Implementation, feasibility and optimization of a web- and mobile-supported semi-automated shoulder rehabilitation...

Shoulder disorders are common and cause pain and functional limitation. Conventional rehabilitation is often hindered by variable assessment, high therapist time demands, and inconsistent adherence. Digital rehabilitation models may reduce variability by standardizing data capture and linking quantified deficits to exercise prescription. However, evidence from routine clinical practice on digitally supported and semi-automated assessment-to-prescription workflows remains limited. This study assessed the feasibility and optimization of a web- and mobile-supported semi-automated shoulder rehabilitation system (WMSSR).

We implemented WMSSR in a prospective descriptive longitudinal interventional study conducted in routine clinical practice, without randomization or a control group. Between March and September 2025, 129 consecutive patients with non-traumatic shoulder pain or dysfunction persisting for more than 3 months were included. WMSSR standardized data collection and individualized exercise prescription based on 24 clinical tests grouped into eight functional clusters. The primary outcome was change in total clinical deficit score from baseline to discharge, analyzed using paired t tests, 95% confidence intervals, and effect sizes, with nonparametric sensitivity analysis. To optimize the assessment battery, tests were ranked using baseline deficit prevalence and magnitude together with responsiveness indices (SRM) and within-subject effect size metrics (paired Cohen's d₍z₎ and Wilcoxon r). False discovery rate was controlled using the Benjamini-Hochberg procedure.

The mean total deficit score decreased from 15.4 ± 5.6 at baseline to 1.6 ± 1.2 points at discharge (mean change 13.8 ± 5.5; ≈89% reduction; 95% CI 12.83-14.75; effect size d₍z₎ = 2.513; p < 0.001), consistent across parametric and nonparametric sensitivity analysis. Composite shoulder pain decreased from 4.5 ± 1.6 to 0.4 ± 0.6 on a 0-10 visual analog scale. Episodes were discharged after a mean of 84.9 ± 42.0 days (≈ 2.8 months). Test-battery optimization integrating baseline deficit prevalence and magnitude and responsiveness identified a reduced 11-test battery with high baseline prevalence and large responsiveness indices.

In this single-center prospective descriptive longitudinal interventional study, the web- and mobile-supported WMSSR workflow was feasible to implement in a supervised routine-care setting and was associated with substantial within-episode clinical improvement among patients with non-traumatic shoulder pain or dysfunction who completed the rehabilitation episode.

The study protocol was registered at ClinicalTrials.gov (NCT07301671), retrospectively registered on 26/11/2025.