Sbir phase II : a tool to automate a narrative patient summary of the medical chart for outpatient physicians

Federal Cooperative Agreement Summary

Abstractive Health, Inc. received a $1.25 million Small Business Innovation Research (SBIR) Phase II Cooperative Agreement from the National Science Foundation (NSF) Technology, Innovation, and Partnerships (TIP) program (CFDA 47.084), effective July 1, 2025, through June 30, 2027. The awardee will develop and refine a machine learning-enabled software tool that leverages natural language processing (NLP) to automatically generate clinically relevant narrative summaries of unstructured electronic health record (EHR) data for outpatient physicians. The tool is designed to integrate with leading EHR systems and national health information exchanges, providing physicians with concise, synthesized patient histories in real-time to enhance clinical decision-making and streamline workflows.

The project addresses the critical challenge of reducing physician burden associated with processing voluminous medical records, thereby mitigating risks of medical errors, missed follow-ups, and provider burnout. The technical approach incorporates state-of-the-art NLP models and federated learning structures to ensure data security compliance and adaptability across diverse healthcare environments. The anticipated deliverables include high-fidelity summary outputs, robust EHR system integration capabilities, and real-time operational functionality, positioning the solution as a scalable product for digital health companies and medical institutions with significant commercial potential for improving clinical efficiency and revenue reimbursement processes.

Abstract

The broader impact/commercial potential of this small business innovation research (sbir) phase II project is to revolutionize healthcare delivery by leveraging natural language processing to provide concise, clinically relevant summaries of patients? Medical records. By reducing the burden on physicians the tool could addresses pressing issues like medical errors, patient safety, and provider burnout. Commercially, the proposed approach could provide the potential to streamline clinical workflows, improve revenue reimbursement, and reduce administrative burdens for healthcare providers. Its integration with national health information exchanges and leading electronic health record systems positions IT as a pivotal tool for digital health companies and medical institutions, creating a scalable solution with broad market applicability. The proposed project addresses the critical challenge of reducing the time burden associated with processing unstructured electronic health records while ensuring the accuracy and comprehensiveness of patient care. Physicians often lack adequate tools to quickly synthesize patient histories, which can lead to missed follow-UPS, medical errors, and inefficiencies. The project aims to develop and refine a machine-learning-enabled tool to generate clinically relevant, narrative summaries of medical records, enhancing decision-making and streamlining clinical workflows. The proposed research focuses on natural language processing techniques to analyze broad and unstructured medical data. By integrating state-of-the-art models and a federated learning structure to address data-sharing constraints, the project aims to ensure adaptability across various healthcare environments. Anticipated technical results include high-fidelity summaries, robust integration with electronic health record systems, and real-time capabilities for physicians to access and query patient records. The research scope includes fine-tuning large language models, implementing speech-to-text integrations, and developing retrieval-augmented generation systems for personalized physician queries. Methods involve annotating diverse datasets, employing advanced evaluation metrics, and rigorous testing with medical professionals. This project is expected to produce a scalable, clinically validated tool to potentially enhance physician efficiency, reduce medical errors, and ultimately improve patient outcomes. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the foundation's intellectual merit and broader impacts review criteria.- subawards are not planned for this award.