Asthma and chronic obstructive pulmonary disease (COPD) affect an estimated 8 million people in the United Kingdom [1]. The National Health Service (NHS) spends more than £4.9 billion (US $6.49 billion) treating these conditions annually [2]. Treatment costs and patient numbers are set to grow exponentially in the coming years [3]. Asthma and COPD are distinct conditions, but patients with these conditions experience similar symptoms, including chest discomfort, frequent coughing, and shortness of breath. Current clinical guidelines recommend self-management plans for both conditions [4,5]. The purpose of these treatment regimens is to ensure the best possible quality of life while minimizing the risk of exacerbations [6].

Inhaled medications are recognized to improve quality of life in both asthma and COPD. However, evidence suggests that inhaler adherence is below 50% for both conditions [7,8]. This poor adherence can lead to worsening quality of life and greater burdens on health care systems [9]. Consequently, self-management interventions offer a strategy to provide tailored behavior training, education, support for medication adherence, and symptom monitoring. Self-management interventions are integral to [10] supporting patients with long-term conditions, such as asthma and COPD, to develop an ability to balance lifestyle choices with risk of exacerbations [11]. Improvements can be observed in self-reported quality of life [12,13], reduced emergency visits, and reduced hospital readmissions [14,15].

Adherence to asthma and COPD self-management interventions remains problematic. However, digital platforms provide the potential for professionals to precisely monitor usage and understand how interventions might improve outcomes. For patients, digital interventions can help patients monitor their condition, promote correct medication use, and provide environmental alerts [16,17]. The evidence base for the effectiveness of digital interventions is still at an early stage and needs to demonstrate both improvements in individual health outcomes and infrastructure benefits [18].

Digital health interventions are highlighted by the World Health Organization (WHO) [19] as a mechanism to improve quality, coverage, and equity of health care for all. A total of 96% of the UK population has internet access [20], and approximately 50% use that access for health information [21]. Digital interventions for the management of asthma and COPD have grown in popularity over the past fifteen years [22]. Digitalization has promised much, including supporting structural cost-effectiveness [23,24] and individual personalization [25]. Digital platforms can enable real-time reporting between clinicians and patients while also allowing platform-wide alerts and updates. However, evidence on the effectiveness of digital interventions for asthma and COPD remains unclear.

Systematic reviews and meta-analyses have reported small benefits that are not clearly sustained over longer time periods. Two such reviews each identified 3 studies of questionable quality that produced small and negative effects [26,27]. Meanwhile, narrative reviews suggest frameworks are developed to rigorously assess effectiveness and standardize reporting [28-30]. These reviews also suggest that reporting of different outcome measures makes it difficult to compare interventions. Conceptual models revolve around evidencing aspects of engagement and potential impacts on relevant outcome measures.

Engagement is the process of user investment through interaction with a digital platform [31]. It is dependent on both intervention provision (how and what resources are provided) and users (ability, accessibility, and comprehension), each contributing to engagement itself and what usage data are recorded. Engagement is not restricted to usage, but usage metrics provide a fixed record that can ostensibly be treated as neutral and impartial. Usage has therefore been conceptualized as “objective engagement” [32,33] that enables analysis of engagement and the potential impacts of such engagement. Examples of objective engagement include amount (eg, frequency and duration), breadth (eg, coverage of different aspects of a health condition), and depth (eg, level of detail) of user investment [32,33].

Identifying patterns in usage metadata is foundational to understanding engagement within and between interventions. These patterns offer a rigorous assessment beyond conventional analysis. Rather than comparing nonusers (controls) and users (interventional), another layer is added between nonusers and different types of users (eg, high- and low-intensity users). This shift recognizes the point that while usage is integral to digital interventions, it is not the ultimate aim. Intervention usage is a mechanism to support engagement with target behaviors, which leads to improved health outcomes [34]. Individuals may respond with lower or higher rates of engagement and/or after accessing specific content during critical periods (eg, after an exacerbation). The Analyzing and Measuring Usage and Engagement Data (AMUsED) framework [35] provides a methodology to develop greater understanding of engagement.

