How Artificial Intelligence Is Changing Diabetes Care
Key takeaways:
- New AI tools are emerging in diabetes care and research to analyze data in new ways
- A panel of experts said that the growing availability of these AI tools has the potential to increase the quality of diabetes care.
- Security and privacy concerns persist and should be addressed for wider adoption of the new technology.
The emergence of artificial intelligence in healthcare has put diabetes care at the precipice of tremendous change, potentially offering people more control over their health and greater access to new options to manage their glucose levels and all aspects of their lives.
Such topics were at the forefront of an expert panel discussion on the opening day of the ADA 2024 Scientific Sessions conference in Orlando, Florida, asking the question, “How will Artificial Intelligence Change Clinical Practice?” The answer was unequivocal: AI promises to deliver significant improvements in diabetes care for those who avail themself of digital tools, such as apps, pumps, AID systems and CGMs.
For those now using continuous glucose monitors, the machine learning revolution is already under way, according to the expert panelists. The improved time in range that most people with CGMs experience from tracking their glucose levels is the basic building block of future AI, data-driven advancements.
The future of diabetes data science
While AI has meaningful uses across health care in general — using pattern recognition, large language models, expert systems and decision support — diabetes care is uniquely positioned to benefit from AI advances since technology has made it so quantifiable, said Dr. Boris Kovatchev of the University of Virginia.
“AI applications are rapidly entering healthcare,” Kovatchev said. “Generative AI and large language models are at the forefront of this trend. Diabetes is one of the best quantified human conditions. Hence, diabetes care is making rapid progress with numerous applications.”
Advances that he listed include:
- Detection and prediction of events, classification and tracking disease progression,
- AI powering decision support systems,
- AI-driven neural network automated insulin delivery,
- AI-augmented clinical trials – “a most promising future,” he said.
Of particular note, Kovatchev presented results from the Virtual DCCT Project to illustrate how AI can augment clinical trials by reproducing virtual CGM traces for each DCCT participant to provide insights into how CGM metrics could relate to the risk for chronic complications and severe hypoglycemia in the historic study data.
Not only did the AI trial reproduce A1C outcomes of the groups in the treatment groups in the original DCCT trial, but Kovatchev showed how the AI-augmented study of the decades old trial was able to measure a “simulated” Time in Range that was significantly higher (60%-70%) in the “virtual” intensive treatment group compared to the “virtual” conventional treatment group.
AI and diabetes devices
Dr. Peter Jacobs, director of the Artificial Intelligence for Medical Systems lab at Oregon Health & Science University, predicted a future in which a person with diabetes might be able to use an AI-enabled “automated hormone delivery system.”
This hypothetical future system would take much of the guesswork and mental labor out of daily management by including:
- Automatic detection and dosing of insulin and pramlintide for meals
- Automatic detection and prevention of DKA with a ketone monitor
- Automatic detection of exercise and adjustment of insulin and glucagon to avoid hypoglycemia
- Automatic pattern detection to forecast and solve future problems
- Automatic adjustment of insulin dosing in response to cyclical events, such as weekends, menstruation, or illness.
- Digital twin-based decision support on medication choices, exercise guidelines and nutrient intake
- Automated detection of pump occlusion, sensor failures and other system faults
- Alert system and user interface that learns to optimally satisfy the user
This future world is not yet here, but it might not be far off, said Jacobs, who has been working with his team to develop advanced control systems for the delivery of insulin and glucagon, among other uses of AI assistance in diabetes care and research.
“There’s a lot of new technologies coming out right now that could significantly help people,” he said. “We’re at a time right now where you have an explosion of accurate sensors, an explosion of computation and access to all these tools. Combining computation with sensors, it's a tremendous opportunity and an exciting time for patients.”
Connecting teams to deliver more powerful care
The third panelist at the session, Dr. Mudassir Rashid of the Illinois Institute of Technology, spoke about his research into how the use of AI impacts multidisciplinary team approaches to healthcare. The complexity of diabetes care and its need for coordination between healthcare teams make it well suited for AI enhancements, he said.
Rashid said the safe use of AI must include protections for data security and privacy as well as efforts to build trust in AI systems. In addition, he said, it is important to be aware of the potential for bias of algorithms and have clinicians ensure proper diagnosis, care and patient outcomes.
“People should be excited about the fact that it’s going to empower them,” he said of the emergence of AI technology in diabetes care. “It’s going to give them a lot more knowledge and insights about the chronic disease outside of the clinical setting, so they can get information outside of seeing their doctor. And this can also inform their diabetes treatment and care.”
Security and privacy
Such data, Rashid explained, can be analyzed by AI tools and then be available for the patient’s health care team and caretakers to glean specific insights about the individual to offer them better care.
“There are some concerns with data privacy and security, but the best way to overcome them is to demonstrate to patients that there are safeguards,” he said. “It’s their data, it belongs to the patients, so more has to be done in securing their privacy. I’m optimistic that these are technical and regulatory challenges that we can overcome. This is going to substantially improve care for people with diabetes.”
Learn more about AI, diabetes data and the future of diabetes care: