In a groundbreaking development at Mass General Brigham, researchers have created an innovative AI tool that meticulously analyzes electronic health records to help healthcare providers identify cases of long COVID—a puzzling condition that includes a range of persistent symptoms like fatigue, chronic cough, and brain fog following an infection with SARS-CoV-2. The findings, published in the journal Med, suggest that many more individuals than previously recognized might be suffering from this debilitating condition.

“Transforming Diagnosis: The Power of AI”

“Our AI tool has the potential to transform what has been a foggy diagnostic process into something clear and actionable, empowering clinicians to understand this complex condition better,” said Dr. Hossein Estiri, senior author and head of AI Research at the Center for AI and Biomedical Informatics at Mass General Brigham, alongside his position as an associate professor at Harvard Medical School. “With this advancement, we may finally grasp the true nature of long COVID and how best to treat it.”

Long COVID, also known as Post-Acute Sequelae of SARS-CoV-2 infection (PASC), manifests in a diverse array of symptoms. In their research, Estiri and his team defined long COVID as not just any lingering health issue, but as a specific diagnosis that must be clearly associated with COVID-19 and persisted for at least two months within a year following the infection.

“Precision Phenotyping: The Game-Changer”

The team harnessed data from nearly 300,000 de-identified patient records across 14 hospitals and 20 community health centers within the Mass General Brigham system. Rather than relying on standard diagnosis codes, the AI employs a novel approach termed "precision phenotyping," allowing it to delve into individual patient records to pinpoint symptoms and conditions connected to COVID-19. This capability extends to tracking symptoms over time to differentiate them from pre-existing medical conditions. For example, if a patient displayed shortness of breath, the AI could determine if it stemmed from chronic issues like heart failure or asthma instead of long COVID.

Dr. Alaleh Azhir, an internal medicine resident at Brigham Women’s Hospital and co-lead author of the study, emphasized the significance of this tool: “Physicians often grapple with complex symptom webs and extensive medical histories, balancing busy workloads. An AI-powered tool that can systematically disentangle this information could be revolutionary.”

“Unveiling the Hidden Epidemic”

The AI tool not only aims to enhance clinical outcomes but also seeks to mitigate biases that might skew current long COVID diagnostics. The study found that approximately 22.8% of patients could potentially have long COVID—far higher than the previously estimated 7%. This new, broader approach offers a more vivid representation of the long-term effects of the pandemic, aligning with alarming national trends.

The researchers claimed their tool improved diagnostic accuracy by 3% over traditional ICD-10 codes and was notably less biased. The identified cases reflected the demographic diversity of Massachusetts, advocating for the inclusion of underserved communities typically overlooked in clinical assessments.

“Addressing Limitations: A Path Forward”

Despite the promise of the AI tool, the study noted certain limitations. The health records utilized may not capture the full spectrum of long COVID symptoms. Additionally, prior worsening of existing conditions might not be reflected in the algorithm. The recent decline in COVID-19 testing also complicates identifying when patients first contracted the virus, influencing the study’s focus primarily on the Massachusetts population.

Future research intends to validate the algorithm within specific patient cohorts, such as those with COPD or diabetes. The researchers also plan to release the algorithm for public use, allowing healthcare providers globally to implement this tool in their practices.

Moreover, this research paves the way for further inquiry into the genetic and biochemical nuances of long COVID’s varying subtypes. “We are now closer than ever to answering critical questions regarding the extensive burden of long COVID, questions that have remained elusive until now,” concluded Estiri.

With this innovative tool on the horizon, the fight against long COVID may soon enter a pivotal new chapter!