Artificial intelligence in home-based serious illness care: a scoping review of applications supporting quality palliative care

Introduction

Background

Serious illnesses are conditions with elevated mortality risk that significantly impair daily functioning or quality of life and often impose a substantial burden on family or friend caregivers (1). In the United States, nearly one in four adults (approximately 63 million people) serve as family caregivers, with the majority providing care at home for individuals with serious or complex health conditions (2). Home care is critical because it offers personalized support, comfort, and continuity in a familiar environment, which can reduce patient stress and improve quality of life (3). Given that much of serious illness care occurs in the home, this environment plays a pivotal role in shaping care delivery and outcomes, making it a critical focus for improving serious illness care.

Artificial intelligence (AI) encompasses technologies that enable machines to simulate human learning, reasoning, and decision-making (4). AI-driven approaches, including machine learning, natural language processing, and clinical decision-support systems, have the potential to deliver proactive, individualized, and efficient strategies for serious illness care (5). These technologies can promote continuity of care, optimize symptom management, facilitate psychosocial support, and strengthen assistance for family caregivers. Through remote monitoring, predictive analytics, and personalized interventions, AI has the potential to enhance safety, reduce hospitalizations, and improve quality of life for adults with serious illness and their family caregivers.

Rationale and knowledge gap

The growing population of individuals with serious illness, coupled with persistent shortages and burnout among palliative care providers and nurses delivering primary (or “non-specialty”) palliative care, underscores the urgent need for innovative solutions to support home-based care (6-8). AI has emerged as a transformative technology in healthcare, offering promising opportunities to bridge gaps in access and delivery of palliative care, particularly in resource-limited settings (9).

Despite increasing interest in leveraging AI for home-based serious illness care, current evidence remains fragmented and largely conceptual (10,11). Few empirical studies have examined how these technologies can be integrated into palliative care workflows or their real-world impact on continuity of care, symptom management, and caregiver support. Moreover, the scope and application of AI in palliative care remains largely in clinical and outpatient settings with poor characterization of potential applications to the home, highlighting a critical knowledge gap (12-15). The National Consensus Project (NCP) Palliative Care Quality Domains provide an established framework for defining high-quality palliative care across settings, including home-based care, and offer a useful lens for interpreting how emerging AI tools align with core palliative care principles.

Objective

Given these gaps, this scoping review aims to systematically map existing evidence on AI applications supporting home-based care for adults with serious illness and their caregivers, with a focus on their potential role in enhancing palliative care delivery. Specific objectives are to:

• Identify and map the literature on AI technologies used in home-based serious illness care;
• Categorize the types of AI-driven tools and applications (e.g., machine learning, natural language processing, decision-support systems);
• Examine the palliative care domains addressed by these technologies (e.g., structure and processes, physical, psychological, social, cultural); and
• Highlight knowledge gaps and research priorities related to integrating AI into home-based palliative care.

We present this article in accordance with the PRISMA-ScR reporting checklist (available at https://apm.amegroups.com/article/view/10.21037/apm-2025-1-139/rc).

Methods

Eligibility criteria

Inclusion criteria required studies focus on: (I) adult patients (≥18 years) with a serious illness; (II) family caregivers of adult patients with a serious illness; or (III) adult patient-caregiver dyads. Serious illness populations were defined according to the World Health Organization’s leading causes of mortality between 2000 and 2021, including coronary artery disease, stroke, cancer, lower respiratory infections (e.g., influenza, pneumonia), chronic obstructive pulmonary disease, Alzheimer’s disease and related dementias, diabetes, kidney diseases, and cirrhosis or liver disease (16).

Articles were excluded if they focused on acute illness, end-of-life or hospice care, bereavement, caregiver self-care, paid or formal caregivers, or populations outside the specified serious illness conditions. We excluded hospice and end-of-life-focused studies because their goals and outcome priorities (e.g., comfort in the last weeks to months of life, supporting bereavement) differ meaningfully from the review’s focus on ongoing home-based serious illness self-care and caregiver contributions early in the illness course. This boundary aligns with the NCP guidelines’ distinction between palliative care across the illness trajectory (Domains 1–6) and hospice care as the patient approaches the end of life (Domain 7), as well as recognized definitions that differentiate palliative care from hospice.

Eligible articles were original research focused on AI as a tool or measure to support patient self-care or caregiver contributions to patient care in the home setting. Self-care was defined as activities aimed at maintaining physical and emotional stability, monitoring and tracking symptoms to detect deterioration, and managing symptoms through care adjustments, help-seeking, and coping strategies (17). Caregiver contributions to patient self-care were defined as supporting maintenance, monitoring, and management activities (18).

Articles were excluded if AI or machine learning were not used as a tool or measure (i.e., used solely for administrative or research purposes) or if they occurred outside the home setting (e.g., assisted living, long-term care, acute care). Additional exclusions included unpublished studies, conference abstracts, dissertations, summary reports, editorials, protocols, book chapters, and review articles. Articles where AI was implemented exclusively through clinician-led or institutionally monitored approaches (e.g., remote monitoring managed by healthcare professionals) were also excluded, as the review focused on self-care and family caregiver-led home care rather than clinician-led care. The review protocol is available upon reasonable request.

