Longitudinal comparison of existing frailty indices to predict complications following adult spinal deformity surgery

This study aimed to evaluate and compare the predictive ability of four distinct frailty indices for perioperative and postoperative outcomes in patients undergoing adult spinal deformity (ASD) surgery.

A total of 917 ASD patients with baseline frailty classification and follow-up data available in the analytic registry extract were analyzed. The frailty instruments included the ASD-FI, the mASD-FI, the 5-item modified frailty index (mFI-5), and the FRAIL Scale. Patients were categorized according to the published strata for each instrument. Outcomes included operative burden, in-hospital resource utilization, perioperative complications, long-term complications, reoperation, clinical recovery, and exploratory cost-utility measures. Logistic regression was used for binary outcomes and linear regression-based association analyses for continuous outcomes, with models adjusted for age, BMI, CCI, and surgical invasiveness or complexity. Sensitivity analyses examined collinearity, models with and without CCI, distribution-aware analyses for LOS and SICU duration, and model performance metrics for key binary outcomes. The Benjamini-Hochberg false discovery rate correction was applied to primary outcomes.

Severely frail patients had worse baseline disability and radiographic deformity across all instruments. All four indices were associated with operative complexity, LOS, and SICU duration. The FRAIL Scale was the only index significantly associated with SICU admission (OR = 1.3, p = .011) and hospital-acquired conditions (OR = 1.6, p = .005), and it also showed one of the strongest associations with major perioperative complications. At 2 years, mASD-FI showed the strongest associations with mechanical complications (OR = 1.58, p < .001) and reoperation (OR = 1.55, p < .001), whereas ASD-FI showed the strongest association with broader long-term complication burden and LOS. Sensitivity analyses showed low multicollinearity, but discrimination for key binary outcomes was modest. Cost-utility findings were exploratory.

Frailty was associated with worse outcomes across the ASD surgical course, but the indices were not interchangeable. In this cohort, the FRAIL Scale performed well for index-admission and HAC-related outcomes, whereas mASD-FI showed the strongest signal for long-term mechanical complications and reoperation. Frailty assessment may be useful in ASD surgery, with instrument selection tailored to the outcome of interest.