Identifying meaningful outcome measures is also essential to assessing effectiveness and engagement. Just as more usage does not necessarily indicate an improved outcome, different outcomes may be influenced by different usage. Furthermore, each user may expect different outcomes from the same intervention. For example, more physically able users may not require exercise features of an intervention but may benefit from personalized alerts. The concept of “effective engagement” attempts to identify patterns of engagement that correlate with a particular outcome, typically represented as behavior change or improved health outcomes [34]. Usage is analyzed in combination with the outcome, identifying optimal patterns within specific interventions. Effective engagement attempts to identify a specific amount of time, specific modules, or a critical event that statistically signals advancement toward a particular outcome. This provides conceptual scaffolding for more rigorous analyses of individual interventions and comparisons between interventions. The concept is demonstrated by Duckworth et al [36], who identified that user reports of increased reliever medications pre-empted reports of a decline in health. These data were used to suggest that digital interventions could prompt users to review and report their health when medication increases are detected [36], for example, that users could be encouraged to recognize and respond to a decline in health.

Health disparities are differences in diagnosis, treatment, and outcome among patient subpopulations that are avoidable, unnecessary, and unjust [37]. These disparities can be the result of systematic, historical, and social injustices [38-41], including ageism, racism, and sexism. Disparities describe risks to patients with specific demographic characteristics, including experiencing symptoms at younger ages and with greater severity [42-46]. Structurally, health care services are used at increased rates, placing greater burdens on health care providers [47,48]. An example within asthma and COPD is the adjustment for “race” in spirometry readings, which sets lower expectations for lung health on the basis of crude categorizations of ethnicity, potentially leading to systematic underdiagnosis and undertreatment [49-52]. To understand how disparities manifest on digital platforms, it is necessary to recognize existing, disease-specific health disparities. Such recognition serves to better monitor how disparities may alter and/or reproduce in digital interventions [39]. For example, results of meta-analyses examining chronic conditions, including asthma and COPD, showed that minoritized ethnic populations benefited from digital interventions, while older adults and women did not [53].

Research on digital health interventions also needs to account for a digital divide. The concept of digital divide describes disparities in access to and engagement with digital platforms [54]. The Office for National Statistics [21] suggests minoritized ethnic groups, older adult groups, and women as more likely to lack internet access. Populations without internet access overlap those vulnerable to asthma and COPD inequalities. However, this does not necessarily mean disparities will be similarly mirrored among those able to engage with digital health interventions. A systematic review investigating engagement with patient-facing digital technologies identified no correlations among a narrow field of demographic characteristics [55]. A real-world study concluded that age, geographical location, and wealth were not barriers to using a digital COPD intervention [56]. It is currently unclear how disparities manifest on digital platforms [29,57-59]. While some suggest needs and resources will drive engagement [55], evidence is required by disease and population.

To summarize, it is important to identify and track subpopulations that are at risk of disease-specific disparities. However, researchers should be encouraged to add complexity and consider more intersectional approaches [60]. Intersectionality is the idea that a person or people are more complex than any single demographic characteristic, and research should account for multiple characteristics. Comprehensive, good-quality data are crucial to achieving this goal. Such data would enable policymakers and digital developers to identify specific vulnerabilities across heterogeneous populations and respond with more tailored strategies [61].

This scoping review aims to explore gaps in knowledge regarding engagement with digital interventions for the self-management of asthma and COPD. This includes a detailed consideration of heterogeneous disease subpopulations for a more comprehensive understanding. To accomplish this, the following objectives were set:

A scoping review methodology was selected because the objective was to describe and categorize the data that studies report and analyze rather than to evaluate intervention effectiveness. This scoping review was not registered. The search strategy followed PRISMA-S (Preferred Reporting Items for Systematic Reviews and Meta-Analyses Literature Search Extension) guidance [62] (see Checklist 1).

To summarize the Population, Concept, and Context (PCC) framework (Table 1) [63], empirical studies were included based on four criteria: (1) participants had a diagnosis of asthma or COPD; (2) the intervention supported self-management via a digital platform and reported usage data; (3) a clinically validated outcome measure was reported; and (4) the study reported any demographic characteristics.

Five academic databases were used, selected based on their relevance to the subject areas and pilot searches. The systematic search was originally conducted in February 2023, after consultation with a librarian, and updated in April 2026. No additional filters were applied (eg, humans, age groups, study design, or publication status, date, language, or publication type). Please see Multimedia Appendix 1 for further details on the search strategy.

The selection of databases and search terms was agreed upon by all authors after being reviewed by a senior librarian at the University of Southampton. The selection was based on authority, efficiency, relevance, and the ability to control the search. This ensured we captured relevant medical and technological journals. Search strategies combined controlled vocabulary (eg, MeSH) with free-text keywords tailored to each database.

One author (MR) and one reviewer completed title and abstract screening. Discrepancies were discussed and resolved remotely with a third reviewer. Full-text screening was conducted by one author (MR), and one reviewer screened 59.5% of records. During full-text screening, articles were reviewed and excluded in 2 phases using criteria developed by all authors. The 2-phase design reflects an approach used by Nouri et al [55] and was systematized by adhering to the AMUsED framework [35]. Phase 1 identified empirical research that reported all 3 types of data (demographic characteristics, clinical outcome measures, and usage). During phase 2, the research team identified and reviewed the range of reported usage measures; 4 comparable measures were selected based on their clinical relevance and objectivity (ie, numerical data without subjective user input).

Supplementary searches of ClinicalTrials.gov and International Standard Randomized Controlled Trial Number (ISRCTN) were undertaken to identify completed studies of digital self-management interventions for asthma and COPD. ClinicalTrials.gov was searched using structured fields (condition: “asthma,” “Chronic Obstructive Pulmonary Disease,” and “COPD”; other terms: “digital,” “mobile,” “smartphone,” “app,” “web,” “telehealth,” “telemedicine,” and “self-management”). ISRCTN was searched using paired keyword combinations of respiratory condition terms with digital-intervention terms due to platform constraints. ISRCTN does not support Boolean nesting or field-specific searching; therefore, paired keyword combinations were required to ensure comprehensive retrieval. Both registries were restricted to completed studies, with no date limits applied. Titles and summaries were screened for relevance (see Multimedia Appendix 1 for full details). Only completed studies were included because the review required empirical reporting of usage, demographic characteristics, and validated outcome measures, which are not available for ongoing trials.

To confirm, online resources or websites, handsearching, citation chasing, and author contact were not undertaken, as these were not part of the planned search methodology for this scoping review. Finally, truncation and wildcard operators (eg, *) were used where appropriate (see Checklist 1).

All data extraction processes were discussed among all authors, with regular updates. Descriptive statistical data were extracted and reviewed first by MR. Table 2 provides a list of data items for extraction. Data for descriptive thematic analysis were extracted by MR; line-by-line coding was performed in NVivo (Lumivero LLC). Descriptive themes were developed in NVivo and through thematic mapping (see Multimedia Appendix 2 for an overview of the process). This thematic mapping aligns with Joanna Briggs Institute (JBI) guidance [63] for collating and summarizing scoping review results.

The National Institutes of Health (NIH) quality assessment tool [64] was used to appraise methodological rigor, focusing on aspects such as sample selection, measurement validity, and risk of bias. Although scoping reviews do not typically require quality appraisal, this additional step provides contextual insight into the methodological strengths and limitations of the included evidence base.

The results section primarily reports descriptive information, such as demographic characteristics, types of outcome measures, and types of usage data. These data respond to the objectives of the study, identifying current reporting practices in research on digital interventions for the management of asthma and COPD. Following JBI guidance for scoping reviews [63], we analyzed the evidence by mapping patterns across usage, outcomes, and demographic characteristics to identify consistencies, contradictions, and areas where evidence was absent. This enabled us to interpret how reporting practices shape the field and where conceptual or methodological gaps persist.

