Summary: A new AI tool identified long COVID in 22.8% of patients, a much higher rate than previously diagnosed. By analyzing numerous health records of nearly 300,000 patients, the algorithm identifies long COVID by distinguishing symptoms related specifically to SARS-CoV-2 infection rather than pre-existing conditions.
This AI approach, known as “precision phenotyping,” helps clinicians differentiate long COVID symptoms from other health conditions and can improve diagnostic accuracy by about 3%.
Key facts:
- AI-powered precision phenotyping: Identifies long COVID only after excluding other causes of symptoms in health records, improving diagnostic accuracy.
- Broader representation: Algorithmic diagnoses reflect the demographic profile of Massachusetts, addressing biases found in traditional diagnostic codes.
- Research potential: The algorithm could advance future research on the genetic and biochemical factors of long COVID subtypes.
Source: Harvard
While previous diagnostic studies suggested that 7 percent of the population suffered from long COVID, a new AI tool developed by Mass General Brigham revealed a much higher figure of 22.8 percent, according to the study.
The AI-powered tool can sift through electronic health records to help clinicians identify long COVID cases. This often mysterious illness can encompass a litany of persistent symptoms, including fatigue, chronic cough and brain fog after SARS-CoV-2 infection.
The algorithm used was developed by extracting de-identified patient data from the clinical records of nearly 300,000 patients across 14 hospitals and 20 community health centers in the Mass General Brigham system.
The results, published in the journal MedRxivcould identify more people who should receive care for this potentially debilitating disease.
“Our AI tool could transform a fuzzy diagnostic process into something precise and focused, giving clinicians the power to make sense of a difficult condition,” said lead author Hossein Estiri, head of research on AI at the Center for AI and Biomedical Informatics of Learning. Healthcare System (CAIBILS) at MGB and Associate Professor of Medicine at Harvard Medical School.
“Through this work, we may finally be able to see long COVID for what it really is – and, more importantly, how to treat it. »
For the purposes of their study, Estiri and colleagues defined long COVID as a diagnosis of exclusion also associated with infection. This means that the diagnosis could not be explained in the patient’s unique medical record but was associated with COVID infection. Additionally, the diagnosis had to persist for two months or more during a 12-month follow-up window.
The new method developed by Estiri and colleagues, called “precision phenotyping,” sifts through individual records to identify symptoms and conditions related to COVID-19 to track symptoms over time to differentiate them from others diseases.
For example, the algorithm can detect whether shortness of breath results from pre-existing conditions like heart failure or asthma rather than long COVID. Only when all other possibilities have been exhausted would the tool flag the patient as having long COVID.
“Physicians are often faced with having to wade through a tangle of symptoms and medical histories, not knowing exactly which threads to pull, while balancing a busy workload. Having an AI-powered tool that can methodically do this for them could be a game-changer,” said Alaleh Azhir, co-senior author and internal medicine resident at Brigham and Women’s Hospital, a founding member of the health system. Mass General Brigham. .
The new tool’s patient-centered diagnostics may also help mitigate biases built into current diagnostics for long COVID, said the researchers, who noted that diagnoses with the official ICD-10 diagnosis code for Long COVIDs tend toward those with easier access to health care.
The researchers said their tool is about 3 percent more accurate than data captured by ICD-10 codes, while being less biased. Specifically, their study demonstrated that the individuals they identified as having long COVID reflect the broader demographic makeup of Massachusetts, unlike long COVID algorithms that rely on a single diagnosis code or individual clinical encounters , skewing the results in favor of certain populations such as those with better access to care.
“This broader reach ensures that marginalized communities, often sidelined in clinical studies, are no longer invisible,” Estiri said.
Limitations of the study and AI tool include that the health record data the algorithm uses to account for long COVID symptoms may be less complete than the data doctors enter into post-visit clinical notes.
Another limitation was that the algorithm did not capture possible worsening of a prior condition that could have been a long-standing symptom of COVID. For example, if a patient had COPD that worsened before developing COVID-19, the algorithm could have removed episodes even if they were long-term indicators of COVID.
The decline in COVID-19 testing in recent years also makes it difficult to identify when a patient may have first contracted COVID-19.
The study was limited to patients in Massachusetts.
Future studies could explore the algorithm in cohorts of patients with specific diseases, such as COPD or diabetes. The researchers also plan to make this algorithm publicly available so that doctors and health systems around the world can use it with their patient populations.
In addition to opening the door to better clinical care, this work could lay the foundation for future research into the genetic and biochemical factors behind different subtypes of long COVID.
“Questions about the true burden of long COVID – questions that have so far remained elusive – now seem more within reach,” Estiri said.
Funding: Support was provided by the National Institutes of Health, National Institute of Allergy and Infectious Diseases (NIAID) R01AI165535, National Heart, Lung, and Blood Institute (NHLBI) OT2HL161847, and National Center for Advancing Translational Sciences (NCATS) UL1 TR003167, UL1 TR001881. , and U24TR004111.
J. Hügel’s work was partially funded by a grant from the IFI program of the German Academic Exchange Service (DAAD) and by the Federal Ministry of Education and Research (BMBF) as well as by the German Foundation for search (426671079).
About this AI and long COVID research news
Author: MGB Communications
Source: Harvard
Contact: MGB Communications – Harvard
Picture: Image is credited to Neuroscience News
Original research: Free access.
“Precision phenotyping for the selection of research cohorts of patients with post-acute sequelae of COVID-19 (PASC) as a diagnosis of exclusion» by Hossein Estiri et al. MedRxiv
Abstract
Precision phenotyping for the selection of research cohorts of patients with post-acute sequelae of COVID-19 (PASC) as a diagnosis of exclusion
Scalable identification of patients with post-acute sequelae of COVID-19 (PASC) is challenging due to lack of reproducible precision phenotyping algorithms and suboptimal accuracy, demographic bias, and underestimation PASC diagnosis code (ICD-10 U09.9). ).
In a retrospective case-control study, we developed a precision phenotyping algorithm to identify research cohorts of PASC patients, defined as a diagnosis of exclusion. We used longitudinal electronic health record (EHR) data from more than 295,000 patients from 14 hospitals and 20 community health centers in Massachusetts.
The algorithm uses an attention mechanism to exclude aftereffects that previous conditions can explain. We conducted independent chart reviews to refine and validate our precision phenotyping algorithm.
Our PASC phenotyping algorithm improves accuracy and prevalence estimation and reduces bias in identifying patients with long COVID compared to diagnosis code U09.9.
Our algorithm identified a PASC research cohort of more than 24,000 patients (compared to approximately 6,000 using diagnosis code U09.9), with an accuracy of 79.9% (compared to 77.8% using diagnosis code U09.9).
Our estimated PASC prevalence was 22.8 percent, which is close to national estimates for the region. We also provide an in-depth analysis describing clinical attributes, encompassing identified persistent effects by organ, comorbidity profiles, and temporal differences in PASC risk.
The PASC phenotyping method presented in this study provides superior accuracy, accurately measures PASC prevalence without underestimating it, and has less bias in identifying Long COVID patients.
The PASC cohort derived from our algorithm will serve as a springboard to delve deeper into the genetic, metabolomic, and clinical intricacies of Long COVID, overcoming the constraints of recent PASC cohort studies, which have been hampered by their limited size and available outcome data.