Computers Predict Diabetes with 94.9 Percent Accuracy

"Currently we do not have sufficient methods for predicting which generally healthy individuals will develop diabetes," says Akihiro Nomura of Kanazawa University, author of a new study that used artificial intelligence (AI) as a clinical tool to predict diabetes.

Diabetes mellitus is a chronic condition that affects how the body metabolizes glucose. Patients with Type 1 and Type 2 diabetes have too much sugar in their blood, a condition that can result in serious health complications. A CDC report estimates that over 100 million Americans have either diabetes or prediabetes. However, the problem, as Nomura points out, is that healthcare providers do not have a good way of foreseeing who to focus clinical and lifestyle interventions on to prevent the future onset of diabetes — until now.

Machine learning is a type of AI in which computers can “learn”, without programming, to make connections out of large datasets. Making sense of massive amounts of data is often too complicated and overwhelming for the human brain to achieve. Additionally, machine learning systems get “smarter” as they are exposed to more data, creating an ever-evolving system that is better equipped at recognizing subtle patterns. In this study, the researchers fed the computer system over half a million health checkup records from over 139,000 participants over a decade. About half of these individuals did not have diabetes, while around 5,000 did go on to be diagnosed with the disease.

According to Nomura, "Using machine learning, it could be possible to precisely identify high-risk groups of future diabetes patients better than using existing risk scores." 

"In addition, the rate of visits to healthcare providers might be improved to prevent the future onset of diabetes." Impressively, the machine learning system predicted the future development of diabetes with an accuracy of 94.9 percent.

The researchers’ next steps involve clinical trials to assess the robustness of the system among larger cohorts and how pharmaceutical interventions such as statin treatment can change patient outcomes.

Sources: Medical XPress, Journal of the Endocrine Society.