1151-P: Predicting Adverse Events in Patients with DKA or HHS

Abstract

Background: Management of diabetic ketoacidosis (DKA) in pediatric patients is well established, but optimal management for hyperglycemic hyperosmolar state (HHS) and mixed DKA/HHS is uncertain. Aims: Develop a model to identify patients at increased risk of adverse events and identify management differences associated with adverse outcomes. Methods: Retrospective chart review of 5,346 admissions from 3,104 unique patients with type 1 diabetes (T1D) or T2D at a large children’s hospital from 1/1/2010-7/1/2022 requiring IV insulin to manage DKA (pH <7.3, glucose > 200 mg/dl) and/or HHS (glucose >600 mg/dl). An ICU stay >48 hours (h) was a surrogate for severe adverse events. Risk of death or ICU stay >48 h (high risk [HR] group) was modeled using generalized estimating equations on a training dataset of 3,979 admissions from 2010-2019. Admissions 2020-2022 were the test dataset. Discrimination was assessed with receiver-operator-characteristic (ROC) curves. Using this predictive model, we selected a control (C) group at similar risk of adverse events to identify outcome-related management differences. Results: There were 810 ICU admissions; 90 were >48 h. There were 8 diabetes-related deaths, 5 in 2020-2022. In the multivariate model, risk factors for severe adverse events included initial pH, glucose, absolute value of the deviation of glucose-corrected serum Na from 140, and a diagnosis of T2D. Discrimination was excellent; area under the ROC curve was 0.938 and 0.964 in the training and test datasets respectively. HR and C groups were similar for pH, glucose and sodium on admission, but altered mental status was more frequent in the HR group. Glucose and sodium corrected more slowly in the HR patients. Conclusions: The prediction model accurately identified patients at increased risk of prolonged ICU stay or death. Rates of death and prolonged ICU stay increased during the Covid pandemic, correlated with more severe clinical presentation. Slower correction of metabolic abnormalities was associated with prolonged ICU stay. Disclosure M.Yousif: None. S.Adhikari: None. P.C.White: Consultant; Provention Bio, Inc.