Patient charts tell physicians why a patient may be prone to type 2 diabetes
Watching your diet and glucose levels is part of living with type 2 diabetes. Artificial intelligence may soon help your doctor see your risk for diabetes before you can.
Artificial intelligence won’t tell you not to eat that second cookie. But it could help your doctor know if there’s a good chance you may develop diabetes in the next year, which could make dessert less appealing.
Robert Menzie, lead data scientist on Sanford Health’s advanced analytics group, came up with an AI algorithm that can show the probability someone has or might develop type 2 diabetes in the near future.
It looks at the past five years of a patient’s medical history and compares them to similar patients who were diagnosed with type 2 diabetes. The AI provides a risk score that the person’s personal care physician can then use to dig deeper.
“What it’s really doing is computing all that information so that the provider can do what we really want them to do: spend time with the patient. Because that’s what AI can’t do. It can tell you a lot from their electronic medical record. But the provider is the only one who can tell you what’s the home life like, what’s the patient been feeling for the last 30 days,” Menzie said. “Those socioeconomic components make up way more of a person’s health than what’s in the EMR.”
The goal is to identify high-risk patients so the provider can diagnose diabetes as soon as possible and get it under control to prevent more serious problems later such as heart disease, vision loss and kidney disease.
More than 34 million Americans, or about 1 in 10, have diabetes, and at least 90% are type 2, according to the Centers for Disease Control and Prevention.
‘More done quicker’
Besides a risk score, the algorithm also informs the provider of the factors behind it. That gives the physician some insight into how to diagnose and treat the diabetes.
“The provider doesn’t have to look through a panel of 1,500 patients. The AI can give them (the) top 10, top five that are at most risk and the reasons why,” Menzie said.
The computer looks at around 800 different values, he said. “If you ask a provider to look at all of those, they’re going to laugh at you.
“The computer can figure that out for you. A doctor can, if they spent a lot of time in front of the computer. But we can have it in front of them in 30 seconds versus 30 days,” Menzie said. “Everything we make the computer do the doctor could have done. We just use the computer because we can get a lot more done quicker.”
Diabetes education
Scott Boyens, M.D., a Sanford Health family physician who specializes in sports medicine, serves as provider champion for the project. He gave input to Menzie such as what information is important to include, whether the data points made sense and how the data should be presented.
“Diabetes is huge. Anything we can do to intervene earlier is going to benefit patients in the long run. What’s exciting about this tool is it does, in a lot of cases, pick up patients who weren’t necessarily on your radar,” Dr. Boyens said.
He recently identified one patient using the AI tool who didn’t immediately show signs of diabetes.
“It now changes the conversation to what can we do to intervene in any of these risk factors to make a difference and head this off at the pass,” Dr. Boyens said.
He said it also gives him a visual aid to educate patients and back up earlier conversations about risk factors.
“It gives me some credibility to what I’ve been saying,” Dr. Boyens said. “You can’t do anything about age and gender. But what can you do about these particular data points to decrease your risk? I’ve been telling you you’re at an increased risk. And now I have a tool to show you why and how are we going to change that.”