The Pulse
TriNetX says human-validated clinical-note facts are now available to researchers
TriNetX says its clinical-note facts are available to network members through the TriNetX LIVE platform and downloadable datasets. The company reports that its checks include human review, confidence scores and comparisons with expected cli
AI.info Team · Written with AI · fact-checked ·
Clinical-note facts added to TriNetX tools
TriNetX announced that NLP-extracted facts from clinical notes are now available in its TriNetX LIVE platform and downloadable datasets. The company says the data are drawn from more than 800 million clinical notes and reports, a collection it expects to reach about 1 billion by the end of 2026.
TriNetX says clinical notes contain details such as disease stage, performance status and biomarker results that complement structured records, but are hard to extract reliably. The company says it uses AI alongside clinical and informatics expertise, with pipelines trained on a large corpus of annotated reference notes. Notes-derived data are linked to structured electronic health record data for the same de-identified patients, according to the release.
Validation and monitoring
The release describes three safeguards: human oversight, clinical corroboration and continuous monitoring. Clinical experts review and correct the information tagged by AI, and the models are retrained using those corrections. TriNetX says each module must reach a precision of 0.90 or higher against human-annotated notes. The company also says every pipeline is tested against those notes and each extracted fact receives a confidence score.
For clinical corroboration, TriNetX says extracted facts are tested against expected clinical patterns, not only model metrics. The company cites survival curves based on extracted Eastern Cooperative Oncology Group (ECOG) performance status, which showed worse survival for patients scoring 2 and higher. TriNetX says it monitors model performance as new healthcare organizations join the clinical-notes program and new types of notes arrive.
Reported results and coverage
TriNetX reports that notes-derived facts increased smoking-status coverage from 22% to 67%. In a lung cancer cohort study, it says the facts produced six times more patients than structured data alone. The release also says the data surfaced ECOG and Karnofsky performance status for 292,000 oncology patients for whom those details were absent from structured fields.
The company says current coverage includes social history, such as smoking and alcohol status, cardiology, pulmonology, general oncology and biomarkers, and liver disease, among other areas. The data are available to TriNetX network members.
“Pointing AI at a chart is easy. What sets TriNetX apart is how we do it: with scientific discipline and human expertise across a global federated network where data stay within the institutions that generated them,” said August Calhoun, CEO of TriNetX. Calhoun added, “That means clinical notes can inform evidence that stands up to scrutiny as AI becomes part of everyday research.”