Information sources

In collaboration with a medical librarian and researchers with AI/machine learning expertise, a search was conducted using a combination of keywords along with database-specific search operators, wildcards, and proximity algorithms to perform searches across PubMed (including Medline via the National Library of Medicine), CINAHL Complete (EBSCOhost), Embase (Elsevier), Scopus (Elsevier), Web of Science (Clarivate Analytics), and APA PsycINFO (via ProQuest). Adoption of AI in home-based care accelerated around 2020, driven largely by the COVID-19 pandemic, which increased demand for digital health solutions. Advances in machine learning and conversational agents, combined with regulatory flexibility and increased investment in digital health, enabled rapid deployment of AI tools during this period (19). To capture this surge in innovation and reflect current technological capabilities, this review included studies published from 2020 through 2025.

Search strategy

Figure 1 shows the primary search string and Table S1 illustrates how the primary search string was translated and performed across the six databases. The inclusion criteria for article eligibility were defined based on population, intervention, exposure, and key study characteristics.

Figure 1 Primary search string for database search. AI, artificial intelligence; CAD, coronary artery disease; CHF, congestive heart failure; CKD, chronic kidney disease; COPD, chronic obstructive pulmonary disease; CVA, cerebrovascular accident; LRI, lower respiratory infection.

Selection process

The search yielded 2,285 references across six databases: PubMed (n=320), CINAHL Complete (n=104), Embase (n=665), Scopus (n=555), Web of Science (n=584), and APA PsycINFO (n=57). All citations were exported to EndNote 21, where 494 duplicates were removed. The remaining 1,791 unique citations were imported into Covidence for screening and data extraction. Within Covidence, the first-listed reference was retained when duplicates occurred. Covidence identified an additional 264 duplicates, and one duplicate was removed manually, leaving 1,526 references for title and abstract screening.

Figure 2 presents the PRISMA 2020 flow diagram of the systematized review process. During the title and abstract screening, 1,460 references were excluded. Each citation was independently screened by two of the three reviewers (C.X. and A.C.B.), with disagreements resolved by a third reviewer (A.N.). Sixty-six references were retrieved for full-text review; however, 13 could not be obtained, leaving 53 articles for eligibility assessment. Of these, 24 met the inclusion criteria. Full-text screening followed the same dual-reviewer process (C.X., A.C.B., and A.T.), with unresolved conflicts adjudicated by a fourth reviewer (A.N.). Twenty-nine articles were excluded for the following reasons: not exclusively focused on serious illness populations (n=11), AI not used as a tool or measure (n=11), clinician-led or institutional-monitored use of AI tool or measure (n=5), and setting outside the home (n=2). The remaining 24 articles were included in the final analysis.

Figure 2 PRISMA 2020 flow diagram. AI, artificial intelligence.

Data extraction and synthesis

Data extraction was conducted in Covidence using a customized template developed by the review team (20). Three reviewers (C.X., A.C.B., and A.T.) independently extracted data from each article and consolidated findings through discussion. When additional confirmation of data relevance was required, a fourth reviewer (A.N.) was consulted to ensure accuracy and consensus. Data items were determined by the study objectives. Extracted information included study aims, sample characteristics, conceptual or theoretical framework, methods, measures, key findings related to AI use, and strengths and limitations of the AI application (Table S2).

Results of syntheses

Study characteristics

Twenty-four articles were included in the review: 5 qualitative, 10 quantitative, 4 mixed methods, 4 multiple methods, and 1 intervention. The 24 studies represented a total of 770 patients, and 231 family caregivers. Several studies also included “Other” participants, such as those involved in human-subject-exempt research (e.g., data mining of social media posts by family caregivers on Reddit) or laboratory-based caregiving simulations. Sample sizes varied between qualitative (8–33 participants), quantitative (2–20,000 participants), mixed-methods (10–70 participants), multiple (2–50 participants), and intervention studies (150 participants). Studies were from the United States (n=7), Singapore (n=3), South Korea (n=3), United Kingdom (n=2), Canada (n=1), Egypt (n=1), Italy (n=2), Germany (n=1), Latvia (n=1), the Netherlands (n=1), New Zealand (n=1), Norway (n=1), Peru (n=1), and Taiwan (n=1). One study reported its location broadly as Europe. Seventeen articles focused on Alzheimer’s and related dementias, 3 on cancer, 2 on type 2 diabetes, and 2 on stroke. Four articles identified a theoretical framework: Theory of Emotion Practice, Transtheoretical Model and Self-Determination Theory, Graph Barlow Twins Model, and Model of Challenges and Coping Behaviors After Stroke. Table S2 provides a summary table of included articles.

Study syntheses

A three-step content analysis (Table 1) was used to classify AI applications supporting home-based serious illness self-care and caregiving (42). Step 1: AI Tool Unit Identification captured distinct AI functions (e.g., chatbot, predictive model, dashboard) as individual tool units. Step 2: Category Naming grouped these units by primary purpose (e.g., Emotional Care, Predictive Care, Interface Design). Step 3: Theme Identification synthesized conceptual patterns within categories, linking tools to broader caregiving and health contexts (e.g., Emotional and Relational Dimensions, Predictive and Proactive Care). Qualitative content analysis of the included studies identified six themes: personalization and contextual adaptation, multimodal and accessible interfaces, emotional and relational dimensions, predictive and proactive care, daily routines and care ecosystems, and equity and access. To enhance interpretability for palliative care clinicians and researchers, each theme was mapped to the eight National Consensus Project (NCP) Palliative Care Quality Domains to illustrate alignment with established palliative care principles (Table 2) (43). Each theme is described in detail below.

