Objective This study aimed to assess the clinical utility of combining the Controlling Nutritional Status （CONUT） score with the Prognostic Nutritional Index （PNI） for evaluating anemia risk in elderly colorectal cancer patients and to establish a risk prediction model. Methods A total of 661 elderly colorectal cancer patients treated at Xinjiang Uygur Autonomous Region People's Hospital from July 2018 to March 2025 were included in this retrospective study. Patients were categorized into anemic and non-anemic groups and randomly assigned to a training set and validation set at a 7∶3 ratio. The XGBoost algorithm was applied to develop a predictive model for anemia risk， and its performance was assessed using the receiver operating characteristic （ROC） curve. SHAP value visualization， and other methods. Results Among the 661 patients， 257 （38.9%） were diagnosed with anemia. Compared with the non-anemic group， patients in the anemic group had significantly lower levels of PNI and albumin， but higher CONUT scores and blood urea nitrogen levels. Additionally， the anemic group had higher proportions of tumor diameter ≥ 5 cm， poorly differentiated tumors， and stage Ⅲ?Ⅳ disease （all P < 0.05）. The XGBoost model demonstrated good discriminatory ability， with an AUC of 0.897 （95%CI：0.868 ~ 0.925）. SHAP value analysis identified PNI， CONUT score， albumin， blood urea nitrogen， TNM stage， tumor differentiation， and tumor size as major contributing variables. PNI and albumin were protective factors， whereas CONUT score， blood urea nitrogen， and tumor-related features were risk factors. Conclusion Nutritional indicators such as PNI and CONUT score， along with tumor characteristics， can effectively predict the risk of anemia in elderly patients with colorectal cancer. The XGBoost-based predictive model demonstrates high discriminatory power and good interpretability， providing valuable support for early screening of high-risk patients and guiding individualized nutritional interventions and anemia management.