Dementia is a progressive, complex neurodegenerative condition with a multitude of types and clinical presentations. It is the leading cause of death in the United Kingdom [1] and is projected to have the highest global proportional rise (264%) in suffering associated with a need for palliative care [2]. Dementia is characterized by a complexity of care needs, which advance as the condition progresses [3]. These needs span multiple domains of health care, and multimorbidity is common [4]. Symptoms such as pain and breathlessness [5,6] cause significant distress for the person and their caregiver and increase toward the end of life [7].

Over half of the people with dementia (58%) die in care homes in England [8]. They are the main providers of end-of-life dementia care, with the average life expectancy on admission for a resident with dementia being 1 to 2 years [9]. The term care home in the United Kingdom refers to both residential and nursing homes. These differ with regard to the provision of input from health care professionals, with nursing homes providing additional access to 24-hour on-site nursing care. Care homes require the resources to deliver multidisciplinary care to meet these advancing and acute dementia-specific needs [10]. Providing access to good quality, continuous care throughout the dementia trajectory is essential. This can be achieved by integrating care homes with primary care, palliative care, and dementia care teams to enable multidisciplinary and specialist input on vital care processes [11].

Care needs change and develop over time and cause considerable distress if left unmet [6]. A comprehensive assessment by a multidisciplinary team to review medical, functional, mental, and social abilities is essential for this population with complex needs [12]. Interprofessional collaboration is also required to share clinical expertise and experience, best inform complex clinical decisions concerning treatment for multimorbidities, and deliver appropriate care [13]. Integrating these processes across services is widely acknowledged to improve person-centered treatment outcomes for older people with complex needs [11] and reduce detrimental transitions between settings occurring in the final years of life, such as unplanned hospitalization [14].

To integrate services and deliver continuous and coordinated care, established methods of communication are required to share information between systems about residents’ care needs and outcomes. A way to facilitate this is through the use of eHealth [15], which is defined as “health services and information delivered or enhanced through the internet and related technologies” [16], encompassing an array of interventions that enable care to be delivered remotely.

The COVID-19 pandemic has had major implications for the use of eHealth in care homes. In response, former barriers to the integration of health and social care services in England have been liberalized, with restrictions on information sharing between services relaxed [17]. The pandemic highlighted the lack of systems for efficiently sharing information between services nationally and internationally [18,19]. This compromised the delivery of care to meet the escalating health care needs and residents’ quality of life. Innovation was required to communicate between services for the comprehensive assessment of symptoms and management of care and outcomes. eHealth provides a way to do this [20].

There is little evidence to guide the design and implementation of eHealth resources to manage symptoms and concerns for people with dementia in care homes. The key task at hand is to understand how we can scale-up eHealth interventions to embed them in routine care. Despite positive findings regarding their benefit [21], eHealth interventions are yet to attain widespread implementation in care homes, and the way of effectively achieving this remains unclear. This study aimed to explore factors that influence the implementation of eHealth interventions to support assessment and decision-making regarding care and treatment for people with dementia in care homes.

The review is reported in accordance with the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-analyses; Multimedia Appendix 1) guidelines. A systematic review, using narrative synthesis and thematic analysis, was conducted to identify common facilitators of and barriers to eHealth implementation in care homes. The synthesis followed the guidance of Popay et al [22], which provides a specific direction for reviews concerned with the implementation of interventions. The protocol for this review was registered on PROSPERO (International Prospective Register of Systematic Reviews; registration number CRD42020184587). Before registering this review, a search of PROSPERO was conducted to ensure that no similar reviews were underway. Our final search was performed in the week the review was registered (June 1, 2020). Originally, the scope of the review also included collecting data on intervention effectiveness and key components; however, given the breadth of evidence relating to these outcomes, the findings regarding effectiveness will be reported in a second review.

A total of 4 databases (Embase, PsycINFO, MEDLINE, and CINAHL) were searched for studies published in English from January 2000, with the final search being on April 28, 2020. The year 2000 was chosen as the cutoff year to exclude eHealth that may be outdated in the context of today’s technological advancements. A search strategy was developed with the help of an information support specialist and informed by scoping the literature for the types of eHealth interventions currently in use (see Multimedia Appendix 2 for the full search strategy). A combination of Medical Subject Headings terms and keywords was used to develop a strategy based on the following concepts: dementia AND care homes AND eHealth AND assessment OR decision-making. The search strategy was complemented through reference chaining and citation tracking using Google Scholar, following the initial identification of studies from the database search. The eligibility criteria were developed using the PICO (population, intervention, control, and outcomes) acronym, following recommendations on its suitability and enhanced sensitivity when conducting qualitative systematic reviews [23]. The criteria are outlined in Textbox 1.

