Assessing the Performance of National Surgical Quality Improvement Program Surgical Risk Calculator in Elective Spine Surgery: Insights from Patients Undergoing Single-Level Posterior Lumbar Fusion.

INTRODUCTION: The American College of Surgeons-National Surgical Quality Improvement Program Surgical Risk Calculator is a tool developed to use 21 individual patient characteristics to make predictions for occurrence of 13 general and 2 procedure-specific outcomes. The goal of this study was to evaluate the performance of the Surgical Risk Calculator in predicting outcomes in patients receiving posterior lumbar fusion.

METHODS: American College of Surgeons-National Surgical Quality Improvement Program Participant Use File for 2015 was queried for patients with age ≥18 years undergoing single-level posterior lumbar fusion (PLF) surgery. Individual patient characteristics were entered into the online risk calculator interface to retrieve the predicted estimated risk for perioperative outcomes and complications. Following this, predictive performance was analyzed by computing Brier score, c-statistic, and sensitivity values for all observed outcomes.

RESULTS: A total of 2808 patients undergoing single-level PLF were included in the analysis. Overall, a very low incidence of 30-day postoperative complications was observed with the procedure (0.9%-6.3%). Poor predictive performance was found for all outcomes, including readmissions (c-statistic = 0.63; sensitivity = 15.28%; Brier score = 0.048) and returns to operating room (c-statistic = 0.56; sensitivity = 21.05%; Brier score = 0.032). The best performance was observed for venous thromboembolism (c-statistic=0.66: Brier score = 0.008), although sensitivity was poor (3.85%) on account of low incidence. Predictive performance for length of stay revealed good agreement between observed and predicted values with the exception of prolonged predicted hospital stays (>3.5 days).

CONCLUSIONS: This study assesses the performance of the risk calculator for a homogenous population of patients undergoing a single-level PLF. Although the calculator did not fare well in predicting most outcomes, results need to be interpreted in the context of the low incidence rate of such outcomes.