Following JBI guidance [63], we conducted a descriptive thematic analysis of extracted data to collate and map key concepts. Coding followed Thomas and Harden [65] for a structured approach, restricted to semantic, descriptive themes consistent with scoping review methodology. All text after the “Results” section was treated as data and imported into NVivo for analysis (ie, Discussion, Implications, Limitations, and Conclusion sections). After familiarization through multiple readings of the data, codes were categorized into initial descriptive codes; these codes were collated into broader conceptual descriptive themes. These descriptive themes are presented as a narrative in the Results. Co-authors reviewed the themes, including the conceptual descriptive themes generated.

Finally, patient and public involvement (PPI) sessions were organized to discuss themes that developed in the results and how to interpret them. This follows best practice in health research and encourages building community partnerships among populations [66-68]. The PPI participants were members of the Priory Road Group based in Hampshire, England. Members included individuals with a diagnosis of asthma or COPD, those with caregiving experience, or health care professionals. The purpose was to discuss themes, missing variables, and prioritization. To accomplish this, preset matrix scoring activities were used, a visual participatory tool that identifies and prioritizes a range of categories [69]. The preset categories were identified from the results; one activity discussed themes of engagement, and another activity discussed demographic characteristics (see Multimedia Appendix 3). Outcomes of PPI sessions were integrated into the Discussion section to provide context and real-world relevance.

The evidence is reported in adherence to PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Reviews) guidelines [70] (see Checklist 2). The PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) 2020 flow diagram (Figure 1) shows that 2464 unique records were identified through searches of 5 academic databases and 2 registries, with 27 records included in the final analysis. Records were managed in EndNote during screening and deduplication. The software identified duplicates that were then verified by an author (MR). Additional duplicates were identified and checked by an author (MR) and reviewers. The 27 [13,25,56,71-94] included studies investigated 26 unique digital interventions (one intervention was evaluated in 2 separate studies). Eighteen [13,71-73,75-77,82-85,87-93] studies focus on asthma and 9 on COPD [25,56,74,78-81,86,94]. Table 3 provides an overview of study characteristics, focusing on extracted data. The NIH Quality Assessment tool [64] raised few concerns (Multimedia Appendix 4).

All studies [25,56,71-84,86-94] were conducted in the United States or Western Europe except for one study [85] from Malaysia (see Table 3). The studies contained a minimum of 15 and a maximum of 899 participants, with a total of 4019 participants (n=1421 control; n=2598 intervention). Twelve studies [25,71,73,75,76,80,82,86,87,89,91,94] were full RCTs, and 8 [13,74,77,79,81,84,92,93] were feasibility or pilot adaptations of an RCT model. Additionally, 6 [13,25,72,74,85,90] were mixed methods and 5 were real-world studies [72,78,83,85,88]. Nineteen interventions investigated mobile phone apps, and 8 were web-based interventions. This is largely reflected by the year of the publication; the 6 studies [71,73,79,84,87,93] published before 2017 were all web-based.

Full-text screening identified 284 studies that investigated relevant digital health interventions; the majority of these did not report any usage measures (n=206, 72.5%). Four comparable usage measures were identified by adhering to the AMUsED framework [35]: (1) frequency of access, (2) in-app time, (3) number of logins, and/or (4) studies that reported the use of modules (ie, usage of any specific components such as exercise or inhaler technique). Studies that did not report any of these 4 measures were excluded (n=27). An average of 1.63 comparable usage measures were reported by the included studies, with the most frequently reported measure being the number of logins (Figure 2). Most studies reported at least 2 usage measures (mean 2.15, SD 0.74). In addition to the 4 comparable measures, 9 studies [25,56,71,73,75,79,82,91,93] reported at least one other usage measure.

Outcome measures were categorized into three groups to standardize the diversity of validated instruments: (1) condition-specific self-reported measures (eg, Childhood Asthma Control Test [C-ACT] and Asthma Control Questionnaire [ACQ]); (2) clinical and physiological measures (eg, forced expiratory volume [FEV] and oxygen saturation); and (3) other self-reported measures (eg, patient activation measure and wider quality of life measures). Each of these categories contains directly comparable instruments with a full description listed in Table 3.