Studies identified from the search were exported to Endnote [24]. Duplicates were identified and removed. All titles and abstracts were screened by 1 researcher (JG). Approximately 20% were randomly selected for blinded double screening by 2 independent researchers (TA and EY) as a calibration process to test the application of the eligibility criteria. Once an agreement of 90% was confirmed between the reviewers, the eligibility criteria were applied to all identified studies. This was undertaken to screen for eligibility at the title and abstract screening and again at the full paper review. Discrepancies were resolved through discussion.

The quality of publications was appraised using the Critical Appraisal Skills Programme (CASP) tool [25], which was appropriate for the study design. It was conducted by 1 author (JG) and reviewed independently by 2 authors (CE and ND). Where the design was not amenable to the CASP, alternative tools were used, including the Mixed Methods Appraisal Tool (MMAT) [26] and the Joanna Briggs Institute (JBI) critical appraisal tools [27]. No studies were excluded only on the basis of their appraised quality; rather, it was conducted to help understand and describe the studies.

A standardized data extraction tool was developed in Microsoft Excel and informed by the review questions. Extracted data included study aim, country of origin, design, population of interest, setting, eHealth intervention (type, components, and summary), methods of data collection and analysis, outcomes regarding intervention implementation and conclusions, implications, and limitations. Implementation data were extracted from both the results and discussion sections to capture relevant findings relating to the authors’ observations of why an intervention was or was not effective.

A deductive thematic analytic approach was undertaken, underpinned by the adapted version of the Consolidated Framework for Implementation Research (CFIR) [28]. The original CFIR [29] comprises 5 broad theory-based constructs and 39 subconstructs within these: intervention characteristics, inner setting, outer setting, individual characteristics, and implementation process. It is a comprehensive and practical guide for assessing the potential factors that influence implementation. A recent adaptation of the framework added a sixth construct [28], patient needs, acknowledging the importance of person-centered care in health care interventions and the paucity of attention it frequently receives in implementation frameworks. Given the nature of this review, the adapted CFIR was used.

Preliminary synthesis involved tabulation to organize the findings and compare data across different studies. The data from each study were then mapped onto the framework. Common themes that arose across the studies in line with the subconstructs of the framework were then synthesized to form a narrative regarding facilitating elements of and inhibiting barriers to successful implementation. Where data did not align to a subconstruct, an inductive thematic analytic approach was undertaken to avoid biasing data and identify gaps in the framework when applying it to this context. The themes were then developed to identify additional constructs to adapt the framework to this specific context.

A total of 1055 papers were screened by title and abstract, of which 128 (12.13%) full-text articles were assessed; of the 128 articles, 29 (22.7%) met the eligibility criteria for the review (Figure 1 [30]). The included studies reported 27 unique interventions that aimed to facilitate comprehensive assessment of care home residents or improve decision-making surrounding care and treatment, published between 2000 and 2020. The sample sizes ranged from 5 to 4171 for residents and 6 to 609 for carers. Approximately 31% (9/29) of studies omitted the number of participants.

Of the 29 articles, the included studies comprised 7 (24%) randomized controlled trials, 7 (24%) quasi-experimental studies [31-37], 6 (21%) qualitative studies [38-43], 3 (10%) descriptive studies [44-46], 2 (7%) mixed methods studies [47,48], 2 (7%) cohort studies [49,50], and 2 (7%) cross-sectional studies [51,52].

All studies included adults with dementia, focusing specifically on requirements for people with dementia (11/29, 38%) [31,32,38-40,43,47,49,50,53,54] or within a mixed population (18/29, 62%) [33-37,41,42,44-46,48,51,52,55-59]. The average percentage of residents with dementia was 52.8% in the mixed population studies that reported the proportion (6/29, 21%) [35,37,44,46,55,56]. Approximately 41% (12/29) of studies did not delineate the proportion of residents with dementia [33,34,36,41,42,45,48,51,52,57-59] (see Table 1 for a summary of study characteristics).

