Ethics of artificial intelligence prognostication in palliative care: perspectives from a national survey of palliative care physicians

OBJECTIVES: Artificial intelligence (AI) tools have the potential to transform access to and delivery of palliative care globally-an urgent unmet need-in part by early identification of patients experiencing serious illness-related suffering. Although global consensus is emerging about the importance of ethical principles for AI, such as respecting autonomous choice, promoting patient well-being, reducing bias, trust, and more, little is known about how palliative care physicians perceive these issues. This study aims to improve understanding of palliative care physicians' perspectives about the ethics of AI-based prognostication. METHODS: A national, cross-sectional survey of palliative care physicians in the USA between January 2024 and March 2025 (N=2500; overall response rate=32.6%). RESULTS: Overall, 64% (n=342/534) of physicians felt use of AI-based prognostication was at least somewhat ethically challenging; female physicians had 2.06 times the odds of feeling that way (95% CI 1.34 to 3.19; p=0.001). Most physicians (81.4%, 429/527) were at least a little concerned that AI-based prognostication could lead to an overemphasis on time until death in decision-making; older physician age was associated with this concern. Physicians reported moderate levels of concern with individual ethics issues, but 32.3% (170/527) thought AI could positively affect patient trust and 56.0% (294/525) thought AI could help them practise the way they see best. CONCLUSION: Palliative care physicians see the potential of AI-based prognostication to improve palliative care and also express ethical concerns. Implementation of AI requires context-specific ethical guidance responsive to palliative care, where relationships and communication are paramount.