Designing for Adoption: Implementing Digital Twins and Decision Support After ACL Reconstruction

Musculoskeletal (MSK) conditions are the leading cause of disability in the United States, affecting more than 126 million adults and accounting for over $200 billion annually in direct and indirect costs.20 Despite this burden, clinical decision-making in MSK rehabilitation remains largely reactive and population-based, relying on isolated assessments and coarse thresholds rather than integrated, individualized evidence. Anterior cruciate ligament (ACL) injury, which disproportionately affects adolescents and young adults and requires a series of time-sensitive decisions regarding rehabilitation progression and return to activity, represents a high-impact and time-bounded model for studying MSK recovery and long-term joint health. More than 200,000 ACL injuries occur annually, and up to 30% of individuals experience reinjury or develop early post-traumatic osteoarthritis within a decade of reconstruction.21-23 Although ACL recovery generates rich longitudinal data, these data are rarely synthesized to inform individualized recovery trajectories or to identify patients who are deviating toward poor outcomes early enough to intervene. As a result, clinicians lack practical tools to distinguish expected from concerning recovery patterns and to tailor rehabilitation strategies accordingly. Emerging digital twin–inspired approaches that use data-driven models to represent and update individual recovery trajectories over time offer a promising framework for addressing this gap by transforming routinely collected data into clinician-centered decision support. Still, their application in MSK rehabilitation has not yet been developed, evaluated, or integrated into real-world clinical workflows. As Technical Director of the ACL Reconstruction and Rehabilitation Outcomes Workgroup (ARROW), I have contributed to prototype tools that underscore both the feasibility of this approach and the critical unmet need for scalable, workflow-compatible decision support. The objective of this study is to address the lack of individualized, data-informed decision support in ACL rehabilitation by developing and evaluating a digital twin–informed clinical decision support tool for young people recovering from ACL injury. The overarching hypothesis of this proposal is that integrated, longitudinal representations of recovery can be translated into clinician-centered decision support that is both interpretable and actionable within real-world rehabilitation workflows. A secondary hypothesis is that the feasibility, usability, and sustainability of such tools depend on stakeholder-informed design and implementation strategies. This work will generate preliminary evidence and formal training in implementation science to support a future R01 proposal. We will test these hypotheses through the following specific aims: AIM 1: Develop interpretable digital twin models using the ARROW dataset. We will use multi-modal, longitudinal data from pediatric and young adult patients recovering from ACL reconstruction to build preliminary digital twin representations that prioritize interpretability, temporal stability, and clinical relevance.Hypothesis: A Bayesian digital twin–inspired model will produce stable and reproducible estimates of individual recovery trajectories, with explicitly quantified uncertainty, allowing meaningful characterization of how recovery evolves within and across individuals. AIM 2: Evaluate the implementation determinants of an ACL clinical outcomes dashboard. We will conduct a theory-driven implementation evaluation of an existing cohort-comparison ACL outcomes dashboard to systematically identify barriers, facilitators, and contextual factors influencing clinical decision support adoption in ACL rehabilitation settings. Using the Consolidated Framework for Implementation Research and Proctor implementation outcomes frameworks, we will assess acceptability, appropriateness, feasibility, and usability among multidisciplinary clinicians. Hypothesis: Implementation determinants identified for cohort-based decision support will reveal workflow integration requirements, interpretability needs, and clinician reasoning patterns that are generalizable to more complex longitudinal decision support tools, including digital twin approaches. AIM 3: Identify stakeholder-informed strategies for future integration of digital twin tools into clinical workflows. We will conduct semi-structured focus groups with clinicians and other implementation stakeholders to explore barriers, facilitators, and desired adaptations for real-world use of digital twin–informed decision support. Qualitative analysis will inform tool refinement and future implementation strategies. Hypothesis: Stakeholders will identify specific workflow integration points and delivery preferences that can guide future implementation efforts. Overall Impact: By centering on ACL injury and applying rigorous implementation science frameworks, this project will enable the translation of longitudinal recovery data into clinician-centered tools that support personalized, time-sensitive decision-making within routine care workflows, shifting digital twin research from technical feasibility toward clinical impact. This award will provide the candidate with a comprehensive and novel skill set in implementation science, human-centered design, and translational data science, positioning them to lead future NIH R-series studies that rigorously evaluate digital twin–guided interventions aimed at reducing reinjury risk, preventing early osteoarthritis, and improving long-term function across the lifespan.