The full search yielded 2667 de-duplicated publications, 175 full-text publications were assessed for eligibility, and the reference lists of 62 publications were searched for additional relevant literature, which identified five additional publications. Sixty-seven publications were included in the final qualitative synthesis. Using the MMAT-version 2018, 56 of 67 studies were rated 100%, and 11 studies were rated 80%. The study exclusion process can be found in Fig. 1, and details of study characteristics and findings can be found in Table 1, found below, and Table 1 in Additional file 3.

PRISMA diagram illustrates the process used to identify eligible studies

The 67 included studies were published between 1996 and 2021; more than half were published between 2016 and 2021 (40/67; 59.7%). These studies reported 58 unique interventions, which were classified into one of eight types of interventions for informal caregivers of PwD based on the most prominent intervention components. This stratification was performed to examine the implementation characteristics of EBIs with clear commonalities to enhance the review’s usability. Multicomponent interventions (e.g., the combined use of case management, support groups, and eHealth tools) (18/67; 26.9%) [84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101] were most common, followed by eHealth (15/67; 22.3%) [40,41,42,43,44,45,46,47,48,49,50,51,52,53,54], psychoeducation (12/67; 17.9%) [60,61,62,63,64,65,66,67,68,69,70,71], care coordination and case management (6/67; 8.9%) [75,76,77,78,79,80], support interventions (5/67; 7.4%) [102,103,104,105,106], respite care (5/67; 7.4%) [55,56,57,58,59] exercise (3/67; 4.4%) [72,73,74], and occupational therapy (3/67; 4.4%) [81,82,83]. Studies originated mostly from the USA (36/67; 53.7%), followed by The Netherlands (11/67; 16.4%), the UK (9/67; 13.4%), Australia (4/67; 5.9%), Portugal (2/67; 2.9%), and India, Israel, Poland, Germany, Canada (each n = 1). The most common study designs were pre-posttest studies (38/67; 56.7%), followed by descriptive qualitative studies (20/67; 29.9%) and parallel convergent mixed-methods design (9/67; 13.4%).

Twenty-one articles were explicitly guided by an implementation framework (21/67; 31.34%). Ten unique frameworks were used, including adaptive implementation model [90, 102, 103, 105, 106], multimethod assessment process (MAP)/reflective adaptive process (RAP) [46], reach, efficacy, adoption, implementation, and maintenance (RE-AIM) [83, 98,99,100], Medical Research Council Framework [44, 45, 89], Fixsen and Blasé Implementation Process Model [67, 95], Consolidated Framework for Implementation Research [48], Leontjevas process evaluation model [45, 53], process evaluation model by Reelick and colleagues [74], Lichstein’s treatment implementation model [84], and normalization process theory [88].

Several constructs were frequently included within these frameworks. Intervention characteristics, including quality and validity of evidence, were prevalent considerations made prior to implementation [44, 45, 48, 53, 83, 88,89,90, 98, 100, 102, 103]. All ten frameworks included constructs relating to implementation setting factors, including both internal (e.g., resources) and external (e.g., government policy) to the implementing organization, and the implementation process, including planning, program adoption, implementation execution, and sustainment. Iterative and reflexive monitoring and (re-)evaluating implementation strategies and outcomes were also components of all included frameworks (see Table 2 in Additional file 3 for details).

The barriers and facilitators to implementation were mapped based on the domains (and constructs) of the CFIR, including intervention characteristics, outer setting and inner setting of the implementing organization (e.g., nursing home), characteristics of individuals, and process of implementation, which allowed for systematic examination of the contextual variables.

Intervention characteristics domain presented barriers to implementation, including lack of relative advantage (4/67; 6%), poor adaptability (12/67; 17.9%), and unsuitable design quality and packaging (25/67; 37.3%). New interventions are hindered by high market saturation and are less likely to penetrate organizations due to the presence of similar “usual care” programs [75, 98, 100, 105]. The EBI user’s poor digital literacy hindered use, as did the interventions’ complicated user interface designs, fragmented information, complex language, and unsuitable components that fit poorly with users’ capabilities [40, 47, 53, 54, 75, 98, 100, 105].

The outer setting domain presented barriers to implementation, including patient needs and resources (24/67; 35.8%), such as implementing agencies’ lack of awareness surrounding influential cultural nuances that deter caregivers from seeking external support (e.g., filial piety) [92, 105], and caregivers’ personal circumstances, including insufficient personal finances, time constraints, poor digital literacy, and adequate information to confidently participate [41, 55, 59, 74, 89, 92, 106]. Additionally, an intervention is less likely to be positively received if introduced to caregivers at an inappropriate stage. For instance, introducing occupational therapy to caregivers immediately following a PwD’s dementia diagnosis creates confusion; alternatively, engaging caregivers in a support program at a later stage in the care trajectory will be less effective since they need communication training and decision-making guidance beginning in early stages [61, 62].

Barriers to implementation under external policy and incentives (15/67; 22.4%) include lack of care coordination and continuity within less developed health systems [77,