Acquired brain injury (ABI), including stroke and traumatic brain injury (TBI), is a globally prevalent condition affecting >135 million people [1]. ABI represents a significant and growing global burden of disease [2,3] with a high need for rehabilitation service support [1].

Cognitive-communication disorders are common sequelae of ABI [4,5]. Cognitive-communication disorders arise when underlying impairments in cognitive skills, such as memory and organization, manifest in a person’s communication [6]. Evidence-based treatment for these communication difficulties is provided by speech-language pathologists [5]. Author RM recalls the following from her living experience:

As communication is a bilateral and multilateral exchange between “communication partners,” close others such as friends and family can also be affected by communication changes after ABI. For example, if a person with ABI experiences difficulty staying on topic or remembering details to include during conversations, their communication partners have been found to adjust their own communication to compensate for these communication behaviors [7]. In doing so, communication partners can positively or detrimentally affect the communication rehabilitation of the person with ABI [7-9]. The protocol for this study included a firsthand account of potentially negative impacts of a close other’s sense of disempowerment during communication after ABI [10]. Conversely, author MTM recalls positive experiences of being able to support a friend:

Communication changes after ABI can have pervasive and long-term psychosocial effects on both the person with ABI and their close others. As a result of communication changes after ABI, people with ABI can experience negative impacts on their social participation [11,12], relationships [12,13], employment [14,15], and mental health [16]. Inversely, close others can experience altered and challenging relationships with the person with ABI [13], increased burden, and reduced quality of life [17,18]. Wider communities are also economically affected by the burden of informal care, including reduced workforce participation by caregivers [19].

The internationally recommended best practice approach to managing communication impairment after ABI is to provide communication partner training (CPT) to both the person with ABI and the people who communicate with them after ABI [20,21]. However, the substantial and growing burden of ABI makes it challenging to meet this recommendation at a global scale, for both health care systems generally [2,3] and the speech-language pathology profession specifically [22]. For example, in 2 recent national surveys of speech-language pathologists supporting people with ABI, a minority of 122 Australian speech-language pathologists self-reported using published, evidence-based programs to train familiar (11/110, 10%) and unfamiliar (10/75, 13%) communication partners of adults with stroke [22], with comparable proportions of 169 speech-language pathologists surveyed in the United Kingdom reportedly using published programs to train familiar (27/136, 19.9%) and unfamiliar (12/87, 14%) communication partners of adults with TBI [23].

Current research also indicates a need for neurorehabilitation professionals to support people with ABI in the effective use of social media [24], as communication difficulties can extend beyond face-to-face conversations to digital interactions. Potential neurorehabilitation support therefore extends to cyber safety training for people with ABI, with exploratory research indicating that more than half (n=54, 53.5%) of 101 surveyed Australian and New Zealand neurorehabilitation providers had clients who had been affected by cyberscams [25]. However, there is currently a lack of effective interventions supporting digital interactions through social media after ABI [25,26].

To address these needs, the “Social Brain Toolkit” [27] was developed as a novel suite of 3 targeted communication interventions delivered via the internet. It includes a self-directed web-based training program, “social-ABI-lity,” specifically developed to address an identified need [24,26] to support people with ABI to use social media, and 2 stand-alone yet complementary CPT programs, “interact-ABI-lity” and “convers-ABI-lity.” interact-ABI-lity and convers-ABI-lity were derived from previously published, evidence-based [28,29], face-to-face and telehealth-delivered CPT programs targeting cognitive-communication disorders, “TBI Express” [30] and “TBIconneCT” [31]. A key differentiation from these previous programs is that convers-ABI-lity has a fully web-based service delivery model incorporating self-directed learning modules and weekly telehealth sessions between a person with ABI, communication partner, and speech-language pathologist. interact-ABI-lity is a short, self-directed learning program solely for the communication partner and features advice relating to aphasia (language impairment), dysarthria (weakness in speech muscles), augmentative and alternative communication (communication via methods other than natural speech, such as facial expression or a voice output device), and cognitive-communication difficulties (communication difficulties due to cognitive impairments). A comprehensive comparison of each intervention is presented in Figure 1.

