Prediction of 30-day readmission in diabetes management using Machine learning

Area of Science:

• Medical Informatics
• Machine Learning in Healthcare
• Diabetes Management

Background:

• Diabetes mellitus is a chronic condition requiring complex care, often leading to frequent hospital readmissions.
• Predicting 30-day readmissions is crucial for proactive patient management and reducing healthcare burdens.
• Identifying high-risk patients enables targeted interventions to improve outcomes.

Purpose of the Study:

• To develop and evaluate machine learning models for forecasting 30-day readmissions in diabetes patients.
• To compare the performance of Logistic Regression, Decision Tree, Random Forest, and XGBoost algorithms.
• To identify the most effective model for predicting diabetes-related readmissions.

Main Methods:

• Utilized a dataset of 352 records from a diabetes specialty clinic in Varanasi, India.
• Developed prediction models using Logistic Regression, Decision Tree, Random Forest, and XGBoost.
• Assessed model performance using precision, recall, F1-score, and AUC-ROC metrics.

Main Results:

• XGBoost achieved the highest precision (0.84), recall (0.87), and F1-score (0.85).
• Random Forest demonstrated a superior AUC-ROC value of 0.94, indicating strong detection capabilities.
• XGBoost offers superior prediction accuracy, while Random Forest is more robust against overfitting in smaller datasets.

Conclusions:

• Machine learning models, particularly XGBoost and Random Forest, can effectively predict 30-day diabetes readmissions.
• Model selection is critical, with XGBoost excelling in overall accuracy and Random Forest being suitable for smaller datasets.
• These predictive capabilities empower healthcare providers to implement timely interventions, enhancing patient care and reducing readmission rates.