Validation of a Diabetes Subtype Classification Model Using Data from U.S. Adults Before and After the COVID-19...

Area of Science:

• Endocrinology
• Diabetes Research
• Computational Medicine

Background:

• Five distinct diabetes subtypes have been identified, but accessible models for their classification are limited.
• The impact of the COVID-19 pandemic on diabetes subtype distribution remains unclear.
• Previous research has linked COVID-19 to an increased risk of new-onset diabetes.

Purpose of the Study:

• To develop and validate a machine learning model for identifying diabetes subtypes.
• To analyze changes in diabetes subtype distribution before and after the COVID-19 pandemic.
• To investigate the prevalence of severe diabetes subtypes in recent years.

Main Methods:

• Trained classification models using electronic health records (EHRs) from 2010-2019 at the University of Alabama at Birmingham (UAB).
• Applied the trained model to EHR data from 2020-2024 at UAB for retrospective cluster analysis.
• Validated findings using data from the National Health and Nutrition Examination Surveys (NHANES) from 2015-2023.

Main Results:

• The classification model achieved high accuracy with 98% specificity and 93% sensitivity.
• A significant shift in type 2 diabetes subtype distribution was observed in both UAB and NHANES cohorts.
• The proportion of severe insulin-dependent diabetes increased from 42% to 61% (UAB) and severe insulin-resistant diabetes increased from 31% to 40% (NHANES).

Conclusions:

• The developed model effectively facilitates diabetes subtype identification.
• There appears to be an increasing trend in severe diabetes subtypes in recent years, potentially linked to the pandemic.
• Further investigation is warranted to elucidate the underlying causes of this observed increase.