Application of novel subgroups of Chinese inpatients with diabetes based on machine learning paradigm

Weihao Wang; Zhi Chen; Sen Wang; Fei Chen; Mingqun Deng; Qi Pan; Lixin Guo

doi:10.1016/j.dsx.2022.102556

Application of novel subgroups of Chinese inpatients with diabetes based on machine learning paradigm

Diabetes Metab Syndr. 2022 Jul;16(7):102556. doi: 10.1016/j.dsx.2022.102556. Epub 2022 Jun 29.

Authors

Weihao Wang¹, Zhi Chen², Sen Wang², Fei Chen³, Mingqun Deng¹, Qi Pan⁴, Lixin Guo⁵

Affiliations

¹ Department of Endocrinology, Beijing Hospital, National Center of Gerontology, Institute of Geriatric Medicine, Chinese Academy of Medical Sciences, Beijing, PR China.
² School of Information Technology and Electrical Engineering, The University of Queensland, Brisbane, 4072, Australia.
³ College of Life Sciences, University of Chinese Academy of Sciences, PR China.
⁴ Department of Endocrinology, Beijing Hospital, National Center of Gerontology, Institute of Geriatric Medicine, Chinese Academy of Medical Sciences, Beijing, PR China. Electronic address: panqi621@126.com.
⁵ Department of Endocrinology, Beijing Hospital, National Center of Gerontology, Institute of Geriatric Medicine, Chinese Academy of Medical Sciences, Beijing, PR China. Electronic address: glxwork2016@163.com.

PMID: 35792469
DOI: 10.1016/j.dsx.2022.102556

Abstract

Background and aims: Six variables were used to determine five diabetes subgroups in European, Chinese and US populations in previous studies. This study aims to make new classification method of diabetes easier to use in clinical settings.

Methods: Clinical data of 1152 hospitalized diabetic patients were collected and built a highly accurate model based on machine learning paradigm.

Results: We visualized the confusion matrix of the classification model. The diagnose accuracy of five clusters (MOD, MARD, SIRD, SIDD and SAID) were 95%, 100%, 99%, 96% and 100%. An online tool (uqzhichen.uqcloud.net) was set up according to the cluster data based on machine learning paradigm. Six variables (age when diagnosed, HbA1c, BMI, HOMA2-β, HOMA2-IR and GADA) were needed to input in this diagnose system and then a highly accurate subgroup result was showed.

Conclusions: This is a stable and accurate online diagnose system to identify five new subgroups of diabetes based on machine learning paradigm.

Keywords: Classification; Diabetes; Machine learning; Web-based tool.

MeSH terms

China / epidemiology
Diabetes Mellitus* / diagnosis
Humans
Inpatients*
Machine Learning