Digital health interventions and applications targeting neurological conditions continue to proliferate with modern technologies [32,33]. However, the demonstration of their clinical effectiveness is insufficient to guarantee their successful implementation [34]. Digital health implementation can be highly complex [35], with growing concern in relation to complex challenges such as equity of access [36]. Despite a need for guidance, there is currently a paucity of empirical evidence to guide the process of tailoring implementation strategies [37-39]. This gap is compounded by misalignment between general implementation frameworks and the specific challenges of implementing digital health [40]. Therefore, there is currently an urgent need for guidance to inform the development of tailored, evidence-based implementation strategies that align with the burgeoning need for digital health.

It has been recommended that researchers consult stakeholders to identify implementation barriers and tailored strategies [37]. Therefore, this study aimed to coproduce this implementation knowledge with stakeholders [10] to support the deployment of implementation strategies: (1) at intervention launch, (2) iteratively during initial pilot (convers-ABI-lity) and 12-month implementation-effectiveness hybrid studies (interact-ABI-lity and social-ABI-lity) [41], and (3) in future versions. The Australian development team includes speech-language pathologist researchers LT, RR, and MB from the University of Sydney and EP and MM from the University of Technology Sydney, with advisory and steering committees comprising people with living experience of ABI and clinicians supporting them. Partner organizations include technology vendor Changineers, community partner Brain Injury Australia, and funding partner icare New South Wales. The coproduction of implementation knowledge was conceptualized and led by author MM, whose relevant experience and qualifications include a clinical background in speech-language pathology with specialized training in accessibility, prior experience of coauthorship with stakeholders, qualitative research and peer-review experience, and membership of a high-risk human research ethics committee.

Given the aforementioned medical and psychosocial complexity of ABI [1,3,11-16]; the complex nature of health care systems as a complex adaptive system (CAS) [42,43] facing the growing global burden of ABI [1-3]; and the complexity of implementation itself [44,45], specifically digital health implementation [35,46-49], researchers should anticipate that the implementation, scale-up, and sustainability of the Social Brain Toolkit might be highly complex. It has been suggested that the field of complexity science can offer essential insights to support health [50] and implementation [44] research contending with real-world complexity. However, given that researchers have used a wide range of conceptualizations of complexity [51], with interchangeable use of related terms and concepts [52-54], a clear definition of complexity was required for this study [51,54]. In this study, digital health implementation was conceptualized as a CAS in its own right, in which multiple “agents” (eg, implementation considerations and strategies) followed local rules that adapted to each other (eg, an implementation strategy reducing the impact of an implementation consideration). System “attractors” (eg, a new implementation strategy or consideration) could vary according to parameters such as the number, strength, and diversity of their connections, with some stability gained from the constraints of other agents but also lost as a result of the evolving, emergent nature of local rules. Despite conceptualizing implementation as a CAS, it must also be acknowledged that this study adopts the “least radical” conceptualization of complexity [51], in which complexity is conceptualized as varying between the “simple” (straightforward, predictable, and with few components), “complicated” (with multiple interacting components or issues), or “complex” (dynamic, unpredictable, and not easily disaggregated into constituent components) [55], rather than more radically viewed as something omnipresent at all times. This least radical conceptualization of complexity served as a logic pathway for the use of tailored implementation strategies [56] as recommended in implementation science, whereby implementation strategies could facilitate implementation by theoretically managing or reducing the complexity of implementation considerations from “complex” to merely “complicated” or “simple.”

As described in the published protocol [10], the broad aim of this study was to support the implementation of the Social Brain Toolkit by coproducing implementation knowledge with people with ABI, close others, clinicians, and leaders in digital health implementation. The specific aims of the study were to (1) prioritize theoretically informed implementation targets for the Social Brain Toolkit; (2) understand the nature of these priorities; (3) coproduce targeted implementation strategies; and (4) explore potential interrelationships between implementation considerations and strategies as a CAS, as a post hoc aim formulated in response to emergent complexity within the data set.

