Early prediction of delirium in a pediatric cardiac intensive care unit: A pilot study

Abstract

Abstract Pediatric delirium is associated with increased risk of mortality, prolonged time to extubation, and increased length of stay. If the risk of delirium could be identified early, then preventative strategies could be targeted. Therefore, we studied the performance of an early delirium prediction model in children admitted to a pediatric cardiac intensive care unit using variables extracted from the electronic medical record 24\xa0h after admission. This single center, retrospective study extracted the presence of fifteen previously defined variables associated with delirium in children to create a model to predict the risk of delirium anytime during a child s stay. Delirium was considered present if a patient had at least one Cornell Assessment of Pediatric Delirium score of 9 or greater, required an antipsychotic or had an ICD-10 diagnosis of delirium. Repeated measures logistic regression between the preselected variables associated with delirium and the clinical diagnosis of delirium was performed. Variables significant in the unadjusted univariate analyses were entered into a multivariable model for adjustment. The final model contained four predictors of delirium: serum albumin below 3\xa0g/dl, blood transfusion, dexmedetomidine, and mechanical ventilation. The model demonstrated good discriminative ability with an area under the receiver operating characteristics curve of 0.74. This model requires additional study and validation in a larger subset of subjects but could aid in early identification of high-risk patients and facilitate targeted delirium prevention strategies.