This AI tool detects more cases of long Covid

Artificial intelligence (AI) has been quickly adopted in healthcare as a tool to help diagnose patients. Some programs work to determine the type of care patients need or inform patients of their risk for Alzheimer’s disease.

Now, a new tool out of Somerville, Massachusetts-based Mass General Brigham (MGB), a health system that includes Massachusetts General Hospital and Brigham and Women’s Hospital, is using AI and analyzing de-identified electronic medical records to help clinicians determine if patients have or have ever had long Covid, the system announced in a release on November 8. The results of the study were published in health journal Med last week.

Long Covid occurs in some patients who were previously diagnosed with Covid-19, and can cause symptoms like fatigue, chest pain, digestive issues, brain fog, and headaches, according to the CDC. As of October 26, the World Health Organization estimates that there have been a total of 777+ million Covid cases throughout the world, 103+ million of which were diagnosed in the US.

Diagnosing the condition can be difficult, as there is neither a lab test that detects long Covid nor a set of specific symptoms. This has led experts, including researchers at MGB, to believe that long Covid is underdiagnosed and therefore undertreated.

The researchers developed a method called precision phenotyping, and used data from the National Center for Health Statistics from June 2022 to October 2023 to test their algorithm and its 79.9% accuracy rate.

How it works. MGB investigators wrote in the study that they used the AI tool to sift through clinical records of 295,000 patients at 14 hospitals and 20 community health centers in the system. The team looked for symptoms and diagnoses related to Covid and monitored if the symptoms persisted for two months or longer within the year that the patient was diagnosed, while also ruling out links to another medical condition.

In the study, the tool determined, for example, whether a patient’s chest pain or shortness of breath was related to another condition like asthma. If there weren’t other recorded underlying conditions, the patient was deemed a long Covid candidate.

Among those identified as having long Covid, the study found that 71.4% were white, 10.4% were Black, 6.6% were Hispanic, and 64.5% were female.

Fewer patients have tested for Covid-19 recently compared to when the virus first emerged, and along with this, other factors affecting the completeness of medical records made it difficult to pinpoint when patients contracted the disease.

Going forward, the researchers plan to release the tool publicly so other clinicians can use it with their patients’ medical data, according to the press release.

“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,” senior author Hossein Estiri said in the release. “With this work, we may finally be able to see long Covid for what it truly is—and more importantly, how to treat it.”