Health-AI Decision Support Tools

Healthcare AI-systems often do not account for the socio-technical complexities present in clinical workflows, and fail in implementation regardless of their technical merit. This research focuses on employing methods from human-computer interaction (HCI), machine learning, and medicine to design AI interfaces that are usable in real-world settings. The objective of my work is to build tools that are co-designed with end-users, like clinicians, to support their unique needs, such as addressing the complexities of embedding AI into clinical workflows. Our ongoing project analyzes the deployment of a ML-assisted decision support tool developed and deployed internally at Northwestern Medicine. The tool draws from electronic medical records (EHR) to support the identification of patients needing triage for advanced heart failure (AHF). Our goal is to understand the complexities of collaborative development and deployment of this system and generalize our findings to help the future of clinical-AI.

Related publications: