A validation study of normoglycemia and dysglycemia indices as a diabetes risk model

TL;DR

This study evaluates glucose indices derived from OGTT data to classify diabetes risk, demonstrating their effectiveness in distinguishing normoglycemic and dysglycemic individuals and supporting their use in diabetes risk assessment.

Contribution

It introduces and validates the use of Peak glucose concentration and glucose removal rate indices for diabetes risk classification using a support vector machine.

Findings

Classification success rate of 83% for distinguishing patient groups

Identification of population clusters related to diabetic conditions

Validation of Ackerman's model for glucose-insulin regulation

Abstract

In this work, we test the performance of Peak glucose concentration ($A$) and average of glucose removal rates ($\alpha$), as normoglycemia and dysglycemia indices on a population monitored at the Mexico General Hospital between the years 2017 - 2019. A total of 1911 volunteer patients at the Mexico General Hospital are considered. 1282 female patients age ranging from 17 to 80 years old, and 629 male patients age ranging from 18 to 79 years old. For each volunteer, OGTT data is gathered and indices are estimated in Ackerman's model. A binary separation of normoglycemic and disglycemic patients using a Support Vector Machine with a linear kernel is carried out. Classification indices are successful for 83\%. Population clusters on diabetic conditions and progression from Normoglycemic to T2DM may be concluded. The classification indices, $A$ and $\alpha$ may be regarded as patient's indices and used to detect diabetes risk. Also, criteria for the applicability of glucose-insulin regulation models are introduced. The performance of Ackerman's model is shown.