A groundbreaking artificial intelligence algorithm developed by researchers at Mass General Brigham has demonstrated remarkable accuracy in identifying undiagnosed cases of long COVID through electronic health records (EHRs), achieving a precision rate of 79.9%.
The innovative tool, detailed in a recent publication in the journal Med, leverages precision phenotyping to analyze patient records and track COVID-19-related symptoms over time. Trained on de-identified data from nearly 300,000 patients across 14 hospitals and 20 community health centers, the algorithm distinguishes long COVID manifestations from similar conditions such as asthma and heart failure.
Advanced Detection Capabilities
The AI system's sophisticated approach surpasses current diagnostic methods based on International Classification of Diseases (ICD-10) coding by approximately 3%. More significantly, it demonstrates reduced bias compared to traditional diagnostic approaches, potentially addressing healthcare inequities in long COVID diagnosis and treatment.
"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," explains Dr. Hossein Estiri, senior author and head of AI research at the Center for AI and Biomedical Informatics of the Learning Healthcare System at Mass General Brigham.
Prevalence and Impact
The research reveals concerning statistics about long COVID's prevalence. While the CDC estimates that 7.5% of US adults (approximately 24.75 million individuals) experience long COVID symptoms, the new algorithm suggests the actual rate could be significantly higher at 22.8%. This figure aligns more closely with the National Center for Health Statistics' estimate of 24% for Massachusetts.
Long COVID manifests through various symptoms, including:
- Extreme fatigue
- Shortness of breath
- Chest pain
- Memory problems
- Sleep disturbances
- Heart palpitations
- Dizziness
Future Applications and Accessibility
The research team plans to make their algorithm open-access, enabling widespread deployment across healthcare systems. Future studies will explore the tool's application in specific patient populations, such as those with chronic obstructive pulmonary disease (COPD) or diabetes.
"With this work, we may finally be able to see long COVID for what it truly is – and more importantly, how to treat it," Dr. Estiri notes, emphasizing the algorithm's potential to enhance understanding and treatment of this complex condition.
The tool's ability to identify patients who might otherwise go undiagnosed, particularly in marginalized communities, represents a significant advancement in addressing healthcare disparities. This broader scope ensures more comprehensive detection and treatment opportunities for all affected populations.
