We conducted a systematic review of randomized controlled trials (RCTs), quasi-experimental designs (pre-post studies), quantitative studies, and qualitative studies, following the guidelines from the Centre for Reviews and Dissemination [18].

We included articles if they met the following criteria: (1) the intervention was aimed at informal caregivers (defined as a family member or friend providing unpaid care) of people with dementia, (2) the intervention was a digital intervention delivered via the internet, and (3) the article considered a specific intervention and provided a description of this.

We excluded articles if (1) the intervention was telephone or telehealth based, (2) the interventions solely used Skype or another means of online calling, (3) the intervention had a large face-to-face component, (4) results or outcomes of the intervention were not reported, (5) the intervention was focused on the person with dementia, or (6) the study was not published in a peer-reviewed journal.

As our interest was in digital technologies that could be used by caregivers without input from health professionals, we excluded telephone-based support and those interventions with a large face-to-face component.

We conducted a systematic literature search in CINAHL, the Cochrane Library, EMBASE, MEDLINE, PsycINFO, and Web of Science for articles published between January 1990 and April 2018. We selected 1990, as this was the period when the internet, including email, started to develop in commercial and public settings.

We tracked citations using Google Scholar, and hand searched reference lists for any additional relevant articles, in addition to hand searches of relevant journals. We identified literature reviews on the topic and checked them to ensure that our search identified relevant articles. Search terms and index terms (Medical Subject Headings) were identified from the initial scope of the literature. We added synonyms or abbreviations that we felt were appropriate to the search terms. Multimedia Appendix 1 shows an example search strategy from MEDLINE. We contacted experts in the field. We also included gray literature, including reviews of websites, in the scoping work but not in the review.

Article titles and abstracts were screened and excluded if they did not meet the inclusion criteria by 2 reviewers (JH and ND). We rapidly appraised non-English language articles, using their English abstracts, to ensure that we did not exclude any important articles. Articles considered relevant or where insufficient information was supplied in the abstract and title were read in full by 2 reviewers (JH and ND). Two reviewers enhanced the validity and reliability of the selection procedure [18]. Any disagreement between reviewers or uncertainty about inclusion of articles would have been decided by a third reviewer, although this was not required. Figure 1 shows a Preferred Reporting Items for Systematic Reviews and Meta-Analyses flowchart detailing the selection process.

We appraised the literature for quality using the Critical Appraisal Skills Programme tools, using different tools for varying study design, qualitative studies [19], and RCTs [20], and an adapted version of the Critical Appraisal Skills Programme toolkit for quantitative designs [21]. We excluded no studies based on the results of their quality appraisal; rather, we used the study appraisal to develop discussion of the included studies.

We developed a standardized data extraction tool to examine the included studies. We extracted data on study design, intervention details and components, duration of the intervention, participant characteristics, outcome measures, key findings, and the conclusions drawn by study authors. As the review included both quantitative and qualitative studies, and quantitative designs were heterogeneous, we could not pool quantitative data to conduct a meta-analysis. Therefore, we performed a narrative synthesis, using tabulation to organize the studies and a thematic analysis to categorize and group the studies. Two reviewers independently coded all studies (JH and ND). The 2 reviewers met to discuss each of their coding frames, discuss any disagreements, and develop a refined coding frame. Using the refined coding frame, all studies were coded by 1 reviewer (JH), and a selection of articles (50%) were blindly coded by a second reviewer (ND) and checked for agreement. Any disagreement was discussed and a third researcher would have been consulted if agreement could not be reached, although this was not required.

We included 40 articles [22-61] addressing 31 different interventions. There were 9 RCTs [22-30] (Multimedia Appendix 2), 7 quasi-experimental studies [31-37] (Multimedia Appendix 3), 4 qualitative studies [38-41] (Multimedia Appendix 4), and 20 studies with mixed or other methods [42-61] (Multimedia Appendix 5). All statistics in the multimedia appendices are reported as per the original articles.

All interventions aimed to address the needs of informal caregivers of people with dementia. One intervention also provided support for professional caregivers [26] and 2 provided support for people with dementia [41,49]. Some interventions limited their inclusion population to address specific caregiver needs, including 3 bilingual websites that addressed the needs of caregivers from minority ethnic groups [46,56,61]. Some restricted their intervention to informal caregivers experiencing stress [23,36,47], burden [24,29,57], depression [24,57], or anxiety [24,57].