Personalization and contextual adaptation

Personalization was a dominant feature across AI tools and frequently cited as essential for fostering engagement and trust across diverse cultural contexts. This theme aligns primarily with NCP Domains 1 (Structure and Processes of Care), 4 (Social Aspects of Care), and 6 (Cultural Aspects of Care), reflecting the role of personalization in culturally inclusive assessment, family-centered care, and individualized care planning.

Bosco et al. evaluated Lola, a multimodal app for Black American caregivers of persons with Alzheimer’s and related dementias (32). It leverages a large language model to deliver culturally relevant health literacy through text, voice, and visual outputs, adapting responses to caregiver input and enabling inclusive access via multimodal interaction. Caregivers valued the representation of other Black caregivers and context-specific information, noting that cultural tailoring enhanced usability and adoption. Similarly, Yoon et al. examined the EMPOWER app’s FAMILY module housed in a smartphone app for diabetes care and integrated machine learning algorithms to generate personalized behavioral nudges to family members based on patient’s behaviors such as diet and medication adherence (38). Participants emphasized the importance of aligning nudges with family dynamics and cultural norms to avoid caregiver burden and preserve autonomy. Also, Ana, a chatbot designed for Peruvian dementia caregivers used natural language processing to interpret caregiver queries and respond with culturally appropriate phrasing and tone, embedding local language nuances and caregiving norms into its conversational design (21). Li et al.’s chat-based interaction design using GPT-4 demonstrated a chain-of-thought reasoning process in AI responses to dementia-related questions (38). Across these tools, AI was not limited to automation; it also incorporated context-aware, culturally sensitive, and behavioral-response interactions designed to elevate caregivers’ engagement and trust in the tools.

Multimodal and accessible interfaces

Accessibility was identified as a critical factor influencing usability, with multimodal interfaces offering solutions for diverse literacy and technology skills. This theme aligns with NCP Domains 1 (Structure and Processes of Care), 4 (Social Aspects of Care), and 6 (Cultural Aspects of Care), as multimodal design supports communication, caregiver education, and accessibility across diverse cultural, linguistic, and literacy contexts.

Bosco et al. reported that caregivers appreciated Lola’s integration of voice, text, and visual outputs, which allowed users to ask questions of the chatbot and receive culturally tailored responses in multiple formats (32). The large language model powering Lola dynamically generated answers, adapting them into spoken, written, or visual formats to accommodate older adults and those with limited digital literacy. Similarly, Schmitz & Becker found that caregivers preferred rule-based chatbots delivering short, clear, multimedia content to support learning, while Yoon et al. highlighted the benefits of AI-enabled visual tools and gamification in app-based motivational interviewing for diabetes management (34,36). However, barriers persisted: caregivers in Espinoza et al. reported challenges with the overheating of the device and connectivity issues, and Tiersen et al. identified usability and cost as major hurdles for adopting smart home systems (21,37).

Incorporation of emotional and relational dimensions

Several studies incorporated emotional and relational features into AI tools. This theme aligns with NCP Domains 1 (Structure and Processes of Care), 3 (Psychological and Psychiatric Aspects of Care), and 4 (Social Aspects of Care), reflecting attention to caregiver emotional well-being, relational roles, and appropriate integration of technology into care processes.

Smriti and Huh-Yoo focused on “emotionally charged tasks”, referring to deeply meaningful caregiving interactions such as comforting a loved one, managing identity-related care roles, or shaping persons’ habits (22). Caregivers resisted delegating these intimate responsibilities to AI, emphasizing that the technology should not “handle” the emotion itself. Instead, they preferred AI support for routine tasks (e.g., reminders, scheduling, monitoring) to reduce care burden. Espinoza et al. developed Ana, a chatbot featuring a “vent” function for emotional expression and empathetic language (21). While Ana did not generate or feel emotions, its natural language processing allowed caregivers to type or speak frustrations and receive contextually appropriate, supportive responses, a feature of caregivers valued for stress relief and practical advice. Similarly, Gomaa et al. integrated a chatbot into a chemotherapy symptom management system, which participants described as helpful for coping with emotional distress during treatment (23). Collectively, these emotional support features did not replace human empathy, but provided AI-mediated safe spaces for expression, routine assistance, and context-sensitive language that mimicked empathetic responses.

Predictive and proactive care

Predictive algorithms were common across several studies to enable early detection of health risks and support proactive care. This theme aligns with NCP Domains 1 (Structure and Processes of Care), 2 (Physical Aspects of Care), and 4 (Social Aspects of Care), highlighting the role of predictive analytics in symptom monitoring, care coordination, and caregiver support.

