Session (CTI)

ISSUE: Traditional Patient-Reported Outcome (PRO) instruments often miss the nuance of lived patient experience. Free-text entries and direct patient voice offer richer insights but have been underused due to analytical complexity and scalability challenges. Advances in AI—particularly large language models (LLMs) and conversational AI—now enable scalable capture, interpretation, and communication of patient narratives. Conversational agents provide interactive dialogue that elicits deeper context, reduces missingness, and generates structured summaries for clinicians, regulators, and decision-makers. This panel will examine how AI amplifies patient voice across development phases, improves interpretability for diverse stakeholders, and addresses risks such as bias, hallucination control, transparency, and regulatory validation. OVERVIEW: AI-enabled free-text analysis and direct patient voice complement structured PRO data, enabling scalable extraction of symptom narratives, patient priorities, and quality-of-life context that traditional instruments may miss. Practical frameworks for operationalization, validation, and regulatory acceptance of LLM-assisted analysis are essential for integrating narrative responses within eCOA platforms. This session offers an integrated view from real-world use, methodological standards, and future applications. Talk 1: Case study of LLM-based extraction and coding of an open-ended patient preference measure, covering workflow, timelines, quality checks, and integration into evidence packages. Talk 2: Methodological rigor—model performance metrics, bias detection, construct validity, version control, human review, and reproducibility—using oncology symptom capture examples. Talk 3: Conversational AI for dynamic engagement, reduced missingness, and richer lived-experience capture, alongside safety, hallucination control, and privacy considerations. Each talk runs 12 minutes, followed by 20 minutes of audience interaction. Attendees will gain actionable frameworks, validated benchmarks, and practical guidance to responsibly integrate AI-driven narrative analysis in PRO/eCOA research.