Bridging the Digital Divide: Real-World Efficacy of Commercial and Open-Source Automated Insulin Delivery Systems in a Diverse, Low-Resource Pediatric Population

BACKGROUND: Achieving optimal glycemic control in type 1 diabetes (T1D) is challenging. Automated insulin delivery (AID) systems have improved outcomes, yet data on open-source systems-particularly in low-resource, low technological literacy settings-remain limited. This study compared the efficacy and safety of an open-source system with a regulatory-approved system in a heterogeneous real-world pediatric cohort. METHODS: This longitudinal observational study included 61 children with T1D treated at Shaare Zedek Medical Center: 30 using the regulatory-approved Medtronic 780G and 31 using the open-source AndroidAPS AID system. Socioeconomic status (SES) was determined by residential address. Glycemic outcomes were compared between AID use and prior sensor-augmented pump therapy. Primary endpoints included changes in time in range (TIR), average glucose, glucose management indicator (GMI), and hypoglycemia/hyperglycemia duration. Safety outcomes included rates of severe hypoglycemia and diabetic ketoacidosis (DKA). RESULTS: Both groups demonstrated significant glycemic improvements after transitioning to AID, including an increase in %TIR (56.6 ± 14.1 vs 69.3 ± 8.8) and reduction in average blood glucose (173.3 ± 31.7 vs 149.9 ± 18.3), GMI and % time in hyperglycemia. No significant differences in glycemic outcomes were observed between groups, although a trend toward greater reduction in average blood glucose and GMI was observed in the regulatory-approved group. No episodes of severe hypoglycemia or DKA occurred in either group. CONCLUSIONS: In a heterogeneous real-world pediatric cohort including a low-resource population with limited technological access, both AID systems significantly improved glycemic outcomes without compromising safety. These findings suggest AID benefits transcend socioeconomic barriers, supporting patient autonomy in system selection.