A retrospective study in Shenzhen Hospital, China analyzed mechanically ventilated ICU patients who received enteral nutrition for the first time from January 2022 to December 2024. The cohort was divided into a modeling set (664 patients) and a validation set (284 patients). In the modeling cohort, 300 patients (45.18%) developed refeeding syndrome (RFS). Independent risk factors were age ≥ 60 years, NRS-2002 score ≥ 3 points, SOFA score ≥ 10 points, APACHE II score, albumin before feeding < 30 g/l (p < 0.05). Using LASSO regression and logistic analysis, a nomogram prediction model was created. The model achieved an AUC of 0.859 (95% CI: 0.815–0.903) in the modeling cohort and 0.832 in the validation cohort. The calibration curve showed a good agreement between predictions and reality. The model enables identification of ICU patients at high risk of RFS for early screening and intervention.