Rethinking the Algorithm: AI and the Human Experience – Ashby Dialogues Speaker Matthew Engelhard of Duke University

Toward Interpretable, Timely, Context-Aware Clinical Decision Support

Rethinking the Algorithm: AI and the Human Experience, part of the 2025-2026 Ashby Dialogues Speakers’ Series

---

AI can monitor electronic health records for early signs of chronic conditions, delivering timely and context-aware clinical recommendations to help improve patient outcomes.

Matthew Englehard, M.D., Ph.D. of Duke University will present new methods to predict chronic disease risks and deliver clear, actionable recommendations at the right time.

Abstract: Early recognition of chronic conditions is critical to ensure patients receive timely, appropriate care, in turn promoting improved outcomes and long-term well-being. Passive, AI-based surveillance of routine electronic health records (EHRs) provides information about diagnostic trajectories that can inform early actions supporting diagnosis and optimal disease management. However, determining (a) when to take action and (b) what action to take depends not only on predicted risk, but also on the broader clinical context, including evolving health trajectories, anticipated follow-up patterns, competing risks, and uncertainty over time. Further, clinicians already face an unmanageable volume of alerts, therefore clinical decision support (CDS) systems must deliver information at the right time, in the right format, and in a way that integrates seamlessly with existing workflows. In this talk, I will describe steps toward development of CDS systems that combine interpretable longitudinal prediction from complex EHR data with sequential decision-making methods designed to optimize alert timing based on clinical context. In addition to predicting specific risks, the resulting systems aim to generate context-aware clinical recommendations and justify them based on clinician-interpretable evidence from the EHR.