Equating patient-reported outcomes: developing a psychometric crosswalk between Scoliosis Quality of Life Index (SQLI) and Scoliosis Research Society (SRS-22r) to support long-term follow-up research in patients with adolescent idiopathic scoliosis

BACKGROUND: Despite widespread use of patient-reported outcome measures (PROs), research on measures in adolescent idiopathic scoliosis (AIS) is limited due to scarce large, long-term datasets. Our site began routinely collecting PRO data from patients with AIS in 2007, providing a robust longitudinal dataset. Initially, we used the adolescent-validated Scoliosis Quality of Life Index (SQLI) before transitioning to Scoliosis Research Society (SRS-22r) in 2010. Despite having over 2,400 pre-transition SQLI assessments, inability to convert SQLI to SRS-22r scores has limited our ability to use early data for long-term outcomes research. Recognizing this, we collected SQLI and SRS-22r concurrently, creating a unique opportunity to develop and validate a cross-walk between the two. The primary study aim is to establish such cross-walks, enabling the integration of historical SQLI data into contemporary longitudinal studies and meta-analyses. METHODOLOGY: Using a common persons linking design, this study analyzed SQLI and SRS-22r data from adolescents with idiopathic scoliosis collected during routine clinical care. Calibration and validation samples were formed from patients with concurrent administrations, supplemented by single-instrument visits. Assumptions for linking were evaluated through internal consistency, inter-measure correlations, confirmatory factor analysis, and sex-based invariance testing. Linear equating, equipercentile equating, and Rasch-based true score equating were evaluated using bias, RMSE, correlations, and Bland-Altman plots. RESULTS: All domains met recommended criteria for linking, with strong correlations and low average errors. Linear equating consistently outperformed alternative methods, yielding minimal bias and stable error across score ranges. Linked scores demonstrated high agreement with observed SRS-22r scores for pain, self-image, and mental health, and moderate agreement for function. Linking of change scores resulted in somewhat lower correlations and higher mean error; however, bias remained small, suggesting that longitudinal group-level comparisons are supported. CONCLUSIONS: Crosswalks between historical SQLI and SRS-22r data were established, enabling their integration in long-term studies. While most domain linkages are strong, the function domain shows moderate quality. Aggregated transformed SQLI scores can reliably represent SRS-22r scores, allowing valuable historical data to inform treatment outcomes. This advancement enables more comprehensive longitudinal outcomes research by integrating patient-reported outcomes with clinical measures.