Dementia is a neurodegenerative syndrome in which the individual can experience loss of memory, problem-solving, and other cognitive abilities that interfere with daily living [1], as well as loss of independence, self-esteem, and autonomy [2]. The prevalence of dementia is increasing due to the rapid growth of an aging population worldwide. Approximately 50 million people are living with dementia, with an average new case of dementia every 3 seconds, which is projected to double every 20 years [3].

Life expectancy following a diagnosis of mild dementia ranges from 2 to 7 years, while for moderate to advanced dementia, it ranges from 1.5 to 2.5 years [4]. On the basis of the World Health Organization report in 2019, deaths due to dementia doubled between 2000 and 2016, which makes dementia the seventh leading cause of death worldwide [5]. Given that dementia is irreversible with no curative treatment, a palliative care approach can improve the quality of life among this population. However, many people with dementia receive palliative care less often and experience a high symptom burden at the end of life [6]. Previous studies indicate that people with advanced dementia have profound functional deficits that makes them dependent on full-time care, eventually leading to more intensive medical treatments against their preferences and wishes, such as cardiopulmonary resuscitation, placement on a ventilator [7,8], and nasogastric feeding tube placement, particularly in Asian countries [9]. Hence, there is an urgent need to engage in conversations with people living with dementia about their preferences regarding palliative and end-of-life care and document these decisions.

Making a care-related decision is a complex cognitive process that requires substantial involvement of communication, attention, executive, and memory [10]. Healthy older people or older people at the early stage of dementia are cognitively capable of making decisions about care plans and medical treatments based on their own wishes and preferences. However, older people with advanced dementia may experience memory and language impairments that hinder their ability to remember new information and express their preferences [11]. As a result, depending on the national and local legislation, designated family carers may make end-of-life decisions on behalf of people with advanced dementia. However, substitute decision-making could trigger uncertainty and reactivity due to limited knowledge of the clinical condition and increased complexity in the care plan [12]. Given these challenges, advance care planning (ACP) emerges as a critical framework—an ongoing communication and decision-making process concerning goals and preferences of care between patients, family, and health care providers for future medical treatment and care before decisional capacity is lost [13,14]. By proactively engaging patients and their families in shared decision-making with clinicians, ACP aims to align future care with patient values. The development of structured decision support tools has become a key strategy to facilitate this process, particularly for navigating emotionally and medically complex choices [15].

According to the International Patient Decision Aid Standards Collaboration, decision aids are defined as evidence-based tools designed to help people participate in decision-making about health care options [16]. Decision aids can support shared decision-making by providing evidence-based options with information on the probability of their benefits and risks [17]. Decision aids are most useful in situations where there is not 1 clear option and where the care pathway is highly dependent on individual preferences, values, and personal conditions [18]. Therefore, decision aids may be particularly helpful during end-of-life conversations. The traditional approach of using paper-based decision aids with verbal descriptions is constrained in realistically envisioning future disease states, maintaining consistency between different professionals, and accommodating varying levels of health literacy among patients [19]. Innovative technologies can facilitate the delivery of vivid (using visual images and videos), interactive (navigating the content and responding to interactive questions), dynamic (changing content according to user input and interaction), and tailored (evidence based on individual demographics or clinical conditions) decision support [20].

“Delivering patient decision aids on the internet” refers to the process of using the internet to provide some or all components of a patient decision aid to help individuals involved in the process of choosing between ≥2 appropriate care and life options [21]. The International Patient Decision Aid Standards Collaboration has included “delivering patient decision aids on the internet” as 1 of the 12 dimensions to assess the quality of patient decision aids [22]. According to a previous systematic review synthesizing decision aids to support decision-making in dementia care, 6 of the 10 included studies delivered paper-based decision aids, while the remaining 4 studies delivered video decision aids [23]. To our knowledge, there are currently no reviews that have explored the use of technologies (eg, video, visual aids, web-based decision aids, and mobile health apps) to support the development and delivery of decision aids on the internet for palliative and end-of-life dementia care.

In this review, we propose to synthesize digital decision aids to support decision-making in palliative and end-of-life dementia care. Specifically, this review aimed to summarize the following: (1) what are technologies, components, or procedures used in digital decision aids to support people with dementia or their family carers toward the end of life? and (2) what is the effectiveness of these decision aids on decision-making capacity, quality of palliative care, health care use, and related outcomes?

We conducted the systematic review and meta-analysis following the JBI Manual for Evidence Synthesis on conducting systematic reviews of effectiveness [24]. This review is reported using the principles of the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) statement for reporting of systematic review and meta-analysis [25]. The PRISMA checklist was presented in Multimedia Appendix 1. The protocol was registered with PROSPERO (CRD42024621321).