This study was epistemologically, ontologically, methodologically, and axiologically oriented from a critical realist research paradigm [57]. Critical realism would (1) ontologically support the examination of underlying mechanisms beyond empirical observation; (2) enable an epistemologically pluralist stance in which researchers can value complexity science alongside implementation science and experiential knowledge alongside academic knowledge; (3) allow for pragmatic and pluralist methodologies, including both implementation science and complexity science approaches; and (4) axiologically support the wider emancipatory values of coproducing Open Science [58], implementing evidence-based practice, and improving health outcomes for people with ABI and their communities.

Therefore, this study was theoretically guided by the Nonadoption, Abandonment, Scale-up, Spread, and Sustainability (NASSS) framework [55]. The NASSS framework was selected for its comprehensive inclusion of 7 interacting domains of digital health implementation, scalability, and sustainability; systematic itemization of the degree of complexity in each of these 7 domains; specific focus on digital health; and alignment with author MM’s living experience of using technology to implement and scale speech-language pathology services.

As shown in Table 1, participants were a purposive, maximum variation sample (N=35) of adults with living experience of ABI, communication partners, speech-language pathologists supporting people with ABI, and individuals with experience implementing digital health. Out of 10 individuals who had living experience of ABI, 2 (20%) also had either clinical experience of ABI or digital health implementation experience. In addition, 2 (40%) out of 5 clinicians and 1 (10%) out of the 10 people with ABI had prior experience with TBI Express [30] and TBIconneCT [31] and were thus able to directly compare their delivery and content with those of the potential or prototype versions of convers-ABI-lity and interact-ABI-lity (Figure 1).

People with living experience of ABI (10/35, 29%) were required to be discharged from hospital, participating at least 6 months after sustaining an ABI, and with adequate capacity to consent to study participation. An adapted consenting process [59], published as supplementary materials of the study protocol [10], was used screen capacity to consent. Participants with living experience of ABI were required to be based in Australia, where the Social Brain Toolkit was developed and where they could be financially reimbursed for their living experience expertise. People with ABI were recruited through tailored electronic flyers posted with alternative text on social media, as well as organizational websites, email distribution lists, and snowball recruitment.

Communication partners (11/35, 31%) were self-identified as individuals who interacted with a person with ABI at least once a week and who had not sustained an ABI themselves. Communication partners were also required to be based in Australia, where the Social Brain Toolkit was developed and where they could be financially reimbursed for their living experience expertise. They were also recruited through tailored electronic flyers, organizational websites, email distribution lists, and snowball recruitment.

Clinicians (5/35, 14%) were self-identified as qualified professionals practicing with a caseload of which at least 20% included people with ABI. No other restrictions were applied, including restriction on international participation. Clinicians were recruited through tailored electronic flyers, organizational websites, and email distribution lists.

Individuals experienced in digital health implementation (9/35, 26%) were required to have a research or industry track record in digital health implementation and were able to participate internationally. Eligible individuals were recruited via direct email to publicly listed contact details on university website profiles, researcher networks, and snowball recruitment.

Participant demographic information is described in detail in Table 1 and was previously published in the protocol [10].

The study received human research ethics approval with commendation from the University of Technology Sydney (ETH20-5466).

All participants provided informed written consent. In recognition of their living experience expertise, people with ABI and their close others were reimbursed at the rate stipulated by Health Consumers New South Wales [74]. All data were stored and transferred securely. Identifying audio and video data were deidentified as much as possible and stored separately from identifying information. Collaborative autoethnographic data were attributed to authors by name only when discussed and agreed upon in writing during the authorship process.

Across the top four prioritized domains of (1) condition, (2) technology, (3) value proposition, and (4) adopters of the Social Brain Toolkit, the coproduction of implementation knowledge yielded a total of 48 unique digital health implementation considerations and 52 unique, tailored implementation strategies for both developers and clinicians. Collaborative autoethnography identified both the benefits and challenges of coproducing implementation research within traditional academic systems. Post hoc network analysis yielded 172 unique relationships between the 100 implementation considerations and strategies and identified the potentially most impactful implementation strategies within the network.

The collaborative autoethnography included reflections on the methods of the study from the perspective of the coauthors with living experience of ABI, close others, and clinicians. Reflections on the methods of the study were previously published in the study protocol [10], with academic writing interleaved with quotations from people with living experience to communicate the equal value of both academic and experiential knowledge. The following results continue this reflection from the final stages of the research.

