Knee osteoarthritis (OA) is a progressive and debilitating disease characterized by pain, disability, and poor health-related quality of life, with its incidence on the rise. For advanced stages, knee arthroplasty surgery is the standard treatment, intended to relieve symptoms and restore function and joint mobility. Despite the common use of arthroplasty, the spectrum of relevant and multifaceted [1] knee arthroplasty surgery outcomes is not universally positive [1-5]. Approximately 1 in 5 patients is not satisfied after surgery [5], and upward of 70% experience continued walking deficits years after surgery [2]. These mixed results reflect the fact that treatment decisions are often based on limited indicators, such as pain severity or radiographic changes, rather than a multidimensional understanding of the patient. Evidence shows that combinations of factors, including pain, gait function, joint structures, and patient expectations, better predict recovery [6,7], yet these predictors are rarely integrated systematically into clinical workflows.

Objective measures of function, particularly gait analysis outcomes, provide critical insights into OA progression and surgical response [8,9]. Deficits in knee joint mechanics are strongly linked to patient-reported pain, functional outcomes [7], and even implant migration and early failure [10-12] following knee arthroplasty surgery. However, traditional gait analysis has required costly, complex laboratory systems, limiting uptake in clinical practice. Recent advances in computer vision, wearable sensors, and machine learning now make it feasible to capture prognostic gait features more readily in real-world clinical environments [13-16]. Broad investment in advanced computing and artificial intelligence (AI) augments the potential of such systems, enabling the integration of biomechanical biomarkers with patient-reported outcome measures, activity levels, joint morphometrics, and demographics for clinically meaningful, individualized risk predictions and decision support [17].

A major gap in translating these technological advances into practice lies in implementation. Despite their promise, clinical decision support (CDS) tools of this nature seldom progress beyond the research stage due to poor usability, lack of alignment with clinician and patient needs, and organizational constraints that hinder adoption [18,19]. These challenges highlight the importance of early, systematic engagement with stakeholders to ensure that CDS tools are designed with contextual, organizational, and human factors in mind.

This protocol describes a participatory qualitative study to inform the development of a multidimensional knee OA decision support tool. Guided by implementation science [20] and design thinking principles, the study will iteratively investigate clinical workflows (phase 1), specify stakeholder values and user requirements (phase 2), and test early prototypes for usability and acceptance (phase 3). By situating technical innovation within a stakeholder-driven design process, this study aims to inform how to bridge the persistent gap between digital health potential and real-world clinical adoption for knee OA management.

Our long-term goal is to develop an objective, evidence-driven decision support tool that promotes consistent, enhanced patient outcomes both during the surgical wait period and after knee arthroplasty surgery. This protocol outlines the qualitative methods guiding the participatory design and development of a CDS tool prototype, which will form the foundation for future research into clinical implementation.

Drawing from literature on technology adoption and sustainability in health care [21,22], this protocol deliberately moves beyond the typical solution-driven approach, whereby system intelligence and technical capacity are the primary focus of design. Instead, it grounds these components in human experiences, offering a more empathetic approach that balances system capabilities with diverse stakeholder perspectives. In the context of implementation, where stakeholder perspectives must span many levels of health care, it is necessary that methodologies be designed to capture, organize, and integrate inputs from multidisciplinary team members and end users with varying roles and levels of expertise.

To accomplish this, our protocol follows a holistic framework guided by the Centre for eHealth and Wellbeing Research (CeHRes) Roadmap [20] and enriched by principles of design thinking [23]. Our methodologies expand on the CeHRes Roadmap by incorporating complementary theoretical constructs known to influence implementation success, scalability, and sustainability, including organizational readiness, health literacy, and contextual resilience [21]. This protocol is grounded in three core principles derived from these frameworks: (1) active engagement and oversight by a multidisciplinary team of key opinion leaders, clinicians, designers, and engineers; (2) iterative evaluation cycles with user-driven feedback loops to ensure that tool functionality reflects the lived experiences and expertise of unique stakeholder groups; and (3) ongoing consideration of adoption and scalability across all stages of research to support future uptake.

