Factor analysis of the SRS-22 outcome assessment instrument in patients with adult spinal deformity.

ABSTRACT

PURPOSE: Designed for patients with adolescent idiopathic scoliosis, the SRS-22 is now widely used as an outcome instrument in patients with adult spinal deformity (ASD). No studies have confirmed the four-factor structure (pain, function, self-image, mental health) of the SRS-22 in ASD and under different contexts. Factorial invariance of an instrument over time and in different languages is essential to allow for precise interpretations of treatment success and comparisons across studies. This study sought to evaluate the invariance of the SRS-22 structure across different languages and sub-groups of ASD patients. METHODS: Confirmatory factor analysis was performed on the 20 non-management items of the SRS-22 with data from 245 American English-, 428 Spanish-, 229 Turkish-, 95 French-, and 195 German-speaking patients. Item loading invariance was compared across languages, age groups, etiologies, treatment groups, and assessment times. A separate sample of SRS-22 data from 772 American surgical patients with ASD was used for cross-validation. RESULTS: The factor structure fitted significantly better to the proposed four-factor solution than to a unifactorial solution. However, items 14 (personal relationships), 15 (financial difficulties), and 17 (days off work) consistently showed weak item loading within their factors across all language versions and in both baseline and follow-up datasets. A trimmed SRS (16 non-management items) that used the four least problematic items in each of the four domains yielded better-fitting models across all languages, but equivalence was still not reached. With this shorter version there was equivalence of item loading with respect to treatment (surgery vs conservative), time of assessment (baseline vs 12 months follow-up), and etiology (degenerative vs idiopathic), but not age (< vs ≥50 years). All findings were confirmed in the cross-validation sample. CONCLUSION: We recommend removal of the worst-fitting items from each of the four domains of the SRS-instrument (items 3, 14, 15, 17), together with adaptation and standardization of other items across language versions, to provide an improved version of the instrument with just 16 non-management items.