Upon appraising the outcomes of this implementation research, author RM, who has living experience of ABI, shared the following reflection:

Likewise, author J-CE provided the following conclusion as both a clinician and person with living experience of ABI:

Stakeholders reported experiencing personal satisfaction in coproducing new knowledge, with author RM adding the following:

Author J-CE, who has both living and clinical experiences of ABI, shared a similar reflection:

Sharing the living experience of ABI in the writing stage was identified to be an emotional experience. From their living experience as a close other of a person with ABI, author MTM notes that their reflection was “a very difficult passage to write, with so many memories and feelings.”

Reflecting from the researcher perspective of facilitating coproduction, MM offers the following four key learnings for fellow researchers wishing to engage in coproduction:

1. I found that coproduction required continual courage and perseverance, as it requires a researcher to overcome the inherently elitist traditions of academia, procedural inertia where there is a lack of precedent, and personal vulnerability in doing the unconventional. I particularly experienced this when interleaving first-person accounts of living experience with traditional academic writing, trying to secure and provide payment for living experience expertise, and asking for this expertise to be published as equivalent to an academic degree. I found it helpful to find solidarity at conferences with other researchers who were coproducing their research, and to view setbacks within our collective, longer-term aim to normalize the democratization of knowledge. I also found co-authors with living and clinical experience were a constant touchstone and encouragement during the isolation and struggle of driving implementation efforts [45], and was greatly encouraged to find allyship in journal editors who were committed to inclusive and innovative publishing practices which recognize living experience expertise. Production Editor Hannah Reinhardt communicated: I agree with you strongly that the expertise of those with living experience should be acknowledged in the research community, and this is a cause that is of personal importance to me. I was therefore thrilled to have the opportunity to raise this with JMIR's management team, in hopes of expanding our affiliation regulations. I am fortunate enough to work with an excellent group of people who are always happy to take the time to review our processes and work on innovative solutions, so this is very much a shared achievement.

2. As CR expressed in the study protocol [10], coproduction is a deeply human, relational process, and it is only in the context of these relationships that its ethically and politically complex nature [75] can be managed. Insofar as it can be known in advance, clearly communicating expectations upfront, such using accepted criteria for authorship [76] to determine authorship expectations, was helpful to plan and achieve authentic collaboration. However, coproduction often contravenes a research tradition of obtaining a clearly replicable scientific procedure. I found it important to empower coauthors to make informed choices within its complexities. For example, although the collaborative autoethnographic method [66] prescribes anonymization of co-author contributions in the interests of ‘confidentiality,’ author decisions and indeed enthusiasm to be identified by name, with informed discussion of the potential personal and professional risks of doing so, was potentially a less paternalistic approach than anonymizing all coauthors like participants. An additional example was the need for a trusting relationship in which to raise unforeseen challenges, such as receiving predatory conference and journal emails.

3. As my clinical background was not in ABI, my coauthors provided invaluable expertise in ABI that far exceeded what I could read in academic literature and gave me an invaluable sense of purpose in my work. Coproduction also transformed my perspective of the research process, particularly the many places patient and community voices are missing from it. As ABI does not discriminate, the final author lists are a unique collective of people from all walks of life, including film and television, professional sport, health, and education. As RM and J-CE mentioned, each person brought a wealth of personal and professional experience in addition to their living experience of ABI, and this diversity of knowledge and experience was an asset to addressing the complexity of implementation. However, as shown by a markedly lower response rate in member-checking and the ‘division of labor’ across publications for communication partner co-authors, close others stood out to me as potentially requiring more proactive inclusion and financial support to avoid their underrepresentation or absence from research.

4. Researchers may benefit from allocating additional time when planning and scoping coproduced projects, as a key challenge I observed was the incongruity between the natural timeline of relationship and the tight deadlines of academic research.

The post hoc network analysis yielded a total of 172 unique relationships between the 100 implementation considerations and strategies (Figure 3). The VNA provided an explicit visualization of the complexity reported by leaders in digital health implementation, one of whom remarked that the implementation domains were “all going back to the same problem that I said before...Essentially, I don’t think your questions...are all independent; I think they’re all interrelated” (Digital health implementation leader 7).