Bijlani et al. developed a self-supervised Graph Neural Network model to detect anomalies in the home activity of people living with dementia, achieving high recall and generalizability while providing explainability through sensor-level attention (24). Cho et al. applied machine learning to predict behavioral and psychological symptoms of dementia using multimodal inputs, including actigraphy and caregiver diaries, enabling individualized management strategies (28). HekmatiAthar et al. demonstrated that agitation episodes could be forecasted 30 minutes in advance using environmental sensor data, creating opportunities for timely intervention (29). Similarly, Homdee et al. used machine learning models to predict breakthrough cancer pain based on environmental factors, achieving correlations as high as r=0.90 (30). Tawfik et al.’s ChemoFreeBot demonstrated real-time symptom monitoring to support self-care and maintenance through interactive guidance, with high user ratings for usability (94%) and informativeness (88%), but more moderate performance in navigation (70%), perceived empathy and realism (72%), and error handling (76%) (44).

Integration into daily routines and care ecosystems

AI was also evaluated for its ability to optimize daily care workflows to promote sustainability and caregiver adoption. This theme aligns with NCP Domains 1 (Structure and Processes of Care) and 4 (Social Aspects of Care), emphasizing continuity, coordination, and sustainability of care within caregivers’ everyday environments.

Nap et al. assessed a Decision Support System that consolidated multiple assistive technologies into a single dashboard, improving caregiver coordination, and enabling early detection of changes (25). Epalte et al. evaluated the Vigo chatbot app designed to support cardiovascular treatment recovery. The system delivered structured modules and incorporated gamification strategies to enhance users’ knowledge, skills, and motivation. Outcomes were assessed across multiple time scales (daily, weekly, and monthly), demonstrating its capacity to support recovery processes. The study also reported substantial variability in user engagement over time, along with a relatively high attrition-to-completion ratio in this longitudinal, multi-phase design, which followed participants for up to 52 days (45). Similarly, Andrushevich et al. tested Home4Dem, a sensor-based system with real-time alerts for dementia care, that caregivers perceived as supportive, though they noted challenges with data quality and sensor maintenance (40). Tiersen et al. emphasized the importance of aligning smart home technologies with daily routines to reduce anxiety and enhance usability (37). However, interoperability (e.g., different devices, smart home sensors/platforms, or health apps) and reliability (e.g., consistency and accuracy in detecting or predicting symptoms, sending alerts, or functioning without frequent errors) were significant barriers (7,24,28,30,33,37).

Equity and access

Concerns about equity and access were common across studies. Cultural tailoring and multimodal design addressed some disparities, but barriers such as cost, connectivity, and digital literacy persisted. This theme aligns with NCP Domains 1 (Structure and Processes of Care), 4 (Social Aspects of Care), 6 (Cultural Aspects of Care), and 8 (Ethical and Legal Aspects of Care), reflecting how access, resources, and delivery structures shape equitable engagement with AI tools.

Espinoza et al. reported connectivity issues and device (i.e., smartphones running a WhatsApp-integrated chatbot app) overheating as barriers for caregivers in Peru (21). Yoon et al. found that older or less tech-savvy patients benefited most from family-based modules embedded in a smartphone app (EMPOWER with FAMILY module), though adaptation still depended on digital literacy and consistent device access (38). Tiersen et al. identified cost and usability as key determinants of adoption of smart home technology, and Parmanto et al. emphasized the need for low-resource AI models deployable on basic devices (e.g., wearables, ambient/home sensors, and tablet interface) to improve accessibility (37,41).

Discussion

This review of AI applications supporting home-based care for adults with serious illness and their caregivers revealed six overarching themes: personalization and contextual adaptation, multimodal and accessible interfaces, emotional and relational dimensions, predictive and proactive care, daily routines and care ecosystems, and equity and access. Collectively, these findings illustrate both the promise and limitations of AI in advancing palliative care access and delivery. Personalization emerged as a critical feature, with culturally tailored AI tools improving trust and usability. Multimodal interfaces, combining voice, text, and visual elements, helped overcome digital literacy barriers and enhance engagement, particularly among older adults. Inclusion of emotional and relational features, such as empathetic chatbot responses and peer support functions, provided encouragement and reduced caregiver isolation. Predictive and proactive capabilities enabled early identification of symptom changes and care needs, while integration into daily care workflows promoted usability and adoption by caregivers. Finally, challenges with equity and access persisted, including affordability, connectivity, and ethical issues such as privacy and bias.

Articles reviewed in this analysis demonstrate the potential of culturally tailored AI interventions, consistent with palliative care’s emphasis on individualized, person-centered support (43). Tools such as Lola and Ana exemplify personalized AI systems that leverage context-specific data to meet the needs of diverse caregiver populations (21,32,46). These findings illustrate how personalization can be achieved through culturally relevant content and language adaptation to enhance trust, autonomy, and sustained engagement. Personalization appears to be a critical component for equitable technology adoption within palliative care contexts.

Multimodal design emerged as a key feature for accessibility and usability. Systems integrating voice, text, and visual components have been shown to improve usability among older adults by reducing cognitive load and error rates (47). Bosco et al.’s findings with Lola, which combined audio and visual outputs, align with research demonstrating how audio-visual redundancy enhances social presence and engagement among older users (32,48). These insights suggest that multimodal delivery, paired with streamlined onboarding, is particularly relevant for home-based palliative care, where technology must accommodate varying abilities and minimize caregiver burden.

