This work explores the prevention of hypoglycaemia in children with Congenital Hyperinsulinism (CHI). Children with CHI experience severe, unpredictable and often asymptomatic episodes of hypoglycaemia. Hypoglycaemia refers to a blood glucose below the normal range and carries with it the potential for significant brain injury if left uncontrolled. Prevention is a complex process involving action based on prediction. Most existing prediction work focuses on short term forecasting of future glucose values using Machine Learning processing of continuous glucose monitoring (CGM) data. However, CGM in CHI is insufficiently accurate to reliably inform glucose forecasts. Furthermore, the advance warning provided by short term prediction is insufficient to be proactive in prevention. An alternative approach, described here, uses retrospective review of CGM data to identify periods of repeated risk and proactively target action to these areas. Currently available methods for CGM review are complex and lack actionable outputs. Here, actionable outputs are generated by aggregation of CGM data into discrete chunks to provide a personalised, easy to interpret, visualisation of weekly hypoglycaemia risk. We combined this visualisation with persuasive technology to ensure predictions were actioned. This resulted in patients performing more targeted glucose checks and commenting on how their behaviour had proactively changed, ultimately achieving a 25% reduction in hypoglycaemia in free-living conditions. The real world relevance and impact of our work can be appreciated in the associated graphical abstract.
|Date of Award
|31 Dec 2023
- The University of Manchester
|Paul Nutter (Supervisor), Simon Harper (Supervisor) & Indraneel Banerjee (Supervisor)
- behaviour change
- continuous glucose monitoring