Advancing the development, dissemination, and sustainable implementation of evidence-based screening and treatment with trustworthy genai and responsible AI workforce

The National Center for Advancing Translational Sciences at the National Institutes of Health awarded a $2.136 million Cooperative Agreement to Weill Medical College of Cornell University on September 1, 2026, under the NCATS program (CFDA 93.350).

The award funds development of a trustworthy, scalable framework for automatically translating clinical practice guidelines (CPGs) from free-text format into computer-interpretable guidelines (CIGs) using generative AI, and subsequent implementation and validation of those CIGs across clinical sites. The project focuses on three clinical domains: headache disorders, acute stroke, and postpartum depression. The recipient will develop a dynamic framework to translate CPGs to CIGs using genAI, establish automated verification systems for factual completeness and correctness in translations, and implement the resulting CIGs across Clinical and Translational Science Award (CTSA) sites to analyze scalability, portability, and clinical practice adherence. The work leverages electronic health record systems at Cornell, Columbia, University of Pittsburgh, UT Southwestern, and Virginia Commonwealth University to examine guideline implementation and identify opportunities to fast-track promising clinical and translational science.

Place of performance is New York, New York. The period of performance runs from September 1, 2026, through August 31, 2028. The assistance type is a Cooperative Agreement, which funds the development phase (UG3) and implementation phase (UH3) of this translational science initiative.

Abstract

Clinical practice guidelines (cpgs) are often written in free-text format, making them challenging to be machine-readable and difficult to implement, standardize, and evaluate within clinical systems. This also hampers their ability to make timely revisions in response to the ever-growing knowledge and application of personalized medicine. The rapid advancements in generative artificial/augmented intelligence (genai) hold immense potential to automatically translate text-based CPG into standardized computer-interpretable guidelines (cigs), which can be shared to establish multi-site patient cohorts, validate the generalizability and scalability of the CPG, and identify and fast-track promising clinical and translational science (CTS). This study aims to develop a trustworthy, scalable, portable, and ethical framework for automatically translating cpgs into cigs (UG3) and implementing cpgs (UH3) with the assistance of genai. While the methodologies are applicable across various domains, we will focus on screening and treatment for (1) headache disorders, (2) acute stroke, and (3) postpartum depression. We will then implement the cigs across ctsa sites to analyze their scalability and portability, as well as examine adherence in clinical practice to cpgs. These analyses will help guide the future directions of CTS. We have assembled a team with complementary skills and expertise in both technical, clinical, and ethical aspects, massive EHR systems (cornell, columbia, upitts, UT southwestern, and virginia commonwealth university), and abundant resources to tackle this problem for medical genai. Our specific aims are: (1) develop a dynamic framework to translate cpgs to cigs using genai, (2) automate verification of factual completeness and correctness in CPG-CIG translation, (3) implement cpgs and validate their portability and adherence, (4) continuously improve, scale, and sustain CPG implementation through the responsible AI strategy and evaluation (raise) framework, and (5) reliably supervise genai-augmented CPG with a human workforce. This project is a collaborative effort that combines our expertise and resources from individual sites into a collective program. We aim to translate and implement cpgs, ensuring that the potential benefits of genai are realized to influence the development of future cpgs to accommodate translation while minimizing possible risks. This project also serves as the starting point for developing the genai-based decision support systems and the workforce to manage them. More importantly, IT engages multiple levels of stakeholders to influence genai implementation and training efforts in a longitudinal and contextual manner, fostering a shared responsibility and commitment to the project's success.