The included studies varied in their quality. In general, the CASP criteria for the experimental studies identified consistent reporting of clear and focused aims, with appropriate methodologies to address the research questions. Weaknesses were related to small sample size, evidence of selection and attrition bias, poor description of analysis, and nonblinding. Descriptive studies were generally well-reported (average 86% of MMAT criteria met) and quasi-experimental studies (average 77.7% of JBI criteria met). The weaker study design used mixed methods (average 50% of MMAT criteria met) and cross-sectional studies (average 56.2% of JBI criteria met). Weaknesses pertained to the management of confounding factors and the integration of qualitative and quantitative data. Multimedia Appendix 3 [31-59], details the quality appraisal results for each study using the respective appraisal checklist.

Table 2 details the respective CFIR constructs and subconstructs that were identified as important determinants for implementation. The number of subconstructs identified per study ranged from 1 to 13, with a median of 6. No major differences in implementation requirements were identified between studies with a specific focus on requirements for people with dementia and studies reporting on dementia within a mixed population (Multimedia Appendix 4). Findings for each CFIR construct and respective subconstructs is presented in turn. Constructs that were identified in ≤2 studies are not presented as a narrative because of insufficient data.

This refers to aspects of the eHealth intervention that might affect implementation success in care homes and includes findings relating to intervention complexity, adaptability, and cost.

This refers to the extent to which resident needs are known, prioritized, and pursued by a care home. A total of 3 novel subconstructs were identified inductively.

The outer setting is concerned with external influences on intervention implementation. This was least considered construct across the studies, with the main focus on the ‘External policy and barriers’ subconstruct. External barriers include policies on the medical liability of telemedicine [53], licensing requirements that do not allow physicians to consult across different parts of the country [49], and issues around reimbursement policies [52-54]. External financial support acts as an incentive to circumvent the additional cost barriers to implementation [40,53,54].

The most commonly considered construct across the studies was the inner setting, referring to the internal characteristics of a care home. These focused on the 2 main subconstructs of implementation climate (compatibility and learning climate) and readiness for implementation (leadership, available resources, and access to knowledge).

This pertains to end users’ individual beliefs, knowledge, and attitudes toward eHealth and implementation. The findings focused on 3 main subconstructs.

This concerned reporting on the stages of the implementation process that affect its success. A novel subconstruct was inductively identified to capture data related to engaging end users and other stakeholders.

This review investigated the key facilitators and barriers that influence the implementation of eHealth interventions to support comprehensive assessment and decision-making for people with dementia in care homes. Our findings inform a model of eHealth implementation in long-term care for people with dementia (Figure 2). We identified 4 novel subconstructs (denoted in the figure by an asterisk) required to adapt the CFIR to the care home context and use of eHealth to enhance integrated care for people with dementia. This modification enables the framework to accommodate the context-specific findings identified in this review and enhances its use when applied to implement eHealth in the care home context.

The 3 novel subconstructs—resident experience, clinical benefit, and person-centered care—were identified within the patient needs construct. No subconstructs have previously been delineated here, and the framework was previously not nuanced enough to capture the data and implementation requirements for this population. Clinical benefit was the most commonly identified theme, with the capacity to minimize burdensome transition between care settings as a key facilitator of implementation. This is echoed in the literature that highlights the adverse impact of burdensome care transitions, both on the resident [61] and on costs [62]. Although outcomes were not the focus of this review, eHealth interventions are more likely to be embedded if they can improve the delivery of integrated health care where it is required, enhance integrated care, and improve cost-effectiveness in the National Health Service [63].

Findings around the potential dehumanization of care reiterate a concern raised previously in relation to eHealth implementation [64]. Although the aim of eHealth is to enhance care delivery, entirely omitting in-person contact has clear disadvantages. Striking the right balance between delivering care face to face and virtually is crucial. Future research should focus on delineating the components of health care that must be delivered in person and identifying those that are more amenable to remote delivery to ensure that care quality is not compromised.

The engaging subconstruct of the implementation process was also modified for the purpose of this review. None of the proposed subcategories of engaging (opinion leaders, formally appointed internal implementation leaders, champions, and external change leaders) were suitable for capturing data around engaging end users. The importance of involving end users throughout the implementation process is a key finding and is consistently advocated in the literature to help close the gap between research production and clinical practice [65,66]. Therefore, a novel subcategory titled end users was added to the CFIR. Contradictory findings regarding staff and resident preferences on the design and use of eHealth devices are an important indication that there is no one-size-fits-all approach to eHealth implementation. This inconsistency in preferences highlights the critical need to involve end users during intervention development to accommodate a range of requirements and enhance the likelihood of sustained implementation. Progressing the CFIR to include a subcategory suitable for data pertaining to all stakeholders’ input increases its relevance and applicability for this context.

