Objective To investigate the prevalence of water-loss dehydration on admission among older patients with hip fractures and to develop a machine learning-based risk prediction model. Methods A retrospective cohort study was conducted in older patients with hip fractures admitted to the Third Hospital of Hebei Medical University between January and December 2024. Ten machine learning algorithms were used to develop prediction models. Model performance was evaluated using the area under the receiver operating characteristic curve (ROC-AUC), accuracy, sensitivity, specificity, and F1 score. Shapley additive explanations (SHAP) were applied to interpret the optimal model. Results A total of 529 patients were included, of whom 165 patients (31.2%) had water-loss dehydration. The random forest model showed the best overall performance, with ROC-AUC values of 0.862 [95% confidence interval (CI) (0.825, 0.899)] in the training set and 0.831 [95%CI (0.762, 0.899)] in the internal validation set. In the internal validation set, the accuracy, sensitivity, specificity, and F1 score were 0.744, 0.800, 0.718, and 0.661, respectively. SHAP analysis showed that time from injury to admission, history of diabetes, and history of previous surgery made the greatest contributions to model predictions. Conclusions Water-loss dehydration is common on admission among older patients with hip fractures. The random forest model demonstrated good performance in internal validation and may support early risk stratification for dehydration in this population. Further external validation is warranted.