Most interventions aimed to address the needs of caregivers providing care to people with all stages and types of dementia. One limited the intervention to spousal caregivers of people with mild cognitive impairment or “mild” dementia [44], 1 was limited to caregivers of people who were housebound with dementia [25], and 2 interventions also included people caring for those with other forms of neurodegenerative disease [36,39,40,52].

Although all interventions were primarily internet based, some had supplementary telephone-delivered components; for example, 1 provided a telephone number on their website for caregivers to contact a health care professional [32], and 1 included monthly telephone calls with caregivers [45].

A large number of different outcome measures were used in the studies. Outcomes included data on the usability of the interventions, as well as impacts on well-being, quality of life, burden, competence, physical health, and mental health. A wide range of validated and nonvalidated rating scales were used to assess the impact of the interventions. Qualitative results mainly focused on usability of the interventions and included observation of use [39-41,44,49], free-text surveys [38,41,49,51,55,61], interviews [38-41,46,47,47,53,54,59,60], and focus groups [42,45,56].

The articles were of variable quality. Sample sizes ranged from 5 to 299, with many studies being pilot or feasibility studies. A problem with possible selection, performance, detection, or attrition bias was identified in many of the studies; many studies had small convenience samples, high attrition rates, and poor descriptions of randomization in trials and of data collection; and in some cases data collection was completed by nonblinded members of the study team.

We identified 5 themes as key components of the interventions: peer support; contact with a health or social care provider; provision of information; decision support; and psychological support.

Unlike previous reviews in this area, this review explored the key components of internet-based interventions to support family caregivers of people with dementia. We identified a broad variety of interventions, which focused on providing peer support, engaging with health and social care professionals, and providing information, decision support, and psychological support. Although effectiveness was not a focus of this review, some multiple-component interventions showed promise in reducing stress, anxiety, and depressive symptoms for family caregivers and in increasing self-efficacy [44,57]. However, as with previous reviews [15-17], the limited number of high-quality RCTs, as well as the multiple-component nature of many interventions, makes it difficult to report which aspects of the interventions were effective.

Peer support was a key component of many of the interventions discussed. Caring for someone with dementia has often been described as not only a lonely role but also one in which there is a great deal of uncertainty. The peer support components of the interventions identified in this review aimed to target these feelings and were described positively by many participants, but no significant effect for peer interaction and social isolation was found [45]. However, qualitative data in this review suggest that peer support offered a form of socialization. Previous evidence is mixed on whether use of the internet reduces or enhances loneliness [63,64], but this review suggests that internet-based peer interaction may have a benefit for family caregivers. However, it is evident that the way that peer support is delivered is important, with opportunities for group interaction or videoconferencing being more beneficial than public-facing forums and private messaging functions.

The qualitative data suggest that interactions with health professionals are viewed positively; however, it is unclear whether this positivity was linked to the provision of contact online or whether caregivers may prefer this interaction face-to-face. The mix of professionals providing support in the studies suggests there is a lack of consensus on who is best to deliver professional support. This may reflect ambiguity caregivers feel about who is the most appropriate person to talk to when they need advice.

The provision of information was often at the core of interventions, and this supports findings from previous research where most caregivers preferred to receive information online rather than in paper format [65].

This review demonstrated that interventions that focused solely on decision making were, in general, not favored by family caregivers. However, decision-making tools were viewed more positively when they were used alongside other components, such as peer support. Decision making is often left to family caregivers when the person with dementia no longer has capacity, making this a difficult and challenging time for family caregivers. However, results from this review suggest that face-to-face meetings may be required to make decisions, and internet resources are only used as a method of preparation for discussions. This adds to our understanding of barriers to making decisions, which include a lack of information, poor communication, difficult dynamics and conflict within families, and limited emotional and practical support [66-70].

Interventions including online psychological support showed some of the most promising findings, with individual studies reporting significant reductions in caregiver stress, strain, depressive symptoms, and anxiety, in addition to increases in self-efficacy [23,26,28]. Although studies of both professionally guided and self-guided interventions indicated a positive outcome for participants, including caregiver mental health outcomes, they were quasi-experimental (pre-post studies), feasibility studies, and small RCTs, suggesting these conclusions should be made with caution.

Qualitative evaluations of the interventions demonstrated positive views from most caregivers toward internet-based support interventions, although it is clear that not all would benefit from such interventions. It may be that the internet is most beneficial for those who are classified as most vulnerable (ie, more stressed) [22].

