The protocol for this systematic review was registered a priori on PROSPERO under the registration ID number CRD42024628168 in December 2024. It adhered to the guidelines outlined in the Cochrane Handbook for Systematic Reviews of Interventions [31], and it also followed the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) checklist [32]. The PICO (population, intervention, comparison, outcome) strategy was used to formulate the research question and search strategy [33].

This study aimed to address the following research questions (RQs):

The primary objective of this review is to identify, synthesize, and categorize the factors that influence adherence to DHTs, including both barriers and facilitators across diverse populations and settings. In addition, the review aims to examine the theoretical foundations and tools that have been applied to predict adherence. By integrating these elements, the review seeks to provide a comprehensive understanding of how adherence to DHTs can be improved and inform the development of more targeted interventions and more accurate adherence models, highlighting existing gaps in the current evidence.

In December 2024, data were extracted from 4 different electronic databases, namely PubMed, PsycINFO, Scopus, and IEEE Xplore. The search strategies combined controlled vocabulary (MeSH [Medical Subject Headings] terms in PubMed) and free-text keywords related to (1) models, frameworks, and determinants (eg, tool, measure, questionnaire, survey, instrument, scale, theoretical model, framework, conceptual model, determinant, predictor, barrier, and facilitator); (2) adherence and related terms (eg, adherence, compliance, persistence, nonadherence, and dropout); and (3) DHTs (eg, digital health, eHealth, mHealth, telemedicine, and health apps). Boolean operators (AND, OR) and truncation were used. The full search strings for each database, including applied syntax, are provided in Multimedia Appendix 1. Filters were applied in all databases to include only studies published in the last 5 years and written in English, Portuguese, or Spanish. Additionally, the search was restricted to studies involving humans. The search was not updated after December 2024.

After duplicate removal, all extracted studies were independently assessed by 2 reviewers, TF and LA, following a 3-step process: title screening, abstract screening, and full-text review. The screening process was conducted using the Rayyan platform [34], which allows blinded, independent screening. Disagreements between reviewers were resolved through discussion and, when necessary, with input from a third reviewer (EC). The agreement between the 2 independent reviewers was assessed using Cohen Kappa (κ) coefficient. Decisions to include or exclude studies were coded as binary (include=1, exclude=−1). The Kappa value was calculated for the 3 screening phases. In the title phase, the resulting value was 0.22, indicating only fair agreement according to Landis and Koch’s [35] classification. In the abstract screening phase, the κ value was 0.62, indicating substantial agreement between reviewers, and the full-text phase was 0.42, indicating moderate agreement.

When assessing the titles, studies that met one or more of the exclusion criteria listed below were not considered:

During the abstract and full-text analysis, the selection was based on the criteria presented in Textbox 1.

The methodological quality of the included studies was independently evaluated by 2 reviewers (TF and LA) using the Joanna Briggs Institute (JBI) Critical Appraisal Tools, with specific validated checklists applied according to each study design [36]. The JBI Critical Appraisal Tools were selected for this review, as they are designed to be used across a variety of study types within systematic reviews, enabling consistent quality appraisal, and are widely adopted in health research [37].

Each study was assessed using the appropriate JBI checklist, with items scored as 1 (“Yes”) or 0 (“No,” “Unclear,” or “Not applicable”). The overall quality score was calculated as the proportion of “Yes” responses relative to the total number of items. Based on these scores, studies were classified as high quality (≥75%), moderate quality (50%‐74%), or low quality (<50%). A summary of the final quality assessment scores and categorizations is provided in Multimedia Appendix 2.

Data from the selected studies were extracted by TF and LA using a structured Microsoft Excel spreadsheet, capturing key details such as authors, year of publication, country, study design, adherence or adherence-related concept, validated theories, models, or frameworks, population characteristics, type and purpose of the DHT, factors influencing adherence, and how adherence or adherence-related concept was defined or measured.

Given the inclusion of different study designs, the data extraction process varied accordingly:

These data are summarized in Multimedia Appendix 3.

Factors influencing adherence were initially identified through an inductive thematic analysis [38], allowing themes to emerge naturally from the data without imposing a predefined framework. Both reviewers (TF and LA) were involved in all stages of the analysis. They independently coded the data, identified patterns, and grouped similar factors into descriptive categories and subcategories. Any discrepancies were discussed and resolved through consensus, ensuring the reliability and validity of the categorization process. While the initial coding was data-driven, the development and refinement of the final categories were also informed by theoretical models and frameworks commonly used in the field, such as TAMs and health behavior theories. This combined approach allowed the categories to remain grounded in the data while also aligning with established theoretical constructs, thereby enhancing the clarity and interpretability of the findings.

