Recognizing hypoglycemia in children through automated adverse-event detection

Abstract

Background: Automated adverse-event detection using triggers derived from the electronic health record (EHR) is an effective method of identifying adverse events, including hypoglycemia. However, the true occurrence of adverse events related to hypoglycemia in pediatric inpatients and the harm that results remain largely unknown.

Objective: We describe the use of an automated adverse-event detection system to detect and categorize hypoglycemia-related adverse events in pediatric inpatients.

Methods: A retrospective observational study of all hypoglycemia triggers generated by an EHR-driven surveillance system was conducted at a large urban children's hospital during a 1-year period. All hypoglycemia triggers were investigated to determine if they represented a true adverse event and if that event followed or deviated from the local standard of care. Clinical and demographic variables were analyzed to identify subpopulations at risk for hypoglycemia.

Results: Of the 1254 hypoglycemia triggers produced, 198 were adverse events (positive predictive value: 15.8%). No hypoglycemic adverse events were identified via the hospital's voluntary incident-reporting system. The majority of hypoglycemia-related adverse events occurred in the NICU (n = 123 of 198 [62.1%]). A total of 154 (77.8%) of the 198 adverse events hospital-wide and 102 (83%) of the 123 adverse events in the NICU occurred in patients who were receiving insulin therapy.

Conclusions: Hypoglycemia is common in hospitalized children, particularly neonates and those who receive insulin. An EHR-driven automated adverse-event detection system was effective in identifying hypoglycemia in this population. Automated adverse-event detection holds great promise in augmenting the safety program of organizations who have adopted the EHR.