In developing an internet intervention for family caregivers, several issues need to be addressed. Questions of privacy and security were highlighted [59,60], reflected in the contrast of public versus private messaging approaches and password-protected websites. The details discussed by many on the websites are very personal and emotional topics. Sillence and colleagues discussed a series of factors that influence the mistrust and trust of health websites [71]. The design of the site contributed to most of the reasons for rejecting and mistrusting a website, including complex and busy layout, corporate look, and irrelevant content. However, the reasons for selecting and trusting a website were more focused on the content of the website, including unbiased information and personalized content.

Another issue is complexity. Some caregivers found functions such as private messaging, decision aids, and login screens complex, which affected their use. Using familiar-sounding language [71] and a strong iterative approach, in which the intervention undergoes multiple cycles of development and optimization [72], with future interventions are two ways to help overcome this challenge Tailoring can reduce the quantity of information and resources caregivers must review, and caregivers may be more motivated to use an intervention they feel is applicable to their circumstances. This review found that where interventions were not personalized, caregivers found this frustrating and their needs were not met [47]. Finally, there is the question of internet literacy and access to the internet: the digital divide [73]. There appears to still be a gap between those who use or can use the internet and those who don’t, with a study in 2015 highlighting that almost all adults over 70 years of age had difficulty using the intervention [41]. Many of the studies included in this review consisted of participants who were predominantly younger caregivers, whereas many people caring for someone with dementia are more likely to be older. Reducing the complexity of interventions, supporting access with potential support from health professionals, and highlighting the benefits of such interventions to understand their potential value may aid in bridging the divide. For most of the studies, this digital divide was ignored, as a requirement for participation was computer literacy [61], and observational studies assessed the usability of the interventions with caregivers who had already received training in using the website.

This review demonstrated the need for high-quality research to evaluate the effectiveness of internet-based interventions for caregivers of people with dementia, in particular larger phase 3 trials. Importantly within these studies, it would be useful to describe the interventions in more detail and to understand which aspects of the interventions are used more than others and provide the most benefit. Future research should also focus on which aspects of the interventions are most beneficial for different groups—for example, adult children compared with spouses—and how the interventions can best be delivered to address issues such as the digital divide. Future research would also benefit from including theoretical considerations of how interventions are thought to provide support to caregivers.

This review identified a gap in the development of interventions targeting specific stages of the dementia trajectory. Many of the interventions in this review were broad and generic to the entire dementia trajectory. However, the needs of family caregivers vary at different stages of the disease and transition points; for example, around the end of life of the person with dementia, caregivers face specific challenges around decision making and management of difficult symptoms. Future interventions and research should address these different stages when developing digital interventions to support family caregivers [49].

Similarly to previous reviews in this area [15-17], comparison between studies was difficult, as the interventions used were complex and varied, with wide-ranging study designs and outcome measures. The review was also limited by the quality of some of the studies and the methods employed. There were relatively few RCTs from which to derive effectiveness data. Many of the studies were feasibility and pilot studies, so we were unable to draw definitive conclusions surrounding effectiveness and acceptability. For many of the studies, there were high levels of dropouts and for some interventions participants made limited use of some of the components of the interventions, therefore making it difficult to draw conclusions [46,49]. Few studies provided information on the effectiveness of individual components of the interventions, and some studies explored only usefulness and usability with reference to the design and layout of the interventions, which on the whole were not well described. This is helpful only to an extent because, to develop or build on existing interventions, there needs to be an understanding of which elements have a positive effect on family caregivers and so should be included in new interventions.

Our literature search was limited by including only peer-reviewed publications, and there may have been several other interventions that were being practically used and applied but not published via academic routes. However, the search of the academic literature was thorough and we used a rigorous search strategy, updated before publication.

This review has built on previous literature by identifying the core components of interventions for family caregivers, which will be useful for future intervention development. As our inclusion criteria were much more comprehensive, this review provides a larger evidence base than previous reviews. Unlike previous reviews, we have particularly considered how caregivers are supported with decision making through Web-based interventions and we included data from many qualitative studies, providing richer information on how the interventions were perceived and valued by caregivers.

The evidence base for internet-based interventions for caregivers of people with dementia remains limited. Although this review recognizes that for some caregivers, a face-to-face intervention may be preferred, our findings highlight the promising potential of digital interventions to support caregivers, which warrants further development and testing.