Establishing and clinically validating a machine learning model for predicting unplanned reoperation risk in colorectal cancer

Abstract

BACKGROUND

Colorectal cancer significantly impacts global health, with unplanned reoperations post-surgery being key determinants of patient outcomes. Existing predictive models for these reoperations lack precision in integrating complex clinical data.

AIM

To develop and validate a machine learning model for predicting unplanned reoperation risk in colorectal cancer patients.

METHODS

Data of patients treated for colorectal cancer (n = 2044) at the First Affiliated Hospital of Wenzhou Medical University and Wenzhou Central Hospital from March 2020 to March 2022 were retrospectively collected. Patients were divided into an experimental group (n = 60) and a control group (n = 1984) according to unplanned reoperation occurrence. Patients were also divided into a training group and a validation group (7:3 ratio). We used three different machine learning methods to screen characteristic variables. A nomogram was created based on multifactor logistic regression, and the model performance was assessed using receiver operating characteristic curve, calibration curve, Hosmer-Lemeshow test, and decision curve analysis. The risk scores of the two groups were calculated and compared to validate the model.

RESULTS

More patients in the experimental group were ≥ 60 years old, male, and had a history of hypertension, laparotomy, and hypoproteinemia, compared to the control group. Multiple logistic regression analysis confirmed the following as independent risk factors for unplanned reoperation (P < 0.05): Prognostic Nutritional Index value, history of laparotomy, hypertension, or stroke, hypoproteinemia, age, tumor-node-metastasis staging, surgical time, gender, and American Society of Anesthesiologists classification. Receiver operating characteristic curve analysis showed that the model had good discrimination and clinical utility.

CONCLUSION

This study used a machine learning approach to build a model that accurately predicts the risk of postoperative unplanned reoperation in patients with colorectal cancer, which can improve treatment decisions and prognosis.

Keywords: Colorectal cancer; Postoperative unplanned reoperation; Unplanned reoperation; Clinical validation; Nomogram; Machine learning models

Core Tip: This study developed a machine learning model to predict unplanned reoperations in colorectal cancer patients, using data from two hospitals over two years. It employed support vector machine, least absolute shrinkage and selection operator, and extreme gradient boosting for feature selection and logistic regression to identify key risk factors. The model showed good predictive accuracy, validated by receiver operating characteristic curves, calibration curves, and decision curve analysis. Key predictors included age, gender, prior surgeries, and nutritional status. This predictive tool aims to enhance clinical decision-making, reduce reoperation rates, and improve patient outcomes in colorectal cancer care.