TriNetX Introduces Human-Validated Fact Extraction for Enhanced Clinical Research

TriNetX's Innovative Approach to Clinical Data Extraction

In a significant advancement for clinical research, TriNetX® has introduced a method of extracting facts from clinical notes that is both human-validated and grounded in artificial intelligence (AI). This innovative approach aims to ensure the reliability of extracted data, which is crucial for researchers relying on accurate clinical information.

For researchers, the reliability of data is paramount. With a large volume of clinical notes available, reaching a billion by the end of 2026, the challenge has been to extract meaningful insights from this wealth of information. TriNetX has responded to this challenge by leveraging AI, particularly Natural Language Processing (NLP), while maintaining the involvement of clinical experts to validate the information. This dual approach allows for the extraction of details that might be challenging to obtain otherwise, such as disease stage, overall health status, and biomarker results.

The new capabilities available on the TriNetX LIVE™ platform not only provide researchers with a vast database for analysis but also incorporate structured data from Electronic Medical Records (EMRs) of the same anonymized patients. This creates an enriched dataset that can enhance cohort studies and facilitate the development of clinically relevant phenotypes.

TriNetX has put in place a rigorous framework with three critical guarantees for the facts derived from clinical notes:
1. Human Oversight: Clinical experts evaluate and correct the labels assigned by AI, ensuring that the models are continuously improved based on these corrections. Each module must achieve a precision of at least 0.90 when compared to human-annotated notes.
2. Clinical Confirmation: The extracted facts undergo verification against expected clinical patterns, not just through model indicators. For instance, survival curves based on Eastern Cooperative Oncology Group (ECOG) performance scores have shown that patients with a score of 2 or higher have a lower survival rate.
3. Continuous Monitoring: TriNetX actively monitors the performance of its models as new healthcare organizations join the initiative, enhancing the models to adapt to emerging trends and variations.

The initial results from this groundbreaking method have been promising. For example, the extraction of smoking status from clinical notes has improved from 22% to an impressive 67% coverage, allowing the inclusion of six times more patients than structured data alone in a study focused on lung cancer. Furthermore, this approach has provided ECOG and Karnofsky performance indices for 292,000 oncology patients, information that was not previously captured in structured fields.

As August Calhoun, CEO of TriNetX, stated, “Pointing AI at a graph is easy. What sets TriNetX apart is its rigorous scientific methodology and human expertise within a globally federated network where data remains with the originating entities.” This emphasizes the critical role that validated clinical notes play in supporting robust evidence in research as AI integrates into daily discovery processes.

TriNetX's capabilities currently encompass social histories (such as smoking and alcohol consumption), cardiology, pulmonology, general oncology, biomarkers, liver diseases, and more, all now accessible to its network members. Further information can be found at trinetx.com/data/clinical-notes.

In summary, TriNetX is pioneering a new era of clinical data utilization, merging the strengths of AI and human validation to enhance the reliability of clinical research findings. As the field of health data continues to evolve, such innovations are essential for improving health outcomes through accurate and comprehensive data insights.

About TriNetX

TriNetX, The Global Truth Engine for Better Human Health®, operates on a secure and interconnected network of over 14,200 clinical sites in more than 20 countries. By combining global data integrity with a unique mixture of AI and human intelligence, TriNetX accelerates discovery and contributes to improved health outcomes globally. For more on TriNetX, visit trinetx.com and follow them on LinkedIn for updates.