Assessment of ACS NSQIP’s Predictive Ability for Adverse Events After Major Cancer Surgery

Daniel Borja-Cacho MD, Helen M. Parsons MPH, Elizabeth B. Habermann MPH, PhD, David A. Rothenberger MD, FACS, William G. Henderson MPH, PhD, Waddah B. Al-Refaie MD, FACS
Healthcare Policy and Outcomes
Volume 17, Issue 9 / September , 2010

View full article HTML | View full article PDF | Download Citation

Abstract

Background

The American College of Surgeons National Surgery Quality Improvement Program (ACS NSQIP) has improved operative outcomes in the USA. However, its applicability to oncologic resections at ACS NSQIP hospitals has not been fully explored. We assessed the ability of factors currently collected by ACS NSQIP to predict adverse operative events after major cancer surgery.

Methods

Using pre- and intraoperative factors gathered by the 2005–2008 ACS NSQIP, we constructed logistic regression models to determine their ability to predict 30-day mortality, prolonged length of stay (LOS), major complications or increased number of complications in 15,709 patients who underwent major cancer surgery at 211 hospitals. We assessed each model’s predictive ability using the c-index.

Results

While the mortality rate was relatively low (2.5%), nearly 24% of patients experienced major adverse events. However, up to 43% of patients with prolonged LOS did not have any major complication captured by NSQIP. Furthermore, our model predicting complications showed poor overall predictive ability compared with those predicting mortality and LOS (c-index <0.67 versus 0.80 and 0.73, respectively). When stratified by procedure, the complication model’s predictive ability remained less accurate than models predicting 30-day mortality or prolonged LOS. These results remain unchanged after additional sensitivity analyses.

Conclusions

Current ACS NSQIP variables show low predictive ability for major complications after major oncologic resections. Addition of some disease- and operation-specific variables may be an important consideration in the further evolution of the NSQIP to allow for more accurate predictions of adverse outcomes for major oncologic resections.

Go to Issue Contents