A LASSO-based nomogram incorporating diabetes mellitus, age, admission NIHSS score, and exclusive NGT feeding demonstrated strong discrimination for predicting 30-day nasogastric tube dependence in acute ischemic stroke patients, with AUC of 0.924 in training, 0.920 in internal validation, and 0.935 in external validation.
Four independent predictors of 30-day NGT dependence were identified: Diabetes Mellitus (DM), age, admission NIHSS score, and exclusive NGT feeding.
The nomogram demonstrated strong discriminative performance across training, internal validation, and external validation sets.
Among three machine learning models compared, the Gradient Boosting Machine (GBM) demonstrated the highest accuracy with an AUC of 0.918.
A total of 852 AIS patients requiring NGT feeding were included and allocated to training and internal validation sets.
Prolonged NGT dependence in AIS patients contributes to adverse clinical consequences, motivating early identification of at-risk patients.
This research studied patients who suffered acute ischemic strokes (brain strokes caused by blocked blood vessels) and needed feeding tubes inserted through their nose and into their stomach because they had difficulty swallowing. The researchers wanted to find a way to predict early on which patients would still need these feeding tubes 30 days after their stroke. Using data from 852 patients, they developed and tested a prediction tool called a nomogram — essentially a scoring chart that doctors can use at the bedside. The study found that four factors were the most important in predicting whether a patient would remain dependent on a feeding tube after 30 days: the patient's age, how severe the stroke was at admission (measured by a standard neurological scale called the NIHSS), whether the patient was only receiving nutrition through the feeding tube (not eating anything by mouth), and whether the patient had diabetes. The prediction tool performed very well, correctly distinguishing patients who would and would not remain tube-dependent about 92–94% of the time across different groups of patients tested. This research suggests that clinicians caring for stroke patients who need feeding tubes could use this nomogram to identify high-risk individuals early, potentially allowing for more targeted interventions such as intensive swallowing therapy or closer nutritional monitoring. The tool's strong performance across multiple patient groups suggests it could be practically useful in clinical settings, though it would benefit from further validation in diverse populations before widespread adoption.
Shang Q, Lei Y, Chen H, Liu Z, Sun M, Wang S, et al.. (2026). LASSO-based nomogram and machine learning models for predicting 30-day nasogastric tube dependence after acute ischemic stroke.. Frontiers in neurology. https://doi.org/10.3389/fneur.2026.1811560