A meta-analysis was not performed due to the substantial heterogeneity of the included studies. The studies varied significantly in their design (qualitative, quantitative, mixed methods, and reviews), populations (including patients with different conditions, health care professionals, and the general public), types of DHT (including apps, telehealth, and wearable devices), and the adherence-related concepts investigated. Furthermore, measurement approaches differed widely (including using validated scales, usage metrics, or qualitative findings). This conceptual and methodological variability limited the feasibility of statistical synthesis. A systematic mapping of these differences is presented in Multimedia Appendix 3. Therefore, a structured narrative synthesis was conducted to summarize and categorize the findings.

A total of 4031 studies were identified through searches in 4 databases (PubMed, PsycINFO, Scopus, and IEEE Xplore) after removing duplicates. Following the screening process, which applied predefined inclusion and exclusion criteria, 61 studies [6,8,25,29,30,39-94] were deemed eligible and included in this systematic review. The detailed selection process is illustrated in the PRISMA flow diagram (Figure 1).

Table 1 summarizes the characteristics of the included studies, while Multimedia Appendix 3 presents each study individually in detail. The majority of studies were quantitative (n=32; [39,42,44-46,48,49,53,54,58-67,71-75,78,80-82,87,88,90,93]), followed by qualitative studies (n=15; [41,47,51,52,57,68,70,76,77,79,83,85,86,89,94]) and 4 studies [43,50,55,69] used mixed methods approaches. Among the qualitative studies, the most commonly used data collection method was semistructured interviews (n=7; [41,47,51,70,76,77,83]), followed by focus groups (n=2; [52,79]), in-depth interviews (n=2; [86,94]), questionnaires (n=2; [68,85]), pre-post interviews (n=1; [89]), and think-aloud interviews (n=1; [57]) to identify factors predicting adherence. Additionally, this review also included 6 systematic reviews [6,8,29,30,56,84] and 4 scoping reviews [25,40,91,92]. The studies included in this review cover various geographical regions. The majority were conducted in Europe (n=25; [43-45,49-51,53,57,58,62,65-67,69,71,74-77,79,81,83,85,88,90]), followed by North America (n=14;[39,42,48,52,54,59,61,63,70,72,73,78,80,89]), Asia (n=7; [46,47,55,60,86,93,94]), Oceania (n=4; [41,68,82,87]), and Africa (n=1; [64]). Additionally, several studies involve multiple countries worldwide (n=10; [6,8,25,29,30,40,56,84,91,92]).

The studied concepts included adherence (n=20; [6,8,29,30,40,48,51,53,55,57,58,64,68,74,75,79,81,91-93]) and adherence-related concepts, such as engagement (n=11; [39,41,42,44,52,61,63,69,78,82,87]), adoption (n=11;[25,46,47,50,59,60,65,73,76,77,83]), acceptance (n=9; [43,56,70,84-86,88,90,94]), compliance (n=4; [45,49,60,71]), desirability, acceptability, adherence (n=2; [66,67]), intention to continue use (n=1; [54]), continued use (n=1; [89] persistence (n=1; [80], discontinuation of the use (n=1; [72]), and dropout (n=1; [62]).

The majority of the included studies focused on patients or people with a specific medical condition (n=36; [29,39-44,46-49,51,53,55-57,60-67,70-72,75,79,81,83,84,86,93,94]), with patients with diabetes being the most frequently studied group (n=8; [43,44,46,47,61,63,66,67]). Other common populations included patients with cancer (n=6; [29,40,49,51,60,81]), individuals with cardiovascular conditions (n=5; [39,53,55,75,93]), and patients with chronic conditions (n=5; [48,62,71,79,83]). Several studies included adults in general (n=14; [8,52,54,59,68,69,74,80,82,85,88-90,92]) and health care professionals (n=9; [25,50,56,58,73,76,77,86,94]), such as general practitioners, physicians, psychotherapists, psychiatric prescribers, and other health care providers. One study [65] also included caregivers.

A variety of DHTs were examined in the included studies. The majority focused on mHealth apps (n=31; [8,40-52,53,57-60,65,68-71,74,76,77,79,81,87-91]), followed by telehealth solutions (n=7; [54,56,66,67,75,78,80]), text message–based interventions (n=6; [42,43,55,63,64,82]), and web-based programs (n=3; [41,72,84]). Additionally, some studies investigated conversational agents (n=1; [39]) and cell phone–based interventions (n=1; [86]). A few studies explored broader categories of digital health, including general mHealth solutions (n=3; [25,61,92]), DHTs (n=1; [83]), eHealth (n=2; [6,30]), and digital health interventions (n=2; [29,62]). Fewer studies focused on digital devices, such as wearables (n=2; [85,93]), an ingestible microsensor (n=1; [73]), and a digital pillbox (n=1; [94]). The identified DHTs serve various health care purposes that can be categorized into chronic disease management, health promotion, education and disease prevention, mental health and well-being, telemedicine and remote monitoring, and condition-specific digital interventions.

A part of the studies did not report a clear definition or measure of adherence or the adherence-related concept (n=22; [25,29,41,46,47-51,52-55,56,57,59,68,70,75-77,79,83,86,94]). Among those that did, definitions and measurements varied widely. Some studies assessed adherence through frequency and duration of app or system use, while others focused on interaction with digital content, response rates to questionnaires, completion thresholds, or patterns of sustained engagement. Several studies also relied on self-reported data.

The quality appraisal revealed that 25 of the included studies [39,45-47,53,54,56-60,64-67,73,75,76,78,80,81,87,88,90,93] were classified as high quality and 36 [6,8,25,29,30-41,42-49,50,51,52,55,61-63,68-72,74,77,79,82-86,89,91,92,94] as moderate quality (Multimedia Appendix 2). No studies were rated as low quality, indicating an overall satisfactory methodological standard across the body of evidence.

Several theories, models, and frameworks were used across the included studies to explain or support findings related to DHT adherence. The most frequently cited was the UTAUT (n=5; [55,56,76,86,94]), followed by the technology readiness and acceptance model (TRAM; n=2; [46,47]), HBM (n=2; [56,76]), theoretical framework of acceptability (TFA; n=1; [43]), PMT (n=1; [56]), technology readiness (TR; n=1; [56]), SCT (n=1; [56]), TAM (n=1; [56]), TAM2 (n=1; [56]), theory of interpersonal behavior (TIB; n=1; [56]), theory of planned behavior (TPB; n=1; [56]), theory of reasoned action (TRA; n=1; [56]), DOI (n=1; [56]). A specific tool was also developed and validated for assessing patients’ desirability, acceptability, and adherence to telemedicine in diabetes, the QtelemeDiab [66,67].

Sustained adherence to DHTs remains a major challenge, limiting their long-term effectiveness and impact. Research in real-world settings is still limited, especially regarding the factors that influence it. Gaining a better understanding of these factors is crucial to ensuring that DHTs reach their full potential. To address this, this systematic review aimed to identify, analyze, and categorize the factors influencing adherence to DHTs, while also exploring the theoretical foundations and tools applied to predict it. A comprehensive search across 4 databases identified 61 peer-reviewed studies [6,8,25,29,30,39-94] meeting predefined inclusion criteria. Data were extracted and thematically analyzed.

Regarding the factors influencing adherence (RQ1), the findings highlight a complex and multifaceted range of factors, which were categorized into 4 key domains, such as personal factors (sociodemographic characteristics, health status, and user characteristics), technology and intervention content factors (infrastructure and accessibility, user experience and performance, and content and features of the intervention), social and support system factors (family and informal support and health care professional support), and contextual factors.

It is important to take into account that some results, in particular sociodemographic characteristics, were inconsistent across studies, with age, gender, level of education, and ethnicity yielding mixed outcomes on adherence. These inconsistencies may be due to variations in study populations, contexts, types of DHTs, or the way adherence was defined and measured. This also suggests that sociodemographic factors likely interact with other personal, social, contextual, or technological variables, rather than exerting a uniform influence across settings. It is also important to consider the strength of evidence across studies of varying methodological quality. Among the 61 included studies [6,8,25,29,30,39-94], 25 [39,45-47,53,54,56-60,64-67,73,75,76,78,80,81,87,88,90,93] were rated as high quality and 36 [6,8,25,29,30-41,42-49,50,51,52,55,61-63,68-72,74,77,79,82-86,89,91,92,94] as moderate quality. Several factors, such as sociodemographic characteristics, digital literacy, motivation, perceived usefulness, and caregiver support, were consistently supported by high-quality studies, indicating strong evidence for their association with outcomes. In contrast, factors related to the content and specific features of the intervention were mainly reported in isolated or moderate-quality studies, and therefore, these associations should be interpreted with caution.

The findings of this review, together with current literature, indicate strong interactions both within and across domains, reinforcing the need for a systems-level perspective on adherence. For instance, an individual’s digital literacy (a personal factor) can significantly shape how they interact with a DHT’s interface (a technological factor), which in turn influences adherence, and this is particularly relevant in older populations (personal factor) [100]. Similarly, social support from health care professionals can help mitigate low user confidence or reduce perceived complexity of the intervention [51], and intervention features, such as reminders or gamification, can compensate for low intrinsic motivation [101].

