In this systematic review, qualitative studies were summarized using a narrative synthesis [21]. The process followed the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-analyses) framework [22] (checklist [23] in Multimedia Appendix 1). Only studies with ethical approval were included to avoid encouraging eHealth interventions in which study participants may have harmed their physical or mental health. The review was registered on July 25, 2021 (PROSPERO International Prospective Register of Systematic Reviews; CRD42021261844).

Assisted by 2 university librarians, we searched PubMed, PsycINFO, and CINAHL electronic databases in February 2021; we updated the search results in July 2021 (Figure 1). The search terms were (“acceptability” OR “engagement” OR “usability” OR) AND (“digital service” OR “eHealth” OR “mHealth”) AND (“behavior change” OR “physical activity” OR “diet”). A full overview of the search terms is listed in Multimedia Appendix 2. The search was limited to papers published in English between 2016 and 2021, given that the rapidly progressing nature of health-promoting technology [24] likely lowered the relevance of older publications (ie, outdated technology). All identified studies were imported to review management software (Covidence systematic review software, Veritas Health Innovation) that automatically removed duplicates. Three authors (JB, YW, JR) independently screened titles and abstracts to determine whether papers would be included in the second screening phase. Any disagreements were discussed. In the second screening phase, the full texts were independently screened by 2 authors (JB, JR) to determine the final selection of papers.

The selection criteria (Table 1) were based on the PEO (Population, Exposure, Outcome) framework [25]. We included papers describing qualitative or mixed methods studies that presented qualitative data on acceptability, engagement, and the usability of digital behavior change services for the promotion of physical activity, healthy diets, or both, consistent with current World Health Organization guidelines [4,6]. We did not include papers that focused on interventions that targeted sedentary behaviors only.

Data extraction was performed by JB and reviewed by JR in accordance with Guidance on the Conduct of Narrative Synthesis in Systematic Reviews [21]. The procedure included three steps: (1) tabulation to provide detailed information of all studies in a common table (Multimedia Appendix 3), including reference details, study design, population, exposure, outcome, and quality; (2) clustering to organize the findings into groups relevant to the research aim (participants’ subjective opinion clustered as facilitators, barriers, and suggested improvements); and (3) translation to explore similarities and differences between the studies using thematic analysis (ie, identify the most relevant and important themes and concepts across the studies in an inductive manner, hence, without predefined themes to guide the analysis) [21]. The thematic synthesis included 3 steps, described by Thomas et al [26] as “coding of text line-by-line; the development of ‘descriptive themes’; and the generation of ‘analytical themes.’”

Quality appraisal was performed in accordance with the process described in The Swedish National Agency for Medical and Social Evaluation method book [27]. Quality criteria (authors’ affiliated departments, study design, study theory, recruitment, data collection, data analysis methodology, relevance to the study aim, coherence, sample size, and results) were assessed as high, medium, or low.

A total of 6308 papers were identified; 427 duplicates were removed, 5881 titles and abstracts were screened for eligibility, and 110 full-text articles were read. The final sample comprised 23 papers (Figure 1).

Study characteristics are outlined in Multimedia Appendix 3. Papers were published between 2016 and 2021, with the majority published after 2018 (n=18). The studies were undertaken in the United Kingdom [28-33], Italy [34], Belgium [35,36], Finland [37,38], Portugal [39], Sweden [40], Germany [41], the United States [42-46], Australia [47-49], and Canada [50]. Of the included studies, 11 were qualitative studies [32,34,35,37,40,42-44,46,47,50], and 12 were mixed methods studies [28-31,33,36,38,39,41,45,48,49]. Most studies included participants of both sexes [28-41,43,44,46,47,49,50], except for 1 study [48] with men only, and 2 studies [42,45] with women only. One study did not report sex distribution [39]. In total, sex was reported for 417 women, 309 men, and 1 nonbinary individual. In the studies with both men and women, most often, women were overrepresented. In the 23 studies, 769 participants between 18 and 75 years (mean age range 34-62 years) were included. Ten studies included healthy adults [28,31,34-36,38-40,46,49,50], and 13 included adults with obesity or who were overweight [29,30,33,37,42,45,47,48], adults with type 2 diabetes [43], or adult cancer survivors [29,32,44].

More than half of the studies focused on apps [28,29,31,32,34,35,38,40,42,44,45] or websites [36,41] promoting physical activity, while the rest involved diet-promoting apps [30,33,49], or diet- and physical activity–promoting apps [39,43,46-48,50] or websites [37]. The most commonly applied behavior change techniques according to behavior change technique taxonomy [51] were (presented in descending order) feedback and monitoring (category 2) [28-31,33,35-42,44-48], goals and planning (category 1) [29-31,35,36,38,40,44,47,48,50], social support (category 3) [28,31,37,44,47,48], and rewards and threats (category 10) [29,31,33,38,44,48]. The services were tested for up to a 1-year period, with the majority (n=18) being tested for ≤4 months. One study [44] did not report the intervention period’s duration. Of the 23 studies, 20 studies used semistructured interviews [29-41,44-50], 2 studies used think-aloud interviews [34,39], 5 studies used focus group discussions [30,33,42,43,45], and 3 studies used web-based questionnaires [39,41,49].

