Development and validation of an interpretable machine learning model to predict high-risk patients with acute small bowel obstruction