AI-based Clinical Decision Support Systems
Our goal is to identify the safest and most effective ways to integrate these models into existing clinical processes, thereby minimizing administrative burdens and enabling more focused patient care.
Overview
Building on our foundational work in LLM safety and benchmarking, we are spearheading the development of AI-driven Clinical Decision Support Systems (CDSSs) designed to ease the workload on physicians and improve patient care. Our goal is to identify the safest and most effective ways to integrate these models into existing clinical processes, thereby minimizing administrative burdens and enabling more focused patient care. Recognizing the strict limitations imposed by HIPAA and related privacy regulations, our team relies on a combination of anonymized and synthetic datasets for initial testing and validation. This approach allows us to thoroughly evaluate system efficacy and security without risking patient confidentiality.
Focus
A major focus of our research involves creating and refining diverse clinical scenarios—from intraoperative guidance to real-time problem resolution and evidence-based treatment planning—that challenge our models to handle incomplete or ambiguous patient data. By ensuring that these systems refrain from offering definitive answers when pertinent information is unavailable, we reduce the risk of misdiagnosis or improper treatment recommendations. This careful handling of uncertainty, coupled with robust explainability features, is key to fostering the trust of medical professionals and encouraging widespread model adoption.
Testing
To gauge the real-world utility of our AI-CDSS, we conduct extensive testing of workflow integration, evaluating how effectively these models deliver timely, context-relevant information at the point of care. Our performance metrics extend beyond simple accuracy or precision: we collect both quantitative and qualitative data, including expert evaluations, user satisfaction surveys, and direct clinician feedback. Through this rigorous, evidence-based approach, we strive to establish a new standard for CDSS design and implementation—one that not only accelerates clinical decision-making but also strengthens physician-patient relationships, optimizes resource allocation, and upholds the highest ethical and professional standards in modern medicine.
