Establishing a pediatric acute stroke protocol: experience of a new pediatric stroke program and predictors of acute stroke.

Abstract

Pediatric stroke is among the top 10 causes of death in pediatrics. Rapid recognition and treatment can improve outcomes in select patients, as evidenced by recent retrospective studies in pediatric thrombectomy. We established a collaborative protocol involving the vascular neurology and pediatric neurology division in our institution to rapidly diagnose and treat pediatric suspected stroke. We also prospectively collected data to attempt to identify predictors of acute stroke in pediatric patients. IRB approval was obtained to prospectively collect clinical data on pediatric code stroke activations based on timing metrics in resident-physician note templates. The protocol emphasized magnetic resonance imaging over computed tomography imaging when possible. We analyzed performance of the system with descriptive statistics. We then performed a Bayesian statistical analysis to search for predictors of pediatric stroke. There were 40 pediatric code strokes over the 2.5-year study period with a median age of 10.8 years old. 12 (30%) of patients had stroke, and 28 (70%) of code stroke patients were diagnosed with a stroke mimic. Median time from code stroke activation to completion of imaging confirming or ruling out stroke was 1 h. In the Bayesian analysis, altered mental status, hemiparesis, and vasculopathy history were associated with increased odds of stroke, though credible intervals were wide due to the small sample size. A trainee developed and initiated pediatric acute stroke protocol quickly implemented a hospital wide change in management that led to rapid diagnosis and triage of pediatric stroke and suspected stroke. No additional personnel or resources were needed for this change, and we encourage other hospitals and emergency departments to implement similar systems. Additionally, hemiparesis and altered mental status were predictors of stroke for pediatric acute stroke activation in our Bayesian statistical analysis. However credible intervals were wide due to the small sample size. Further multicenter data collection could more definitively analyze predictors of stroke, as well as the help in the creation of diagnostic tools for clinicians in the emergency setting.