In all studies, a theoretical framework was used to support the purpose of the study; papers included information on authors with relevant professions, recruitment methods, data collection and analysis, and results: 10 studies were appraised as high quality [32,34,35,37,40,42,46,47,49,50], 11 studies were appraised as medium quality [29-31,36,38,39,41,43-45,48], and 2 studies were appraised as low quality [28,33] (Multimedia Appendix 3).

In this systematic review, we explored adults' perceptions of the acceptability, engagement, and usability of eHealth and mHealth services that focus on physical activity, healthy diets, or both in the primary and secondary prevention of noncommunicable diseases. The results showed that users value eHealth and mHealth services, but considerations need to be taken account to maintain engagement. Users preferred services to be (1) interactive and integrated; (2) varying and multifunctional; (3) easy, pedagogic, and attractive; (4) individualized and customizable; and (5) reliable. By taking these findings into account, we believe that adherence to eHealth and mHealth services could be significantly improved.

Users underlined the need for variation. This user view was supported by Dennison et al [52], who reasoned that new and updated content increases mHealth app users’ motivation and engagement. Users also valued a service that is composed of several behavior change techniques. There was some disagreement about the effectiveness of behavior change techniques and about the number of behavior change techniques that should be employed. A meta-analysis [53] reported that intervention effectiveness increased when more behavior change techniques were included. In contrast, Kelders [54] underscored the importance of matching user and intervention characteristics rather than applying several behavior change techniques. Our findings indicate that users preferred individualized services. Users also valued a straightforward and easy-to-use service, which is aligned with the less-is-more strategy for effective human–computer interactions [55]. A service offering multiple behavior change techniques tailored to users’ preferences and activities can enhance user engagement. Broekhuizen et al [56] confirmed that tailoring to an individual’s needs was beneficial in digital health behavior change interventions. On the other hand, tunneling the content and basing it on presumed stereotypical activities should be avoided as presumed assessments may mislead the tailoring process. Analytical and artificial intelligence–based methodologies that use input from the app or captured by external devices, could improve user individualization without increasing user burden [57]. Our findings also show that users valued an interactive service enhanced with bidirectional communication and indicative support, especially when the behavior change techniques were used for goals and planning (category 1 [51]). Evidence supports this finding, showing that it is crucial with indicative support to set realistic and achievable goals when minding motivation and engagement [58]. Some evidence stated the importance of applying well-established behavior change techniques when designing health-promoting technology, with suggestions to include self-monitoring and goal setting as support for physical activity and dietary behavior changes [59-61]. It is well evidenced that self-monitoring and goal setting appear to enhance the behavior change process and increase the intervention effect [59-61]. This review shows self-monitoring of several health parameters, goal setting, social networking, gamification, and push notifications were valued as behavior change techniques. Social networking, gamification, and push notifications were appreciated to gain support and enhance motivation, although some users found these behavior change techniques to be inappropriate or even annoying when not carefully adjusted to personal schedules, motives, and interests. There were conflicting opinions about social networking and gamification. For instance, some studies reported that social networking and gamification favored usability (ie, the efficacy and satisfaction of the service) [62,63], whereas others felt it was not essential for long-term behavior changes [18,64]. However, this difference in findings reflects users’ individual preferences, which suggests that there is a need to offer a customizable and flexible device to provide a personalized and dynamic service that follows the varying attitudes, values, and schedules of users. Some users expressed privacy concerns when sharing personal data, which could be an issue when tailoring is used. Individualization and anonymity have been discussed as a problem in eHealth elsewhere [38,65]. Finally, our results suggest that time efficiency may be another crucial factor that is particularly challenging in monitoring dietary habits. This view was supported by Peng et al [65], who reported that ease of use and time efficiency was significant for long-term engagement to mHealth apps.

Our results are based on newly published studies, which is a strength given the rapidly progression of eHealth and mHealth. Another strength is that we included studies with healthy adults and adults with a medical history and from a wide age range. Thus, the results can be generalized as users' perceptions may vary with age and the purpose of the service being used. We also included studies with participants of both sexes and from several countries. However, a slightly larger number of women (417 women compared with 309 men) were included in this review, limiting the generalizability to male populations. Studies were from diverse high-income countries strengthening international generalization. Yet, few studies were from low- and middle-income countries which restricts generalization to low- and middle-income countries. Most of the studies included in this review had recruited participants interested in using health-promoting services. This recruitment bias could limit the generalizability of our results because such individuals are likely to have a level of motivation that is higher than that of the general population. Also, we included studies in which individuals used the digital service for free, which may affect the expectations and perceptions of the service compared with the real world, where consumers pay for services. Most studies had an intervention period of only 4 months, which is a limitation because users’ perceptions are likely to change; however, one with a 12-month intervention period did not report any deviating results. All studies were included regardless of assessed quality; however, none of the studies assessed with low quality added anything new or distinctive to the results. Also, this review only included qualitative studies; quantitatively measured aspects were not considered. Researcher bias is a potential limitation in analyzing qualitative data. However, this limitation is less of an issue, because in our review, 2 authors independently analyzed and discussed the findings.

Our findings from the synthesis of studies on the optimization of digital services to promote physical activity and healthy diets represent users’ perceptions of acceptability, engagement, and usability and show that eHealth and mHealth services provide value but need to be tailored to make them personalized, dynamic, easily manageable, and reliable. These findings can be useful in improving the user value when receiving support by digital services for behavior change to promote healthy lifestyles and increase adherence to eHealth services.