Sensor Data Analytics to Complement Sparse and Incomplete Medical Records for Diabetes Disease Management

Rudy Raymond, Naoki Nakashima, Yasunobu Nohara, Sozo Inoue,
International Workshop on Pattern Recognition for Healthcare Analytics
(Not Available)
(Not Available)
Diabetes mellitus is considered one of the main chronic diseases, and uncontrolled diabetes can lead to various complications that trigger other chronic dis- eases. Disease management for diabetes is therefore important to reduce the total healthcare cost. Unfor- tunately, managing diabetic patients is often difficult due to their sparse and incomplete medical records. Many patients drop out during treatment, and each pa- tient might require different treatment. On the other hand, the widespread use of mobile devices with var- ious sensors and instant communication capability has enabled healthcare providers to collect and monitor pa- tients’ condition. In this paper, we study the role of sensor data analytics to complement sparse and incom- plete medical records for diabetes disease management. We test various machine-learning techniques on real- world datasets of diabetic patients, and show that sen- sor datasets can be used to improve the precision of methods identifying high-risk patients.

Data Files