Related Experiment Videos
Predicting Failure to Rescue After Surgery for Degenerative Cervical Myelopathy: An ACS-NSQIP Analysis (2015-2020)
Sean T Kelleher1, Jerick Kim2, Shubhang Bhalla3
1Lewis Katz School of Medicine at Temple University, Philadelphia, PA 19140.
Background Context:
Failure to rescue (FTR), defined as 30-day mortality among patients who experienced a major complication, has emerged as a quality metric in spine surgery. Recent FTR analyses report high discrimination using prediction models combining pre- and postoperative variables into a single model. Whether this performance reflects true preoperative risk stratification or the inclusion of peri-mortality events as predictors remains unknown.
Purpose:
To identify predictors of FTR in patients undergoing surgery for degenerative cervical myelopathy (DCM) and compare the discrimination of preoperative-only versus combined preoperative-postoperative models.
Study Design:
Retrospective cohort study using the American College of Surgeons National Surgical Quality Improvement Program (ACS-NSQIP) database from 2015-2020.
Patient Sample:
19,108 adult patients undergoing cervical decompression and/or fusion for DCM. 938 (4.9%) experienced a major complication and formed the FTR cohort.
Outcome Measures:
FTR, defined as 30-day all-cause mortality among patients who experienced a major complication within 30 days of surgery.
Methods:
Two multivariable logistic regression models predicted FTR in the complicated cohort: Model A (preoperative variables only) and Model B (preoperative variables plus composite postoperative complication categories: cardiopulmonary, septic/wound, and other). Internal validation used 1,000 bootstrap iterations with optimism correction.
Results:
Of 19,108 DCM surgical patients, 938 had a major postoperative complication, and 48 (5.1%) experienced FTR. FTR patients demonstrated higher Risk Analysis Index (RAI) scores (median 25 in FTR vs. 22 in rescued; p=0.005), diabetes prevalence (54.2% vs. 32.4%; p=0.003), and preoperative septic shock (12.5% vs. 3.3%; p=0.004). Model A identified RAI (OR 1.07 per point; 95% CI 1.03-1.12), diabetes (OR 2.18; 95% CI 1.18-4.01), and preoperative septic shock (OR 3.22; 95% CI 1.21-8.56) as independent predictors (corrected C-statistic 0.670). Model B identified cardiopulmonary complications as the strongest predictor (OR 35.4; 95% CI 8.27-151.6), while RAI (OR 1.06; 95% CI 1.02-1.10; p=0.003) and diabetes (OR 2.01; 95% CI 1.08-3.72; p=0.027) remained independent predictors. Model B significantly outperformed Model A (corrected C-statistic 0.816; bootstrap ΔC 0.119, 95% CI 0.042-0.189; DeLong p<0.001).
Conclusions:
Among DCM patients who experienced a major complication, frailty, diabetes, and preoperative septic shock were associated with FTR with modest discrimination. Discrimination improved significantly with the addition of postoperative complications, particularly cardiopulmonary events. These findings support frailty-based preoperative risk stratification and identify cardiopulmonary surveillance as a key FTR mitigation target.