Additionally, several theoretical foundations and tools (RQ2) were identified across the studies reviewed to explain adherence or adherence-related concepts. In general, the results indicate that technology acceptance theories, such as the TAM, UTAUT, TRAM, and TAM2, can provide a strong theoretical foundation to examine adherence-related behaviors. The frequent use of UTAUT in the reviewed studies may be attributed to its broader scope compared to TAM, as it offers a more holistic framework for studying acceptance. UTAUT proposes 4 core determinants (performance expectancy, effort expectancy, social influence, and facilitating conditions) [22].

However, some included studies acknowledged that these theoretical foundations remain insufficient on their own. Studies emphasized the need to expand these models by incorporating additional constructs, contextual factors, or previously unexamined external variables to better account for the complexity of adherence in digital health contexts [46,47] or that the use of these models could have potentially missed some findings that did not fit into this predetermined framework [86,94]. In fact, none of these models were originally designed specifically for DHTs; rather, they emerged from organizational and commercial contexts aimed at explaining initial technology uptake, not sustained engagement or long-term adherence in health care settings [22,102]. As such, they often fail to capture the complexity of health care systems, including patient-provider relationships, individual health motivations, and the influence of clinical environments [103,104]. In the same way, the TR index only considers 4 dimensions (innovativeness, optimism, insecurity, and discomfort) that collectively explain technology propensity and usage [105]. Additionally, the DOI, also identified in the included studies, offers a broader perspective on how innovations spread across populations [99]. This highlights the need for more comprehensive and health-specific models to better predict adherence behaviors.

On the other hand, some health-specific models like HBM and PMT mentioned in the included studies incorporate motivational and behavioral components relevant to health contexts [95,96]. However, these tend to be limited in addressing technological usability, personalization, and other factors inherent to digital health interventions. In addition to these, other models and theories related to health behavior were also mentioned in the included studies, such as SCT, TPB, TFA, TRA, and TIB. These frameworks bring important behavioral and cognitive elements, such as self-efficacy, intention, habits, and social norms, and are often used to understand and improve treatment adherence [106].

Only 1 study [67] proposed a specific tool (the QtelemeDiab) for assessing desirability, acceptability, and adherence to telemedicine among patients with diabetes. Although promising, the tool is limited to a specific population and type of technology, and its generalizability remains uncertain.

Notably, despite the relevance of theoretical guidance in understanding adherence, only 10 [43,46,47,55,56,66,67,76,86,94] of the included studies used any theoretical model or tool to explain their findings on adherence-related behaviors, highlighting a significant gap in the current literature. Overall, these findings underscore a critical gap. Most existing models focus on initial acceptance or psychological constructs, failing to fully address the complex, dynamic nature of adherence in digital health. There is a clear need for more integrative, health-specific, and context-sensitive models that combine behavioral, technological, and clinical dimensions. One possible direction for future model development would be to integrate components from both TAMs (such as UTAUT) and health behavior theories (such as HBM), as used in 1 included study [77], ensuring that dimensions identified in this review are systematically incorporated. This systems-level approach would enable more accurate predictions of adherence behaviors across populations, DHT types, and real-world settings.

Additionally, although this review included studies addressing adherence and related concepts such as engagement, adoption, acceptance, continued use, persistence, and dropout, it became evident that these terms were not consistently defined across the studies. Definitions of adherence and adherence-related behaviors varied, as did the methods and thresholds used to measure them. Moreover, several studies did not report any explicit definition of adherence. Among those that did, most conceptualized adherence in terms of intended use, actual use of functionalities as intended, and sustained use over an extended period of time, which goes in line with the definition adopted and defended throughout this systematic review, as explained previously. The persistent existing lack of consensus highlights the need for all researchers to adopt a common and clear definition and standardized measurement approaches for adherence and related behaviors. Indeed, these terms are sometimes used interchangeably in the literature, with one sometimes being used to define another, a concern previously highlighted by other authors [13,107].

This systematic review offers important insights for developers, health care professionals, and policymakers aiming to improve adherence to DHTs. For developers, DHTs should be designed with a user-centered approach that considers the diverse and evolving needs and characteristics of users. This includes incorporating features that enhance usability, personalization, and accessibility while also addressing technical barriers. Previous literature has also highlighted the value of cocreation, particularly co-design, in the development of DHTs [108]. While user-centered design considers users’ needs throughout the process, co-design goes further by actively involving users and stakeholders (eg, family, caregivers, and health care professionals) as partners during all stages of design. This ensures that solutions are grounded in real-life experiences and expectations. Co-design activities should be adapted to the characteristics of the target population and may include techniques, such as personas, user scenarios, ethnography, observations, interviews, and usability testing. Practical strategies, including incorporating features, such as reminders, gamification, or tailored feedback, can support adherence, particularly for users with lower intrinsic motivation or limited digital literacy. Integrating insights from both technology acceptance and health behavior models can better inform the design, deployment, and evaluation of interventions.

