Predicting surgical risk: exclusion of laboratory data set maintains predictive accuracy

Area of Science:

• Surgical outcomes research
• Predictive modeling in healthcare
• Health services research

Background:

• Predictive accuracy of acuity models plateaus with over 10 variables.
• The utility of laboratory data in surgical outcome prediction models is questioned.

Purpose of the Study:

• To test the hypothesis that excluding laboratory data yields a reliable predictive model.
• To evaluate the impact of laboratory data exclusion on the predictive accuracy of surgical outcome models.

Main Methods:

• Utilized the American College of Surgeons National Surgical Quality Improvement Program data (2005-2008).
• Developed logistic regression models using 86 preoperative variables to predict 30-day morbidity and mortality.
• Compared model performance using area under the receiver operating characteristic (AUROC) values and unpaired t-tests.

Main Results:

• Decreasing the number of variables to 10 or 5 did not significantly alter the AUROC for mortality and morbidity.
• Excluding laboratory values resulted in a slight, statistically insignificant decrease in AUROC for morbidity and mortality.
• Surgical outcome prediction models demonstrated comparable accuracy even without laboratory data.

Conclusions:

• Predictive models for surgical outcomes can achieve similar accuracy without incorporating laboratory data.
• Simplifying predictive models by excluding laboratory data is feasible without compromising reliability.
• This finding supports the development of more streamlined and potentially cost-effective predictive tools in surgery.