Study Shows LLMs Can Extract PD-L1 Biomarker Details from EHRs

A recent study published in the peer-reviewed journal AI in Precision Oncology announced in a press release highlights the capability of large language models (LLMs) to extract PD-L1 biomarker details from electronic health records (EHRs). This research, conducted by Aaron Cohen, MD, and colleagues from Flatiron Health and NYU Langone School of Medicine, utilized open-source LLMs to extract seven specific biomarker details related to PD-L1 testing from a nationwide EHR-derived database.

PD-L1 biomarker testing is crucial for guiding cancer treatment decisions, yet accessing these results can be challenging due to the unstructured nature of laboratory reports. The study found that LLMs, when fine-tuned with high-quality labeled data, could accurately extract complex PD-L1 test details despite variations in cancer type, documentation, and time.

This research underscores the potential of AI in improving healthcare by efficiently processing vast amounts of medical data, thereby enhancing patient care and saving valuable time for healthcare providers.