Pediatric Long COVID Subphenotypes: An EHR-based study from the RECOVER program, 2024, Lorman et al.

[SNT Gatchaman]

Now published - link here
____________________

Preprint
Pediatric Long COVID Subphenotypes: An EHR-based study from the RECOVER program
Vitaly Lorman; L Charles Bailey; Xing Song; Suchitra Rao; Mady Hornig; Levon Utidjian; Hanieh Razzaghi; Asuncion Mejias; John Erik Leikauf; Seuli Bose Brill; Andrea Allen; H Timothy Bunnell; Cara Reedy; Abu Saleh Mohammad Mosa; Benjamin D Horne; Carol Reynolds Geary; Cynthia H Chuang; David A Williams; Dimitri A Christakis; Elizabeth A Chrischilles; Eneida A Mendonca; Lindsay G Cowell; Lisa MocCorkell; Mei Liu; Mollie R Cummins; Ravi Jhaveri; Saul Blecker; Christopher B Forrest

Pediatric Long COVID has been associated with a wide variety of symptoms, conditions, and organ systems, but distinct clinical presentations, or subphenotypes, are still being elucidated. In this exploratory analysis, we identified a cohort of pediatric (age <21) patients with evidence of Long COVID and no pre-existing complex chronic conditions using electronic health record data from 38 institutions and used an unsupervised machine learning-based approach to identify subphenotypes. Our method, an extension of the Phe2Vec algorithm, uses tens of thousands of clinical concepts from multiple domains to represent patients' clinical histories to then identify groups of patients with similar presentations. The results indicate that cardiorespiratory presentations are most common (present in 54% of patients) followed by subphenotypes marked (in decreasing order of frequency) by musculoskeletal pain, neuropsychiatric conditions, gastrointestinal symptoms, headache, and fatigue.


Link | PDF (Preprint: MedRxiv) [Open Access]

 

[SNT Gatchaman]

The goal of this study is to identify subphenotypes in a large cohort of pediatric patients with evidence of non-MIS-C Long COVID. […] The foundation of our method is a concept embedding model trained from the clinical facts of 9.8 million patients to produce high-dimensional numeric representations of over 70 thousand unique diagnosis, procedure, and medication concepts. We then apply this model to represent and cluster the clinical trajectories of a cohort of pediatric patients with evidence of Long COVID.

There were 17,525 children and adolescents at 38 medical institutions identified as having evidence of Long COVID by a rules-based algorithm on the basis of a Long COVID diagnosis or post-acute evidence of Long COVID-associated diagnoses

A plurality of patients in the long COVID cohort were in the age 16-20 group (30.4% overall) and a majority were female (54.5%). Patients in this cohort were more likely to have been infected with SARS-CoV-2 during the November 2021-February 2022 period, coinciding with the Omicron wave, than in other time periods. Moderate and severe acute COVID-19 presentations were uncommon (4.9% and 3.2%, respectively).

Finally, a subphenotype (“Fatigue”, representing 5.0% of patients) was characterized by statistically significantly greater proportions of both fatigue and malaise diagnoses (41.7%) as well as Long COVID diagnoses (63.5%); diagnoses of chest pain, arrythmias, and respiratory signs and symptoms were common in this subphenotype as well.

The fatigue subphenotype was somewhat more heterogeneous; in addition to fatigue, cardiac diagnoses (chest pain and arrythmias), headaches, musculoskeletal pain, neuropsychiatric symptoms, and POTS like symptoms such as dizziness and giddiness were relatively more common, as well as non-specific Long COVID diagnoses. Although these more common groups of diagnoses did not always occur in the same sets of patients, this constellation of diagnoses is suggestive of myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS). Although specific diagnostic codes for ME/CFS exist, and a new ICD-10CM code was introduced on 1 October 2023, the disease remains very likely to be under-diagnosed, particularly in children. In addition, because clinical criteria for ME/CFS require symptoms to persist for a minimum of 6 months from onset before assigning the diagnosis, the use of a 28-to-179-day observational window following the index infection in this study made it impossible to strictly meet the 6-month criterion for establishing an ME/CFS diagnosis.

While fatigue was the most commonly reported Long COVID feature in some studies, we found cardiorespiratory presentations to be the most common subphenotype, with the fatigue subphenotype above representing only about 5% of patients. However, diagnoses of fatigue were present across multiple subphenotypes (particularly the headache and musculoskeletal pain subphenotypes, in addition to the fatigue subphenotype). This suggests that fatigue often presents not in isolation but in combination with other aspects of Long COVID and may be present across multiple Long COVID manifestations. Other prospective studies may be able to capture fatigue more reliably than EHR data sources.

 

[rvallee]

They did health record studies on conditions literally characterized, hell defined, by basically not fitting into any typical health record concepts and usually not even coded where it does, or not coded properly, or miscoded. Again. Another one.

Good grief. They're just pissing the whole thing away.

 

[SNT Gatchaman]

Now published —

Pediatric Long COVID Subphenotypes: An EHR-based study from the RECOVER program
Vitaly Lorman; L. Charles Bailey; Xing Song; Suchitra Rao; Mady Hornig; Levon Utidjian; Hanieh Razzaghi; Asuncion Mejias; John Erik Leikauf; Seuli Bose Brill; Andrea Allen; H Timothy Bunnell; Cara Reedy; Abu Saleh Mohammad Mosa; Benjamin D Horne; Carol Reynolds Geary; Cynthia H. Chuang; David A Williams; Dimitri A Christakis; Elizabeth A Chrischilles; Eneida A Mendonca; Lindsay G. Cowell; Lisa McCorkell; Mei Liu; Mollie R Cummins; Ravi Jhaveri; Saul Blecker; Christopher B. Forrest; on behalf of the RECOVER Consortium

Pediatric Long COVID has been associated with a wide variety of symptoms, conditions, and organ systems, but distinct clinical presentations, or subphenotypes, are still being elucidated.

In this exploratory analysis, we identified a cohort of pediatric (age <21) patients with evidence of Long COVID and no pre-existing complex chronic conditions using electronic health record data from 38 institutions and used an unsupervised machine learning-based approach to identify subphenotypes. Our method, an extension of the Phe2Vec algorithm, uses tens of thousands of clinical concepts from multiple domains to represent patients’ clinical histories to then identify groups of patients with similar presentations.

The results indicate that cardiorespiratory presentations are most common (present in 54% of patients) followed by subphenotypes marked (in decreasing order of frequency) by musculoskeletal pain, neuropsychiatric conditions, gastrointestinal symptoms, headache, and fatigue.

Link | PDF (PLOS Digital Health) [Open Access]