Predicting Outcome and Complications in the Surgical Treatment of Adult Scoliosis

In Brief Study Design. prospective, multicenter. Objective. To determine if models for predicting outcome and complications in the setting of adult deformity surgery can be constructed. Summary of Background Data. A classification of adult spinal deformity has been established. The latter categorizes patients radiographically and ties to disability as well as most likely treatment approach. Classification by radiographs alone is insufficient to guide treatment and models for predicting outcome are necessary to establish algorithms. Methods. This is a longitudinal study of adult patients with spinal deformity. All subjects were drawn from a prospective multicenter database. Baseline, 1-year and 2-year radiographic and health related quality of life measures were obtained (SRS, ODI, SF-12). Any peri- or postoperative complications were tabulated. Binary and logistic regressions were applied to create models of outcome and complications. Results. Predictive models of reaching a minimal clinically important difference had an area under the ROC curve above 0.82 and correct classification from 75% to 81%. Development of a complication was correctly classified in 71% of cases. Patients most likely to reach threshold improvement through surgery: older, lower apex deformity, greater disability, higher deformity score by classification. Patients most likely to develop complications: high sagittal balance grade and fusion to sacrum. Conclusion. By applying the classification of adult deformity and considering baseline health status, models for predicting outcome can be established. Further work will refine these models and permit clinical validation. This effort will help establish reliable clinical models such that treatment algorithms can be established for the challenging patients suffering from adult spinal deformity. This is a longitudinal study of adult spinal deformity patients treated surgically. Binary and logistic regressions were applied to create models of clinical outcome and complications (correct prediction up to 81%). Most likely to reach threshold improvement: lower apex deformity, greater disability, and higher deformity score by classification.