Seven clusters of trajectories of patients with type 2 diabetes mellitus identified within the framework of the DIACARE project

The TIC Salut Social Foundation has participated in the study ‘Longitudinal deep learning clustering of Type 2 Diabetes Mellitus trajectories using routinely collected health records’, published this October in Journal of Biomedical Informatics. The article details how, based on data taken from the Information System for the Development of Research in Primary Care (SIDIAP), the trajectory of type 2 diabetes mellitus was analysed using artificial intelligence, taking into account factors such as comorbidities, medical treatments, glycosylated haemoglobin (HbA1c) changes over time, and other important variables.

The study is part of the DIACARE project, coordinated from 2017 to 2021 by the TIC Salut Social Foundation with the collaboration of Novo Nordisk and Grupo Pulso. The aim of the project was to examine new ways of monitoring patients with type 2 diabetes mellitus, taking a holistic approach to all of its characteristics, as current evidence suggests that it is a heterogeneous and complex metabolic disease, and that multiple factors affect its trajectory.

Analysis of the data of 11,028 patients with type 2 diabetes mellitus resulted in seven longitudinal clusters of patient phenotypes with different clinical trajectories. The results confirm the heterogeneity of the disease, although more clinical and genetic data would be needed to obtain more precise clustering.

The project coordinator at the TIC Salut Social Foundation, Elisenda Reixach, stated that, as mentioned in the recently-published paper, “innovative methodologies such as unsupervised deep machine learning will help develop precise and individualized strategies to improve the management and prognosis of type 2 diabetes”.

Bibliographic reference

Manzini E, Vlacho B, Franch-Nadal J, Escudero J, Génova A, Reixach E, Andrés E, Pizarro I, Portero JL, Mauricio D, Perera-Lluna A. Longitudinal deep learning clustering of Type 2 Diabetes Mellitus trajectories using routinely collected health records. J Biomed Inform. 2022 Oct 8;135:104218. doi: 10.1016/j.jbi.2022.104218. Epub ahead of print. PMID: 36216232.