Model Development Cohort Definition: 3,000 patients used to identify risk factors and build the prediction mode.

External Validation Cohort Definition: 2,000 independent patients from multiple centers to test the model's accuracy.

Measurement Tool: Standardized endoscopic gastric residual volume grading system ： Grade 0：Stomach completely clean, no visible residue； Grade 1：Small amount of clear liquid (<1.5 mL/kg or <100 mL), no solid components； Grade 2：Liquid ≥1.5 mL/kg (or ≥100 mL) or small solid fragments (<5 mm)； Grade 3：Large amount of liquid (>200 mL) or significant solid residue (≥5 mm fragments) Grade 4：Visible undigested food chunks or intact food pieces Data Collection: Two physicians independently assessed gastric residue (liquid/solid), measuring volume using a graduated suction bottle, with video recordings archived for third-party review.

Statistical Methods: The incidence (%) of gastric retention (≥Grade 2) was calculated. Multivariate logistic regression analyzed risk factors, and ROC curves evaluated the model's predictive performance (AUC).

The standardized Gastroparesis Cardinal Symptom Index (GCSI) was used to quantify the following core symptoms:

Nausea (frequency & intensity): 0 (none) to 5 (severe, persistent daily).

Vomiting (episode frequency): 0 (none) to 5 (≥3 times/day).

Postprandial fullness (duration): 0 (none) to 5 (>4 hours).

Early satiety (food intake limitation): 0 (normal eating) to 5 (eating <25% of usual amount).

Statistical Analysis: The correlation between GRV grades (0-4) and the total GCSI score was assessed using Spearman's rank correlation coefficient.