Caregiver preferences for AI that supports rather than replaces emotional interactions echo recent evaluations of GPT-4-based chatbot prototypes, which identified needs for emotional validation, trust, and safe spaces for expression (49). While the system of Gomaa et al. relied on rule-based or keyword-triggered responses rather than “true” generative AI, their ability to mimic empathetic phrasing highlights progress in contextual support for emotional distress (23). However, consistent with broader literature noting that generative AI often struggles with contextual nuance in emotionally charged interactions, these findings suggest that if Gomaa’s system achieved contextual responsiveness, it represents an important advance but still requires cautious interpretation regarding the limits of AI in replacing human empathy (50). These findings suggest that AI chatbots may serve as complementary emotional support tools rather than substitutes for human connection in palliative care. Involving palliative care clinicians and researchers in the design process is essential to ensure these tools foster psychological safety and meet the unique emotional demands of caregiving. To support this balance, co-design should involve members of the interprofessional palliative care team (e.g., physicians, nurses, social workers, psychologists, therapists) as well as patients and family caregivers, to ensure emerging tools foster psychological safety and reflect real-world emotional and relational needs.

Our finding that several AI tools focused on predictive and proactive care reflects a broader movement towards predictive analytics in home health. Evidence indicates that AI-driven remote monitoring and risk stratification can reduce readmissions, improve health outcomes, and streamline workflows (51,52). These operational successes parallel our findings of highly accurate forecasting tools but also highlight a persistent gap between algorithm performance and clinically meaningful impact (24,28,29). Future research should prioritize outcome-based trials embedded in real-world palliative care settings to determine whether predictive capabilities translate into clinically meaningful improvements in patient and caregiver outcomes.

Embedding AI into existing workflows is essential for adoption. Reports highlight common barriers such as technical infrastructure gaps, cost constraints, and workflow disruptions that align with a system size, strategy, and long-term goals (53,54). A qualitative study with nurses in home care further reveal concerns about AI’s impact on relational and adaptive caregiving practices (55). These findings emphasize the need for participatory, adaptive design approaches that align AI tools with the realities of home-based care and preserve caregiver-patient relationships. Future research should consider that AI tools, while intended to enhance serious illness care, may inadvertently exacerbate existing disparities, as barriers in access to technology, connectivity, and digital literacy can reinforce the very gaps these systems aim to close (12). Designing AI tools that complement rather than disrupt these relational dynamics ensures technology supports the holistic, person-centered approach that defines palliative care.

Equitable deployment of AI in home-based care remains an urgent challenge. This review highlights key infrastructure barriers, noting that the benefits of these technologies remain limited by the digital cognitive gap, connectivity challenges, and the cost of necessary hardware. Frameworks such as Johns Hopkins University’s Digital Health Care Equity Framework emphasize multidimensional equity, including access, cultural relevance, and affordability (56). Similarly, recent commentaries caution against reliance on generative AI without addressing social determinants of health and design bias (57). Our synthesis reflects these concerns, noting persistent barriers such as connectivity issues in low-resource settings, literacy gaps, and the risk of excluding historically marginalized caregiver populations. In palliative care, equity and inclusivity are core principles embedded within the NCP Cultural, Social, and Ethical Domains, meaning AI tools must be affordable, culturally resonant, and available in multiple languages to ensure that technology enhances rather than limits access to compassionate care. Interdisciplinary collaboration, with input from clinicians, data scientists, behavioral experts, policy experts, and community partners, will be essential to address these persistent equity challenges and develop equitable AI interventions.

Limitations

This review has several limitations. First, most studies focused on feasibility or usability rather than clinical or psychosocial outcomes, leaving gaps in understanding the real-world impact of AI tools on palliative care delivery. This substantial variation in technological maturity, from early-stage prototypes to feasibility testing and limited implementation trials, has important implications for clinical interpretation and real-world applicability, as many reported benefits reflect preliminary testing rather than demonstrated clinical impact (35). Thus, usability, effectiveness, and scalability may differ markedly as systems progress toward real-world deployment in palliative care (58). Most included studies were conducted at a pilot stage, with substantial variability in sample sizes and outcome evaluation designs. For example, some studies employed longitudinal assessments to examine day-, week-, and month-level outcomes (45), whereas others relied solely on pre- and post-intervention measures (44). This methodological heterogeneity limits the ability to determine the optimal timing and effectiveness of similar system designs. This highlights a need for more rigorous, outcomes-focused research in the future. Second, rapid advancements in AI technology mean that findings may not fully capture emerging applications or evolving ethical considerations, underscoring the need for ongoing evaluation. Third, the analysis was restricted to studies published in English, which may exclude relevant evidence from non-English sources. Fourth, studies varied widely in design, sample size, and outcomes, limiting comparability across AI tools. Finally, only one study was published in a palliative care journal, suggesting a scarcity of research on home-based AI tools explicitly grounded in principles of palliative care, which may limit the ability to generalize findings to palliative care contexts (30). Very little research on AI has been published in palliative care-specific journals, suggesting a scarcity of emerging AI tools directly informed by fundamental principles of palliative care.

Future directions

Future research should evaluate clinical and psychosocial outcomes, particularly in real-world palliative care settings. Greater emphasis on studies explicitly grounded in palliative care principles is needed to enhance relevance and applicability in this context. Finally, ongoing research must keep pace with rapid AI advancements while proactively addressing emerging ethical and implementation challenges to ensure safe and equitable integration into care delivery.

