Predicting major complications and discharge disposition after adult spinal deformity surgery

Background Context: Several spine-specific comorbidity indices are available to help risk-stratify patients before they undergo invasive spine procedures. Studies of patients with adult spinal deformity (ASD) typically use the Charlson Comorbidity Index (CCI), which is not specific to spine patients. Purpose: To compare the CCI with the Seattle Spine Score (SSS), the Adult Spinal Deformity-Comorbidity Score (ASD-CS), and the Modified 5-Item Frailty Index (mFI-5) and identify which tool more accurately predicted major perioperative complications and discharge disposition after ASD surgery. Study Design/Setting: Retrospective review. Patient Sample: Patients with ASD who underwent spinal arthrodesis of at least four levels at a single institution. Outcome Measures: Self-reported measures include SSS, ASD-CS, and mFI-5. Functional measures include the CCI. Methods: We retrospectively reviewed records of 164 patients with ASD who underwent spinal arthrodesis of ≥ four levels from January 2008 to February 2018 at our U.S. academic tertiary care center and who had available Oswestry Disability Index values. To assess the predictive ability of the comorbidity indices, we created five multivariable logistic regression models, with the presence of major complications and discharge disposition (home or inpatient rehabilitation) as the primary outcome variables. The base model used validated demographic and surgical factors that were predictors of complications and outcomes in those with ASD and within the broader spinal literature. The other four models used the base model along with one of the four indices. The predictive ability of each model was compared using goodness-of-fit testing, with higher pseudo-R² values and lower Akaike information criteria (AIC) values indicating better model fit. Results: Thirty-one patients (19%) experienced major perioperative complications, and 68 (42%) were discharged to inpatient rehabilitation facilities (vs home). The model using the SSS had the highest pseudo-R² value and lowest AIC value for both major complications and discharge disposition. The mFI-5 had a similar predictive ability. The models using the CCI and ASD-CS were weaker predictors. Conclusions: Compared with the CCI and the ASD-CS, the SSS and the mFI-5 were strong predictors of major complications and discharge disposition after ASD surgery. These results suggest that the SSS and the mFI-5 are preferable to the CCI for clinical risk stratification and outcomes research in patients undergoing ASD surgery.