As outlined in Table 4, 14 (51.9%) studies [25,56,73,75-77,82,83,86,88,90,92-94] reported one type of outcome measure (9 disease-specific self-reported and 5 physiological). Almost half the studies [13,71,72,74,78-81,84,85,87,89,91] (48.1%) reported at least 2 types of outcome measure. Four studies [74,81,84,91] reported all 3 types of outcome measure. Twenty studies [13,72-79,81,82,84,85,87,89-94] (74.1%) reported a disease-specific outcome measure, with one study [82] reporting at baseline only.

As displayed in Figure 3, all studies reported data on 3 characteristics, including age, disease severity, and sex. Comorbidities [72,80] and health literacy [71,74,77,89] were reported in less than 20% of studies. A total of 2587 out of 4022 (66.7%) participants were identified as female. Sex distribution was not explicitly reported for the control groups in 2 studies [25,94]. Thirteen studies (48.1%) reported either a measure of socioeconomic status (SES) or a proxy of SES (ie, education, employment, or income). Ethnicity was reported in 8 studies [13,72,77,85,87,89,92,93] (29.6%), although this was dichotomous in 3 studies [13,77,87] with participants being described as either African American or not, Black or non-Black, and White or other. Across all studies, a total of 705 out of 4022 participants (17.5%) could be identified as belonging to an ethnically underrepresented group. This was concentrated in 3 studies [87,89,91] that accounted for 81.8% (n=577) of such participants. There was a lack of consistency in reporting many characteristics, with most being reported no more than twice.

A total of 10 (37.0%) studies [72,74,76,77,82,83,88,89,91,92] combined different types of data in analysis (demographic, outcome, and usage). One study reported only statistically significant findings. These results are displayed in Figure 4.

Usage and outcome measurements were combined in 6 studies [76,77,83,88,89,91] (22.2%). One categorized usage into levels of usage (eg, high- and low-intensity users) but did not find any significant improvements, although low-intensity users were more likely to complete outcome measures. Three studies [76,77,88] suggested that more usage improved asthma outcomes. Additionally, one study [88] indicated a greater reduction in exacerbations compared to controls but not a greater reduction in health care visits.

Demographic characteristics and usage measures were combined in 6 studies [74,76,82,89,91,92] (22.2%). Significant differences were associated with usage and age (n=1) [82], ethnicity (n=1) [91], digital literacy (n=1) [89], and sex (n=2) [76,92]. Those aged 50 years and older were associated with increased usage. One study [91] reported that usage rates among African Americans were lower compared to other ethnic groups. One study [89] suggested that higher digital literacy correlated with greater usage. Two studies [76,92] reported higher usage among female participants who were more likely to complete interventions, using them more often and for longer periods. One study [89] also noted that low- or nonusage was associated with poor health literacy. Another study reported increased usage when a physician administered the intervention.

Three studies [72,89,91] (11.1%) analyzed demographic characteristics and outcome measures in combination. One [72] suggested improved breathing capacity for those with a range of characteristics, including smokers (vs nonsmokers), males (vs females), and those educated to a high school level (vs college level and above). Additionally, patients with worse asthma control improved their symptoms the most. This study also confirmed no significant differences to ACQ according to ethnicity, comorbidities, education, sex, and smoking. These results were supported by a second study [89] that found no significant differences in outcome and demographics (age, education, ethnicity, and sex). Finally, one study [91] suggested that outcome was moderated by race, but this was caveated by lower engagement among that group.

No studies performed a combined analysis of all 3 data types (demographic, outcome, and usage).

Descriptive thematic analysis yielded 49 codes, 5 descriptive themes, and 3 conceptual descriptive themes (see Figure 5 for an overview). Descriptive themes focused on factors that potentially impacted engagement with digital interventions for asthma and/or COPD; these included (1) features/modules within digital intervention, (2) demographic characteristics, (3) efficacy of the intervention, (4) health care support, and (5) personal motivation. These 5 descriptive themes provided the foundation for 3 overarching conceptual descriptive themes. The following 3 conceptual descriptive themes integrated underlying theoretical frameworks to interpret findings across the included studies.

