56. Predictive models identify patient and surgical variables that synergistically produce an optimal outcome following adult spine deformity (ASD) surgery

Abstract

<h3>BACKGROUND CONTEXT</h3> Identifying the components needed for an optimal adult spinal deformity (ASD) surgical outcome could help inform surgeons and improve ASD treatment. However, defining the components that create an optimal outcome for ASD surgery is challenging because outcome metrics depend on multiple perspectives. A patient may receive a technically successful complication-free surgery but may report dissatisfaction with the treatment outcome. Conversely, a patient may incur multiple postoperative complications with high societal cost but report satisfaction with the surgical outcome. <h3>PURPOSE</h3> (1) Define an "optimal" ASD surgical outcome by integrating outcome metrics from multiple perspectives, and (2) identify the patient-specific and surgical components of ASD treatment that surgeons can employ to improve ASD surgery by creating a model that predicts a multi-perspective "optimal" surgical outcome. <h3>STUDY DESIGN/SETTING</h3> Prospective, multicenter analysis. <h3>PATIENT SAMPLE</h3> ASD patients enrolled into a prospective multicenter study. <h3>OUTCOME MEASURES</h3> Scoliosis Research Society-22r questionnaire (SRS-22r), Oswestry Disability Index (ODI), postoperative complications. <h3>METHODS</h3> Surgically treated ASD patients prospectively enrolled into a multicenter study from 2009-2018 were assessed at minimum 2-year follow-up for optimal outcome defined as (1) no major postoperative complication or complication requiring surgery, (2) patient reached MCID for ODI and SRS-22r subscore, and (3) patient satisfied and indicates would have the surgery again. Demographic, radiographic, PROM and surgical variables were assessed for associations with optimal outcome. Multivariate regression models were built based on level of upper instrumented vertebra (UIV) to identify variables that created a best fit predictive model for optimal outcome by R2 maximization and AIC/BIC minimization. <h3>RESULTS</h3> Of 1291 patients, 788 (mean 3.5 years follow-up), were eligible for study and evaluated. Optimal outcome patients (OP; n=196) had less preoperative opioid use (47.5% vs 56.8%) and fewer histories of prior spine surgery (65.4% vs 77.3%) than nonoptimal outcome (NO; n=592), respectively (p 0.05). Creation of the best fit predictive model for optimal outcome demonstrated synergy between several modifiable variables including preoperative BMI and opioid and tobacco use, final SVA and scoliosis, and use of supplemental rods and PJF prophylaxis. Refining the model for specific surgeries based upon UIV demonstrated increased synergistic impact of the modifiable variables and predictive accuracy (thoracolumbar UIV R2=0.41; upper thoracic UIV R2=0.77). <h3>CONCLUSIONS</h3> No single surgical or radiographic variable is independently predictive of an a priori defined multiperspective optimal outcome following ASD surgery. However, predictive modeling identified preoperative BMI, opioid and tobacco use, final SVA and scoliosis, and use of supplemental rods and PJF prophylaxis as variables surgeons can optimize and/or employ that act synergistically to predict an optimal ASD surgical outcome. Future research will focus on development of predictive models that highlight the synergistic effects of patient specific and interventional variables to improve surgical outcomes. <h3>FDA DEVICE/DRUG STATUS</h3> This abstract does not discuss or include any applicable devices or drugs.