Posted in | News | Medical Robotics

AI Tool for Long COVID Detection Reduces Bias and Expands Patient Reach

Researchers from Mass General Brigham have created an AI-driven tool to analyze electronic health records, assisting clinicians in spotting cases of long COVID—a complex condition marked by lasting symptoms like fatigue, persistent cough, and cognitive issues following SARS-CoV-2 infection.

AI Tool for Long COVID Detection Reduces Bias and Expands Patient Reach

Image Credit: Dmitry Demidovich/Shutterstock.com

The findings could help identify more individuals who may need treatment for this potentially debilitating condition. The number of cases revealed by the tool also indicates that long COVID may be significantly underdiagnosed.

Our AI tool could turn a foggy diagnostic process into something sharp and focused, giving clinicians the power to make sense of a challenging condition. With this work, we may finally be able to see long COVID for what it truly is—and more importantly, how to treat it.

Hossein Estiri, PhD, Head, AI Research, Center for AI and Biomedical Informatics of the Learning Healthcare System (CAIBILS), Mass General Brigham

Long COVID, or Post-Acute Sequelae of SARS-CoV-2 infection (PASC), can manifest in a wide array of symptoms that linger well beyond the initial illness.

For their study, Estiri and the research team defined long COVID as a diagnosis that could not be easily explained by other factors in a patient’s medical history and had to be clearly linked to a prior COVID-19 infection. To qualify, symptoms needed to persist for at least two months within a year of the initial infection.

To develop this AI tool, the team analyzed anonymized records from nearly 300,000 patients across Mass General Brigham’s network of hospitals and health centers. Rather than relying on a single diagnostic code, the AI uses an innovative approach called “precision phenotyping.”

This approach allows the tool to dig deep into individual patient records, identifying specific symptoms and tracking how they evolve over time. For instance, the tool can recognize if shortness of breath is due to an underlying condition like asthma or heart failure rather than long COVID. Only when no other plausible cause remains does the tool flag a patient as likely having long COVID.

Physicians are often faced with having to wade through a tangled web of symptoms and medical histories, unsure of which threads to pull, while balancing busy caseloads. Having a tool powered by AI that can methodically do it for them could be a game-changer.

Alaleh Azhir, MD, Study Co-Lead Author and Internal Medicine Resident, Brigham Women’s Hospital

According to researchers, this patient-centered approach may help address biases inherent in current long COVID diagnostics.

They noted that patients identified using the official ICD-10 diagnostic code often tend to be those with easier access to healthcare. While prior studies have estimated that around 7 % of the population suffers from long COVID, this new method suggests a significantly higher figure—22.8 %. The researchers believe this more accurately reflects national patterns, offering a clearer view of the pandemic’s long-term impact.

The team found their tool to be about 3 % more accurate than relying solely on ICD-10 codes and notably less biased. Unlike conventional diagnostics that often lean toward certain groups (such as those with better healthcare access), the individuals flagged by this tool represent the demographic diversity of Massachusetts more closely. This more inclusive approach helps to prevent any bias in the results that can arise from using single diagnostic codes or isolated clinical encounters.

Estiri added, “This broader scope ensures that marginalized communities, often sidelined in clinical studies, are no longer invisible.

The study and AI tool have a few limitations. For instance, the health record data used in the algorithm may not be as thorough as the notes physicians add after patient visits, which could capture additional nuances of long COVID symptoms.

Another challenge is that the algorithm does not account for the worsening of pre-existing conditions, which might indicate long COVID. For example, a patient with COPD who experienced increased symptoms after COVID-19 could have been excluded by the tool if those symptoms predated the infection.

Additionally, the decline in COVID-19 testing over recent years complicates identifying when a patient first contracted COVID-19. The study’s findings are also limited geographically, focusing solely on patients in Massachusetts.

Looking forward, the researchers plan to test the algorithm on patients with specific conditions, like COPD or diabetes. They also intend to make the algorithm publicly available, allowing physicians and healthcare systems worldwide to assess long COVID within their patient populations.

Beyond immediate clinical applications, this work may also provide a basis for research into genetic and biochemical factors underlying the various subtypes of long COVID, potentially unlocking new insights into this complex condition.

Estiri concluded, “Questions about the true burden of long COVID—questions that have thus far remained elusive—now seem more within reach.

Journal Reference:

Azhir, A. et. al. (2024) Precision phenotyping for curating research cohorts of patients with unexplained post-acute sequelae of COVID-19. Med. doi.org/10.1016/j.medj.2024.10.009

Tell Us What You Think

Do you have a review, update or anything you would like to add to this news story?

Leave your feedback
Your comment type
Submit

While we only use edited and approved content for Azthena answers, it may on occasions provide incorrect responses. Please confirm any data provided with the related suppliers or authors. We do not provide medical advice, if you search for medical information you must always consult a medical professional before acting on any information provided.

Your questions, but not your email details will be shared with OpenAI and retained for 30 days in accordance with their privacy principles.

Please do not ask questions that use sensitive or confidential information.

Read the full Terms & Conditions.