Developing Clinical Decision Support System using Machine Learning Methods for Type 2 Diabetes Drug Management.

Area of Science:

• Artificial Intelligence
• Machine Learning
• Endocrinology

Background:

• Artificial intelligence and machine learning (AI/ML) can automate diabetes management, improving care equity and standards.
• Developing AI/ML tools for diabetes drug management is crucial for Type 2 diabetes patients.

Purpose of the Study:

• To create a clinical decision support system (CDSS) using machine learning for diabetes drug management.
• To predict appropriate diabetes drug classes for Type 2 diabetes patients based on clinical variables.

Main Methods:

• Utilized electronic health records from an Endocrinology clinic, analyzing 1671 prescriptions from 940 Type 2 diabetes patients.
• Employed random forest algorithms to build decision trees for predicting diabetes drug classes.
• Input variables included patient demographics, biochemical parameters (HbA1c, glucose), clinical factors, and diabetes complications.

Main Results:

• Individual drug class prediction accuracy ranged from 85% to 99.4%.
• Multi-drug prescription accuracy achieved 72%, with potential for higher clinical relevance due to interchangeable drug options.
• This study represents a significant advancement in developing AI-driven diabetes care support systems.

Conclusions:

• The developed machine learning model shows high accuracy in predicting diabetes drug prescriptions.
• This AI/ML approach has the potential to enhance the quality and accessibility of diabetes care.
• Further development of this CDSS could transform diabetes management for Type 2 diabetes patients.