Editors' ChoiceDiabetes

Outsourcing glucose management to AI

See allHide authors and affiliations

Science Translational Medicine  07 Oct 2020:
Vol. 12, Issue 564, eabe8120
DOI: 10.1126/scitranslmed.abe8120

Abstract

An automated decision support tool could expand access to intensive insulin therapy for patients with type 1 diabetes.

Although continuous glucose monitors and insulin pumps are now common, less than one-third of patients with type 1 diabetes achieve their diabetes management goals. To help patients reach their targets, frequent adjustments of insulin pump settings by a skilled endocrinologist are needed. However, the task of analyzing weeks of glucose and insulin data are time consuming, and skilled physicians are in short supply, particularly in rural environments. For skilled physicians to see more patients or primary care physicians to assist with managing pump settings, decision support systems are needed. Nimri et al. enrolled 108 patients with type 1 diabetes aged 10 to 21 years into a 6-month multicenter, randomized controlled trial to see whether an automated decision support system, DreaMed Advisor Pro, could provide recommendations for adjusting insulin pumps as effectively as skilled endocrinologists.

Overall, the recommendations generated by the decision support system were found to be just as effective as those coming from skilled endocrinologists. Both kept patients in the target glucose range about half the time while minimizing hypoglycemic readings over a 24-week period. At the 12-week time point, patients in both groups experienced a drop in their glycated hemoglobin, a key measure of long-term blood glucose management. In terms of safety, the decision support system was again on par with the physician-treated group, with all three of the severe adverse events observed occurring in the physician-treated group.

Not only was the decision support system found to be non-inferior to skilled physicians, but the 13 physicians who used it found the tool to be simple, reliable, and safe while also saving them time by providing recommendations for therapy adjustments that were clear, well-reasoned, and easy to communicate to patients. Whether the tool will prove easy to use by general practitioners will need to be evaluated in a larger study. If successful, the Advisor Pro will join a growing list of automated decision support tools poised to improve the quality and equity of care patients receive across different health systems while alleviating pressures caused by physician shortages in key specialties.

Highlighted Article

View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article