Background: Palliative care improves the quality of life of people living with life-limiting conditions and their families; however, global access remains constrained by workforce shortages and late referrals. Artificial Intelligence (AI) has been proposed as a scalable solution for optimising the identification of needs, supporting clinical decision-making, and enhancing care delivery. However, real-world evidence of the application of AI in palliative care remains sparse, particularly regarding its impact on quality of life, quality of care, and associated practical, technical and ethical challenges. Methods: A scoping review was conducted following the Joanna Briggs Institute methodology and the PRISMA-ScR checklist. Five databases (the ACM Digital Library, CINAHL, Cochrane Central, PubMed and Web of Science) were searched between September and October 2025. Studies reporting the use of AI to facilitate or enhance palliative care delivery in adults were eligible. Four reviewers independently screened the records, and two reviewers extracted the data using Covidence software. A narrative synthesis was then performed. Results: Fifteen studies, published between 2021 and 2025, were included. Fourteen originated from Global North settings (USA 5, Germany 2, Japan 2, Taiwan 2, UK 1, Spain 1 and Cyprus 1) and one from Iran. Conceptually, AI applications fall into three domains: (1) early identification of palliative care needs, (2) symptom assessment and management and (3) clinical decision support for care conversations. Fifteen studies (100%) reported or discussed quality of care outcomes, most commonly prognostic performance, usability and referral/conversation rates, and only two (13.3%) directly addressed quality of life. Effectiveness was consistently positive, with four randomised controlled trials demonstrating superiority over usual care in referrals, advance care planning, pain control and quality of life domains. Practical barriers were centred on workflow integration and resource demands, while technical limitations include data quality, generalisability, and interpretability. Ethical discourse is underdeveloped, with major gaps in the principles of AI governance. Conclusions: AI shows potential to improve prognostic accuracy, trigger earlier involvement of palliative care specialists and support symptom management. However, this evidence is geographically skewed, methodologically immature and ethically underdeveloped. Future research must prioritise diverse global settings, patient-reported quality of life outcomes, participatory co-design and systematic ethical governance to ensure equitable implementation.

Diagnosis coding is a large-scale multi-label task in which each clinical encounter may require one or more coding labels from a large label space. Conventional top-k and threshold-based classifiers provide practical coding suggestions but do not directly characterize uncertainty over alternative coding sets. This study proposes sparse Mahalanobis conformal scoring for set prediction in diagnosis coding under extreme multi-label classification (XMC), intended for coding-assist workflows that require compact and reviewable coding suggestions. A sparse XMC model first generates candidate coding labels for each encounter. Candidate label sets are then constructed from the sparse proposal space and scored using a diagonal Mahalanobis nonconformity function calibrated on held-out data. Empirical conformal p-values are assigned to candidate sets, and downstream decision rules are used to obtain a final coding output from the retained region. The framework was evaluated using outpatient EHR data from a tertiary-care hospital, comprising approximately 8.0 million visits from 2018 to 2025 and up to 12,829 diagnosis labels. The primary SMaCS output achieved Micro-F1 close to the strongest threshold-based comparator and the highest exact match ratio among flexible-size decision rules. Compared with the other nonconformity scores, the Mahalanobis score produced a smaller retained region with fewer distinct labels, while preserving the same point-prediction performance. Additional analyses examined conformal region validity, robustness to label-frequency thresholds, code-depth performance, label-frequency subgroups, sample cardinality, department-level variation, and confidence–credibility stratification. Our results suggest that sparse Mahalanobis conformal scoring provides a useful framework for uncertainty-informed outpatient coding set prediction, while also highlighting the importance of candidate-space adequacy in extreme multi-label diagnosis coding.