RESEARCHERS have developed a precision phenotyping algorithm that significantly advances the identification of long COVID, or post-acute sequelae of COVID-19 (PASC), in patient records. This innovative approach, led by Alaleh Azhir, Massachusetts General Hospital, and Jonas Hügel, Brigham and Women’s Hospital, Boston, MA, was designed to optimize research on the long-term effects of COVID-19, was validated using electronic health records from a large Massachusetts healthcare system and shows promise for identifying diverse PASC patient cohorts with greater accuracy.
A key feature of this algorithm is its “attention mechanism,” which enables it to differentiate between chronic conditions resulting from COVID-19 and pre-existing conditions, enhancing accuracy and inclusivity. According to the study, which analyzed over 295,000 patient records from multiple hospitals and community health centers, the algorithm identified a PASC cohort with 79.9% precision, outperforming the commonly used ICD-10-CM code U09.9 in both accuracy and inclusivity.
The algorithm’s focus on precision offers substantial benefits: it reduced demographic biases and accurately identified rare, long-term COVID-19 symptoms, such as vision loss and new diabetic complications. The study found that an estimated 22.8% of patients studied met the criteria for long COVID, closely aligning with regional estimates. This refined approach could help construct high-quality PASC research cohorts, which will aid in exploring genetic, metabolomic, and clinical aspects of long COVID.
Lead researcher Azhir and team suggested that these cohorts will act as a springboard to uncover the full impact of COVID-19’s lingering effects, supporting healthcare systems in more targeted and effective patient management.
Reference: Azhir A et al. Precision phenotyping for curating research cohorts of patients with unexplained post-acute sequelae of COVID-19. Med. 2024.
Anaya Malik | 2024
