AI-generated patient-friendly discharge summaries to empower patients

BackgroundPatients often struggle to fully understand their discharge letters after inpatient hospital stays, which are often replete with domain-specific medical terminology. Large Language Models (LLMs) offer a promising solution by creating patient-friendly discharge summaries. However, direct evaluations of patient comprehension of such summaries have been limited. MethodsThis study explored whether AI-generated patient-friendly discharge summaries improve patients self-reported understanding of their medical condition. We provided patients at the night before discharge with AI-generated summaries of their discharge letters and recorded how this increased their understanding of their medical condition and procedures during their stay using an 11-item survey. ResultsAmong hospitalized patients recruited at two departments in our tertiary care hospital (n=20), most (90%) reported better understanding after reading the AI-generated summary, including those who initially felt well-informed. Notably, 90% (18/20) found the summary more helpful compared to their discharge consultations, and older patients (above 69 years) showed particular interest in receiving such summaries for future hospital stays. ConclusionOur findings highlight the value of LLMs in improving patients comprehension of their medical condition. Larger studies are warranted to guide the implementation of patient-facing AI-generated content into healthcare.