DiaClusT
OverviewDiabetes mellitus (DM) is a metabolic disease based on insulin resistance or deficiency, characterized by chronically elevated blood glucose levels and associated with an increased risk of severe concomitant and secondary diseases. There are two main forms, DM type 1 and type 2, the latter of which can have highly heterogeneous manifestations, so that a more refined classification can contribute to more individualized treatment of high-risk patients with a potentially worse inpatient outcome.This study applies machine learning-based clustering to identify type 2 diabetes subphenotypes in a large German inpatient cohort and to assess the reproducibility of the clusters reported by Ahlqvist et al. (2018, The Lancet Diabetes & Endocrinology). For this purpose, data from Leipzig University Hospital was requested via the Data Integration Center.
Programme: Medical Data Science
LHA ID: 91TYCK33FX-2
Public web page: Not specified
Human Diseases: No Human Disease specified
Health Atlas - Local Data Hub/Leipzig PALs: No PALs for this Project
Project Coordinators: No Project coordinators for this Project
Project created: 8th Jul 2026
Related items
As part of the DiaClusT project, we developed a multinomial logistic regression classifier to assign patients to distinct type 2 diabetes clusters based on their clinical characteristics. The current approach, including age, sex, BMI, HbA1c, and triglyceride levels, achieved an overall accuracy of 92%, with sensitivity, specificity, and F1-score all exceeding 0.87. The model’s robust performance highlights the potential of incorporating cluster assignment into routine admission workflows, ...
Creator: Ulrike Holdgrün
Submitter: Ulrike Holdgrün
Model type: Linear equations
Model format: R package
Environment: Shiny
https://orcid.org/0009-0007-0838-6277
Welcome to the Department for Medical Data Sciences at the University Medical Center in Leipzig. We are an interdisciplinary team focusing on medical data sciences in order to increase the value for the patients. In this way, we are mainly working with medical data. Our focus is twofold. On the one hand, we are working on different facets of computing infrastructures allowing us to prepare and analyze medical data, such as for distributed computing. Secondly, we focus on the analysis of medical ...
Projects: DiaClusT, Synthetica: generation and evaluation of synthetic data, Prediction model of Length of Stay and probability of rehospitalization, LEUKO-EXPERT
Web page: https://www.uniklinikum-leipzig.de/einrichtungen/medical-data-science/en/team