As another leader in digital health implementation acknowledged, implementation findings were revealed to be interdependent across all domains and subdomains: “I don’t mean to scare you, but there’s layers and layers of complexity” (Digital health implementation leader 1). All considerations and strategies were revealed to be connected in a network, except for a singular external connection between the “complexity of technological requirements” (consideration 2N) and a targeted strategy to “tailor supply models to the technological complexity of each tool” (strategy 2.8).

By contrast, strategy 4.1, to “conduct user or persona testing,” represented the highest number (22) of outgoing connections. As 64% (14/22) of these connections targeted the considerations in the stakeholders’ top priority domain of the “condition” of ABI (considerations 1A-1N), this strategy represented both high impact and high stakeholder priority. With fewer than half (9/22, 41%) as many outgoing connections, ensuring a responsive web design (strategy 2.1) was the second most influential strategy, as responsive design allowed the intervention to be available on various mainstream device types (strategy 1.17) and enabled access to a myriad of technological features (considerations 2A-2G), including options that potentially benefit people with physical impairments (consideration 1I). The third most influential strategy, with 6 outgoing connections, was ensuring that the tools were compatible with alternative technology access methods, such as eye gaze, voice, or touch, as opposed to only standard mouse and keyboard peripherals (strategy 1.6). This strategy had the identified benefits of facilitating technology access for people with ABI with sensory (consideration 1J), physical (consideration 1I), and communication (consideration 1B) impairments, and facilitating the input of patient-reported measures (consideration 2J) and progress data (consideration 2K) by people with ABI with accessibility requirements. It was also noted that these accessibility requirements should be considered in tandem with the need to avoid conveying a sense of penalty and judgment (strategy 1.11), such as standard speech-to-text devices being unable to positively acknowledge dysarthric speech. The fourth most influential strategy, and the first strategy to focus on close others, was strategy 4.10, to “facilitate buddy or peer support for close others.” This had the potential to address the 5 identified assumptions and requirements of close others (considerations 4D-4H), including participating in CPT and overcoming performance anxieties around communicating effectively with someone with ABI.

Within the identified network, there were 6 implementation strategies that each influenced 4 other strategies or considerations. These included suggesting that speech-language pathologists refer people with ABI and their communication partners to the tools (strategy 3.3), providing positive feedback of progress (strategy 4.5), ensuring that tools are available on mainstream device types (strategy 1.17), providing a setup checklist (strategy 4.2), streamlining the number of steps for users (strategy 4.3), and providing reminders to close others (strategy 4.11). Of the 90 remaining nodes, there were 9 (10%) nodes with 3 outgoing connections, 24 (27%) nodes with 2 outgoing connections, and 31 (34%) nodes with 1 outgoing connection. A total of 26 nodes received only incoming relationships, including 25 considerations and the aforementioned strategy (2.8) of tailoring supply models.

This study aimed to support the implementation of the Social Brain Toolkit by coproducing new implementation knowledge with people with ABI, close others, clinicians, and leaders in digital health implementation. The first specific aim was to identify stakeholders’ implementation priorities from the NASSS framework [55]. Stakeholders prioritized the investigation of the (1) target condition, (2) technology, (3) value proposition, and (4) adopters of the Social Brain Toolkit. The second specific aim was to identify implementation considerations in each of the top 4 prioritized domains. Stakeholders identified a total of 48 unique digital health implementation considerations across all domains and subdomains. The third specific aim, to coproduce implementation strategies tailored to these considerations, yielded 52 unique, tailored implementation strategies across all domains and subdomains for both developers and clinicians. A collaborative autoethnography revealed both the benefits and challenges of coproducing new implementation knowledge within traditional academic systems. The fourth post hoc aim was to explore potential interrelationships between the identified implementation considerations and strategies. A post hoc network analysis presented a network of 172 unique relationships between the 100 identified considerations and strategies, with highly connected strategies, or “attractors” within the CAS offering data-driven prioritization guidance for implementation effort.