Developing an understanding of effective engagement could help optimize the design of digital interventions for asthma and COPD. However, this requires being able to identify patterns of engagement through the use of reported measures; for example, comparable usage data and contextually defined outcome measures. This review identified 4 comparable usage measures (frequency of access, in-app time, number of logins, and use of modules) and a range of clinical outcome measures that are commonly reported. Improved reporting of usage data would make quantitative synthesis more feasible. This would require each study to attempt to identify patterns of engagement that lead to effectiveness or to provide datasets with participant-level data. Included studies rarely combined the necessary types of data in analysis that would help identify patterns of engagement. This lack of reporting limited our ability to explore how health inequalities might manifest on digital platforms, despite digital interventions offering a unique opportunity to detect disparities that may not be visible in traditional care pathways. Only 6 studies performed analysis that combined demographic characteristics with outcome or usage data.

Current reporting practices fundamentally limit the ability to understand how engagement drives effectiveness, particularly for underserved populations. Assessing patterns within and between interventions would move us from broad, macro-level claims of efficacy to more precise understandings. More detailed and open reporting of demographic characteristics, particularly those related to recognized health disparities, would encourage analysis of how engagement among subpopulations may impact outcomes [101]. Even where digital infrastructure exists, without comprehensive datasets, developing evidence-based tailored interventions that induce behavior change with multiple strategies will be difficult. Systematic review and meta-analyses of individual participant data provide a method for this type of analysis (eg, Struik et al [102] and Jolliffe et al [103]). This discussion goes into more details followed by a consideration of the limitations.

Across the included studies, usage measures were highly heterogeneous, despite the availability of established frameworks such as AMUsED [35] that encourage more systematic reporting. Full-text screening showed that most research on digital interventions for asthma and COPD reported no usage measures at all (n=206, 72.5%). Among the 27 included studies, 4 meaningful and comparable usage metrics were identified, providing a basis for comparison between interventions: 14 studies [13,56,71,73,75,78,80,82,83,85,86,88,89,91] (51.9%) reported only one comparable measure (mean 2.15, SD 0.74), 9 [25,72,74,77,81,84,90,92,94] (33.3%) reported 2, and 4 [76,79,87,93] (14.8%) reported 3. Nine studies [25,56,71,73,75,79,82,91,93](33.3%) reported at least one noncomparable measure (mean 1.6, SD 0.85). Although these represent positive reporting practices, the dominance of simple metrics, typically logins or time spent in the app, provides only a minimal record of interaction and offers a partial view of engagement. Usage data were rarely conceptualized in terms of amount, breadth, or depth, and few studies linked usage patterns to specific intervention components [34,104]. This narrow reporting is striking given that digital interventions routinely collect rich metadata, yet only a small subset is made visible in publications. As a result, opportunities to identify clinically meaningful engagement patterns, such as increased inhaler-related activity preceding symptom deterioration, remain limited, and the development of effective engagement models is constrained by the absence of detailed component-level usage data [36].

From an intersectional perspective, restricted usage reporting also limits the ability to examine whether engagement varies across demographic subgroups or in relation to layered vulnerabilities. Only a small number of studies combined usage with demographic characteristics, and even fewer explored how engagement might differ across intersecting characteristics such as age, sex, ethnicity, SES, or digital literacy. Without richer usage data, these patterns remain obscured. To address this, future research should adopt more comprehensive and theory-informed usage reporting, specifying which components were accessed, how frequently, and at which points in the disease trajectory. Treating usage data as a form of metadata would also allow alignment with established standards such as the findable, accessible, interoperable, and reusable (FAIR) principles [105], supporting more comprehensive reporting practices. This would enable more nuanced analyses of engagement, facilitate the identification of disparities, and strengthen the evidence base needed to tailor digital self-management interventions to diverse patient populations.

