Predictive Analytics for Determining Extended Operative Time in Corrective Adult Spinal Deformity Surgery

BackgroundMore sophisticated surgical techniques for correcting adult spinal deformity (ASD) have increased operative times, adding to physiologic stress on patients and increased complication incidence. This study aims to determine factors associated with operative time using a statistical learning algorithm.MethodsRetrospective review of a prospective multicenter database containing 837 patients undergoing long spinal fusions for ASD. Conditional inference decision trees identified factors associated with skin-to-skin operative time and cutoff points at which factors have a global effect. A conditional variable-importance table was constructed based on a nonreplacement sampling set of 2000 conditional inference trees. Means comparison for the top 15 variables at their respective significant cutoffs indicated effect sizes.ResultsIncluded: 544 surgical ASD patients (mean age: 58.0 years; fusion length 11.3 levels; operative time: 378 minutes). The strongest predictor for operative time was institution/surgeon. Center/surgeons, grouped by decision tree hierarchy, a and b were, on average, 2 hours faster than center/surgeons c-f, who were 43 minutes faster than centers g-j, all P < 0.001. The next most important predictors were, in order, approach (combined vs posterior increases time by 139 minutes, P < 0.001), levels fused (