In phase 1, we will recruit orthopedic surgeons responsible for evaluating patient appropriateness for knee arthroplasty surgery. These clinicians will participate in individual interviews conducted either at their clinic site or remotely for convenience. They may also optionally participate in field observations of clinical consultations, which will be conducted at their clinic site. A purposive sampling strategy will be used to recruit participants across a range of experience levels, from surgical fellows to senior clinicians. Clinicians and fellows who are not directly involved in perioperative care or decision-making for knee OA will be excluded.

Building on phase 1 outcomes, phase 2 will use respondent-driven sampling to re-engage previous participants and expand recruitment to additional stakeholders whose perspectives are relevant to tool development. Eligible participants will include allied health professionals, health care administrators, and policy-level decision-makers involved in clinical workflow design or technology adoption, whereas industry representatives or individuals not directly connected to clinical implementation will be excluded. Additionally, patients with clinical diagnoses of knee OA who are being evaluated or scheduled for knee arthroplasty surgery will be recruited during OA clinic visits, and those with prior knee arthroplasty surgery may also be included if they are awaiting revision surgery or undergoing surgery on the contralateral limb. Participants from phase 2 will be invited to continue into phase 3, with supplementary recruitment as necessary. To minimize travel burdens, interviews and usability testing in these phases will be offered both in person at participants’ workplaces and virtually using videoconferencing software when appropriate.

This study was reviewed and approved by the Nova Scotia Health Research Ethics Board (file 1031703; approval date: July 28, 2025). Eligible patients will first be asked for verbal consent to be contacted by a co–principal investigator during a clinical encounter and, if interested, will complete written informed consent electronically via REDCap (Research Electronic Data Capture; Vanderbilt University) prior to participation. Clinical staff and decision-makers will be recruited through email invitations and will similarly provide written informed consent before participation. All identifying information will be used solely for scheduling and coordination and will be stored securely in accordance with institutional privacy and confidentiality protocols. Study data will be deidentified prior to analysis. No financial compensation or other incentives will be provided to participants for their involvement in this study.

Estimating sample sizes in qualitative research a priori is inherently challenging due to its evolving and emergent nature. For health care research, Hamilton and Finley [22] advise recruitment of 5 to 10 participants in key roles within a single health care setting. However, saturation often extends beyond this range [24-28]. Acknowledging the role of methodologies, researcher expertise, study objectives, and other factors in reaching saturation, Mthuli et al [29] developed the “define, explain, justify, apply” tool, which has been adopted as our guiding framework:

Ensuring rigor in qualitative research requires strategies aligned with the study’s methodological approach and objectives [43]. Accordingly, this study integrates measures to enhance validity, reliability, and rigor across all stages of data collection and analysis.

To enhance data richness, interview guides will first be pilot-tested with participant representatives, including clinicians, patients, and organizational decision-makers, to refine question clarity, relevance, and reliability prior to data collection [22]. Following pilot-testing, the guides will remain flexible and may be iteratively adapted during data collection in response to emerging themes. Insights generated during the contextual inquiry phase will also inform subsequent sampling strategies, ensuring that the study remains responsive to participants and to data not initially anticipated within the tool’s scope [44]. This adaptive approach will be combined with recruitment strategies designed to promote continuous participant engagement across compounding phases, enabling continuous checking that the data are contextually grounded.

To enhance trustworthiness, real-time member checking will be used during interviews, wherein participant responses will be summarized or restated to confirm accuracy and encourage clarification or elaboration [43]. Given the lack of standardized guidelines for modifying data following transcription or analyses, full transcripts or completed analyses will not be returned to participants for validation [44].

