Incidence and risk factors of chemotherapy-induced diarrhea in colorectal cancer patients based on a random forest model

Objective This study aims to investigate the occurrence of chemotherapy-induced diarrhea (CID) in patients with colorectal cancer and analyze its associated influencing factors. Based on this, a random forest model was introduced to screen and model potential influencing factors. Combined with feature importance ranking and interpretability analysis, the contribution of each factor to the occurrence of CID was further explored, thereby improving the accuracy of high-risk population identification, supporting future individualized risk assessment and nursing management. Methods This was a retrospective study using convenience sampling. A total of 508 patients with colorectal cancer who received chemotherapy at a tertiary first-class hospital in Gansu Province from February 2025 to September 2025 were selected as the study subjects. Data on general information, disease-related information, and laboratory indicators were collected. The occurrence of chemotherapy-induced diarrhea was used as the outcome variable. Univariate analysis was used to screen relevant factors. Based on this, the random forest recursive feature elimination (RF-RFE) method was applied to select the optimal feature subset, and a random forest-based risk identification model was developed. Five-fold cross-validation was used to evaluate model performance. The SHAP method was used for interpretability analysis of the model. Results Among the 508 patients, some developed CID during chemotherapy. Univariate analysis showed that BMI, ECOG score, hemoglobin, albumin, total bilirubin, serum creatinine, diabetes, permanent stoma, and irinotecan use were associated with the occurrence of CID (P < 0.05). The random forest model selected seven optimal features (total bilirubin, irinotecan use, albumin, ECOG score, BMI, serum creatinine, and diabetes). The model achieved an AUC of 0.88, accuracy of 0.85, sensitivity of 0.65, and specificity of 0.93 in the test set. SHAP analysis indicated that total bilirubin and irinotecan use contributed the most to the occurrence of CID. Conclusion The incidence of CID is relatively high in patients with colorectal cancer during chemotherapy. Its occurrence is closely related to factors such as liver function status, chemotherapy regimen, nutritional status, and underlying diseases. The random forest-based prediction model demonstrated good predictive performance and may provide a potential tool for early identification of patients at higher risk of CID and support future individualized nursing management.