A systematic literature review of 4 health-related databases (PubMed, Embase, CINAHL, and Web of Science) was performed in September 2024. There was no restriction on the search period. The search string was specified as Text Word [internet OR digital OR technology OR technologies OR visual OR web OR video OR audio OR media OR computer OR tablet OR electronic OR telecommunication OR telemedicine OR telephone OR television OR text messaging OR videoconferencing OR mobile phone] AND Text Word [informed choice OR decision-support* OR decision aids OR decision tree OR decision-mak* OR informed decision] AND Title/Abstract [end of life or eol or terminal care or palliative care or advance care planning or hospice care] AND Title/Abstract [memory disorder* or cognition or dementia or Alzheimer or dement* or cogni*]. The search strategy in each dataset is presented in Multimedia Appendix 2. The reference lists of included studies and relevant articles that cited included studies were hand-searched to identify additional eligible studies. After removing the duplicate results, titles and abstracts were screened based on the eligibility criteria independently by 2 reviewers (J Zhong and WL). All full texts were screened independently by 2 reviewers (J Zhong and WL). Any disagreement was resolved through discussion between the 2 reviewers, and a third reviewer (TW) was consulted when needed. Study screening and selection were conducted on Covidence (Veritas Health Innovation Ltd).

Articles were included if they met the following criteria: (1) the focus was on the development and evaluation of a decision support tool on end-of-life dementia care; (2) the decision support tool used technologies (eg, videos, visual aids, and web pages); (3) the decision aid was aimed at older adults, family carers, or health care professionals; (4) the study used an experimental design (eg, randomized controlled trial [RCT], quasi-experimental, and pretest-posttest); and (5) the article was available in full text in English. Articles were excluded if they met any of the following criteria: (1) the decision support tool only provided information in texts without any visual aids or other technologies, (2) the decision was not related to end-of-life dementia care, (3) the full text was not available (eg, conference abstracts), (4) the article was a study protocol without reported results, or (5) the study used a nonexperimental design (eg, cohort study). The decision to focus on the experimental designs was guided by the primary aim to evaluate the effectiveness of digital decision aids, rather than examining their development or theoretical underpinnings.

Informed by the JBI Data Extraction Form for Systematic Reviews and Research Syntheses, 1 reviewer (J Zhong) extracted the data on the author, year, country, study design, sample characteristics, intervention (technology used in digital decision aids, components, or procedures), control, outcome measures, and effect of the intervention. For RCTs, data were obtained for both intervention and control groups. For pretest-posttest studies, data were extracted for the pretest-posttest sample. All data extraction was checked independently by the second reviewer (WL).

Quality of RCTs was evaluated using the Cochrane Collaboration’s Risk of Bias Tool (version 2.0) [26]. Each domain within the tool was assessed to have a low or high risk of bias. The quality of the pretest-posttest studies was evaluated with the Risk of Bias in Non-randomized Studies of Interventions Version 2.0 assessment tool. Each domain was judged to have a low, moderate, serious, or critical risk of bias. When all the domains were rated as having a low risk of bias, the study would receive an overall rating of low risk. If one domain of a study was considered as high risk of bias, the study was classified as having a high risk of bias overall. Two reviewers (J Zhong and WL) independently evaluated the quality of the included studies, and any disagreements were resolved through consultation with a third reviewer (TW) when needed.

Narrative synthesis was conducted based on the coding of the type of technologies (eg, visual aids, videos, and web pages) and outcomes (eg, preferred goal of care). Meta-analyses were conducted based on the extracted data in ReviewManager software (RevMan version 5.0) from the Cochrane Collaboration. Heterogeneity across studies was tested with I² values, which ranged from 0% to 100%, with higher values indicating greater heterogeneity. When I²≥50%, which demonstrates substantial heterogeneity (P<.05), a random effects model was used; otherwise, a fixed effects model was used [27]. A meta-analysis was performed to determine the pooled effect of the intervention using standardized mean differences for continuous outcomes and odds ratio (OR) for dichotomous outcomes with 95% CIs when at least 2 same designed studies assessed the similar outcome.

The searches yielded 1274 records. After removing 316 duplicates, 958 records were screened at the title and abstract stage, and 49 records were screened with full texts. With an additional 5 records from citation searching and reference reviewing, a total of 20 articles were included in the systematic review (refer to the PRISMA flow diagram in Figure 1). This review included 10 studies using data from RCTs [28-37] and 10 pretest-posttest or descriptive pilot studies [11,19,38-45]. Most studies were conducted in high-income countries, including the United States (17/20, 85%, including 3 secondary analyses of a cluster RCT, a pretest-posttest pilot study and an RCT testing the “My & My Wishes,” a pretest-posttest pilot study and an RCT testing a video decision aid, 4 studies testing a video decision aid across different settings, and 2 studies testing an interactive website), Australia (1/20, 5%), and the Netherlands (1/20, 5%), except for 1 in Taiwan (1/20, 5%).

Data extraction of the included studies is presented in Table 1. Most decision aids focused on the decision about the goal of care (life-prolonging care, limited care, and comfort care) in end-of-life dementia care, except for 1 visual aid focusing on 2 decisions—feeding tube placement and drug treatment for dementia [11]. Most decision aids targeted both patients and their family carers [28,34,36,38-45]. Some decision aids also included long-term care staff, case managers, or clinicians in the shared decision-making meetings [34,39,44,45].

The results of the quality appraisal are presented in Multimedia Appendix 3 [11,19,28-45]. Allocation concealment was rated as unclear in 1 RCT [32]. Three RCTs were rated as having a high risk of bias for blinding of participants and personnel, as the interviewer was involved in delivering part of the intervention (eg, providing a verbal narrative) [34,36,37]. Blinding of outcome assessment was rated as high risk in 2 RCTs, as they reported that the data assessor was not blinded to the group allocation [36,37]. Overall, 10 pretest-posttest pilot studies were evaluated using the Risk of Bias in Non-randomized Studies of Interventions Version 2.0. Three studies were rated as moderate risk during the selection of participants into the study because they either recruited participants from a single center or did not report the procedure of participant recruitment with a small sample size [38,39,44]. Three studies were rated as unclear bias regarding deviations from intended interventions because they did not report on the participants’ adherence to the intervention [38,39,44].

Visual aids, videos, web pages, and telehealth were applied in the identified digital decision aids to support decision-making for palliative and end-of-life dementia care.

Due to the difference in resident- or family-rated outcome measures and study design, only 4 RCTs were included in the meta-analyses on 2 outcomes (2 RCTs for each outcome). Results on the preferred goal of care (comfort care) from 2 RCTs were extracted and used in the meta-analysis [36,37]. There was a low degree of heterogeneity (I²=0%; P=.34), likely because both studies were conducted by Volandes et al [36,37] in similar settings. In the sample with 102 participants in the intervention and 112 in the control group, the meta-analysis suggested that the video decision aid was effective in increasing the proportion of participants opting for comfort care (OR 3.81, 95% CI 1.92-7.56) compared to the control group (Figure 2). In addition, results on the recorded DNH orders from 2 RCTs were extracted and used in the meta-analysis [28,30]. There was a high degree of heterogeneity (I²=88%; P=.004); therefore, the random effects model was applied. This high heterogeneity might come from the great difference in sample size across these 2 RCTs. In the sample with 2507 participants in the intervention and 3909 participants in the control group, the meta-analysis suggested that the video decision aid was ineffective in the proportion of recorded DNH orders (OR 1.60, 95% CI 0.70-3.67) compared to the control group (Figure 3).

The review highlighted the use of technologies to support the development and delivery of decision aids on the internet for palliative and end-of-life dementia care. These digital decision aids mainly focused on the decision about end-of-life GOC, including 3 levels of life-prolonging care, limited care, and comfort care. Narrative synthesis demonstrated that digital decision aids were effective in improving the probability of opting for comfort care [19,36,37]; concordance with the primary GOC between patients, families, and clinicians [28,34]; the quality of end-of-life communication and palliative care [28,40]; knowledge of advanced dementia [37,40]; documentation of GOC communication [35]; and decision-making skills [11], as well as decreasing decisional conflict [42]. Meta-analyses supported that video decision aids were effective in increasing the proportion of opting for comfort care but inconclusive for the proportion of DNH orders. However, these results need to be interpreted with caution as only 2 RCTs were included for meta-analysis of each outcome. Pilot studies examining the feasibility and acceptability of decision aids showed that most participants found these decision aids relevant to their needs, easy to use, and were able to complete the intervention sessions [38-41,43-45]. These findings suggested that decision aids were a feasible and acceptable approach to support decision-making in end-of-life dementia care. However, given the variety of outcome measures used across studies, more consistent outcomes need to be evaluated in large-scale RCTs to provide more robust evidence on the effectiveness of decision aids in people with dementia.

