Digital health technologies (DHTs) comprise apps, programs, or software used in the health and social care systems [1]. They are considered to have almost unlimited potential to transform health care interventions and delivery and empower people to take a greater role in their own care and well-being [2,3].

Stroke is one of the leading causes of acquired disability worldwide, with around 12 million people experiencing a stroke each year [4]. Rehabilitation is a complex, multifaceted process [5] that facilitates those with health conditions and disabilities to participate in and gain independence in meaningful life roles [6]. It is considered an essential aspect of health care provision following a stroke [7] as a means to address poststroke impairments, which can involve motor, sensory, and cognitive functions. Changes in the ability to move due to impairment of both movement and sensory function are commonly experienced by people following a stroke [8] and are addressed by physical rehabilitation comprising regular, intensive practice and repetition of movements and tasks [9,10]. Conventional physical rehabilitation often struggles to deliver the intensity required to optimize recovery [11], and over recent years, there has been significant interest in the use of DHTs, such as virtual reality (VR), telerehabilitation, robotics, and activity monitors [12-16], to enhance and increase the intensity of rehabilitation. DHTs can provide a whole intervention or be used as a component of a wider intervention; the term DHT-based intervention has been used within this review to refer to both situations.

For many people who survive a stroke, rehabilitation requires individuals to engage in regular and frequent rehabilitative activities to achieve improvements in function and realize their optimal recovery. This necessitates adjustments to an individual’s behavior [17] over a sustained period of time. Changing behavior is a complex process and is underpinned by a variety of different theories, models, and frameworks [18], such as social cognitive theory [19] or the behavior change wheel framework [20]. Individual activities within a complex intervention that are designed to change behavior can be separated into replicable active components widely referred to as behavior change techniques (BCTs) [21]. Historically, labels applied to BCTs have lacked consensus, resulting in uncertainty and difficulty in comparing interventions. This has been addressed in the behavior change technique taxonomy version 1 (BCTTv1) [22], a classification system of 93 distinct BCTs clustered into 16 groups, which is a well-recognized tool to provide consistency with BCT reporting in interventions. DHTs provide an emerging opportunity to support the behavior change required within physical stroke rehabilitation interventions through facilitators that are embedded within the technology itself that aim to form, alter, or reinforce behaviors [23]. Understanding of this area is limited, with most literature exploring the use of DHTs to support behavior change focused on specific health-related behaviors such as physical activity or healthy eating [24] rather than as a core component of a type of rehabilitation intervention. Motivation is acknowledged to play an integral role in behavior change [25], and it is often assumed that DHTs provide motivation to engage with rehabilitation [26]. However, for this assumption to be realized, the DHTs must be able to support and deliver interventions that facilitate the vital changes in behavior needed to promote prolonged and sustained engagement in stroke rehabilitative activities. Imperative to this is understanding the theories, models, and frameworks that underpin interventions and the BCTs (active components) within the interventions [27-29]. The theories, models, and frameworks alongside the BCTs will be referred to hereinafter as approaches. Within the context of DHT-based physical stroke rehabilitation interventions, approaches to behavior change warrant further investigation.

This scoping review aimed to identify if and how behavior change approaches are incorporated within DHT-based physical stroke rehabilitation interventions. Specifically, it sought to:

A scoping review was completed and reported following established guidelines [30,31] and the Preferred Reporting of Systematic Reviews and Meta-Analyses Extension for Scoping Reviews (PRISMA-ScR; Multimedia Appendix 1) [32]. The protocol was registered with the Open Science Framework [33].

Any published sources that reported a completed primary research study in which a behavior change approach could be identified within a DHT-based physical stroke rehabilitation intervention were included (Multimedia Appendix 2). Physical rehabilitation comprised interventions that addressed an impairment, or sequela of impairment, of sensory function and pain, neuromusculoskeletal and movement-related functions, or voice and speech, as defined by the International Classification of Functioning, Disability, and Health [34]. Completed primary research included all types of studies, both quantitative and qualitative, and no minimum sample size or intervention length was set. The BCTTv1 [22] was used to support the identification of BCTs within the interventions.

A systematic database search was conducted in Embase, MEDLINE, PsycINFO, CINAHL, Cochrane Database of Systematic Reviews, CENTRAL Register of Controlled Trials, and AMED on March 21, 2023. The search was completed in collaboration with an information specialist who provided support with the development of the free text and thesaurus search terms, created the final search, adjusted the searches for the different databases, and ran the search. It consisted of 4 distinct search streams: behavior change, DHT, physical rehabilitation, and stroke, which were then combined (Multimedia Appendix 3). Searches were restricted to the English language (due to review resources) and by date to search from 2001; the date restriction acknowledges the main time period of DHT growth [35], captures sources reported in systematic reviews of DHTs in stroke rehabilitation [12-16], and is reflected in other scoping literature exploring DHTs [24]. Additional sources were identified by hand searching, including scrutiny of the included source reference lists.

