Machine learning-driven prediction of hospital admissions using gradient boosting and GPT-2.

Area of Science:

• Medical Informatics
• Machine Learning in Healthcare
• Clinical Decision Support Systems

Background:

• Accurate prediction of hospital admissions from the emergency department (ED) is vital for optimizing patient care and resource allocation.
• Current prediction methods often rely on limited data types, potentially impacting accuracy.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting hospital admissions from the ED.
• To assess the impact of integrating structured clinical data with unstructured text data on prediction performance.

Main Methods:

• Utilized data from the 2021 National Hospital Ambulatory Medical Care Survey-Emergency Department (NHAMCS-ED).
• Employed a Gradient Boosting Classifier for structured data and a fine-tuned GPT-2 model for unstructured text (chief complaints, injury descriptions).
• Developed a combined model by averaging predictions from both individual models and evaluated using 5-fold cross-validation.

Main Results:

• The combined model achieved 75.8% accuracy, outperforming the structured data model (73.8%) and the unstructured data model (64.6%).
• The combined model demonstrated superior performance with the highest area under the receiver operating characteristic curve (AUC-ROC).
• Sensitivity and specificity for the combined model were 75.8%.

Conclusions:

• Integrating structured and unstructured data with machine learning models significantly enhances the prediction of hospital admissions from the ED.
• This hybrid approach offers a promising strategy for improving clinical decision-making and optimizing emergency department operations.