Savonia article: AI-powered eHealth systems to support diabetics in their daily diet
Diabetes is a demanding disease for patients, and it can be seriously debilitating if not managed well. The International Diabetes Federation has calculated that there are approximately 537 million adults (20-79 years) living with diabetes. The number is projected to rise to 643 million by 2030. In the year 2021, diabetes caused 6,7 million deaths and at least USD 966 billion dollars in health expenditure around the world.
Educating patients and supporting them throughout their health journey is key to effective diabetes management. eHealth technologies and Artificial Intelligence (AI) have proven to positively impact patients’ lifestyle choices in diabetes self-management. AI research has been growing exponentially in the last decade, and new AI-powered eHealth systems have been created to help patients manage their diabetes better. Chaouf’s (2022) narrative literature review shows some examples of how different AI technology is being used now by diabetics to manage their disease.
AI-powered diet self-management and physical activity support
There are several distinct AI-powered eHealth systems for diet self-management. Some systems provide personal recommendations based on the individual’s taste, lifestyle, metrics, and even internal factors like genetics. Some automatically provide a meal’s nutritive and calorific information.
Others educate patients on appropriate dietary behaviors through training, coaching, or playing. There are also AI-powered eHealth systems for physical activity self-management support. Some of these systems automatically record and determine patients’ daily activities, and others educate patients on how to adapt their activities to their blood sugar levels and food intake.
AI also automates low-value tasks for patients, such as automatic food and physical activity recognition performed by smartphone apps. Automating these tasks paves the way for building more comprehensive systems that incorporate additional automation features. An example of this would be to automate insulin boluses calculations based on real time glucose readings from an insulin pump, and the nutritive and calorific information of foods eaten by the patient. Another example is the creation of personalized notifications to exercise based on the activities performed by patients. Reminders to exercise would be sent to the patients while at home rather than while at work. For example, patients taking transportation would be advised to walk instead.
Serious games and virtual assistants helping to take control over diabetes
These systems are based on machine learning and other branches of AI such as semantic and symbolic reasonings. They use various media, serve different use cases, and target multiple user groups. AI opens new opportunities for patient education and at home coaching by targeting challenging patient groups with innovative approaches to usability and engagement. Robots are well suited to help children with type 1 diabetes while “serious games” are a convenient medium for children with type-2-diabetes and elderly people. Indeed, these solutions use recreational strategies such as taking a quiz, playing video games, and role playing.
By doing so, they enable patients to adopt an active role in their education and care, and to learn by practice. Their knowledge of diabetes management is tested, and they are given feedback on their personal dietary and activity behaviors. Virtual assistants are adapted for educating non-tech-friendly populations with specific characteristics, such as some minority groups. These virtual assistants are voice-based and integrate specific information tailored to these unique users in their recommendations. Robots are also a good medium for at-home type-1 diabetes coaching, because they can provide feedback to patients and create social bonding through frequent interactions.
AI creating personalized user experience
AI-powered personal robots generate more identification and engagement with patients. Personal robots adopt their interface and design to match the user’s preferences. A personal food recommendation system would optimize the food effect for every individual based on internal and external factors unique to them. Personalizing serious games to patients’ lifestyle environment and diabetes knowledge would generate more uptake by patients. Personalized notifications would create more engagement with patients.
Ultimately, systems which are personalized according to patients’ profiles would nudge them more effectively towards making appropriate lifestyle decisions.
Today, most AI-powered eHealth systems for diet and physical activity self-management to support diabetics are still in a preliminary research phase with notable shortcomings. However, ever-increasing processing power and software design are driving development of AI systems. In the future, it is likely that more comprehensive systems will integrate and combine a myriad of AI-based features. By harnessing the power of AI to automate tasks, personalize user-experience, and reach specific user segments, the impact on diabetes self-management can be dramatically increased.
Read more: Chaouf K (2022), Utility of artificial intelligence in diabetes self-management, a narrative literature review.
Kenza Chaouf, Graduated Student in Master’s Degree Programme in Digital Health, Savonia University of Applied Sciences
Bryn Lane, Lecturer, MBA (International Business, Finance), CFA (Chartered Financial Analyst), Savonia University of Applied Sciences, Unit of Continuous Learning, Master School, Kuopio
Elisa Snicker, Lecturer, CBC, MSc (Econ and Bus Admin), MSc (Health Sci.), Savonia University of Applied Sciences, Unit of Continuous Learning, Master School, Kuopio