The titles and abstracts of deduplicated sources from database searches and hand searches were independently screened by 2 reviewers, 1 of whom had completed the BCTTv1 web-based training package [36] to inform decisions made around the use of BCTs. Any conflicts were discussed, and if a consensus was not reached, the source was included for full-text screening. Attempts were made to locate a completed study publication from eligible conference abstracts, protocols, and trial registry entries. Full-text sources were screened independently by 2 reviewers, and disagreements were resolved by a third reviewer. Reasons for full-text exclusion were recorded. EndNote X9 software (Clarivate) and the Rayyan web tool software (Qatar Computing Research Institute) [37] were used to facilitate the source selection process.

A review-specific data charting tool was developed and initially piloted using 3 sources by 3 reviewers, and then further developed iteratively throughout the process [30]. Data charting was completed collectively by 2 reviewers. When several sources referred to a single study, these sources were grouped together for data charting, and if a source identified additional sources for further detail of the intervention (eg, a protocol or supplementary material), then this information was also used to support data charting.

The data charting tool was developed with reference to the Template for Intervention Description and Replication (TIDieR) checklist [27] and with a focus on the DHT-based intervention and behavior change approaches (Multimedia Appendix 4 [14,38-40]). In the absence of a recognized predefined taxonomy for DHTs, the DHTs used in the sources were charted iteratively by the type of technology [41] from the information provided about the intervention. Over time, DHT categories emerged and were defined (Multimedia Appendix 4). Discrete BCTs were identified from the intervention detail provided using the BCTTv1 [22] (Multimedia Appendix 5 [42]). A pragmatic decision was made that the single reviewer who had completed the BCTTv1 web-based training package [36] would code the interventions to the BCTTv1. Any areas of uncertainty were discussed in detail among the review team.

In accordance with the aims of a scoping review, formal assessments of methodological quality were not completed [30,31]. Findings were synthesized using descriptive statistics facilitated by SPSS Statistics 28.0.0.0 (IBM Corp) and Microsoft Excel (version 2208; Microsoft Corporation) and presented in text, table, and chart formats. The characteristics of the included sources, specifically participant numbers, age, and time since stroke, and intervention details, were summarized to provide contextual information for the review. Time since stroke was based on a published timeline framework [43], which describes the following phases: acute (1-7 days), early subacute (7 days to 3 months), late subacute (3-6 months), and chronic (greater than 6 months).

The behavior change theories, models, or frameworks underpinning the DHT-based interventions and sources where interventions had already been coded to the BCTTv1 were summarized. The use of individual BCTs, as coded by reviewers from intervention descriptions, was briefly summarized; however, the main focus of the BCT synthesis was completed by grouping the BCTs into the 16 BCTTv1 clusters, in order to provide an overview of their use across the sources and allow comparison with other reviews [44,45]. A cluster was only identified once per source, irrespective of the number of individual BCTs within that cluster. Relationships between BCTTv1 clusters and between DHT type and BCTTv1 clusters were descriptively explored. A relationship map was used to visually represent the strength of the connections between the BCTTv1 clusters, with a thicker line indicating that variables were more frequently reported together. No inferential statistical analysis was used.

From a total of 1973 sources screened, 357 full-text sources were assessed for eligibility, then after grouping sources that referred to a single study, 103 (5%) distinct sources were included in the review [46-148] (Figure 1). Of the 245 sources excluded at full-text screening, 165 (67%) were excluded due to a lack of a behavior change approach.

This scoping review provides a comprehensive overview of approaches used to support changes in behavior in DHT-based physical stroke rehabilitation interventions. Research in this field is in its infancy, with the predominance of studies in this review being described as pilot or feasibility studies with limited participants.

Despite using comprehensive behavior change search terms, only a limited number (103/1973, 6%) of screened sources were included. Over two-thirds of full-text sources were excluded as they did not describe or refer to any behavior change theories, models, or frameworks or BCTs, suggesting that in general, explicit behavior change approaches are not reported as being integral to DHT-based physical stroke rehabilitation.

Only 18 (17%) of the 103 included studies articulated a theory, model, or framework to underpin the intervention, which aimed to change behavior, despite widely published recommendations about the importance of overt use of theory when developing, evaluating, and reporting interventions [27,29], including those related to behavior change [28]. The proportion of studies articulating a behavior change theory, model, or framework in this work is significantly lower than review findings in non-rehabilitation DHT-based interventions that have sought to influence specific behaviors such as physical activity or weight control [24,44]. These reviews have identified up to two-thirds of sources reporting a theory, model, or framework. However, our findings mirror the relative absence of behavior change theories, models, and frameworks in rehabilitation interventions more generally, irrespective of whether they use digital technology [149] or not [45], and it is widely recognized that the complex nature of rehabilitation often results in the essential characteristics of interventions being poorly defined [150]. Consistent with our findings in these other reviews, a variety of theories, models, and frameworks were found to underpin interventions, with social cognitive theory being the most frequently reported [24,44,45,149]. The explicit description of BCTs using the BCTTv1 within DHT-based physical stroke rehabilitation interventions is also poorly reported (2%), despite a significant proportion of the sources being dated after the publication of the BCTTv1 in 2013 [22]. This lack of acknowledgment of behavior change approaches impedes the accumulation of knowledge within this field.