Reporting of clinical and self-reported outcome measures was comparatively consistent across the included studies. Outcome measures were grouped into 3 categories (disease-specific self-reported, physiological, and other self-reported), and many studies (n=13, 48.2%) [13,71,72,74,78-81,84,85,87,89,91] reported at least 2 of these categories. Four studies [74,81,84,91] incorporated all 3, providing a multidimensional basis for assessing the potential impacts of digital self-management interventions. Disease-specific self-reported measures (eg, ACQ and C-ACT) were the most frequently used (n=20, 74.1%) [13,72-79,81,82,84,85,87,89-94], reflecting their central role in asthma and COPD research. Physiological outcomes such as FEV₁ were reported less often, but encouragingly, most studies that included physiological measures reported them alongside other outcome types (n=11, 68.8%) [13,71,72,74,78-81,84,85,91], supporting triangulation across domains.

These reporting practices create a strong foundation for comparability across studies, the potential for meta-analyses, and the ability to examine whether specific usage patterns relate to specific outcomes. However, integration of outcomes with demographic and usage data remains limited, meaning that potential disparities in intervention effectiveness are still difficult to identify. For example, it remains unclear how subpopulations may experience differential improvements in symptom control, lung function, or self-management confidence.

Strengthening outcome reporting further will require clearer justification for outcome selection and greater alignment with patient priorities [55,106,107]. Analyzing outcomes alongside demographic and usage data would also enable researchers to identify disparities and understand which populations benefit most. Overall, current reporting practices provide a solid platform from which to build a more comprehensive and equitable evidence base for digital self-management interventions.

Reporting of 3 demographic characteristics was common to all studies, including age, disease severity, and sex. On average, studies reported 5.9 demographic characteristics (range 3‐15). However, their use was underwhelming, as Szinay et al [108] reported demographic characteristics of known health inequalities are largely neglected. In our included studies, age was routinely reported, whereas ethnicity, SES, and smoking were often missing. SES, or a proxy of SES, was unreported in half of the included studies (n=14, 51.9%) [25,71,73,75-78,81,83,86-88,92,93]. Similarly, ethnicity was unreported in 82.5% (n=3317) of participants and 70.4% (n=19) of studies; unless studies explicitly focused on ethnicity, participant populations were generally presented as homogeneous. In practice, demographic characteristics were often reported descriptively but rarely used analytically; as such, at best, they imply sample homogeneity. Therefore, little could be accomplished with demographic data.

Improved reporting should be paired with analytic frameworks capable of interrogating how demographic characteristics interact with engagement and outcomes. Greater complexity would encourage discussion that is much more sensitive [97,98]. For example, only 4 studies [72,80,86,88] reported comorbidities despite research evidencing the benefits of managing “treatable traits” that are common across conditions [109]. Existing reporting practices identified in the results, including age, disease severity, and sex, are important, but more characteristics need to be reported and analyzed to develop understanding. A flexible model could be adopted that attempts to identify relevant characteristics at the intervention level. Luna’s [97,98] concept of layered vulnerabilities, like effective engagement, could be applied within specific interventions to support greater comparison between interventions. This would also help identify meaningful demographics, characteristics that go beyond the basics, which might relate to the health care setting, support offered, and the number and type of comorbidities.

Layered vulnerabilities [98] encourage researchers to move away from stereotyping master status identities to generate more nuanced, disease- and intervention-specific taxonomies. The concept contrasts with the idea of “the digital rainbow” [110] and the Prognosis Research Strategy (PROGRESS) framework (place of residence, race/ethnicity/culture/language, occupation, gender/sex, religion, education, SES, and social capital) [111]. Rather than use prescriptive labeling systems, Luna [98] encourages researchers to identify relevant, disease-specific, intersectional disparities as they develop. Luna’s [98] concept could work well with person-centered approaches to statistical analysis [112] as well as participatory and qualitative approaches. Fundamental to achieving this is developing, and reporting a range of diverse and inclusive demographic characteristics.