Digital decision aids could be used in routine care to facilitate the shared decision-making process between people with dementia, family carers, and health care providers. The use of technologies in the development and delivery of decision aids offers an opportunity to incorporate flexible applications (eg, texts, animations, images, audios, videos, games, and social networking tools) to provide health information and inform decision-making [47]. Previous systematic evidence suggested that technologies have been inadequately applied in palliative and end-of-life dementia care [48]. Compared to paper-based decision aids, modern and advanced technologies are unique in that they not only deliver the complex information required for decision-making (such as visual aids and videos) but also integrate personalized preferences and clinical conditions into the ultimate decisions (such as web pages and applications). Four formats of technologies, including visual aids, videos, web pages, and telehealth, were developed in the identified decision aids. External visual support (eg, graphs, pictures, and videos) was helpful in visualizing hypothesized situations for the patient and their carer to make the real decision [11,37]. Web pages were designed to involve dementia care networks and enable the patient to give their individual opinions about dementia-related topics and individual circumstances through specific functions and modules [39,43,44]. Furthermore, telehealth can mitigate barriers related to time scheduling and geographic constraints, allowing patients and family carers to engage in health care decision-making conversations from their homes [41]. These technology-based decision aids are supposed to support the development and delivery of decision support on the internet. However, most decision aids are not yet available on the internet, except for the PREPARE website [46]. To benefit the maximum number of people, researchers and developers need to consider publishing their decision support materials (such as videos, modules, and websites) on the web and advocate for the significance of end-of-life conversations. In addition, implementation studies are needed to generate implementation strategies to facilitate the uptake and integration of digital decision aids in clinical practice for quality palliative and end-of-life dementia care.

Decision-making to administer, withhold, or withdraw life-saving treatment at end-of-life needs to be a step-by-step communication process in which patient autonomy must be respected [49]. Decision aids are feasible for patients and their family carers due to the length of time needed to learn the knowledge and the incorporation of personal values and preferences [23]. Many decision aids were designed specifically for patients with advanced dementia and their family carers in long-term care settings (eg, nursing homes and residential care homes). Considering the gradual loss of communicative and cognitive abilities for decision-making, the engagement of family carers is necessary to represent personal and family values. However, earlier identification of dementia as a terminal illness would allow cognitively capable patients to be more actively involved in making decisions about their future end-of-life care. Therefore, earlier implementation of decision aids needs to be developed and delivered to patients with early dementia. In addition, the summary of the decision and related information could be shared with clinicians to implement personal decisions in the clinical treatment based on the patient’s clinical condition. Practically, a follow-up care plan meeting with clinicians after reviewing the decision aid is needed to share the decision. Furthermore, decisional coaching has not been applied in the identified decision aids. It is important to note that sensitive medical decisions around end-of-life care need to be facilitated by professional guidance. A decision coach could facilitate the decision-making process with the decision aid and eventually encourage the decision maker to share their decisions and preferences with significant others and health care providers [50]. Decisional coaching could be applied in the future development of end-of-life decision aids.

This systematic review was conducted with methodological rigor. We conducted a thorough search and independent full-text screening and quality appraisal. We also included both RCTs and pilot studies to comprehensively explore digital decision aids. Therefore, the narrative synthesis allowed for the comprehensive inclusion of various technologies and various outcome measures. However, due to the differences in outcome measures and study design, only 4 RCTs testing video decision aids were used in the meta-analyses. Therefore, there was a lack of power to support a conclusive finding regarding the effect of video decision aids. Large-scale RCTs are needed to evaluate the effect of these decision aids on GOC concordance, quality of palliative care, decision-making performance, and health care use outcomes. In addition, due to the lack of comprehensive translation capabilities for non-English publications, the restriction to English-language studies is acknowledged as a limitation. Furthermore, while the included studies evaluated the effectiveness of digital decision aids in dementia, the challenges and related strategies to implement these tools in real-world settings have not been explored and synthesized. Future studies need to address contextual factors such as clinician workflows, health literacy, and cost-effectiveness to better introduce and implement these tools to the public.

Digital decision aids offer a feasible and acceptable approach to support the shared decision-making between patients, families, and clinicians. The included studies reported various outcome measures in preferred GOC, quality of palliative care, decision-making performance, and health care use. Although current evidence supported the benefits of digital decision aids in the proportion of opting for comfort care, more large-scale RCTs are needed to include consistent outcome measures to evaluate the effect of digital decision aids for palliative and end-of-life dementia care. Future development and implementation of digital decision aids needs to engage patients with early dementia and family carers in the community rather than only focusing on patients with advanced dementia residing in long-term care settings. Finally, a summary of the end-of-life decision can be shared with clinicians who provide direct caring services to patients with dementia or family carers to facilitate the shared decision-making process for palliative dementia care.