RPD314 - Implementing a Digital Conversational AI Platform to Extend Outpatient Palliative Care: Feasibility Study at Duke

Key Message : Implementing Epilog, an AI conversational platform, in an outpatient palliative care clinic was feasible and innovative, enabling high patient engagement across medical and psychosocial needs. Sharing AI-facilitated goals of care conversations with providers enhanced holistic care. Integrating such digital innovations into workflows is essential to scale supportive palliative care.

Abstract:

Background:
Patients in outpatient palliative care experience high symptom burden and psychosocial needs between visits. 1 AI-enabled digital interventions can extend support, but integration into provider workflows remains limited.2,3

Objectives:
To evaluate the feasibility of implementing Epilog, an AI conversational platform, in an academic outpatient palliative care clinic, focusing on patient engagement and workflow integration.

Methods:
Epilog was implemented under IRB approval in the Duke Outpatient Palliative Care Clinic. All in-person patients were offered participation via QR-code-based consent after their initial visit. From December 2024 to June 2025, 112 patients accessed the consent form and 53 (47.3%) enrolled. Patient-reported data were viewable via a HIPAA-compliant dashboard and providers received secure email links before follow-up visits. Twelve providers participated with iterative co-design over six months to align workflows.

Results:
Fifty-three patients enrolled, including 31 (58.5%) with cancer, 6 (11.3%) with COPD, 2 (3.8%) with dementia, and 14 (26.4%) with other serious conditions. Patients engaged with the platform an average of 8.6 times per week (median 7.2, IQR 4.1–11.0, range 1.5–37). Engagements spanned symptoms, advanced care planning (ACP), daily life, social determinants of health, goals, and legacy topics. 40 patients (75.5%) reported symptoms consistently (at least every two weeks), and 28 (52.8%) engaged in meaningful (ACP) conversations. Clinicians received structured, patient-specific summaries before visits, including symptom trend graphs, patient-posed questions, and stated goals-of-care preferences. Provider feedback indicates the platform enhanced holistic understanding of patient needs and improved ACP, though workflow integration requires further refinement.
Conclusion / Innovation Implications:
This study demonstrates the feasibility of using an AI conversational platform to engage palliative care patients between visits and deliver actionable insights to clinicians. Sustained engagement and uptake were high. Future work will focus on workflow optimization and evaluating impact on care delivery.

References: 1. Bakitas M, Lyons KD, Hegel MT, Balan S, Brokaw FC, Seville J, Hull JG, Li Z, Tosteson TD, Byock IR, Ahles TA. Effects of a palliative care intervention on clinical outcomes in patients with advanced cancer: the Project ENABLE II randomized controlled trial. JAMA. 2009 Aug 19;302(7):741-9. doi: 10.1001/jama.2009.1198. PMID: 19690306; PMCID: PMC3657724.
2. Schenker Y, Abdullah S, Arnold R, Schmitz KH. Conversational Agents in Palliative Care: Potential Benefits, Risks, and Next Steps. J Palliat Med. 2024 Mar;27(3):296-300. doi: 10.1089/jpm.2023.0534. Epub 2024 Jan 12. PMID: 38215235.
3. Finucane AM, O'Donnell H, Lugton J, et al. Digital health interventions in palliative care: A systematic meta-review. NPJ Digit Med 2021;4(1):64;