Construction of a risk prediction model for enteral nutrition feeding intolerance in patients with severe cerebral hemorrhage based on machine learning algorithms

Objective To construct and validate a risk prediction model for enteral nutrition feeding intolerance (FI) in patients with severe cerebral hemorrhage based on machine learning algorithms.

Methods The clinical data of 485 patients with cerebral hemorrhage admitted to the neurological intensive care unit of Northern Jiangsu People's Hospital Affiliated to Yangzhou University from January 2020 to December 2022 were retrospectively analyzed. The patients were randomly divided into training set (n=339) and validation set (n=146) in a 7 to 3 ratio. Five machine learning algorithms were used to construct FI risk prediction models. The receiver operating characteristic (ROC) curve was plotted, and the model with the best predictive performance was selected based on the area under the curve (AUC). A nomogram model was constructed based on the optimal model. The calibration curve and decision curve analysis (DCA) were used to evaluate the calibration and clinical net benefit of the nomogram model.

Results The incidence of enteral nutrition FI in patients with severe cerebral hemorrhage was 38.4%(186/485). Among the five machine learning algorithm models, the Logistic regression model had the best predictive performance(AUC=0.88). The analysis results of the Logistic regression model showed that the use of diuretics, mechanical ventilation, Glasgow Coma Scale score ≤5, vasoactive drugs, and albumin level<35 g/L were risk factors for enteral nutrition FI in patients with severe cerebral hemorrhage. A nomogram model was further constructed based on these five risk factors. The calibration curve analysis showed that the calibration curve fitted well with the ideal curve, indicating a high calibration degree of the nomogram model. The DCA results showed that when the threshold probability was 5% to 73%, the application of the nomogram model for screening could clinically benefit patients.

Conclusion The construction of a nomogram model for predicting the risk of enteral nutrition FI in patients with severe cerebral hemorrhage based on machine learning methods can help to early screen high-risk patients for enteral nutrition FI and timely formulate preventive measures, thereby reducing the incidence of enteral nutrition FI in patients with severe cerebral hemorrhage.