Hope for the best, but prepare for the worst – Diagnostic accuracy of the American College of Surgeons National Surgical Quality Improvement Program – Risk model for patients undergoing abdominoplasty after massive weight loss – Results from a Retrospective Cohort Study

Abstract

Background

This study aimed to validate the American College of Surgeons (ACS) National Surgical Quality Improvement Program (NSQIP) risk calculator for predicting outcomes in patients undergoing abdominoplasty after massive weight loss.

Methods

Patients’ characteristics, pre-existing comorbidities and adverse outcomes in our department from 2013 to 2023 were collected retrospectively. Adverse events were defined according to ACS-NSQIP standards and predicted risks were calculated manually using the ACS-NSQIP risk calculator. Binary logistic regression and the Brier score were used to assess the diagnostic accuracy of the model.

Results

Among the 337 individuals who underwent abdominoplasty, 251 had achieved significant weight loss before surgery. After excluding 46 cases due to incomplete data, 205 cases remained for analysis. There were 20% cases of serious complications, 26.3% of some complications, 10.2% of readmissions, 18.8% returned to the operating theatre, 15.6% of surgical site infections and 0.5% each of pneumonia and venous thromboembolism. Although the calculator predicted a 1.5% discharge rate to nursing or rehabilitation facilities and a 0.1% rate of sepsis, neither outcome was observed. Elevated American Society of Anesthesiologists (ASA) status was significantly associated with a higher complication rate, except for surgical site infections (SSI) (p = 0.06). Additionally, an elevated Body Mass Index (BMI) before post-bariatric surgery and a higher resection weight were both associated with increased rates of return to the operating theatre (p = 0.01) and serious complications (p = 0.01). Predicted complication rates (0.1%-8.6%) underestimated actual complication rates (0.5%-26.3%). The Brier scores did not differ significantly from the null model for any outcomes except for general complications (p = 0.001) and logistic regression models demonstrated low sensitivity (0.0-9.8%) and weak odds ratios (1.28-1.46), indicating limited reliability.

Conclusion

The ACS-NSQIP risk calculator does not reliably predict adverse outcomes in this patient cohort.