Simplifying ICF Assessment for Individuals with Spinal Cord Injury with Lower Limb Dysfunction Using Item Response Theory and Graphical Models.

Abstract

To develop and validate a parameterized assessment tool based on the ICF for individuals with SCI with lower limb dysfunction by applying item response theory (IRT) and graphical modeling (GM). Cross-sectional, multicenter study. Multiple rehabilitation centers in China. 252 participants with SCI and lower limb dysfunction. Not applicable. Participants were assessed using the ICF questionnaire, the LEMS, the SCIM, the SF-12 and the MBI. We utilized Mokken scale analysis and parametric item response modelling to evaluate these items. Furthermore, we employed IRT to develop both non-parametric IRT (npIRT) and parametric IRT (pIRT) models for psychometric measurement. To further investigate the complex relationships among ICF functional items and its interventional implications, we applied Graphical Modeling (GM) to visualize the network structure of functional categories and computed the Overall Benefit Index . The four-parameter logistic model, comprising 38 ICF items, was identified as the optimal model. This model demonstrated good reliability, with a Cronbach's α coefficient of 0.9445 and a latent class reliability coefficient of 0.9708. The estimated person ability parameter (θ) showed significant positive correlations with external measures. The results also confirmed the measurement equivalence of the final model across different genders, levels of self-care ability, clinical stages, and age groups. GM indicated that interventions targeting key functional items, such as d420 (Transferring Oneself) and d415 (Maintaining a Body Position) were expected to produce significant overall functional benefits for patients. The 38-item ICF-based assessment tool demonstrated good structural validity and internal consistency. It holds significant potential for informing functional assessment and formulating individualized intervention strategies for this population.