye Zhang, Anna S. Nordvig, Edward J. Schenck, Elizabeth A. Shenkman, Russell L. Rothman, Jason P. Block, Kristin Lyman, Mark G. Weiner, Thomas W. Carton, Fei Wang, Rainu Kaushal

The post-acute sequelae of SARS-CoV-2 infection (PASC) refers to a broad spectrum of symptoms and signs that are persistent, exacerbated or newly incident in the period after acute SARS-CoV-2 infection. Most studies have examined these conditions individually without providing evidence on co-occurring conditions.

In this study, we leveraged the electronic health record data of two large cohorts, INSIGHT and OneFlorida+, from the national Patient-Centered Clinical Research Network. We created a development cohort from INSIGHT and a validation cohort from OneFlorida+ including 20,881 and 13,724 patients, respectively, who were SARS-CoV-2 infected, and we investigated their newly incident diagnoses 30–180 days after a documented SARS-CoV-2 infection.

Through machine learning analysis of over 137 symptoms and conditions, we identified four reproducible PASC subphenotypes, dominated by cardiac and renal (including 33.75% and 25.43% of the patients in the development and validation cohorts); respiratory, sleep and anxiety (32.75% and 38.48%); musculoskeletal and nervous system (23.37% and 23.35%); and digestive and respiratory system (10.14% and 12.74%) sequelae. These subphenotypes were associated with distinct patient demographics, underlying conditions before SARS-CoV-2 infection and acute infection phase severity.Our study provides insights into the heterogeneity of PASC and may inform stratified decision-making in the management of PASC conditions.

Link | PDF (Nature Medicine)

 

Note, one of the authors is Dhruv Khullar who published a problematic, minimising piece on LC in The New Yorker in 2021. See this post.

 

This looks like classic GIGO. Healthcare records are not reliable for this because they don't record most of the pertinent data. It's basically as reliable as this:

Spoiler: Photo of a botched taxidermy of a lion

[\SPOILER]

Machine learning is not useful for this stuff, it cannot solve unsolved problems, it can only automate labor-intensive-solved-problems.

Don't use healthcare records for this. They are as useless as it gets. Or, ideally, fix healthcare records so they don't filter out most of the data.

 

You need to hide that AI generated horror show behind a content warning spoiler @rvallee Not sure which bit is more disturbing, the tongue, the eyes? Maybe the human teeth, replete with single central incisor?

Agree: GIGO.

 

Sorry for off topic, but had to correct you, because that horror show is not AI generated!
https://www.atlasobscura.com/places/the-lion-of-gripsholm-castle-strangnas-sweden

 

Hehe. Not AI-generated, though, this is an old botched taxidermy. Many exotic animals brought by explorers did not really have much of a before picture, so many early specimens looked terrible. Because they were missing data, was mostly the point I tried to make.

Without a plan for what it actually looks like, it's hard to compare whether what you have is accurate. This study only looked at things that were recorded in the siloes of medicine, with zero records for anything related to chronic illness. And it looks as much like the real thing as that poor lion was made to look like by someone who had never seen a living lion in their life.

(Definitely on topic, @moobar )

 

Wow! That's much more amazing.