Development of a Social Functioning Assessment Using Computerized Adaptive Testing for Patients With Stroke

Objective: To develop a computerized adaptive test of social functioning (Social-CAT) for patients with stroke. Design: This study contained 2 phases. First, a unidimensional item bank was formed using social-related items with sufficient item fit (ie, infit and outfit mean square [MNSQ]). The social-related items were selected from 3 commonly used patient-reported quality-of-life measures. Items with differential item functioning (DIF) of sex were deleted. Second, we performed simulations to determine the best set of stopping rules with both high reliability and efficiency. The participants' responses to the items were extracted from a previous study. Setting: Rehabilitation wards and departments of rehabilitation/neurology of 5 general hospitals. Participants: Patients (N=263) with stroke (47.1% were inpatients). Interventions: Not applicable. Main Outcome Measure: Social-CAT. Results: The unidimensionality of the 24 selected items was supported (infit and outfit MNSQs =0.8–1.2). One item had DIF of sex and was deleted. The item bank was composed of the remaining 23 items. With the best set of stopping rules (person reliability ≥.90 or limited reliability increased ≤.001), the Social-CAT used on average 10 items to achieve sufficient reliability (average person reliability =.88; 81.0% of the patients with reliability ≥.90). Conclusions: The Social-CAT appears to be a unidimensional measure with acceptable reliability and efficiency, and it could be useful for both clinicians and patients in time-pressed clinical settings.