Blogs

Clinical trial matching requires research teams and clinicians to translate complex eligibility criteria into actionable screening decisions across large volumes of fragmented patient and trial data. As AI gains momentum in its potential to accelerate and standardize this process, CHAI dedicated a work group to focus on the use case. Over many collaborative meetings, the work group discussed how successful implementation depends on trustworthy data, transparent reasoning, rigorous governance and evaluation, and workflows that recognize the realities of clinical research.

Today, CHAI is profiling a Best Practice Guide and Testing & Evaluation Framework for AI-supported clinical trial matching solutions, the fourth in a recently announced series of outputs from its Q1-Q2 collaborative work groups.

Over the past several months, clinical researchers, health system leaders, AI developers, informaticists, patient advocates and implementers collaborated to translate real-world experience with AI-supported trial matching into practical, vendor-agnostic guidance. The resources focus specifically on solutions that extract and structure eligibility criteria from clinical trial protocols, interpret structured and unstructured patient data from electronic health records, and identify potential patient-trial matches with transparent, verifiable rationale.

New Clinical Trials Resources

• Best Practice Guide - Clinical Trials (v1.0): A practical, vendor-agnostic guide that equips developers and implementers with consensus-defined best practices for building, deploying and governing AI-supported clinical trial matching solutions. The guide addresses responsible AI considerations across usefulness, usability and efficacy; fairness and bias management; safety and reliability; transparency; and privacy and security.

• Testing & Evaluation (T&E) Framework - Clinical Trials (v1.0): A living framework, hosted publicly on GitHub, that provides literature-backed methods and metrics for evaluating the real-world performance of AI-supported protocol extraction and patient-trial matching tools. The framework includes approaches for measuring matching accuracy, enrollment impact, screening efficiency, explanation quality, fairness and subgroup performance, among other dimensions. Organizations are encouraged to adapt the framework to their own patient populations, workflows and deployment environments as the technology and evidence base evolve.

Work group leads from: Montana State University,, Biotale, Triomics, UT Southwestern, Prompt Opinion, Columbia University

Clinical trial recruitment has long been constrained by fragmented information, labor-intensive screening, and the difficulty of translating narrative eligibility criteria into data that can be reliably compared with a patient's record. While AI can help reduce the time and cost of that process, work group members emphasized that matching is only as reliable as the trial criteria and patient data available to the system. Public registries may be incomplete or outdated, full protocols are not always available, and key patient information often remains buried in unstructured notes or spread across disconnected systems.

The work group's discussions repeatedly returned to the same difficult questions at the center of this dichotomy:

• How can AI aggregate and reconcile trial information when no single source is complete, consistently current or always available?

• How should developers interpret nuanced eligibility criteria when important protocol details or patient data are missing, unstructured or stale?

• How can organizations protect highly sensitive patient data when trial matching tools may ingest demographics, labs, pathology, genomic information, imaging and clinical notes?

• What level of evidence, explanation and uncertainty should accompany each match or exclusion so that clinical teams can independently verify the result?

• How should organizations distinguish fairness in identifying eligible patients from fairness in who is ultimately able to enroll and remain in a trial?

• How can AI-supported matching account for practical realities such as travel, cost, caregiving needs and functional status without using those factors to restrict access?

Read the full guide and testing and evaluation framework here to learn more about the product of CHAI’s collaborative work group conversations specific to clinical trial capabilities.

Hear from our work group participants:

“My background is in human research protections, and I appreciated the chance to contribute that perspective across multiple CHAI work groups, including Clinical Trials, Agentic AI, and Ambient AI,” said Challace Pahlevan-Ibrekic, MBE, CPXP, CIP, Director, Regulatory Affairs of Research Intelligence & Institute of Health System Science at Northwell Health. “The problems we tackled were different, but the thread that linked them all was the same: people genuinely dedicated to setting standards for responsible AI in healthcare. The Best Practice Guides and Testing & Evaluation Frameworks coming out of these work groups will be practical resources for healthcare systems considering these technologies.”

“Clinical trials represent more than scientific discovery. They are a promise of hope for the individuals, families, and communities who make innovation possible. Every participant helps illuminate a path toward better treatments and healthier futures,” said Elizabeth Johnson, PhD, MS-CRM, RN Assistant Professor, Mark & Robyn Jones College of Nursing at Montana State University. “The Coalition for Health AI Clinical Trials Working Group demonstrates the power of multidisciplinary collaboration in developing practical guidance for trustworthy AI across the clinical research ecosystem. From remote patient monitoring and decentralized trials to adaptive study designs and data-informed decision making, the Best Practice Guide and Testing & Evaluation Framework support responsible AI that expands access to innovative therapies for rural, frontier, and historically underrepresented populations. These participants are my neighbors. They are the lighthouses that guide us toward new possibilities, reminding us that the future of medicine is built on trust, collaboration, and the courage of those who choose to participate in research. Together, we can ensure AI strengthens clinical trials while bringing hope and the promise of new cures to communities everywhere.”

These resources reflect CHAI's broader mission to convene the healthcare community around practical solutions to shared challenges. Through collaborative work groups, clinicians, health systems, technology developers, researchers, patient advocates, policymakers and other stakeholders work together to develop consensus-driven guidance that helps organizations deploy AI responsibly and with confidence. CHAI looks forward to seeing these resources adopted, refined and expanded by the broader community as AI-supported clinical trial matching continues to evolve.

Thank you to our members who made this work possible:

• Archana Sah BPharm., MS ( Pharm.), PMP, AS Pharma Advisors

• Navin Maganti, Biotale

• Pawan Jindal, MBBS, MS, Darena Health

• Aamani Budhota, Docomed

• Dr. Parineeta Jaiswal, MBBS, MMCi, MA, Duke University

• Anthony Solomonides, PhD FAMIA FACMI, Endeavor Health

• Ashley M Hopkins, PhD, Flinders University

• Joe Derenzo, PMP, Healthcare Performance Group Inc.

• Ben May, Individual Contributor

• Simran Tiwari, Individual Contributor

• Elizabeth Johnson, PhD, MS-CRM, RN, Montana State University

• Maijd Afshar, MD, MS, University of Wisconsin - Madison