A Minimized Measurement Scheme for Predicting HbA1c Using Discrete Self-Monitoring Blood Glucose Data Within 4 Weeks...

Area of Science:

• Endocrinology and Metabolism
• Biomedical Data Science
• Clinical Diabetes Management

Background:

• Type 2 diabetes (T2D) management requires regular monitoring of glycemic control.
• Hemoglobin A1c (HbA1c) provides a long-term measure of blood glucose levels.
• Current monitoring strategies can be burdensome and costly.

Purpose of the Study:

• To develop a robust model for predicting HbA1c in T2D patients using minimal, irregular blood glucose data.
• To propose an efficient and cost-effective routine blood glucose monitoring scheme.
• To enhance the accuracy and stability of HbA1c prediction models.

Main Methods:

• Utilized two large datasets (2017-2022) with 2432 T2D patients, ~420,000 irregular blood glucose values, and 10,000 HbA1c values.
• Employed a regularized extreme learning machine for data fitting and model development.
• Compared multiple monitoring schemes based on accuracy (MAE, RMSE, R-value) and cost-effectiveness.

Main Results:

• The optimal scheme requires a minimum of seven fasting and seven postprandial blood glucose values within the last four weeks for HbA1c prediction.
• Achieved high prediction accuracy with R = 0.8029 (P < 0.001) and MAE = 0.3181% (95% CI: 0.2666-0.3695%).
• The developed model demonstrated superior accuracy and stability compared to previous studies, particularly in patients with significant glucose fluctuations.

Conclusions:

• An accurate and robust calculation model for predicting HbA1c in T2D patients using limited self-monitoring of blood glucose data has been established.
• The findings provide a new, evidence-based reference for designing practical and scientific blood glucose monitoring schemes.
• This approach offers a cost-effective strategy for routine diabetes care, improving patient management and outcomes.