To what was previously systematically evaluated to be a sparse evidence base [41], the 48 stakeholder-identified implementation considerations in this study contribute the first extensive compilation of considerations for web-based psychosocial interventions for people with ABI. In particular, this same systematic review [41] identified that the complexities within the first priority domain of the “condition” of ABI have traditionally been simplified via participant exclusion criteria to establish a clear value proposition for interventions (stakeholders’ third priority domain). By contrast, stakeholders in this study identified 14 implementation considerations and 17 implementation strategies in the top priority domain, as well as 4 demand-side and 4 supply-side considerations (Table 4) for the Social Brain Toolkit and 13 tailored strategies to directly address these 8 considerations. There was also guidance on how to target the strategies to specific stakeholders, including people with ABI, close others, clinicians, and health services (Textbox 1). Therefore, the study both enacted and supported the review conclusions to collaborate with stakeholders to increase implementation knowledge.

The 52 tailored strategies identified in this study were largely consistent with generic implementation [77] and sustainment [78] strategies but provided more specific implementation knowledge for ABI and digital health. For example, the general recommendation to “develop strategies with patients to encourage and problem solve around adherence” [77] was addressed through 9 unique strategies (4.1-4.9) to support people with ABI and a further 5 strategies (4.10-4.14) to support close others and clinicians. The generic recommendation to “increase demand” [77] was extended with very specific guidance on which value proposition messages were most important to stakeholders and how, when, and to whom these might be communicated. Thus, these strategies may provide more specific guidance for other digital health implementation endeavors to support people with ABI, as well as digital health implementation more broadly [34,35]. Furthermore, the identified implementation strategies were immediately relevant for current and future iterations of the Social Brain Toolkit [41].

Positive final reflections of the reciprocal benefit in coproducing implementation research corroborated previously published coauthor reflections in the study protocol [10]. Reflections on the emotional nature of the writing stage of research were also consistent with the emotion described during data collection [10], supporting the use of trauma-informed principles in coproduced research [79,80]. Researcher experiences appeared to corroborate suggestions that research practices, cultures, and structures encumber the process of coproduction, making it challenging to achieve the aspirations of coproduction [81]. Potentially of particular interest to researchers seeking to coproduce research with people with ABI was the experience of predatory email invitations being sent to all authors, which may warrant proactive warning, education, and follow-up, given the particularity of this type of scam to academia and the known potential vulnerability of people with ABI [25].

In addition to the complex network revealed by the VNA, the findings of this study illuminated a clear way forward [69]. A complexity-informed approach identified that user and persona testing (strategy 4.1) may be the single most impactful leverage point to support digital health implementation for people with ABI, reflecting findings from a recent systematic review highlighting the potential to address the complexities for “adopters” [41]. The fruitful use of VNA in this study directly addresses an identified need in implementation science for data-driven methods to prioritize and select implementation strategies from long lists of possibilities [82], especially given potential time and resource constraints. In addition, a focus on strategy 4.1 to target diverse considerations in the “condition” domain offers specific guidance on how to address a previously identified gap in implementation knowledge for people with ABI who have comorbidities and additional challenges [41], specifically, how to practically address the complexity of ABI, rather than reduce it through participant exclusion criteria. It is possible that analogous synergies may be identified and exploited for other digital health interventions and other populations.

The results also provide much-needed guidance on how to select an appropriate theoretical underpinning for future studies [41,83], as the critical position of user and persona testing in the network highlights the value of user-centered design [84]. The sociotechnical focus of the 2 most well-connected “attractors” within the system highlights the importance of using a technology-specific, as opposed to a generic, implementation framework. Therefore, this study supports and contributes methodological knowledge on how to conduct a sound integration of complexity science and implementation science [44] to prioritize digital health implementation strategies and targets.

