Grant: Project Grant

Type 1 diabetes is a lifelong condition affecting approximately 40,000 children in Canada. Effective diabetes care requires continuous monitoring and timely action to prevent complications. Wearable devices such as glucose sensors and insulin pumps now generate large amounts of data that could help identify early warning signs of risk. Artificial intelligence (AI) can analyze these data to support earlier and more personalized care, but few AI tools are used in routine practice due to poor workflow fit, limited transparency, and limited real-world evidence. Our team has developed Haibu Diabetes, a digital platform that securely brings together data from wearable devices, patient-reported information, and clinical records into a shared dashboard for patients, families, and healthcare providers. The platform is currently being deployed at BC Children's Hospital. The next phase will implement a validated AI model to identify patients at higher risk and deliver these insights directly within clinical workflows. This project will address key barriers to adoption of AI in diabetes care by identifying effective ways to integrate AI-generated risk alerts into real-world workflows, determining the level of explanation clinicians need to trust and act on these alerts without increasing workload, and generating real-world evidence on the benefits and risks of AI-driven risk alerts in routine care. Using interviews, workflow mapping, simulated clinical scenarios, and a pilot implementation study, we will evaluate how AI risk alerts can be used safely and effectively in everyday practice. The results will inform healthcare systems and regulators and support responsible scale-up of Haibu Diabetes, with the goal of improving proactive care for children and families living with type 1 diabetes.