The most salient construct identified in this review was inner setting of the implementing organization. This is concurrent with the existing research on care homes. A systematic review by Goodman et al [67] identified many of the same factors that influence the readiness of care homes to participate in change. These include ensuring compatibility between the intervention and existing care home routine, providing sufficient training and resources, and engaging care home leaders. Both reviews highlight the significance of the organizational context on implementation success and the need to consider context in the planning stage to inform study design rather than retrospectively. This is contrary to a review of determinants for eHealth implementation with informal caregivers in the community [68], where few factors relating to context were identified as important. This indicates that factors influencing implementation may not be consistent across health care settings and stresses the importance of understanding specific contextual factors and tailoring implementation strategies accordingly.

One of the most prominent determinants of implementation identified in this review was the complexity of the intervention, consistent with previous findings highlighting ease of use as the most important facilitator of eHealth adoption. The nonadoption, abandonment, scale-up, spread, and sustainability framework [69] was developed specifically in response to the finding that eHealth, which is categorized as complex is rarely, if ever, successfully embedded in mainstream care. It aims to help innovators and implementers measure and minimize complexity in eHealth and scale-up and sustain innovation, and therefore, it could helpfully be used in this context.

Using the most salient subconstructs of the framework, a conceptual model was developed to highlight the most important factors that influence the implementation of eHealth interventions to enhance integrated care focusing on care processes of comprehensive assessment and decision-making about care and treatment (Figure 2). The findings can be translated into practical recommendations for organizations aiming to embed eHealth within long-term care settings for people with dementia (Textbox 2). The model indicates that for implementation to be successful, eHealth devices must be low cost, simple to use, and tailored to the care home setting and residents. It must be clinically beneficial to the residents, with special consideration of changing, multimorbid dementia-specific needs. Extensive planning and engagement of care home leaders, end users, and champions in the development and implementation process are key to ensuring successful execution. Providing sufficient training and resources to ensure that care home staff feel valued, motivated, and optimistic about a change in practice is crucial to fostering a positive implementation climate.

The CFIR is a comprehensive determinant framework and is chosen to guide analysis owing to its previous application with eHealth interventions [68,70,71]. Generally, the extracted data were amenable to the CFIR, with no data left uncoded. However, 33% (13/39) of the subconstructs had no associated data (Table 2), concurrent with findings from a previous review of eHealth implementation across health care settings [70]. This consistency either suggests that some subconstructs are not relevant to implementing eHealth or highlights a limitation of the evidence base and lack of existing literature for this setting and population. These areas of uncertainty are important for informing future research, which should focus on identifying barriers to and facilitators of the implementation of these underresearched constructs.

A systematic approach to this review allowed for the rigorous and thorough identification of the relevant literature. Evidence synthesis was theory driven and guided by the CFIR, which has been built upon here to increase its relevance in eHealth and long-term care contexts.

Although this review specifies care home residents with dementia as its population of interest, only 38% (11/29) of the included studies had an exclusive dementia population. This was to reflect the real-world heterogeneous populations in care homes and include interventions that were suitable at an organizational level rather than the individual level. Although there was no real difference in factors that influence implementation between dementia-specific and mixed populations, caution must be exercised when extrapolating these findings to homogeneous dementia populations.

Gray literature was not included in this review. This was because, when identified, it provided little data on the factors that influence implementation. This could have excluded relevant data on the eHealth interventions used in care homes. A recent scoping review on telehealth during the COVID-19 pandemic reported a rapid rise in eHealth during the pandemic [72].

To our knowledge, this is the first review to synthesize evidence on the implementation of eHealth interventions focusing specifically on improving assessment and decision-making in care homes for people with dementia. We adapted the CFIR and progressed its applicability for use in this context. We developed a conceptual model to demonstrate the most important factors to consider when designing and implementing an eHealth intervention and translated it into 18 practical recommendations for implementers, innovators, and organizations to implement eHealth for people with dementia in long-term care. Particular focus should be placed on the individual care home setting and on the consideration of resident and end-user needs when developing an implementation strategy for use in this context.