Voice-based AI makes insulin management a breeze for T2D patients

A voice-based conversational artificial intelligence (VBAI) application helps type 2 diabetes (T2D) patients navigate basal insulin titration at home and, in turn, achieve rapid glycaemic control, as reported in a study.

Developed by a team of researchers from Stanford University School of Medicine in Stanford, California, US, the VBAI app was voiced by Alexa, a digital assistant by an e-commerce giant. Through a smart speaker incorporated with the custom VBAI, Alexa recites updated insulin dosing instructions based on recent insulin use and fasting blood glucose values after the prompt “Alexa, check in with clinical trial.” The daily check-in process was a multiturn conversation that took about 2 minutes to complete without requiring a smartphone or computer. 

“The technical application studied here was the use of AI to provide a conversational interface with patients. The actual clinical protocols implemented using this technology were guideline-based approaches to insulin titration with approval or modification by the patient’s physician,” the researchers pointed out.

“The technology was not designed to let the AI independently decide the dose titration. Our effort was to create a digital health tool that allows for AI-assisted medication management without increasing clinician workloads,” they added.

In a trial, VBAI was evaluated against standard of care whereby basal insulin was titrated by the clinician based on an online blood glucose and insulin log that patients had to complete daily. Thirty-two patients (mean age 55.1 years, 59.4 percent women, 50 percent White, mean HbA_1c level 9.6 percent) with T2D requiring initiation or adjustment of once-daily basal insulin were randomly assigned to either the VBAI (n=16) or the standard of care group (n=16) and followed up for 8 weeks.

The use of the VBAI led to quick insulin dose optimization (median, 15 vs >56 days; p=0.006) as well as better insulin adherence (mean, 82.9 percent vs 50.2 percent; p=0.01) compared with standard of care. [JAMA Netw Open  2023;6:e2340232]

Patients in the VBAI intervention group were more likely than those in the standard of care group to achieve glycaemic control (81.3 percent vs 25.0 percent; difference, 56.3 percent, 95 percent confidence interval [CI], 21.4–91.1; p=0.005) and glycaemic improvement (mean change in FBG level, −45.9 vs 23.0 mg/dL; difference, −68.9 mg/dL, 95 percent CI, −107.1 to −30.7; p=0.001) at week 8.

Also, the VBAI significantly alleviated the diabetes-related emotional distress experienced by patients (−1.9 points vs 1.7 points with standard care; difference, −3.6 points, 95 percent CI, −6.8 to −0.4; p=0.03).

“To our knowledge, this study marks the first time a VBAI has been used to autonomously adjust medication doses based on a protocol preapproved by a clinician. These findings suggest that digital health tools can be useful for medication titration and that voice user interfaces can be effective for patient-facing digital technologies,” the researchers said.

“We chose a voice-based interface over the more commonly used smartphone interface because of its potential to improve access, usability, and convenience, especially for older patients with diabetes,” who may have low digital literacy, they added. [BMC Public Health 2012;12:602; J Diabetes 2018;10:600-608; J Med Internet Res 2020;22:e16629; JAMA Netw Open 2022;5:e2237960]

Another advantage of VBAI, according to the investigators, is the potential to mitigate the decline in clinical engagement often seen in chronic disease patients. This is important as diabetes requires lifelong patient involvement, and several studies have shown that sustained patient engagement with digital health devices is challenging. [Open Forum Infect Dis 2022;9:ofac338; NPJ Digit Med 2022;5:195; NPJ Digit Med 2023;6:25]

In the trial, patients in the VBAI group received 7.3 automated insulin titrations over 8 weeks, with the 13 who achieved glycaemic control receiving 8.2 titrations. On average, participants in the VBAI group logged data on 50 of the 56 days they were followed up (89.3 percent), and the 13 participants (81.3 percent) who achieved glycaemic control logged data on 54 of 56 days (96.4 percent).

On the other hand, patients in the standard of care group received a mean of 1.6 titrations over 8 weeks and had a mean of only 1.3 clinic visits, which represented their only opportunity to receive diabetes care during the trial.

“As a result, despite having strict blood glucose and insulin adherence criteria for titration, our VBAI was able to provide frequent titration recommendations,” the researchers said.

“The willingness of participants to follow the VBAI’s instructions and the positive survey results regarding attitudes toward health technology more broadly suggest that patients might be accepting of this model of care delivery,” they added.