Diastat - Model Driven Diabetes Care

Diabetes type 1 is a serious disease that is less prevalent than type 2 diabetes, but still causes large everyday challenges for those who are afflicted. Patients rely on self-treatment in the form of insulin injections. The current project will investigate how these patients can be provided feedback through statistical modeling of the data that are stored in a diabetes diary on their mobile phones.

Background

Approximately 25000 persons have type 1 diabetes in Norway, a number that is rising. The disease is characterized by the destruction of the insulin-producing cells in the pancreas and insulin must be administered by the patients themselves by use of injections or insulin pump.

A patient typically injects insulin 4-5 times per day, usually at the time of meals. How much insulin to inject in any given situation depends on a large variety of factors, such as the carbohydrate content of the meal, current blood glucose, physical activity, stress, and a number of other known or unknown factors. All this information is challenging to balance and analyze.

At NST we have developed a mobile phone application for patients with diabetes, both type 1 and type 2, where patients have been closely involved in design and interface. These types of mobile applications amass a large amount of data on the user, and these data are rarely used in a larger context.

Project description

In 2011 we performed a pilot study on 30 patients with type 1 diabetes, who used the mobile application such that they provided a realistic picture of what data we could expect to obtain in this setting. The data from this study are used to build statistical models for the blood glucose trends.

The results from the study has shown that the application is accepted as a daily tool by most users and more than half used it regularly for over three months. The meal registration unit proves to be too coarse to perform predictive modeling, but the blood glucose can nevertheless be modeled in other ways giving valuable feedback to the user.

Further in the project, we will implement the model-driven feedback to users on a new platform, and perform a randomized control trial to see which effects this feedback has on the patients.

Project partners

University Hospital of North Norway and the University of Tromsø.

Subject areas

Endocrinology, mobile applications, statistics and modeling.