Loading...
An AI-Powered Clinical Decision Support System for Patients with Diabetes: Integrating ML for Readmission Prediction and LLM for Recommendation
Kumar Patel, Mukesh
Kumar Patel, Mukesh
Citations
Altmetric:
Genre
Poster (Research)
Date
2025-05-01
Advisor
Committee member
Department
Health Services Administration and Policy
Permanent link to this record
Collections
Research Projects
Organizational Units
Journal Issue
DOI
https://doi.org/10.34944/5hmt-f779
Abstract
Purpose/hypothesis: The purpose of this study was to design a web-based Clinical Decision Support System (CDSS) to collect data of patients with diabetes from electronic health record systems, perform predictions using machine learning (ML) models, visualize results, and provide personalized recommendations based on important features using a large language model.
Methods: The data comes from the 130-US Hospitals dataset, which includes 101,766 encounters from 71,518 unique patients with diabetes. After preprocessing to address missing values, encode categorical features, and convert ICD-9 diagnoses to ICD-10, we trained several machine learning models, including Random Forest, XGBoost, and CatBoost. The models were evaluated using performance metrics such as accuracy, precision, recall, and F1 score, with an 80% training and 20% testing data split. The CatBoost model, along with LLaMA 3.2 (LLaMA-1B), was then integrated into a graphical user interface (GUI) via Streamlit, enabling real-time clinical decision-making and providing actionable risk predictions.
Results: The CatBoost model outperformed Random Forest and XGBoost, achieving an accuracy of 71.81% and an F1 score of 71.71%. It showed strong performance for predicting readmissions within 30 days (AUC = 0.94). The integration of LLaMA 3.2 (LLaMA-1B) into a Streamlit-based GUI enabled real-time, actionable insights for clinical decision-making.
Clinical Relevance: This work demonstrates the potential of a data-driven CDSS to predict patient readmission risks, providing healthcare providers with real-time insights to improve clinical decision-making. The integration of the CatBoost model and LLaMA-1B into a user-friendly interface enables more accurate risk assessments, which can guide personalized care plans, reduce readmissions, and enhance patient outcomes, particularly in managing complex and high-risk patients.
Description
A poster presented at the College of Physicians of Philadelphia Public Health Student Poster Session, which took place May 1, 2025 in Philadelphia, PA.
Citation
Mohammad Alizadeh, J. et al. (2025, May). An AI-Powered Clinical Decision Support System for Patients with Diabetes: Integrating ML for Readmission Prediction and LLM for Recommendation. Poster presented at the College of Physicians of Philadelphia Public Health Student Poster Session, Philadelphia, PA.
Citation to related work
Has part
ADA compliance
For Americans with Disabilities Act (ADA) accommodation, including help with reading this content, please contact scholarshare@temple.edu