Machine Learning can predict glycaemic control in diabetes patients

Machine Learning can predict glycaemic control in diabetes patients
©shutterstock/Ahmet Misirligul

The risk of poor glycaemic control in type 2 diabetes patients can be predicted using Machine Learning methods, according to new research from the University of Eastern Finland.

The most important factors in predicting glycaemic control are prior glucose levels, duration of type 2 diabetes, and the patient’s existing anti-diabetic medicines.

Over a six-year period, researchers examined patients with type 2 diabetes in North Karelia, Finland. Patients’ glycaemic control was assessed on the basis of long-term blood glucose, HbA1c.

The study analysed data from the electronic patient information system of the Joint Municipal Authority for North Karelia Social and Health Services. The researchers also examined registers maintained by the Social Insurance Institution of Finland, as well as data from Statistics Finland’s open postal code database.

In total, 9,631 people with type 2 diabetes were selected for the study cohorts. The research was carried out in collaboration between the University of Eastern Finland and the University of Oulu, and was funded by the Finnish Diabetes Association, the Strategic Research Council at the Academy of Finland, Kuopio University Hospital (VTR funding), and the HTx project funded by the EU Horizon 2020 programme.

Machine Learning can predict the risk of hyperglycaemia

Through the data, the researchers were able to identify three HbA1c trajectories. Informed by this information, patients were divided into two groups: patients with adequate glycaemic control, and patients with inadequate glycaemic control.

The researchers studied the association of patients’ baseline characteristics, clinical- and treatment-related factors and socio-economic status with glycaemic control, through Machine Learning methods. Over 200 baseline characteristics were included as variables.

The researchers found that, by using data on the duration of type 2 diabetes, prior HbA1c levels, fasting blood glucose, existing anti-diabetic medicines and their number, it was possible to identify patients with a persistent risk for hyperglycaemia. This means that inadequate control could be predicted using data that is routinely collected as part of standard diabetes monitoring and management.

Good glycaemic control is essential to diabetes treatment

Maintaining good glycaemic control is the primary objective of treatment in type 2 diabetes as this can prevent complications associated with the disease. According to the Finnish Current Care Guidelines for Diabetes, a patient’s glycaemic control should be assessed annually, making it possible to predict the long-term trajectory of the disease.

Identifying patients with poor glycaemic control is vital in treating at-risk patients, as it allows doctors to intensify treatment at the right time. Delayed intensification of treatment can increase the risk of complications, which leads to higher costs of care.

“Our findings suggest that heterogeneity in long-term treatment outcomes is predictable with patient’s unique risk factors. This, in turn, offers a useful tool to support treatment planning in the future. However, future studies are needed to obtain even more accurate and personalised predictions,” wrote the authors of the study.

Subscribe to our newsletter


Please enter your comment!
Please enter your name here