Conclusions

This scoping review highlights both the promise and limitations of AI in supporting home-based palliative care for adults with serious illness and their caregivers. Framing the six review themes (personalization and contextual adaptation, multimodal and accessible interfaces, emotional and relational dimensions, predictive and proactive care, daily routines and care ecosystems, and equity and access) within the NCP Palliative Care Quality Domains underscores that current AI innovations primarily address structural and social components of palliative care. Findings highlight the potential for AI tools to enhance person-centered care, reduce caregiver burden, and anticipate care needs. However, persistent barriers such as gaps in digital literacy, infrastructure constraints, and ethical concerns, underscore the need for inclusive, culturally sensitive, and affordable solutions. To meaningfully advance AI in palliative care, future research should prioritize co-design with patients, caregivers, and clinicians; and evaluate relevant outcomes in real-world settings. Aligning AI innovation with core palliative care principles can transform home-based care for individuals with serious illness and their families.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the PRISMA-ScR reporting checklist. Available at https://apm.amegroups.com/article/view/10.21037/apm-2025-1-139/rc

Peer Review File: Available at https://apm.amegroups.com/article/view/10.21037/apm-2025-1-139/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://apm.amegroups.com/article/view/10.21037/apm-2025-1-139/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Kelley AS, Bollens-Lund E. Identifying the Population with Serious Illness: The "Denominator" Challenge. J Palliat Med 2018;21:S7-S16. [Crossref] [PubMed]
2. AARP; National Alliance for Caregiving. Caregiving in the US Research Report. Published July 2025. Accessed December 2, 2025. Available online: https://www.aarp.org/caregiving/
3. Lin CP, Chu WM. Home-Based Palliative Care: Benefits, Challenges, Opportunities and Future Directions in a Super-Aged Society. J Hosp Palliat Care 2025;28:81-8. [Crossref] [PubMed]
4. Stryker C, Kavlakoglu E. What Is Artificial Intelligence (AI)? IBM. Accessed December 2, 2025. Available online: https://www.ibm.com/think/topics/artificial-intelligence
5. Rawal S, Shah AA. AI in Geriatric Medicine and Serious Illness Care. Curr Geriatr Rep 2025;14:19.
6. Frydman JL, McKendrick K, Chen Y, et al. Disparities in the Geographic Distribution of Palliative Care Specialists in 2022. J Gen Intern Med 2024;39:1528-30. [Crossref] [PubMed]
7. Li LZ, Yang P, Singer SJ, et al. Nurse Burnout and Patient Safety, Satisfaction, and Quality of Care: A Systematic Review and Meta-Analysis. JAMA Netw Open 2024;7:e2443059. [Crossref] [PubMed]
8. Zogby CB. Burnout Among Palliative Care Providers. J Am Assoc Nurse Pract 2023;35:676-81. [Crossref] [PubMed]
9. Xie Y, Zhai Y, Lu G. Evolution of Artificial Intelligence in Healthcare: A 30-Year Bibliometric Study. Front Med (Lausanne) 2024;11:1505692. [Crossref] [PubMed]
10. Zhu EM, Buljac-Samardžić M, Ahaus K, et al. Implementation and Dissemination of Home-and Community-based interventions for Informal Caregivers of People Living with Dementia: A Systematic Scoping Review. Implement Sci 2023;18:60. [Crossref] [PubMed]
11. Abdul Latif El Ejel B, Sattar S, Fatima SB, et al. Digital Diabetes Management Technologies for Type 2 Diabetes: A Systematic Review of Home-Based Care Interventions. Cureus 2025;17:e84177. [Crossref] [PubMed]
12. Nikoloudi M, Mystakidou K. Artificial Intelligence in Palliative Care: A Scoping Review of Current Applications, Challenges, and Future Directions. Am J Hosp Palliat Care 2025; Epub ahead of print. [Crossref]
13. García Abejas A, Geraldes Santos D, Leite Costa F, et al. Ethical Challenges and Opportunities of AI in End-of-Life Palliative Care: Integrative Review Interact. J Med Res 2025;14:e73517.
14. Bozkurt S, Fereydooni S, Kar I, et al. AI in Palliative Care: A Scoping Review of Foundational Gaps and Future Directions for Responsible Innovation. J Pain Symptom Manage 2025;70:e394-418. [Crossref] [PubMed]
15. Silva-Ferreira M, Cruz S, Luís MS, et al. Mapping Study on AI-Based Technologies in Palliative Care-A Scoping Study. BMC Palliat Care 2025;24:274. [Crossref] [PubMed]
16. World Health Organization. Global Health Estimates: Leading Causes of Death. Published 2025. Accessed December 2, 2025. Available online: https://www.who.int/data/gho/data/themes/mortality-and-global-health-estimates/ghe-leading-causes-of-death
17. Schulman-Green D, Feder SL, David D, et al. A Middle Range Theory of Self-and Family Management of Chronic Illness. Nurs Outlook 2023;71:101985. [Crossref] [PubMed]
