On a global scale diabetes is a growing health hazard. Diabetes was the main cause of approximately 4.6 million deaths in 2011. It has been estimated, that there are 366 million people with the disease, which is expected to grow to 552 million people by 2030. This development is mainly caused by life style changes, which lead to less physical activity and an increase in obesity.

Diabetes is the cause of large health care expenditures and for 2011 it has been estimated, that globally the health care expenses caused by diabetes amounts to 465 billion USD.

Diabetes leads to reduced quality-of-life in those affected by the disease, which is mainly due to the long-term complications diabetes causes. The risk of developing long-term complications is related to hyperglycemia.

To control the blood sugar levels and prevent hyperglycemia, and thereby lower the risk of long-term complications, diabetes patients often receive insulin treatment. A successful diabetes treatment has been shown to lead to increases in the quality-of-life for diabetes patients. However, insulin treatment is associated with an increased risk of hypoglycemia, which increases mortality and morbidity while reducing quality-of-life for diabetes patients.

In this project an algorithm was developed for retrospective detection of hypoglycemia. This allows diabetes patients to analyze their blood glucose measurements over a period of time and identify possible hyperglycemia episodes. This may aid diabetes patients in optimizing their diabetes treatment.

For developing and testing the algorithm, a dataset with blood glucose measurements from 37 Type 1 diabetes patients was used. The dataset contained on average 21 daily blood glucose measurements pr. patient recorded in a period of 4-5 days. The developed algorithm was constructed using a linear discriminant function based on two features. The features used were based on the analysis of four daily blood glucose measurements spread throughout the day. The result obtained from testing the algorithm gave a sensitivity of 86 % and a specificity of 67 %. The algorithm needs to be tested on other datasets to validate its reliability prior to testing in clinical practice.

To make the algorithm easily accessible for diabetes patients, it was implemented in a smartphone application named HypoDetect. A prototype of the application was developed, which allows users to enter data into the application such as blood glucose measurements. The user can visualize blood glucose measurements on a graph and analyze the entered data for hypoglycemic events using the algorithm.

The HypoDetect application was tested in a usability test, where a Type 1 diabetes patient tested the application in a daily life scenario throughout a period of two days. Overall the usability test resulted in positive feedback from the test subject. However, the test clarified that some aspects of the application has to be improved before it is ready for use in daily life.