JMIR Publications

ABSTRACT

Background:

Clinical guidelines recommend a stepped-care strategy for patients with hip and knee osteoarthritis (OA), starting with non-operative, self-management approaches. However, the implementation of stepped-care remains limited. Digital self-management interventions have the potential to support patients in applying lifestyle advice and self-care strategies, but current tools often provide generic support without long-term continuity. Artificial Intelligence (AI) offers new opportunities to deliver personalized, just-in-time self-management interventions for people with OA. The development of such AI algorithms is limited due to a lack of rich, longitudinal datasets.

Objective:

Therefore, the objective of this study is to develop and evaluate data-driven models that support personalized recommendations regarding the optimal timing of self-care programs for individuals with hip and/or knee OA.

Methods:

This prospective cross-over study aims to include 600 people with hip and/or knee OA, meeting the NICE criteria. The study is registered in ClinicalTrials.gov (registered 20-02-2026, NCT07423858). We aim to specifically include a subset of people with limited digital health literacy (30% of the sample). Participants will be recruited across the Netherlands and use the ArtroseCoach web application, with three self-care programs (i.e., physical activity promotion, weight management, and sleep optimization). Each participant will complete all three 12-week programs and one 12-week control period in a randomized sequence. Participants will be followed for 12 months, with biweekly assessments conducted via the application. The primary outcome for model development will be physical functioning over time. Secondary outcomes will be pain and participation. Additional measures include patient characteristics, physical activity, sleep, psychosocial factors, behavioral determinants, and engagement with the application. These data will inform the development of the data-driven models using supervised (causal) machine learning.

Results:

Inclusion for this cross-over study started in September 2025 and is expected to finish in March 2026. In February 2026, 520 of 600 participants were included.

Conclusions:

This study will provide data-driven models that forecast changes in physical functioning over time and support personalized recommendations on the optimal timing of specific self-care programs for people with hip and/or knee OA. These models will be integrated into a new iteration of a self-management application, ArtroseCoach version 2, to provide personalized support for people with OA across varying levels of digital health literacy. Clinical Trial: The study is registered in ClinicalTrials.gov (registered 20-02-2026, NCT07423858).