Journal of pain and symptom management · 2025
AI in Palliative Care: A Scoping Review of Foundational Gaps and Future Directions for Responsible Innovation.
СтатьяScoping ReviewНазвание и аннотация: Перевод в очереди
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Bozkurt Selen, Fereydooni Soraya, Kar Irem, Chalmers Catherine Diop, Leslie Sharon L, Pathak Ravi, Walling Anne M, Lindvall Charlotta, Lorenz Karl, Parikh Ravi, Quest Tammie, Giannitrapani Karleen, Kavalieratos Dio
Аннотация
BACKGROUND: Artificial intelligenc (AI) holds increasing promise for enhancing palliative care through applications in prognostication, symptom management, and decision support. However, the utilization of real-world data, the rigor of validation, and the transparency and reproducibility of these AI tools remain largely unexamined, posing critical considerations for their safe and ethical integration in sensitive end-of-life settings.
OBJECTIVES: This scoping review systematically mapped the landscape of AI applications in palliative and hospice care, focusing on three key domains: (1) the purposes and data sources of AI models; (2) the methods and extent of model validation and generalizability; and (3) the degree of transparency and reproducibility.
METHODS: A comprehensive search was conducted across multiple databases (e.g., PubMed/MEDLINE, Embase.com, IEEE Xplore, Web of Science, ClinicalTrials.gov) from inception to December 31, 2023. Studies of any design applying AI (including machine learning or natural language processing) in palliative or hospice contexts for adults were included. Two independent reviewers screened studies and charted data on study context, patient population, data type, AI methodology, outcome, evaluation approach, and indicators of model generalizability, transparency and reproducibility.
RESULTS: From 4,747 unique records, 125 studies met inclusion criteria, with over half published in the last three years, predominantly from the United States. Most studies (86%) were retrospective proof-of-concept designs, with few randomized controlled trials (n = 7) or prospective evaluations (n = 6). AI applications primarily focused on mortality prediction (n = 63) in cancer populations (n = 62), followed by advance care planning (n = 18) and symptom assessment (n = 17). Structured electronic health record data were the most common input (n = 67, 54%). Transparency was limited, with only 19 studies (15%) sharing code and 14 (11%) providing data access; none adhered to AI-specific reporting guidelines. Ethical frameworks for evaluation were notably absent.
CONCLUSION: AI in palliative care remains in early development, showing promise in areas such as prognosis and documentation support. However, limited validation, insufficient cross-site testing, and lack of transparency currently limit clinical applicability. Future research should emphasize external validation, inclusion of broader patient data, and adoption of open science practices to ensure these tools are reliable, safe, and trustworthy.
OBJECTIVES: This scoping review systematically mapped the landscape of AI applications in palliative and hospice care, focusing on three key domains: (1) the purposes and data sources of AI models; (2) the methods and extent of model validation and generalizability; and (3) the degree of transparency and reproducibility.
METHODS: A comprehensive search was conducted across multiple databases (e.g., PubMed/MEDLINE, Embase.com, IEEE Xplore, Web of Science, ClinicalTrials.gov) from inception to December 31, 2023. Studies of any design applying AI (including machine learning or natural language processing) in palliative or hospice contexts for adults were included. Two independent reviewers screened studies and charted data on study context, patient population, data type, AI methodology, outcome, evaluation approach, and indicators of model generalizability, transparency and reproducibility.
RESULTS: From 4,747 unique records, 125 studies met inclusion criteria, with over half published in the last three years, predominantly from the United States. Most studies (86%) were retrospective proof-of-concept designs, with few randomized controlled trials (n = 7) or prospective evaluations (n = 6). AI applications primarily focused on mortality prediction (n = 63) in cancer populations (n = 62), followed by advance care planning (n = 18) and symptom assessment (n = 17). Structured electronic health record data were the most common input (n = 67, 54%). Transparency was limited, with only 19 studies (15%) sharing code and 14 (11%) providing data access; none adhered to AI-specific reporting guidelines. Ethical frameworks for evaluation were notably absent.
CONCLUSION: AI in palliative care remains in early development, showing promise in areas such as prognosis and documentation support. However, limited validation, insufficient cross-site testing, and lack of transparency currently limit clinical applicability. Future research should emphasize external validation, inclusion of broader patient data, and adoption of open science practices to ensure these tools are reliable, safe, and trustworthy.