---

Researchers at Mass General Brigham have developed an artificial intelligence (AI) tool designed to identify long COVID cases by analyzing electronic health records (EHRs). This innovative approach, termed “precision phenotyping,” examines individual patient histories to detect symptoms associated with long COVID, distinguishing them from other conditions.

The study, published in the journal Med, suggests that approximately 22.8% of the population may experience long COVID symptoms, a figure significantly higher than previous estimates of around 7%. 

Key Insights:

• Enhanced Detection: The AI tool analyzes EHRs to identify patterns indicative of long COVID, including symptoms like fatigue, chronic cough, and cognitive difficulties. By evaluating symptom progression over time, it differentiates long COVID from other illnesses. 
• Increased Prevalence Estimates: Utilizing this AI-driven method, researchers estimate that nearly 23% of individuals may suffer from long COVID, highlighting a potentially underrecognized public health issue.
• Personalized Care Potential: By providing a more accurate diagnosis, the tool could facilitate tailored treatment plans, addressing the specific needs of long COVID patients. 

Study Overview:

The AI algorithm was developed using de-identified data from nearly 300,000 patients across 14 hospitals and 20 community health centers within the Mass General Brigham system. Unlike traditional methods that rely on specific diagnostic codes, this tool employs precision phenotyping to sift through individual records, identifying symptoms and conditions linked to COVID-19. It tracks symptom progression over time to differentiate long COVID from other illnesses. 

“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,” said Alaleh Azhir, MD , internal medicine resident at Brigham Women’s Hospital.

The AI tool demonstrated approximately 3% greater accuracy than existing diagnostic codes and reduced biases associated with healthcare access disparities. The study’s findings suggest that the prevalence of long COVID may be significantly underrecognized, underscoring the need for improved diagnostic tools and patient care strategies. 

Prevalence Rates: Illustrating long COVID prevalence rates detected by traditional methods (7%) compared to the AI tool (22.8%). This highlights the AI tool’s ability to uncover more cases, offering a more realistic picture of long COVID’s impact.

Long COVID, referred to as Post-Acute Sequelae of SARS-CoV-2 infection (PASC), encompasses a diverse array of symptoms. In their study, Estiri and colleagues characterized it as a diagnosis of exclusion linked to a prior COVID-19 infection. This means that the symptoms could not be attributed to any other underlying condition in the patient’s medical history and had to be associated with a confirmed COVID-19 infection. Furthermore, the symptoms needed to persist for at least two months within a 12-month follow-up period to meet the study’s criteria.

According to the researchers, this new patient-centered diagnostic approach has the potential to reduce biases inherent in current methods for identifying long COVID. They highlight that patients diagnosed using the official ICD-10 code for long COVID are often those with better access to healthcare services.

While prior studies have estimated that around 7% of the population experiences long COVID, this innovative method suggests a significantly higher prevalence of 22.8%. The authors note that this estimate more accurately reflects national trends, offering a clearer and more comprehensive understanding of the pandemic’s long-term impact.

Detection Accuracy: A bar graph comparing the detection accuracy of traditional diagnostic methods (ICD-10 codes) with the AI tool using precision phenotyping. The AI tool shows a slight improvement (78% vs. 75%).

The researchers found that their tool demonstrated approximately 3% greater accuracy compared to the use of ICD-10 codes, with significantly less bias. Their study showed that individuals identified as having long COVID through this method reflected the diverse demographic composition of Massachusetts.

In contrast, traditional algorithms relying on single diagnostic codes or isolated clinical encounters often produce results skewed toward populations with better access to healthcare. As Estiri explained, “This broader scope ensures that marginalized communities, often overlooked in clinical research, are no longer invisible.”

Implications for Healthcare:

AI is emerging as a powerful tool in understanding and managing long COVID, particularly in its ability to predict subtypes by analyzing extensive datasets of patient health records, symptoms, and biomarkers. Using machine learning techniques such as clustering, natural language processing, and deep learning, AI identifies patterns and categorizes patients into subgroups based on their symptoms, disease progression, and response to treatment. This innovative approach offers the potential to revolutionize long COVID care by enabling precision medicine and targeted interventions.

One of AI’s key contributions is its ability to identify symptom-based clusters, or subtypes, of long COVID. By analyzing electronic health records (EHRs), AI can group patients experiencing similar symptoms. For example, some patients may primarily suffer from fatigue, while others experience respiratory issues such as chronic cough or shortness of breath.

Still others may struggle with cognitive symptoms, including brain fog and memory problems. By uncovering these distinct subtypes, researchers and clinicians can better understand how long COVID manifests and develop tailored treatment strategies.

AI also excels in leveraging biomarker data to enhance precision in diagnosing and managing long COVID. Machine learning models trained on genetic, biochemical, and immunological data can identify unique biomarkers associated with specific subtypes.

For instance, persistent inflammation biomarkers might indicate immune dysregulation, while cardiovascular or neurological biomarkers could reveal heart- or brain-related complications. This capability allows treatments to target the root causes of symptoms, improving both accuracy and effectiveness in care.

Another critical application of AI lies in tracking the progression of long COVID symptoms over time. By analyzing time-series data, AI can predict whether certain subtypes are prone to prolonged recovery or recurrent flare-ups. Data from wearable devices, such as heart rate, oxygen levels, and activity patterns, enables AI to assess which patients are at risk for severe or chronic symptoms. These insights support proactive management and early interventions, potentially improving long-term outcomes for patients.

With the identification of subtypes, AI facilitates the development of personalized care plans. For instance, patients with respiratory-dominant long COVID might benefit from pulmonary rehabilitation programs, while those experiencing cognitive symptoms could require neurocognitive therapy or medications targeting brain inflammation. By tailoring treatment plans to each subtype, AI ensures that interventions are both specific and effective, minimizing the one-size-fits-all approach often seen in long COVID care.

Despite these advancements, challenges remain. The effectiveness of AI depends on access to high-quality, diverse datasets, which are essential to ensure that underrepresented populations are not overlooked. Furthermore, standardization is needed in defining long COVID subtypes and developing corresponding treatment guidelines. Finally, integrating AI-driven insights into clinical workflows is critical to translating research findings into practical applications that benefit patients 

As the understanding of long COVID evolves, tools like this AI algorithm are crucial for accurately identifying and managing the condition, ultimately improving patient care and informing public health strategies.

---

Are you interested in how AI is changing healthcare? Subscribe to our newsletter, “PulsePoint,” for updates, insights, and trends on AI innovations in healthcare.