18. Vellone E, Riegel B, Alvaro R. A Situation-Specific Theory of Caregiver Contributions to Heart Failure Self-care. J Cardiovasc Nurs 2019;34:166-73. [Crossref] [PubMed]
19. Keesara S, Jonas A, Schulman K. Covid-19 and Health Care's Digital Revolution. N Engl J Med 2020;382:e82. [Crossref] [PubMed]
20. Innovation VH. Covidence Systematic Review Software. Available online: https://www.covidence.org/
21. Espinoza F, Cook D, Da Re M, et al. Designing AI-powered chatbots for dementia care in Peru: Stakeholder engagement and field observations. Interact Comput 2025; [Crossref]
22. Smriti D, Huh-Yoo J. Bringing Emotions into Practice: A Framework for AI design to Support Emotion Work. In: Companion Proceedings of the 2023 ACM Conference on Computer Supported Cooperative Work and Social Computing; Minneapolis, MN; 2023. doi: 10.1145/3584931.3606988
23. Gomaa S, Posey J, Bashir B, et al. Feasibility of a Text Messaging-Integrated and Chatbot-Interfaced Self-Management Program for Symptom Control in Patients With Gastrointestinal Cancer Undergoing Chemotherapy: Pilot Mixed Methods Study. JMIR Form Res 2023;7:e46128. [Crossref] [PubMed]
24. Bijlani N, Maldonado OM, Nilforooshan R, et al. Utilizing Graph Neural Networks for Adverse Health Detection and Personalized Decision Making in Sensor Based Remote Monitoring for Dementia Care. Comput Biol Med 2024;183:109287. [Crossref] [PubMed]
25. Nap HH, Stolwijk NE, Ipakchian Askari S, et al. The Evaluation of a Decision Support System Integrating Assistive Technology for People with Dementia at Home. Front Dement 2024;3:1400624. [Crossref] [PubMed]
26. Sanchez AA, Lai J, Tighe A, et al. Exploring Home-based Reminder Systems for Daily Routine Support: Usability, Data Insights, and Future Directions. Gerotechnology 2025;24:1-12. [Crossref]
27. Hossain MA, Ray SK, Harris G, et al. An Emergency Response System to Support Early Stage Dementia Patients. In: 2022 International Symposium on Computers and Communications; 2022:1-4. doi: 10.1109/ISCC55528.2022.9912916.
28. Cho E, Kim S, Heo SJ, et al. Machine Learning-based Predictive Models for the Occurrence of Behavioral and Psychological Symptom of Dementia: Model Development and Validation. Sci Rep 2023;13:8073. [Crossref] [PubMed]
29. HekmatiAthar S. Data-Driven Forecasting of Agitation for Persons with Dementia: A Deep Learning-Based Approach. SN Comput Sci 2021;2:326. [Crossref] [PubMed]
30. Homdee N, Lach J, Blackhall L, et al. The Influence of Ambient Environmental Factors on Breakthrough Cancer Pain: Insights from Remote Health Home Monitoring and a Proposed Data Analytic Approach. BMC Palliat Care 2024;23:62. [Crossref] [PubMed]
31. Turetta C, Demrozi, F, Franceschi S, et al. Non-invasive Monitoring of Alzheimer's Patients through WiFi Channel State Information. In: 2023 International Workshop on Advances in Sensors and Interfaces; 2023:103-8. doi: 10.1109/IWASI58316.2023.10164475.
32. Bosco C, Otenen E, Osorio Torres J, et al. Designing a Multimodal and Culturally Relevant Alzheimer Disease and Related Dementia Generative Artificial Intelligence Tool for Black American Informal Caregivers: Cognitive Walk-Through Usability Study. JMIR Aging 2025;8:e60566. [Crossref] [PubMed]
33. Neo JRE, Ser JS, Tay SS. Use of Large Language Model-based Chatbots in Managing the Rehabilitation Concerns And Education Needs of Outpatient Stroke Survivors and Caregivers. Front Digit Health 2024;6:1395501. [Crossref] [PubMed]
34. Schmitz D, Becker B. Chatbot-Mediated Learning For Caregiving Relatives of People With Dementia: Empirical Findings and Didactical Implications For Mulitprofessional Health Care. J Multidiscip Healthc 2024;17:219-28. [Crossref] [PubMed]
35. Li Z, Xie B, Hilsabeck R, et al. Effects of Different Prompts on the Quality of GPT-4 Responses to Dementia Care Question. In: 2024 IEEE 12th International Conference on Healthcare Informatics; 2024:412-7. doi: 10.1109/ICHI61247.2024.00059
36. Yoon S, Tang H, Tan CM, et al. Acceptability of Mobile App-Based Motivational Interviewing and Preferences for App Features to Support Self-Management in Patients With Type 2 Diabetes: Qualitative Study. JMIR Diabetes 2024;9:e48310. [Crossref] [PubMed]
37. Tiersen F, Batey P, Harrison MJC, et al. Smart Home Sensing and Monitoring in Households With Dementia: User-Centered Design Approach. JMIR Aging 2021;4:e27047. [Crossref] [PubMed]
38. Yoon S, Lau R, Kwan YH, et al. Acceptability of an AI-enabled Family Module in a Mobile App for Enhanced Diabetes Management: Patient and Family Perspectives. Digit Health 2025;11:20552076251322654. [Crossref] [PubMed]
