Safe and Explainable AI-Enabled Decision Making for Personalized Clinical Decision Support

What this Trial Is About

Researchers are evaluating artificial intelligence (AI) technology to improve clinical decision support across several medical areas including cardiology, breast cancer, and sepsis. The study aims to address challenges in medical AI such as integrating medical knowledge, making AI recommendations clear to clinicians, and ensuring safety. This observational research is sponsored by Abramson Cancer Center at Penn Medicine and focuses on personalized AI-enabled tools for patient care. The study involves developing and testing machine learning (ML) models that use real-time clinical, demographic, imaging, and molecular data. In cardiology, the AI model predicts impending cardiac arrest and identifies potentially reversible causes using data from hospitalized ICU patients. For breast cancer outpatients, the model predicts short-term mortality and symptom decline, comparing results to established prognostic tools. In sepsis, the model predicts the need for broad-spectrum antimicrobial therapy using structured and unstructured electronic health record data, aiming to support timely diagnosis and treatment. Participants include adults 18 years and older admitted to Penn Medicine hospitals with specific conditions: cardiac patients admitted since 2017, breast cancer patients registered in the Penn Cancer registry, and sepsis patients presenting or admitted since mid-2017. The study monitors the development and benchmarking of advanced learning and explanation algorithms over periods up to 36 months. Researchers will evaluate AI model performance and safety guarantees without administering treatments, with data collected from patient records and monitoring devices throughout the study period ending in November 2028.

Safe and Explainable AI