Development and Pilot Clinical Evaluation of DiaBuddy, a Point-of-Care Decision Support Tool for Children and Adolescents with Insulin-Treated Diabetes.

Background - Optimal management of insulin-treated pediatric and adolescent diabetes, particularly type 1 diabetes (T1D), requires precise insulin dosing and sick-day decision-making; deficiencies in either predispose to hypoglycemia, suboptimal glycaemic control, and diabetic ketoacidosis. We developed DiaBuddy™, a guideline-aligned mobile decision support tool, and evaluated its accuracy versus expert guidance and clinical changes observed during pilot evaluation in children and adolescents with insulin-treated diabetes. Methods - The preclinical validation compared DiaBuddy and family recommendations (N = 37) for 20 insulin-dosing and 20 sick-day vignettes against a pediatric endocrinologist gold standard. The prospective single-arm pilot study explored changes over 12 weeks in 20 children and adolescents with insulin-treated diabetes. HbA1c was the primary clinical outcome; CGM-derived metrics, PedsQL quality-of-life scores, application use, and satisfaction were exploratory secondary outcomes. Results - In the preclinical study, DiaBuddy showed lower absolute relative deviation for basal (5.0 ± 6.7% vs 24.2 ± 25.9%), bolus (6.9 ± 10.9% vs 45.3 ± 48.8%), and combined doses (6.3 ± 9.3% vs 39.0 ± 37.8%; all P < 0.001) than families. DiaBuddy™ achieved ≥90% accuracy across all sick-day domains versus 27-70% for families; and provided a guideline-concordant response in 94.5% of instances in which family responses were erroneous, including all instances involving an incorrect family hospitalisation decision. In the clinical study, HbA1c declined from 9.18 ± 1.99% to 8.48 ± 1.44% (p = 0.049) and the PedsQL score increased from 76.5 ± 8.6 to 89.1 ± 7.1 (+12.6 points, P < 0.001) with no change in CGM metrics. Application satisfaction was high (mean score 44.1 ± 4.1 out of 50) with most wishing to continue using it. Conclusions - DiaBuddy delivered guideline-aligned guidance in vignette testing. In this single-centre, uncontrolled pilot, observed changes in HbA1c and PedsQL are hypothesis-generating and require evaluation in an adequately powered multicentre randomised controlled trial.