Voice-based AI app reduces time to optimal basal insulin dose in type 2 diabetes
Click Here to Manage Email Alerts
Key takeaways:
- Adults with type 2 diabetes reached their optimal basal insulin dose in 15 days with use of an artificial intelligence-based app.
- The AI group was more likely to achieve glycemic control vs. standard of care.
Adults with type 2 diabetes who used an artificial intelligence-based app to manage their basal insulin reached their optimal dose in a shorter amount of time than adults receiving standard of care, according to a study data.
“Voice-based conversational AI, under the supervision of a physician, can be used to help patients with type 2 diabetes quickly achieve basal insulin dose optimization,” Ashwin Nayak, MD, MS, a clinical assistant professor of medicine at Stanford University, told Healio. “Clinicians can leverage technology to help extend and augment care delivery in the home. We introduce the concept of remote patient intervention, which we think of as closing the loop on remote patient monitoring data by making a clinical decision in real time, following a physician-prescribed protocol. Insulin dose management based on real-time blood glucose and medication adherence data reported by a patient is just one example, but you can imagine how this model of care delivery could be used in other diseases.”
Nayak and colleagues conducted the Managing Insulin with Voice AI trial, a randomized, open-label, parallel-group trial conducted at four Stanford University primary care clinics from March 2021 to December 2022. Adults with type 2 diabetes who needed to initiate once-daily basal insulin or required a dose adjustment were randomly assigned, 1:1, to basal insulin management with voice-based AI or standard of care. Adults in the AI group received an Amazon Alexa smart speaker with a custom app designed to assist them at home with basal insulin titration. The AI app determined titration using blood glucose readings and algorithms from the American Association of Clinical Endocrinology and the American College of Endocrinology. The AI group was instructed to check in with the app once a day. The standard of care group had basal insulin titrated by their clinician.
Time to optimal insulin dose from baseline to the date a participant reached their 3-day fasting blood glucose goal was the primary outcome of the study. Glycemic control was defined as a mean FBG of less than 130 mg/dL. Researchers collected mean insulin adherence based on logged data. Participants completed surveys to measure attitudes toward diabetes, health technology and medication adherence at baseline and 8 weeks.
AI improves insulin adherence
There were 32 adults who participated in the study (mean age, 55.1 years; 59.4% women; 50% white). Participants using the AI app reached their optimal insulin dose in 15 days compared with 56 days for the standard of care group (P = .006). Mean insulin adherence was achieved by 82.9% of the AI group vs. 50.2% of the standard of care group (P = .01). Participants in the AI group logged data on 89.3% of days in the study.
“We were pleasantly surprised at how adherent participants were in terms of checking in with their device daily and following its instructions,” Nayak said. “We had very little communication with participants after enrollment, in general. We let them contact us if there were technical issues or clinical questions, but for the most part, participants were on their own for the 8-week study period. So we were really happy to see that, despite the lack of hand-holding, participants in the intervention arm were interacting with their device almost 90% of the days they were followed.”
Glycemic control more likely with AI
Adults using the AI app had greater improvement in diabetes-related emotional distress as assessed in the five-item Problem Areas in Diabetes Scale compared with the standard of care group (P = .03). No differences in self-reported attitude toward diabetes technology or medication adherence were observed.
Of the participants, 81.3% in the AI group achieved their glycemic targets compared with 25% of the standard of care group (P = .005). Mean FBG decreased by 45.9 mg/dL with the AI app from baseline to 8 weeks, whereas those receiving standard of care had a 23 mg/dL increase in FBG (P = .001).
Nayak said researchers are hoping to make technology similar to the app used in the study available to people with diabetes outside of research settings because it has to potential to help those in underserved areas who need high-touch care to achieve glycemic control.
“This study serves as a proof of concept for the use of conversational AI in this new method-of-care delivery,” Nayak said. “Further studies are needed with larger, more diverse patient populations as well as with other clinical indications.”
For more information:
Ashwin Nayak, MD, MS, can be reached at aknayak@stanford.edu.
Collapse
Read more about
Click Here to Manage Email Alerts