Predictive accuracy of the complex 30-day model for colon surgical site infections in emergent operations

Background

Surgical site infections (SSI) following colon surgery continue to pose a significant clinical and financial impact on a healthcare system. The National Healthcare Safety Network (NHSN) utilizes the Complex 30-day SSI model to report the predictive number of infections for a hospital, which is submitted to the Centers for Medicare and Medicaid Services. However, this model does not account for patient acuity. The primary aim of the study was to determine the predictive accuracy of the model for predicting colon SSI in patients undergoing urgent and emergent general surgical procedures.

Methods

The National Surgery Quality Improvement Program (NSQIP) database was queried for variables used in the NHSN Complex 30-day SSI model from 2016 to 2022. Patients ≥18 years old who underwent an urgent/emergent general surgery colon operation were included. Utilizing model variables and coefficients, the predicted number of infections was calculated. A receiver-operating characteristic (ROC) curve was generated and the area under the curve (AUC) calculated to evaluate the model's predictive accuracy.

Results

Of the 476,178 patients included, 51 ​% were female with a mean age of 58 (±24), 15.7 ​% had diabetes, 35.2 ​% had body mass index (BMI) ​≥ ​30mg/kg2 and 61.1 ​% has ASA classification of 3/4/5. The overall infection rate was 4.3 ​%. In patients sustaining an SSI, 16.6 ​% had diabetes, 36.7 ​% had BMI ≥30mg/kg2, and 68.0 ​% had an ASA class of 3/4/5. The ROC curve revealed an AUC of 0.554 (95 ​% CI: 0.5498, 0.5577.)

Conclusions

Our findings demonstrate the Complex 30-day SSI model currently has a predictive accuracy only slightly better than chance in patients undergoing urgent or emergent general surgical colon operations. This may have significant impact on hospitals caring for patients with a higher acuity. Future models should take into consideration the acuity of the operation in efforts to improve the predictive accuracy in this patient population.