DiaBuddy: AI-Enabled Digital Self-Management for Youth With Type 1 Diabetes (DiaBuddy)

Type 1 diabetes mellitus (T1DM) requires continuous daily management, including glucose monitoring, insulin administration, dietary decisions, physical activity, recognition of abnormal glucose levels, and management of diabetes-related risks. For children and adolescents, this management is shared between the child and caregiver, and responsibility gradually shifts toward the child with increasing age and independence. This can create challenges related to adherence, diabetes-related distress, and caregiver burden.

DiaBuddy is an artificial intelligence (AI)-empowered digital self-management intervention developed to support children and adolescents with T1DM and their caregivers. The intervention is intended to complement, rather than replace, routine diabetes care. It provides digital tools for recording and reviewing diabetes-management information, educational content, reminders, motivational features, and caregiver support.

The application includes functions such as glucose logging and trend visualization, insulin and meal/snack reminders, physical activity reminders, hypoglycemia alerts, diabetes safety and diabetic ketoacidosis (DKA) education, educational videos, interactive quizzes, Urdu and English content, points, badges and rewards, caregiver dashboards, child-caregiver account linking, weekly summaries, synchronization, offline functionality where supported, and technical support.

DiaBuddy will also incorporate AI-supported analysis of available longitudinal diabetes-management information, such as glucose readings, insulin information, dietary information, physical activity information, and previous glucose patterns. The AI may provide individualized information concerning glucose trends, dietary considerations, physical activity, anticipated glucose patterns, diabetes-management reminders, and when medical review may be appropriate.

The AI component is designed as a decision-support and self-management tool. It will not independently change or administer insulin doses. Any insulin-related suggestion will remain subject to the participant's or caregiver's existing clinical management plan and treating clinician oversight. DiaBuddy will not replace treating physicians, routine diabetes care, emergency medical services, or clinical decision-making.

The definitive trial follows a preliminary DiaBuddy pilot. The pilot provided information about feasibility, usability, technical issues, questionnaire administration, recruitment, follow-up, and preliminary changes in self-management, caregiver burden, diabetes-related distress, and reported caregiving time. Because the pilot did not include a control group and was short in duration, it cannot determine whether observed changes were specifically attributable to DiaBuddy. The randomized controlled trial is therefore designed to provide a controlled comparison of DiaBuddy added to usual diabetes care versus usual diabetes care alone.

Participants will be followed for six months, with assessments planned at baseline, approximately three months, and approximately six months. Clinical and participant-reported information will be collected throughout follow-up. Laboratory-verified glycated hemoglobin (HbA1c) will be used to assess glycemic control. The study will also monitor relevant diabetes-related events and healthcare utilization, including hypoglycemia, severe hypoglycemia, DKA, emergency department visits, and diabetes-related hospitalization.

Safety monitoring will include potentially harmful AI outputs and technical failures that could affect participant safety in addition to clinical safety events. Participants will be informed that DiaBuddy is not an emergency service and that clinical emergencies should be managed through their treating healthcare team or appropriate emergency services.

Before the trial begins, the randomized controlled trial (RCT) version of the application will be documented, including the permitted and prohibited AI outputs, insulin-related safeguards, human oversight procedures, software and AI version control, handling of erroneous or incomplete information, and procedures for reporting potentially serious AI-related incidents. Changes to the software or AI system during the trial will be documented.

The trial is intended to determine whether adding DiaBuddy to usual diabetes care provides meaningful improvements in diabetes self-management and caregiver outcomes while also evaluating clinical outcomes, safety, usability, and engagement. The longer-term objective is to develop a safe and evidence-based digital self-management platform that may subsequently be evaluated for implementation within routine diabetes care in Pakistan.