Reframing Automation Benefits: A Human-Centered Approach to Expanding the Value of AI-Generated Reports in Healthcare

The use of Large Language Models (LLMs) for generating medical reports is being increasingly explored across medical specialties. However, these solutions often prioritize the perspectives of report authors, leaving the recipients and their needs outside of the scope of consideration. In this dermatology-focused study, we conducted co-design sessions with general practitioners (GPs) (N=12) in Denmark to establish their content preferences in dermatology reports and investigate unmet information needs in specialist-authored reports. We discuss using their preferences as input to generative AI to enhance the usefulness of medical reports. Such AI could support dermatologists by drafting reports and acting as a proxy for GP information needs, thus improving GP efficacy, enhancing patient outcomes, and reducing the overall burden on healthcare systems. Building on this study, we plan to: refine report structures with dermatologists and patients focusing on collaborative decision-making; and investigate non-chat genAI interaction space for automated reporting in safety-critical environments.