With weight loss and management, it appears that the use of artificial intelligence (AI) may be able to manage health outcomes with type 2 diabetes (T2D) patients.
A recent study led by Natalie Stein, MPH, MSc, a professor in the department of human medicine at Michigan State University, examined the use of conversational AI in the form of an automated text-based coaching service, called the Lark Health Coach, for 239 patients with T2D, all overweight or obese (body mass index [BMI] ≥25 kg/m2). The study found that the average weight loss experienced by patients was 2.38% (standard error, 0.69%) of baseline weight.
Two-thirds of adult Americans have either prediabetes or are overweight and at-risk for T2D, and while the US faces a shortage of health care providers, in-person behavioral counseling, while effective, is not always a viable option for patients in need of the service.
The Lark AI Health Coach mimics the empathic health counseling utilized by many health professionals and includes a variety of other features such as a dashboard with the ability to monitor activity, sleep, food, weight, and notes.
Some example conversations with the AI health coach showed the bot asking questions and replying to statements, such as asking the user, “What did you have for breakfast?” and replying with either positive reinforcement of ‘good’ eating habits, or constructive criticism and suggestions for alternatives for ‘bad’ eating habits. For example, when the user replied, “Eggs, bacon, and a side of toast and coffee,” the AI noted a number of times bacon was being consumed weekly and suggested switching to a leaner morning source of protein. When the user replied, “Chicken, veggies, and rice” the AI noted that it was a healthy choice.
Of the 239 that began the study, 159 installed the app, and entered initial height/weight data – 83 recorded a second weight compared to 75 that did not. From there, 70 users were considered “active” enough to analyze, while 13 were judged as “inactive.”
Users averaged 103 sessions over the course of the study, with 75.7% of users (n = 53) experiencing weight loss in any capacity. The mean loss was 2.4 kg for the group, while the percentage of healthy meals consumed increased by 31% (from 51% [n = 414 meals] of total meals logged at baseline to 67% [n = 22]) at 21 weeks.
The app offered an in-app trust survey, which resulted in a 100% response rate, with users averaging 7.9 out of 10 in their satisfaction with the program and 8.3 for their disappointment if not offered the program and 6.73 in their satisfaction with their health outcome.
According to Stein and co-author Kevin Brooks, MS, PhD, the results of this study showed that health coaching AI is “available for real-life applications.” The AI program is currently involved in further investigation into its use with chronic disease management and the weight control, diet quality, medication adherence, and home blood pressure monitoring that comes along with it. One study is planned for individuals with hypertension and another with pre-diabetics.
AI Aids Weight Loss Management in Patients with T2D