39. Lee JY, Yang E, Cho AY, et al. Sleep Efficiency in Community-dwelling Persons Living with Dementia: Exploratory Analysis Using Machine Learning. J Clin Sleep Med 2025;21:393-400. [Crossref] [PubMed]
40. Andrushevich A, Camenzind M, Nieke C, et al. Pilot Experiences on Independent Living in High Age via Non-invasive Multisensory Data Analytics. In: 2024 IEEE International Conference on Human System Interaction; 2024:1-6. doi: 10.1109/HSI61632.2024.10613589
41. Parmanto B, Aryoyudanta B, Soekinto TW, et al. A Reliable and Accessible Caregiving Language Model (CaLM) to Support Tools for Caregivers: Development and Evaluation Study. JMIR Form Res 2024;8:e54633. [Crossref] [PubMed]
42. Bengtsson M. How to Plan and Perform a Qualitative Study Using Content Analysis. Nurs Plus Open 2016;2:8-14.
43. National Coalition for Hospice and Palliative Care. Clinical Practice Guidelines for Quality Palliative Care. 4th ed. 2018. Available online: https://www.nationalcoalitionhpc.org/ncp
44. Tawfik E, Ghallab E, Moustafa A. A Nurse Versus a Chatbot-the Effect of an Empowerment Program on Chemotherapy-related Side Effects and the Self-care Behaviors of Women Living with Brest Cancer: A Randomized Controlled Trial. BMC Nurs 2023;22:102. [Crossref] [PubMed]
45. Epalte K, Tomsone S, Vētra A, et al. Patient Experience Using Digital Therapy "Vigo" for Stroke Patient Recovery: A Qualitative Descriptive Study. Disabil Rehabil Assist Technol 2023;18:175-84. [Crossref] [PubMed]
46. Hird N, Osaki T, Ghosh S, et al. Enabling Personalization for Digital Cognitive Stimulation to Support Communication With People With Dementia: Pilot Intervention Study as a Prelude to AI Development. JMIR Form Res 2024;8:e51732. [Crossref] [PubMed]
47. Tu Y, Luo J. Accessibility research on multimodal interaction for the elderly. Human Aspects of IT for the Aged Population: 10th International Conference, ITAP 2024, Held as Part of the 26th HCI International Conference, HCII 2024, Washington, DC, USA, June 29–July 4, 2024, Proceedings, Part II. Available online: https://doi.org/10.1007/978-3-031-61546-7_25
48. El Kamali M, Angelini L, Lalanne D, et al. Please capitalize the title as: Older Adults' Perspectives on Multimodal Interaction with a Conversational Virtual Coach. Front Comput Sci 2023;5:1125895.
49. Herschbach L, Festl-Wietek T, Stegemann-Philipps C, et al. Evaluation of an AI-Based Chatbot Providing Real-Time Feedback in Communication Training for Mental Health Care Professionals: Proof-of-Concept Observational Study. J Med Internet Res 2025;27:e82818. [Crossref] [PubMed]
50. Miner AS, Milstein A, Schueller S, et al. Smartphone-Based Conversational Agents and Responses to Questions About Mental Health, Interpersonal Violence, and Physical Health. JAMA Intern Med 2016;176:619-25. [Crossref] [PubMed]
51. Poon EG, Lemak CH, Rojas JC, et al. Adoption of Artificial Intelligence in Healthcare: Survey of Health System priorities, Successes, and Challenges. J Am Med Inform Assoc 2025;32:1093-100. [Crossref] [PubMed]
52. Neal E. New Report from Wellsky Show Use of Predictive Analytics Improves Clinical Outcomes, Leads to More Efficient Care. Business Wire. Published June 6, 2023. Accessed December 2, 2025. Available online: https://www.businesswire.com/news/home/20230606005152/en/New-Report-From-WellSky-Shows-Use-of-Predictive-Analytics-Improves-Clinical-Outcomes-Leads-to-More-Efficient-Care
53. Diaz N. What IT Overhauls Cost Health Systems. Becker's Health IT. Accessed December 2, 2025. Available online: https://www.beckershospitalreview.com/healthcare-information-technology/ehrs/what-it-overhauls-cost-health-systems/
54. Wang B, Asan O, Zhang Y. Shaping the Future of Chronic Disease Management: Insights into Patient Needs for AI-based Homecare Systems. Int J Med Inform 2024;181:105301. [Crossref] [PubMed]
55. Karnehed S, Larsson I, Petersson L, et al. Navigating Artificial Intelligence in Home Healthcare: Challenges and Opportunities in Nursing Wound Care. BMC Nurs 2025;24:660. [Crossref] [PubMed]
56. Culli L. Bridging the Digital Divide in Health Care: A New Framework for Equity. Johns Hopkins Bloomberg School of Public Health. Published 2025. Accessed December 2, 2025. Available online: https://publichealth.jhu.edu/2025/bridging-the-digital-divide-in-health-care-a-new-framework-for-equity
57. Kevin MD. How Home-based AI can reduce Health Inequities in Underserved Communities. The Podcast by KevinMD. Podcast audio. Published June 20, 2025. Accessed December 2, 2025. Available online: https://kevinmd.com/2025/06/how-home-based-ai-can-reduce-health-inequities-in-underserved-communities-podcast.html
58. El Arab RA, Abu-Mahfouz MS, Abuadas FH, et al. Bridging the Gap: From AI Success in Clinical Trials to Real-World Healthcare Implementation-A Narrative Review Healthcare (Basel) 2025;13:701.