Machine learning prediction model for postoperative ileus following colorectal surgery

Area of Science:

• Colorectal Surgery
• Machine Learning
• Predictive Analytics

Background:

• Postoperative ileus (POI) is a significant complication after colorectal surgery, affecting approximately 30% of patients.
• Current preventative strategies have limited success in reducing POI incidence.

Purpose of the Study:

• To apply machine learning techniques to identify risk factors for POI in colorectal surgery patients.
• To guide the development of enhanced preventative strategies for POI.

Main Methods:

• A retrospective study of 504 colorectal surgery patients (2018-2022).
• Trained and evaluated multivariate logistic regression, decision trees, radial basis function, and multilayer perceptron (MLP) models.
• Assessed model performance using the area under the receiver operating characteristic (AUROC) curves.

Main Results:

• The MLP model achieved the highest AUROC (0.800), outperforming other models.
• POI occurred in 183 (36%) of the patients.
• The MLP model identified sarcopenia as a significant, potentially modifiable risk factor for POI, alongside known factors like hypokalemia and opioid use.

Conclusions:

• The MLP model demonstrated superior predictive performance for POI compared to traditional methods.
• Machine learning offers valuable insights into ranking predictive variables for POI.
• Preoperative sarcopenia warrants further investigation as a predictive factor for POI.