Study Design: Retrospective cohort study. Objective: To minimize misclassification of the adult idiopathic scoliosis (IS) phenotype. Summary of Background Data: The contribution of IS to health conditions in adults is difficult to study due to challenges in differentiating IS from other spinal conditions. Materials and Methods: We queried a single large health care system to identify all patients with spine imaging who were 50 years or younger of age when one or more scoliosis ICD codes were documented (n=8778). We reviewed a random subset of 2000 patients to identify true IS (Cobb angle ≥10° with rotation and the absence of congenital or acquired conditions). Multivariable logistic regression models were used to identify demographic and clinical characteristics associated with pain, cosmetic concerns, and/or functional limitations. Results: The presence of IS was validated in 44.6% (870/1950) of participants. Spinal asymmetry and the presence of syndromic scoliosis were classified as nonidiopathic. Sex, age, type of imaging, number of visits with one or more scoliosis ICD codes, and total number of unique scoliosis ICD codes were significantly different between groups. The overall performance of models for predicting the presence of IS was poor, with an area under the curve (AUC) of 0.64. Model performance increased when using higher Cobb angle cutoffs (≥15° AUC: 0.66; ≥20° AUC: 0.69). Prior surgery or Cobb angle ≥15° was associated ( P <0.05) with the increased odds of pain, functional limitation, or cosmetic concerns relative to smaller curves (Cobb 10–14°). Conclusion: Only ~45% of the scoliosis cases captured by the inclusive scoliosis phecode were validated as true IS cases. Predictive models for identifying scoliosis (≥10°) based on these variables were suboptimal. Alternative cutoffs for defining the presence of IS (Cobb angle ≥15°) in adults were associated with better predictive performance. Future large-scale database studies of IS in adults should consider adopting elevated Cobb angle cutoffs to enhance the clinical relevance of research findings.
Taylor et al. (Fri,) studied this question.