A custom artificial intelligence (AI) app for insulin management advice can quickly achieve blood sugar control in patients with type 2 diabetes, a small randomized clinical trial has found.
When titrating basal insulin, people who used a voice-based conversational AI application delivered via Amazon’s Alexa at home were able to reach optimal insulin doses faster (over 56 days with standard care). compared with a median of 15 days), Ashwin Nayak, MD, PhD, MSc, Stanford University, California, and colleagues reported. JAMA network open.
AI app users also had significantly better insulin adherence during the 8-week trial (82.9% vs. 50.2% on average, respectively). P=0.01); Significant blood sugar improvement was achieved (mean value 45.9 vs. 23 mg/dL reduction),P=0.001); glycemic control, or fasting blood glucose (FBG) levels below 130 mg/dL, were more likely (81.3% vs. 25%; P=0.005).
“The key insight here is that clinicians can leverage technology to expand and enhance care delivery in the home,” Nayak said. today’s med page. “We have introduced the concept of ‘remote patient intervention,’ which we believe is about closing the loop on remote patient monitoring data by making clinical decisions in real time according to physician-prescribed protocols. .”
“Insulin dose management based on real-time patient-reported blood glucose and compliance data is just one example, but you can imagine how this care delivery model could be used for other diseases as well.” said.
Beyond the blood sugar-improving effects, app users also reported significantly less diabetes-related psychological distress than standard care (-1.9 vs. 1.7 points on composite survey score; P=0.03).
“We are currently working to make technology like this available to patients outside of research settings, as it is a technology that is well suited for healthcare professionals who need high-touch care to control their diabetes. Because we think we can really help patients in underserved areas,” Nayak said.
The researchers developed a custom voice-based AI app to leverage Alexa (Amazon was not involved in this research). The software was equipped with titration algorithms from the American College of Clinical Endocrinology and the American College of Endocrinology and included emergency protocols to address hypoglycemia and hyperglycemia.
Before making the app available to patients, the participant’s diabetes clinician selected an insulin titration protocol. The researchers noted that this technology did not allow the AI to independently decide on dose adjustments.
The same app was then deployed to study participants via Amazon smart speakers, which participants were able to interact with through voice commands and short conversations with Alexa to assist with basal insulin titration at home.
Individuals randomized to the standard care group had their basal insulin dose adjusted by a clinician and were instructed to complete online blood glucose and insulin records daily. They also received an Amazon smart speaker with daily reminders to complete their logs, but did not have access to the AI apps tested.
The study was conducted in 2021-2022 at four primary care clinics at one academic center. Recruitment targeted adults with type 2 diabetes with HbA1c >8%. Those who used insulin pumps or had technological barriers at home, and those who did not speak English were excluded.
The average age of the 32 participants analyzed in the primary analysis was 55 years, and nearly 60% were female. Mean HbA1c at baseline was 9.6%.
Nayak highlighted that 81.3% of AI app users who achieved blood sugar control were highly compliant with the technology, recording data on 54 out of 56 days.
“Given how compliant participants in the intervention group were with the device, we were actually not that surprised that 81% achieved insulin dose optimization and glycemic control,” he said. Ta. “We know clinical protocols work, but the challenge is providing enough support for patients to follow those protocols, and that’s where technology like this comes in.”
Nayak noted how loyal participants were in checking in with their Alexa devices daily and following instructions, even though there was typically “very little” communication with participants after registration. Kani added that his group was pleasantly surprised.
“Despite not being able to hold hands, we were really pleased to see that participants in the intervention group were interacting with the device for almost 90% of the follow-up period,” he said. said.
Limitations of this study include that, except for data collected from electronic medical records, all other variables were self-reported.
The study authors also acknowledged that glycemic control was measured by mean FBG rather than HbA1c because the follow-up period was only 8 weeks.
disclosure
Nayak and other study co-authors reported owning shares of UpDoc.
Primary information
JAMA network open
Source reference: Nayak A, et al. “Using voice-based conversational artificial intelligence for basal insulin prescription management in patients with type 2 diabetes” JAMA Netw Open 2023; DOI: 10.1001/jamanetworkopen.2023.40232.