This study has several strengths, primarily in relation to stakeholder input throughout the research. First, this study provides a multidimensional understanding of digital health implementation from the perspectives of people with ABI, their close others, clinicians, and leaders in digital health implementation, with some strategies identified specifically by certain subgroups (Multimedia Appendix 1). In addition, stakeholder prioritization of the study targets enabled research efforts to be directed toward the implementation questions that are most important to end users. Complementarily, stakeholder input included a discussion of preliminary findings from a prior systematic review [41] and the rigorous integration of the NASSS framework [55] from data collection to interpretation. This arguably managed the “balance to be struck between a paternalistic overemphasis on the research literature and a disregard of evidence...Selection and tailoring methods are guided by the best available theory and evidence, while preserving the benefits of stakeholder engagement and preference” [37].

An additional strength of this study was its methodological pluralism. By drawing on the complexity-informed NASSS framework [35] and methods such as network analysis, additional insights could be gained, which maximized the benefit of the initial list of implementation findings, providing new insights into the nature of and potential prioritization opportunities within digital health implementation for people with ABI. In doing so, this study also contributed an applied example of how a critical realist approach can be used to rigorously manage the conceptual, theoretical, and methodological tensions between complexity and implementation sciences to support the practical implementation of real-world interventions.

This study also has several limitations. First, although a focus on domains 1 to 4 reflected stakeholder priorities, additional implementation considerations and strategies from domains 5 to 7 were not reported, despite there being less research evidence to guide implementation in these domains [41]. With regard to complexity, this limitation is notable, given known interactions between domains 5 to 7 and the results of domains 1 to 4. For instance, professional expectations and regulations in the “wider system” (domain 6) would also interact with the reported results. In domain 6, although “client stories” (strategy 3.2) were directly requested by people with ABI and close others to convey the “value proposition” (domain 3) of the Social Brain Toolkit, the implementation of this strategy is not straightforward when considered in relation to the Code of Ethics of Speech Pathology Australia [85], which stipulates that all advertising “must not use testimonials / reviews. The use of ‘client stories’ may constitute testimonializing, which is prohibited” [86].

Second, given that ABI is especially prevalent in low- and middle-income countries [87], there was a noteworthy lack of representation from these countries, where additional complexities and strategies may have been identified. In addition, the results should be interpreted with consideration of the majority representation of Australian participants. For instance, strategy 3.12 was a recommendation of hierarchical pricing that demands least to no payment from people with ABI and most funding from governments and insurers. The identification of this strategy may have been influenced by the Australian context, in which national Medicare and insurance schemes may increase expectations of government support in a way that differs from countries and markets where self-funding health care is the norm.

Finally, although the exploratory and mechanistic focus of this study is consistent with complexity science and its distinctive quality criteria [50], the identified implementation strategies and considerations and their interrelationships have not yet been empirically tested or observed and would, therefore, be considered lower on traditional health research evidence hierarchies [88]. Therefore, future research might consolidate these preliminary findings by empirically testing the identified strategies, considerations, and interrelationships with a more confirmatory aim. This may also eventually allow for more in-depth network analytical methods in future, such as the weighting of certain links based on a deeper understanding of their importance or the use of alternative measures of centrality when digital health implementation mechanisms are better understood. However, it should be noted that a complexity-informed investigation would not expect the findings of this study to be precisely replicable at 2 points in time, given the qualitative nature of the data and the dynamic nature of complexity. Therefore, research resources may be better directed toward more practical implementation efforts, namely user and persona testing (strategy 4.1) for profiles that were not directly represented in this study.

People with ABI, close others, clinicians, and digital health leaders coproduced new knowledge of digital health implementation considerations for adults with ABI and their communication partners, as well as strategies tailored to address them. Several of the identified strategies may be applicable to other interventions for ABI, as well as interventions for conditions other than ABI, addressing current gaps in digital health implementation knowledge more broadly. Although this study was limited by its focus on 4 NASSS domains and the underrepresentation of some participant demographics, the wealth of immediately actionable implementation knowledge generated is aligned with the prospect of including stakeholders in the coproduction of implementation research, with mutually beneficial outcomes for stakeholders and researchers. A complexity-informed theoretical underpinning and a post hoc network analysis revealed that digital health implementation for people with ABI, their close others, and clinicians is a complex phenomenon. Traditional implementation science methods may therefore benefit from leveraging complexity-informed frameworks such as the NASSS and methods such as VNA to more fully understand, better prioritize, and more effectively address digital health implementation needs.