Are targeted preoperative risk prediction tools more powerful? A test of models for emergency colon surgery in the very elderly.

BACKGROUND

Whether preoperative risk prediction improves with the use of more patient- and procedure-targeted models is unclear. We created a customized preoperative mortality risk prediction score for patients 80 years or older needing an emergency colectomy and compare it with existing, more generic risk assessment methods.

STUDY DESIGN

A targeted mortality prediction model was created using 2007 to 2008 American College of Surgeons National Surgical Quality Improvement Program (ACS NSQIP) data and was validated using 2005 to 2006 data. We constructed a scoring system from the significant predictors identified. The model fit of our targeted score was compared with the American Society of Anesthesiologist's (ASA) score, the Surgical Risk Scale, and the ACS Colorectal Surgery Risk Calculator.

RESULTS

Analyses identified 1,358 and 372 emergency colectomies in the training and validation samples, respectively. Our targeted risk prediction score had a goodness-of-fit p value greater than 0.05 (indicating a good fit) and a c-statistic of 0.77, which represents a significantly better fit compared with the ASA score, the Surgical Risk Scale, and the ACS Colorectal Surgery Risk Calculator c-statistics (0.66, 0.66, and 0.71, respectively). When using the scores to predict mortality with 80% specificity, our targeted risk prediction score was 25% more likely to predict correctly than the ACS Colorectal Surgery Risk Calculator and 33% more likely to predict correctly compared with the ASA score and Surgical Risk Scale.

CONCLUSIONS

Our study presents a validated preoperative mortality score for very elderly patients needing an emergency colectomy. The greater discriminating power of this targeted score indicates that preoperative risk assessment may need to be customized to specific procedures and patient circumstances.