A drug mix and dose decision algorithm for individualized type 2 diabetes management.

Area of Science:

• Artificial Intelligence in Medicine
• Computational Health
• Diabetes Management

Background:

• Current type 2 diabetes (T2D) pharmacotherapy guidelines promote patient-centered care but face practical challenges in outpatient settings.
• Existing data-driven treatment optimization methods lack the ability to account for drug-specific and dose-dependent safety and efficacy variations.

Purpose of the Study:

• To develop and evaluate an AI Drug mix and dose Advisor (AIDA) for optimizing glycemic management in T2D patients.
• To assess AIDA's ability to identify minimal drug mix and dose changes for improved glycemic control while adhering to clinical guidelines.

Main Methods:

• AIDA was developed using electronic medical records from 107,854 T2D patients.
• A predict-then-optimize approach was employed, integrating patient profiles with clinical knowledge-based guidelines.
• Performance was validated on internal, external, and temporal datasets, with qualitative evaluation by endocrinologists.

Main Results:

• AIDA recommendations were estimated to improve glycated hemoglobin (HbA1c) by 0.40-0.68% over standard of care (P < 0.0001).
• Endocrinologists rated AIDA recommendations as mostly reasonable and precise in qualitative evaluations.
• AIDA demonstrated advantages in consistency with practice preferences, guideline adherence, and consideration of renal dosing.

Conclusions:

• AIDA offers a data-driven approach to personalized T2D pharmacotherapy, addressing limitations of current methods.
• The AI tool can provide crucial drug-dose recommendations for improved patient outcomes.
• AIDA has potential as a clinical decision support system at the point-of-care, particularly in resource-limited settings.