Triangulation will occur at multiple stages to verify data accuracy. Inductive thematic analysis will include ongoing comparison, with independently generated codes shared between the 2 researchers to identify discrepancies, examine interpretive assumptions, and iteratively refine the analytical framework. Additionally, triangulating interview data with direct observations of clinician workflows will help corroborate the findings and identify discrepancies between reported and observed practices (eg, clinician-reported vs researcher-observed workflows) [45].

This study protocol describes a participatory, multiphase qualitative approach to developing a digital CDS tool incorporating gait technologies to support clinical decisions for advanced knee OA guided by the CeHRes Roadmap and design thinking principles. By centering the research process on implementation, our methodologies actively challenge designer bias and prior assumptions, incorporating real-world insights into clinical workflows and stakeholder experiences to validate anticipated barriers and uncover unforeseen challenges to adoption and sustainability. In doing so, our approach extends beyond technical development and directly addresses why CDS tools often fail to achieve adoption or sustainability.

Implementing decision support technologies is an inherently delicate process as they introduce workflows and ideologies that may potentially disrupt established health care ecosystems [46,47]. If executed poorly, this can challenge the longevity of CDS tools, possibly resulting in nonadoption, abandonment, or limited scalability [21]. Long-term viability hinges on tools that effectively recognize and balance diverse personal and operational factors [48], usually through design that acknowledges both typical and edge use cases [49,50]. Our approach to this design challenge is to engage clinicians, patients, health care leaders, and decision-makers from the outset, continuously checking that prototype development meets the distinct goals of these user groups at critical decision points in the development process [51].

Research aimed at advancing health care technologies operates in a challenging space where rapid innovation must be balanced with rigorous, evidence-based methodologies. From a methodological perspective, this study contributes to the broader digital health field by demonstrating how participatory design can be operationalized in the context of CDS tool development. Traditional linear models of health technology design, where design, development, testing, and deployment occur in distinct and sequential stages, provide a logical structure but are poorly suited to the rapid iteration required for digital health innovation [52]. Our 3-phase design introduces structured feedback loops that allow evidence from interviews, observations, and usability testing to directly inform functional requirements and software prototypes. This agile framework combines the rigor of qualitative research with the flexibility required for real-world translation, offering a replicable model for testing other CDS technologies where implementation has proved challenging.

In the context of knee OA, this research also carries important clinical implications. A digital CDS tool that incorporates gait biomarkers has the potential to substantially improve both patient- and system-level outcomes. By providing clinicians and patients with objective, individualized prognostic information, the tool could help optimize surgical timing, reduce inappropriate or premature procedures, and support monitoring in the postoperative period. At the health system level, gait-informed CDS may streamline triage, reduce unnecessary clinic visits, and improve resource allocation while also generating large-scale datasets to advance research on OA progression. Future validation studies and implementation planning will build on the outcomes of this study to assess the tool’s impact on surgical decision-making, patient outcomes, and health care resource use.

While these methods were chosen to enhance the overall quality of this study, we acknowledge the limitation that this research is focused specifically on an orthopedic setting in Nova Scotia in Canada’s health care system, which is a specific ecosystem within a publicly funded, resource-constrained environment. This targeted approach enables a deep, context-specific analysis and a tailored implementation strategy but may limit the generalizability of our findings to other health care environments. However, we view this as a necessary first step in establishing a robust, small-scale implementation model that can later be adapted and scaled to more diverse settings.

This protocol sets out an approach to developing a knee OA decision support tool that is tethered to implementation, combining stakeholder perspectives, participatory design, and iterative refinement. Rather than treating technology development and adoption as separate challenges, our methodologies integrate them into a single process, ensuring that translational tools evolve in step with the needs and realities of their intended users. Although initially situated within an orthopedic context in Nova Scotia, this work represents an important step toward leveraging digital innovation to improve decision-making and patient outcomes in OA care while also offering transferable insights into how complex health technologies can progress from promising concepts to sustainable tools.