The aim of this study was to carry out a psychometric analysis, using both Classical Test Theory and Rasch analysis (RA) methods, of the Iowa Level of Assistance Scale (ILAS) administered in patients with recent total hip arthroplasty (THA) or total knee arthroplasty, to examine its metric properties and provide insights for a refined version. A total of 203 patients who had undergone THA or total knee arthroplasty were assessed using the ILAS for assistance needed during functional activities (ILAS-funct) and need for assistive devices (ILAS-dev) before discharge from the orthopedic ward. The responses were psychometrically analyzed using the Classical Test Theory and RA. Cronbach's α was adequate only for group comparisons (ILAS-funct, 0.82; ILAS-dev, 0.79). The two domains (ILAS-funct and ILAS-dev) showed a good correlation. According to RA rating scale diagnostics, ILAS-funct showed two disordered response category thresholds: of the seven different response levels of 'assistance', only five were appreciably discernible. All five ILAS-funct items fitted the model and did not show either local dependence or differential item functioning across age groups or sex. ILAS-dev presented two unused response categories, which precluded Rasch calibration and subsequent analyses. ILAS-funct showed sound psychometric properties, but the rating system of ILAS-funct could be simplified, at least collapsing the response categories 5 (failed) and 6 (not tested). In ILAS-dev, the presence of unused response categories calls for a reconsideration of its scaling options and methods.