For health care professionals, patient adherence can be promoted by actively endorsing DHTs, explaining their benefits, addressing patient concerns, and providing hands-on guidance during initial use. Supportive interactions should consider social and contextual factors. Family or caregiver involvement can improve adherence, but care must be taken to avoid pressure or conflict. For policymakers and health care systems, efforts should focus on enabling environments that support both patients and providers. For instance, providing more robust training and support for health care professionals to enhance their confidence and competence in using DHTs, as well as using incentives for adherence to validated DHTs. This is an already well-discussed topic [109]. Policies should also consider strategies to integrate social support mechanisms safely and effectively into clinical practice. Taken together, these recommendations highlight the importance of a systems-level approach that integrates technological, personal, social, and contextual factors to optimize adherence and ensure that DHTs achieve their full potential in real-world settings.

In summary, future research should aim to establish standardized definitions and metrics for adherence to DHTs to improve consistency and comparability. Furthermore, while this review identified key factors influencing adherence, there is still a need for more research on the interaction between personal, technological, social, and contextual factors. Exploring these relationships in greater depth can inform the development of more comprehensive, integrative predictive models, translating broad findings into targeted strategies for developers, health care professionals, and policymakers to enhance adherence to DHT.

While this review offers valuable insights and is, to our knowledge, the first to systematically examine all factors influencing adherence across diverse populations and types of DHTs, several limitations must be acknowledged. Despite a broad search across 4 databases, the heterogeneity of study designs, populations, and DHTs may limit the generalizability of findings. The exclusion of non-English, Portuguese, and Spanish studies, as well as the restricted time frame, may have led to the omission of relevant studies and introduced potential language bias. Furthermore, restricting the review to peer-reviewed studies and the 3 selected languages may have excluded relevant gray literature and studies from non-English-speaking contexts, which are particularly important in the rapidly evolving field of digital health. This restriction also increases the risk of publication bias, as studies with negative or inconclusive findings are less likely to be published. Although the methodological quality of included studies was appraised using the JBI checklist, no formal assessment of publication or reporting bias was performed. This limits the ability to fully evaluate the presence of selective outcome reporting, thereby affecting the overall transparency and robustness of the findings. In addition, while our focus on statistically significant factors was intended to ensure clinical relevance and methodological rigor, this approach may have contributed to an overrepresentation of positive findings. By excluding nonsignificant findings, which may still hold exploratory value, the synthesis may underrepresent the full spectrum of evidence and inadvertently amplify the perceived strength of certain associations. Additionally, the lack of consistent definitions and standardized measures of adherence across studies complicates comparisons and limits the ability to draw firm conclusions.

Moreover, most included studies were conducted in Europe (n=25; [43-45,49-51,53,57,58,62,65-67,69,71,74-77,79,81,83,85,88,90]) and focused on mHealth apps (n=31; [8,40-52,53,57-60,65,68-71,74,76,77,79,81,87-91]). This may be explained by regional policy initiatives such as the Germany Digital Healthcare Act, which legally enables physicians to prescribe certified digital health applications reimbursed by statutory insurance [110]. Similarly, Belgium’s mhealthBelgium validation pyramid establishes a structured framework for assessing and reimbursing mHealth apps, including several that have achieved top-level funding status [111].

This systematic review offers a comprehensive overview of the factors influencing adherence to DHTs. The findings reveal that adherence is shaped by a complex interplay of determinants, which can be grouped into 4 key categories, such as personal factors, technology and intervention content factors, social and support system factors, and contextual factors. Among these, several stand out as particularly actionable, including digital literacy, perceived usefulness, accessibility, user experience, and social support from caregivers and health care professionals. These factors offer concrete targets for stakeholders. Developers can focus on usability and personalization of DHT, health care providers on guidance and encouragement, and policymakers on improving accessibility and digital literacy. Prioritizing these key determinants can help translate broad insights into targeted strategies to enhance sustained DHT adherence in real-world settings.

Among the theories and models identified, the UTAUT emerged as the most frequently applied. However, these were developed outside the digital health context and lacked specificity to capture key factors influencing long-term adherence to DHTs. The findings underscore the need for more integrative predictive models and frameworks, along with consistent definitions and measurement strategies of adherence, to guide future research and optimize the design and implementation of DHTs.