Researchers at Korea University Guro Hospital (KUGH) have developed a predictive model of type 2 diabetes by combining electronic medical records (EMR) and machine learning technology based on Big Data Analysis.
|Professor Nah Seung-woon|
The research team, led by Professor Nah Seung-woon at KUGH’s Cardiovascular Center, analyzed 8,454 non-diabetic patients with 28 patient information extracted from EMR. The prevalence of type 2 diabetes mellitus during the five-year follow-up period was 4.78 percent, while the type 2 diabetes prediction model had an accuracy of 70 to 80 percent.
Type 2 diabetes is a chronic disease which impairs blood sugar metabolism of the human body and increases the blood sugar level. Type 2 diabetes in particular adversely affects patients with severe cardiovascular disease in both the short and long run. Therefore, it is essential to prevent complications and reduce the incidence of type 2 diabetes by improving lifestyle and medication.
That explains why many studies have suggested a prediction model for type 2 diabetes. However, existing prediction models have limited user convenience and repeatability.
The study’s purpose was to develop a high-performance prediction model of type 2 diabetes using EMR and machine learning that overcame the existing limitations, and to compare the performance of the model with the current statistical methods.
“This study proposes prevention methods before the outbreak of diseases such as type 2 diabetes through machine learning based on the analysis of big data and may suggest optimal personalized treatment method in the future,” Professor Nah said.
The results of the study were published in the latest issue of Yonsei Medical Journal.