Predicting COPD Readmission: An Intelligent Clinical Decision Support System.

Area of Science:

• Artificial Intelligence in Healthcare
• Clinical Decision Support Systems
• Chronic Obstructive Pulmonary Disease (COPD) Management

Background:

• Chronic Obstructive Pulmonary Disease (COPD) exacerbations lead to frequent hospitalizations, increasing healthcare burden.
• Predictive models for medium-term (2-3 months) COPD readmission are underexplored.
• Existing AI tools for COPD management have not focused on readmission prediction.

Purpose of the Study:

• To develop an intelligent clinical decision support system for predicting 90-day hospital readmission risk in COPD patients.
• To address the gap in medium-term readmission prediction for COPD acute exacerbations.
• To create a system with potential for clinical integration.

Main Methods:

• A two-level system combining machine learning (Random Forest, Naïve Bayes, Multilayer Perceptron) and a fuzzy inference-based expert system.
• Utilized a database of over 500 patients with demographic, clinical, and social variables.
• Employed filter-based and Random Forest-supported recursive feature selection for dimensionality reduction.

Main Results:

• Preliminary testing on the dataset yielded an Area Under the Curve (AUC) of approximately 0.8.
• Achieved a sensitivity of 0.67 and a specificity of 0.75 at a selected cutoff point.
• Demonstrated promising performance for early risk identification.

Conclusions:

• The developed AI system shows significant potential for early identification of COPD patients at high risk of readmission.
• Further clinical validation and database expansion are necessary for robust implementation.
• The system's generalization capacity can be improved with larger datasets and rigorous validation.