AI and the safe travel lane of diabetes care

Technology could help lengthen lives, reduce complications

| Author: Eric Baerren | Media Contact: Aaron Mills

Artificial intelligence could extend the lives of people with diabetes by up to a decade, according to a doctor at the Central Michigan University College of Medicine. It also has the potential to improve the quality of those years.

It’s one way the technology has promise when it comes to improving health care.

Dr. Sethu Reddy

Doctors often struggle to keep diabetes patients properly treated, said Dr. Sethu Reddy, the medical school’s senior associate dean of research. The reason is that diabetes is a constant in someone’s life.

“Diabetes is not a once every-three months condition,” he said.

It’s like a self-driven car trying to navigate down a narrow path toward its destination, he said. To arrive at its destination safely, the car must travel a narrow lane. If it drifts, the AI driving the car could correct the course.

Like that car, a person with diabetes has a narrow range of glucose levels to maintain. But those can change during the day depending on what they eat and how they exercise, Dr. Reddy said. AI can potentially keep that car inside its lane more consistently, like a self-driven car.

As an example, Dr. Reddy said that a doctor treating a young person who suddenly develops Type 1 diabetes might take years of trial and error to figure out how to manage that person’s blood sugar. That person is also at risk for complications like blindness and kidney issues.

AI has the potential to figure that out much more quickly and create an insulin adjustment algorithm that improves the patient’s quality of life.

“Imagine being able to get someone back to proper blood sugar in two weeks,” he said. Type 1 diabetes frequently cuts as much as 10 years off of someone’s life expectancy, and causes complications like kidney failure, blindness and neuropathy.

AI has already found utility in diabetes care.

The Food and Drug Administration has approved the use of AI in making certain narrow recommendations for insulin use, Dr. Reddy said. One advantage that AI has is that it can process a massive amount of data very quickly, leading to a timely recommendation.

In terms of insulin use, if a person with diabetes is going out to eat, an AI-powered app could recommend a specific dose of insulin based on the individual, the restaurant’s menu, their physical activity, the time of day and the underlying glucose trend.

Dr. Reddy plans to research other ways that AI could improve diabetes care with CMU medical students. For now, the work will take place over narrow lines of inquiry.

“It’s here to stay,” he said. “It’s soon to be a part of our daily lives.”

He added a note of caution.

AI isn’t ready for prime time, he said. It’ll probably take a few years for it to get fully engrained into health care. The biggest reason is to make sure it can be incorporated safely. Another reason is making sure that a person still ultimately oversees the process.

It will also have widespread implications, from diagnosis to prevention and management of complications. That creates a necessity to avoid passively accepting the technology.

An emerging trend in healthcare is treating a person as a whole rather than individual components. Generative artificial intelligence can take massive amounts of data, scan it and look for answers to questions that a person might not ask.

One model was recently able to predict type 2 diabetes based on a short sample of a patient’s voice, he said.

“We think in straight lines … A to B, B to C and then C to D,” Dr. Reddy said. “Machine learning and generative AI can look for unexpected correlations.”

For instance, doctors with diabetes patients routinely examine the back of their patient’s eyes. Generative AI has the potential to take hundreds of scans, analyze the data and identify someone at risk for heart disease. Or it could help identify people at risk for strokes, allowing doctors to prescribe preventative measures.

Like any new technology, there are ethical concerns. Patients need to control their data, while what data is shared needs to be anonymous.

“We have to balance patient and individual privacy versus the common good,” he said.