Many of the included studies did not combine any types of data in analysis (n=17, 62.9%) [13,25,56,71,73,75,78-81,84-87,90,93,94] and no studies combined all 3 types of data in analysis (demographic, outcome, and usage). As others have found [33,113], few studies analyzed usage measures with clinical/physiological or self-reported outcome measures (22.2%). This missing analysis makes it difficult to determine relationships within interventions and to make comparisons between interventions [33]. This was reinforced through the thematic analysis where differences among subpopulations were typically descriptive. Husain et al [106] similarly suggest that research on digital health disparities is typically descriptive and lacking any theoretical basis. Although we use the term “master status identities,” Husain et al [106] used the complementary term, “single-axis analysis.” A combined analysis may offer a more rigorous understanding of effectiveness, within and between interventions, helping to explain differences among subpopulations rather than simply identifying them [34,35]. Nouri et al [55] suggest reporting and responding to different demographic populations is important to increase uptake, sustain engagement, and identify disparities. We highlight analysis by demographic characteristic as crucial to fulfilling the expectations of inclusive research engagement.

This scoping review focused on digital behavior change interventions, and it is likely that the findings are relevant to broader digital health technology (eg, automated sensors or wearable technology). The 2-phase screening approach produced a focused set of usage measures that generated uniformity, overcoming an issue recognized by Nouri et al [55].

The focus on attempting to identify patterns of engagement among heterogeneous populations was problematic. The National Institute for Health and Care Research (NIHR) [114] reports only 60% of RCTs report ethnicity, and they have not yet started to collate data on the reporting of SES. It is important to track recognized disease-specific health inequalities, and digital interventions present a unique opportunity to identify vulnerable characteristics that emerge within datasets. Focusing on “master status identities” potentially magnifies an issue that requires more nuance [96,97].

Finally, this literature review did not develop understanding of organizational or structural factors. Management, resource allocation, and delivery priorities can influence adoption and uptake of digital health interventions and are significant areas to understand. For example, Ramachandran et al [115] suggest barriers and facilitators at the management level impact the adoption of digital interventions for COPD. Scoping reviews on engagement may be a good format to consider if and how management and intersectional factors can be incorporated into research. The results reported here focus on user engagement, and the themes identified overlap with similar research [116,117].

Existing frameworks such as AMUsED [35] and FAIR [105] have already been highlighted. However, these do not address the identification of health disparities among disease subpopulations. For this, we turned to Luna’s [98] concept of layered vulnerabilities, an intersectional approach that encourages disease- and intervention-specific considerations of demographic disparities. This concept fits well with the [99], an intersectional framework that integrates consideration of health inequalities at all stages of research. However, layered vulnerabilities provide a theoretically informed base that suggests researchers should not wholly rely on existing taxonomies but seek to develop taxonomies through methodology and analysis.

We suggest ways to improve research, with an integrated focus on underserved populations. Rather than a subsidiary research genre, an overarching aim is to bridge the gap between research focused on health inequalities and wider health research. A more concerted effort would encourage greater comprehension of disparities, engagement, and effectiveness for digital health interventions, who is included, how they engage, and who is served. In doing so, evidence-based tailoring options might become apparent. Our recommendations do not rely upon simply increasing “diversity” of samples; indeed, a sample could be homogenous in specific respects (eg, age or ethnicity). We suggest developing meaningful characteristics (eg, delivery site, primary or secondary care, engagement with in-person services, and/or postcode as a marker of environmental exposures). We emphasize two priorities: (1) the identification of meaningful demographics through the expansion of collected characteristics to complicate and complement existing knowledge; and (2) conducting more comprehensive analysis that combines the different types of data. Addressing these priorities in tandem should identify more relevant and sensitive characteristics while encouraging an intersectional analysis.

In summary, a more sensitive intersectional approach to analysis would provide a depth that is currently absent. Disparities should be considered within the epidemiology of the disease and within the matrices of specific digital interventions for a more comprehensive understanding of effectiveness and engagement. Finally, these recommendations are meant to be inclusive rather than prescriptive, not dictating what research should collect, report, or analyze.

This study advances the field by applying an intersectional lens to digital engagement research, highlighting how current reporting practices limit the ability to identify disparities or tailor interventions. Unlike previous reviews that focus on effectiveness, our analysis maps structural gaps in demographic reporting, outcome selection, and usage analytics. These insights offer actionable recommendations for researchers and developers, supporting more equitable design, evaluation, and implementation of digital self-management interventions in real-world settings.