Testing Machine Learning Models to Predict Postoperative Ileus after Colorectal Surgery

Abstract

Background: Postoperative ileus (POI) is a common complication after colorectal surgery, leading to increased hospital stay and costs. This study aimed to explore patient comorbidities that contribute to the development of POI in the colorectal surgical population and compare machine learning (ML) model accuracy to existing risk instruments. Study Design: In a retrospective study, data were collected on 316 adult patients who underwent colorectal surgery from January 2020 to December 2021. The study excluded patients undergoing multi-visceral resections, re-operations, or combined primary and metastatic resections. Patients lacking follow-up within 90 days after surgery were also excluded. Eight different ML models were trained and cross-validated using 29 patient comorbidities and four comorbidity risk indices (ASA Status, NSQIP, CCI, and ECI). Results: The study found that 6.33% of patients experienced POI. Age, BMI, gender, kidney disease, anemia, arrhythmia, rheumatoid arthritis, and NSQIP score were identified as significant predictors of POI. The ML models with the greatest accuracy were AdaBoost tuned with grid search (94.2%) and XG Boost tuned with grid search (85.2%). Conclusions: This study suggests that ML models can predict the risk of POI with high accuracy and may offer a new frontier in early detection and intervention for postoperative outcome optimization. ML models can greatly improve the prediction and prevention of POI in colorectal surgery patients, which can lead to improved patient outcomes and reduced healthcare costs. Further research is required to validate and assess the replicability of these results.

Keywords: artificial intelligence; colorectal cancer; comorbidities; machine learning; neural networks; postoperative ileus.

Figure 1

Demographics—Age (A) and BMI (B) by Gender. Blue: Male, Orange: Female.

Figure 2

Postoperative Ileus by Surgery Type (A) and Surgical Approach (B).

Figure 2

Postoperative Ileus by Surgery Type (A) and Surgical Approach (B).

Figure 3

SHAP Summary Plot for Co-Morbidity Contribution to POI.

Figure 4

Receiver Operating Characteristics for ML Models Predicting POI. Area under the receiver–operator curve for each machine learning (ML) model was calculated. (a) logistic regression model. (b) Random forest model. (c) Bagging classifier Model. (d) AdaBoost (tuned) model. (e) XGBoost (tuned) model. ROC curves for grid and random search tuning for both AdaBoost and XGboost classifier models are similar and so not represented in the figure. ROC curves for stacking classifier were overfitted (as were bagging classifier and random forest models—included as an example) and not represented in the figure. The diagonal red line divides the ROC space. Points above the diagonal represent good classification results (better than random); points below the line represent bad results (worse than random).Any model which is overfitted would not be useful for accurate predictions or conclusions, as it would not discriminate from the training data and defeats the very purpose of machine learning. Thus, based on accuracy, only the highest functioning models were included due to overfitting. For example, panel (b,c) are overfitting [5,9,11,12].

Figure 4

Receiver Operating Characteristics for ML Models Predicting POI. Area under the receiver–operator curve for each machine learning (ML) model was calculated. (a) logistic regression model. (b) Random forest model. (c) Bagging classifier Model. (d) AdaBoost (tuned) model. (e) XGBoost (tuned) model. ROC curves for grid and random search tuning for both AdaBoost and XGboost classifier models are similar and so not represented in the figure. ROC curves for stacking classifier were overfitted (as were bagging classifier and random forest models—included as an example) and not represented in the figure. The diagonal red line divides the ROC space. Points above the diagonal represent good classification results (better than random); points below the line represent bad results (worse than random).Any model which is overfitted would not be useful for accurate predictions or conclusions, as it would not discriminate from the training data and defeats the very purpose of machine learning. Thus, based on accuracy, only the highest functioning models were included due to overfitting. For example, panel (b,c) are overfitting [5,9,11,12].