Checking Diabetes Before It Starts

Type 2 diabetes affects more than 34 million people in the United States and is a leading cause of death. Its prevalence across different areas of the country, however, varies greatly.

A new University at Buffalo study reveals how artificial intelligence can help us better understand these variations and spot future trends, ultimately leading to more effective approaches to intervention, treatment and prevention.

According to lead researcher Zia Ahmed, a senior scientist and research associate professor at UB’s RENEW Institute, the prevalence of Type 2 diabetes between and within states varies substantially (between 2.2% and 28.7% in 2017)—the result of wide-ranging socioeconomic and lifestyle risk factors.

The study mapped the relationship between the disease and six risk factors (poverty, education, obesity, physical inactivity, access to exercise and access to healthy food) at the county level, using machine learning—a subset of AI that involves computers acting intelligently without being explicitly programmed. It drew upon data reported by the Centers for Disease Control and Prevention and the Census Bureau.

A number of spatial modeling studies have correlated Type 2 diabetes with socioeconomic and lifestyle factors in the past. However, the relationship between risk factors and the prevalence of the disease is a complicated one.

The machine learning program the UB team utilized—a geographically weighted random forest model—is capable of handling the nonlinear nature of this relationship, says Ahmed, and thus outperforms existing models.

This is key, Ahmed points out, because a better understanding of how the disease interacts with these risk factors will lead to more effective intervention and treatment approaches. The findings could also lead to more tailored prevention strategies from a policy perspective—which will be critical, says Ahmed, given the projected increase of the disease over the next few decades.