Objective:To investigate the associated factors of bone marrow suppression in elderly patients with colorectal cancer(CRC)undergoing chemotherapy,construct a risk prediction model to identify key determinants of bone marrow suppression,and pro-vide data-driven support for personalized clinical treatment strategies.Methods:The clinical data of 32 elderly CRC patients who expe-rienced bone marrow suppression(study group)and 70 patients who did not experience bone marrow suppression(control group)were retrospectively analyzed.Univariate and multivariate Logistic regression analyses were conducted to identify the factors influencing bone marrow suppression in elderly CRC patients undergoing chemotherapy.A risk prediction model was subsequently developed and valida-ted.Results:The results of the Logistic regression analysis indicated that advanced age,prolonged duration of chemotherapy,a Nutrition-al Risk Screening-2002(NRS-2002)score of≥3,combined chemotherapy,and low serum levels of white blood cell count(WBC),he-moglobin(HGB),and platelet count(PLT)were risk factors for bone marrow suppression in elderly CRC patients(P＜0.05).Based on these findings,an XGBoost model was developed.The top five most influential features identified by the model were prolonged chem-otherapy duration,NRS-2002 score≥3,combined chemotherapy,low serum WBC and HGB levels,followed by advanced age and low PLT levels.Evaluation using the ROC curve demonstrated that the AUC of the XGBoost model for predicting myelosuppression following chemotherapy in elderly CRC patients was 0.905,with a sensitivity of 83.36%and specificity of 92.35%.The Hosmer-Lemeshow test revealed no significant difference between the predicted and actual observed values(P>0.05).Conclusion:Advanced age,prolonged duration of chemotherapy,NRS-2002 score≥3,combined chemotherapy regimen,and low expression levels of serum WBC,HGB,and PLT are independent risk factors for myelosuppression in elderly CRC patients undergoing chemotherapy.The XGBoost model construc-ted based on these factors demonstrates excellent predictive performance.