A Natural Language Processing Framework for Assessing Hospital Readmissions for Patients With COPD.

Area of Science:

• Health informatics
• Clinical data analysis
• Natural Language Processing (NLP)

Background:

• Federal legislation mandates hospitals to reduce readmission rates, particularly for chronic diseases.
• Chronic obstructive pulmonary disease (COPD) is a significant contributor to hospital readmissions and incurs financial penalties.
• Existing readmission prediction models often overlook unstructured clinical notes.

Purpose of the Study:

• To propose a novel framework for predicting hospital readmissions using NLP on clinical notes.
• To develop a component selection strategy for optimizing machine learning algorithms in this context.
• To identify patients at high risk of readmission for COPD.

Main Methods:

• Development of a framework leveraging Natural Language Processing (NLP) to analyze unstructured clinical notes.
• Implementation of a component selection framework to identify optimal data mining and machine learning algorithms.
• Utilizing Naïve Bayes with Chi-Squared feature selection for readmission prediction.

Main Results:

• The proposed framework achieved an Area Under the Curve (AUC) of 0.690 for readmission prediction.
• Naïve Bayes with Chi-Squared feature selection demonstrated effective predictive performance.
• The model maintained fast computational times, indicating practical applicability.

Conclusions:

• NLP analysis of clinical notes offers a viable approach for predicting hospital readmissions, especially for COPD.
• The developed component selection framework aids in optimizing predictive models.
• This approach can assist medical institutions in proactively managing readmission risks and improving patient outcomes.