PreCare: A Web-Based Advance Care Planning Platform Integrating AI Assistants to Enhance Personal Value Exploration, Knowledge, and Decisional Confidence: Usability Study (Preprint)

UNSTRUCTURED Advance care planning (ACP) helps individuals document end-of-life care preferences before losing decision-making capacity. While web-based ACP platforms offer broad accessibility, their static content cannot adapt to users in real time, limiting value clarification, question-answering, and consequence review that clinical consultations typically provide. This study evaluated PreCare, a web-based ACP platform integrated with three AI assistants powered by large language models (LLMs) and retrieval-augmented generation (RAG), assessing its usability and effects on ACP knowledge, readiness, and decisional conflict. A pretest-posttest usability study enrolled 12 adults (5 male, 7 female; mean age 46.1 years). Participants independently used PreCare for 60–90 minutes, completing demographic surveys, pre/posttests, and a semi-structured interview. Outcome measures included self-rated ACP knowledge (4 items), readiness to engage in ACP (4 items), the Decisional Conflict Scale (DCS; 0–100), and the System Usability Scale (SUS). Wilcoxon signed-rank tests assessed pre-to-post changes. Participants used PreCare for an average of 40.2 minutes. PreCare achieved a mean SUS score of 80.63, exceeding the 68.2 website benchmark and indicating excellent usability. All 12 participants said they would recommend PreCare to others (mean 9.2/10). ACP knowledge improved significantly across all four items (P