Digital Interventions for Knee Osteoarthritis: A Review of Personalization Strategies

Background: Knee osteoarthritis (KOA) is a prevalent and growing health concern, driven by aging populations and rising obesity rates. Although clinical guidelines prioritize lifestyle-based conservative treatments, many patients continue to experience significant pain and disability. eHealth interventions may offer accessible treatment options requiring minimal healthcare provider involvement. However, these interventions often suffer from low adherence and once were one-size-fits-all. Recent research highlights that theory-based, personalized digital interventions may improve engagement and outcomes. However, personalization strategies are inconsistently described and not yet analysed, making it difficult to assess their impact on KOA management.

Objective: This review aims to provide a comprehensive overview of personalization strategies used in digital interventions for patients with KOA, using a previously developed categorization framework.

Methods: This narrative review systematically identified and analysed personalization strategies in digital interventions for KOA. Six databases were searched using terms related to KOA, digital interventions, and personalization. Eligible studies included original research describing personalized digital conservative treatments for KOA patients. Data were extracted and analysed using a previously developed framework categorizing personalization strategies by (1) input (e.g., a measurement), (2) personalization method (e.g., decision rules), and (3) output elements (e.g., an achievement goal). All interventions were analysed across these categories and their subcategories. Data discrepancies were resolved through consensus and authors of the included studies were contacted for clarification when required.

Results: From 1.416 identified records, 52 papers describing 32 unique digital interventions for KOA met the inclusion criteria. Across these interventions, 66 distinct input variables were identified, primarily collected via explicit user input (80%) and monitoring devices (18%). With patient preference and performed exercises being the most commonly used input variables. Most interventions (69%) used rule-based personalization methods; 31% relied on healthcare provider decisions, and 19% incorporated self-learning algorithms. All but one intervention utilized personalized content, with feedback being the category that had the most interventions involving individual and dynamic personalization.

Conclusions: This review provides the first structured analysis of personalization strategies in digital interventions for KOA and demonstrates that a categorization framework can support consistent reporting across diverse study designs. While personalization is widely recognized as important, current implementations remain basic and predominantly dependent on patient self-report and healthcare providers involvement. Heavy reliance on explicit user input may hinder engagement due to increased patient burden and data reliability issues, while reliance on providers increases healthcare workload. As this field continues to develop rapidly, future research should determine which personalization strategies most effectively improve usability, engagement, and clinical outcomes in KOA management.