Risk of Unintentional Severe Hypoglycemia in Hospital (RUSHH)
Hypoglycemia, also known as low blood sugar, is when blood sugar decreases to below normal levels. Severe hypoglycemia is the most common acute adverse effect of glucose-lowering medications and can be associated with life-threatening neurologic symptoms (e.g., seizure, coma). It can also prolong hospitalization, since patients require in-hospital monitoring after the event. Since most episodes of severe hypoglycemia are preventable, identifying patients at highest risk will allow for direct changes in management to prevent severe hypoglycemia. For example, a patient’s insulin dose can be decreased or their diet can be supplemented. In addition, if episodes are successfully prevented then patients who otherwise would have remained in hospital for closer monitoring can be discharged from hospital faster.
This project aims to develop a machine learning algorithm to accurately predict the risk of severe hypoglycemia for patients hospitalized at St Michael’s Hospital. We intend to implement this algorithm in clinical practice to alert clinicians to high risk patients. To evaluate the impact of this algorithm on clinical outcomes, an interrupted time-series analysis will be performed to quantify the rate of hypoglycemia on hospital wards, comparing event rates prior to model implementation to that following model implementation.