It is important that both the underpinning theory and BCTs are reported so the mechanisms by which the BCTs elicit change can be better understood [21]. The general assumption that the motivational and captivating aspects of DHTs will promote prolonged and repeated engagement with rehabilitative activities, in particular in those DHTs that incorporate game design [151], risks suboptimal outcomes for patients and wasted investment of time and money if the mechanisms by which the DHT elicits change are not considered.

When exploring which BCT clusters featured within the reviewed DHT-based interventions, the findings relating to the commonly used clusters of feedback and monitoring, goals and planning, and shaping knowledge are consistent with findings from DHT-based interventions to change a specific behavior [44] and non-DHT–based rehabilitation [45]. However, a novel finding in our review was the frequent identification of the reward and threat cluster, although it was noted that all techniques related to reward and none to threat. A large number of studies in this review used VR technology, which frequently incorporates gamified tasks or gameplay. Reward is an integral part of game design theory alongside feedback [152], and so it is perhaps unsurprising that the feedback and monitoring, and reward and threat clusters dominated and an association between these 2 clusters was seen.

Rehabilitation is a process that comprises multiple behaviors and so exploring approaches to change behavior within this context was complicated. There were challenges in searching and screening sources for inclusion as few studies explicitly reported approaches to change behavior, and there is a similarity in the vocabulary used within behavior change and other theoretical approaches (eg, “feedback,” which is used within motor learning). Similarly, only a very small proportion of studies explicitly reported BCTs within interventions. The lack of clear reporting of behavior change introduces the risk that sources may be omitted during both the searching and screening process highlighting the difficulty of comprehensively reviewing this field of work. An inclusive approach to screening reduced the risk of erroneously excluding sources, but it is perhaps inevitable that the sources included reflect those studies that have reported a behavior change approach rather than all studies that have used one.

This lack of clear BCT reporting also posed challenges for intervention coding. The use of the BCTTv1 aimed to ensure the review used a generalizable nomenclature to describe BCTs, and the 1 reviewer who had completed BCTTv1 training coded all the interventions. It is acknowledged that decisions made in the application of the BCTTv1 within the context of the review will have introduced some subjectivity in intervention coding, which will ultimately influence the review findings. Although the coding process could have been made more robust by having a second reviewer trained in the BCTTv1 also code the interventions, regular and extensive discussions between all members of the review team took place with the aim of ensuring consistency with the coding process. Clear documentation as to how the BCTTv1 was used within this review (Multimedia Appendix 5) supports transparency as to the decisions made and the reproducibility of the review.

The absence of a recognized predefined taxonomy for DHTs posed a challenge when categorizing the DHT interventions, acknowledging that the distinction between the categories used to present the results is open to interpretation. A description of how the reviewers interpreted these categories is provided (Multimedia Appendix 4).

Future studies aimed at developing and evaluating DHT-based rehabilitation interventions, including those relating to physical stroke rehabilitation, need to ensure there is explicit recognition and reporting of the specific approaches used to change behavior, articulating both the theory on which the intervention is based and how the intervention plans to deliver the change in behavior using universally recognized terminology. This should be reported as part of a program theory and potential mechanisms of action, which are key parts of developing and evaluating complex interventions [29]. This detailed reporting would further support an understanding of how changes in behavior could be best enabled by DHT-based rehabilitation interventions and how this contributes to changes in patient outcomes. It would also enable further evaluation of the optimal behavioral components of interventions, enabling patients to use and clinicians to deliver the most effective DHT-based rehabilitative interventions. More generally, as the use of DHTs expands, there is an urgent need for some form of taxonomy to categorize and clearly define the different types of DHTs to facilitate consistent reporting, replication, and comparison of DHT-based interventions.

This novel and original review is the first to explore if and how approaches to change behavior are incorporated within DHT-based physical stroke rehabilitation. It demonstrates that a minority of studies report using approaches to change behavior within this context, despite these changes in behavior being vital to meet the demands of rehabilitative activities. Those who do report behavior change often lack the underpinning detail as to how the DHT-based intervention will facilitate these changes. In order for DHT-based interventions to realize their potential within rehabilitation and their impact on patient outcomes, approaches to change behavior must be embedded in the intervention and appropriately reported.