Protocol plans synthesis of caregivers’ views on AI dementia care

A published protocol sets out a planned qualitative evidence synthesis of how family and informal caregivers accept, question and anticipate artificial intelligence-enabled dementia care. The project is intended to examine perceptions of systems used for assessment, monitoring, decision support and personalized care, rather than evaluate whether any particular AI product improves clinical outcomes. Its authors say that adoption may depend not only on technical performance but also on whether caregivers trust the technology and can incorporate it into daily care.

The protocol, registered with the International Prospective Register of Systematic Reviews as CRD420261470957, has not yet produced review findings. Family and informal caregivers are often unpaid spouses, partners, adult children, relatives, friends or neighbors who provide continuing physical, emotional or practical support to a person living with dementia. As cognitive impairment progresses, these caregivers may take on responsibility for tracking health, managing behavioral and psychological symptoms, communicating with professionals and making care decisions.

They are therefore likely to be both end users and important decision-makers when AI-enabled tools enter the home, community or healthcare setting. The protocol describes potential applications including machine-learning decision support, conversational and generative AI, intelligent monitoring, predictive analytics and socially assistive robots incorporating AI. Researchers have described possible benefits of such systems, including better care coordination, more timely decision-making, support for independent living and more efficient management of care.

At the same time, the article notes that caregivers may simultaneously value AI as practical assistance and raise concerns about privacy, data security, accountability, reduced human interaction, the ethics of care and the risk of displacing person-centered relationships. Their expectations may center on AI complementing rather than replacing human caregivers. These remain implementation questions rather than settled effects.

The protocol says existing qualitative evidence is fragmented across different technologies, care settings and study populations. Previous reviews have concentrated on individual categories such as socially assistive robots or supportive mobile applications. That narrower focus cannot necessarily capture technologies that use adaptive learning, predictive models, autonomous functions or data-driven personalization.

The investigators argue that conventional digital interventions and AI-enabled systems create different practical and governance questions, making a broader synthesis necessary. For the planned review, eligibility will be defined with the SPIDER framework, which covers sample, phenomenon of interest, design, evaluation and research type. Eligible studies must include family or informal caregivers of people living with dementia and report qualitative evidence about acceptance, perceptions, attitudes, beliefs, trust, concerns, expectations, preferences, benefits, challenges or experiences associated with AI-enabled dementia care.

Studies confined to healthcare professionals, paid staff or people with dementia without caregiver perspectives will be excluded. Peer-reviewed qualitative studies in English, and mixed-methods studies with separately extractable qualitative data, will be eligible. The researchers plan to search PubMed, Embase, CINAHL, PsycINFO and Web of Science from database inception through the review period.

A librarian will help develop the strategy using the three-step approach recommended by the Joanna Briggs Institute: an initial exploratory search, a database-specific search combining controlled vocabulary and free-text terms, and screening of reference lists. The planned subject areas are dementia, family and informal caregiving, and AI-enabled care. Boolean operators will combine terms, and backward citation searching will be supplemented where necessary with forward searching in Web of Science.

Because the review is limited to peer-reviewed English-language research, it will not systematically search grey literature. Two reviewers will screen titles and abstracts and then full texts independently against the predefined criteria. A pilot exercise will precede formal screening so the team can identify inconsistencies and clarify the rules if necessary.

Disagreements will be addressed through discussion, with a third reviewer consulted if consensus cannot be reached. Reasons for full-text exclusions will be recorded, and the selection process will be presented in a PRISMA 2020 flow diagram. The protocol itself follows the PRISMA-P 2015 statement, while the completed review is expected to follow ENTREQ reporting guidance for transparent synthesis of qualitative research.

Included studies will undergo independent methodological appraisal with the 10-item Joanna Briggs Institute Critical Appraisal Checklist for Qualitative Research. The checklist covers such features as methodological congruity, data collection, analysis, interpretation, researcher reflexivity, ethics and representation of participants’ voices. Quality scores will guide interpretation and confidence judgments rather than serve as automatic exclusion criteria, so a study with methodological limitations may still be included if it meets the review’s scope.

Reviewers will pilot the appraisal process, discuss discrepancies and use a third reviewer when necessary. Relevant study data will be extracted independently with a standardized form, including caregiver characteristics, dementia-care setting, AI technology, study methods and reported qualitative findings. Synthesis will follow the thematic synthesis method developed by Thomas and Harden, allowing recurring ideas to be compared across studies while retaining the context of caregivers’ accounts.

In the first stage, two reviewers will code the extracted findings line by line, initially using inductive methods so that concepts emerge from the data rather than being forced into a pre-existing framework. Similar codes will then be grouped into descriptive themes. Finally, those themes will be interpreted and refined into higher-order analytical themes.

The researchers will maintain an audit trail of coding and analytical decisions, examine differences across technologies and care settings, and consult a third reviewer if disagreements remain unresolved. Confidence in each eventual review finding will be rated with GRADE-CERQual, a framework designed to assess qualitative evidence syntheses rather than individual studies. The assessment will consider methodological limitations, coherence, adequacy of supporting data and relevance to the review question.

Findings will be classified as high, moderate, low or very low confidence, and a Summary of Qualitative Findings table will present each theme, its contributing studies and the reasoning behind its confidence rating. The final review is scheduled to begin its systematic search in October 2026, with screening expected by January 2027, extraction by March, synthesis by May and final results by July. These are planned dates and may shift with the volume of eligible evidence.

The protocol therefore describes a rigorous method for assessing caregivers’ views, not a completed demonstration that AI is accepted, effective or safe in dementia care.

Subject of Research: Planned qualitative synthesis of caregivers’ perspectives on AI-enabled dementia care

Article Title: Family and informal caregivers’ acceptance, concerns and expectations regarding artificial intelligence-enabled dementia care: a protocol for qualitative evidence synthesis

Article References: Wang, F., Li, J., & Shen, Y. (2026). Family and informal caregivers’ acceptance, concerns and expectations regarding artificial intelligence-enabled dementia care: a protocol for qualitative evidence synthesis. PLOS One, 21(10), e0360163. https://doi.org/10.1371/journal.pone.0360163

Image Credits: AI Generated

DOI: 10.1371/journal.pone.0360163

Keywords: artificial intelligence, dementia care, family caregivers, qualitative synthesis, systematic review protocol, technology acceptance, caregiver trust, ethical governance, human-centered care, Family, informal, caregivers