The aims of our study were to describe the spectrum of signs and symptoms of radiologically confirmed stroke in children presenting to a tertiary emergency department and more specifically to determine whether adult stroke recognition tools could be applied in childhood stroke. We found that face, arm or leg weakness and speech disturbance were the most common clinical signs of stroke and that two adult stroke recognition tools, FAST and ROSIER had fair to good sensitivity of 76% and 81% respectively for detection of stroke symptoms. Mode of presentation was influenced by vascular territory with lateralised limb weakness and sensory disturbance being more common in anterior circulation stroke and visual field defects, eye movement abnormalities and limb ataxia being more common in posterior circulation stroke. There was a corresponding difference in the sensitivity of the FAST and ROSIER tools, being lower in posterior circulation stroke.
Overall, this study shows that childhood AIS has similar clinical features to adult stroke . In adults the most common presenting signs of stroke are arm weakness (68-81%), leg weakness (54-73%), facial weakness (45-59%), speech disturbance (45-59%) and sensory loss (36-49%) [9, 16, 18] largely similar to our pediatric stroke data. There is limited data on the presenting clinical features of stroke in children and our study provides in depth analysis of all neurological signs and symptoms at presentation. Previous studies have described clinical findings in terms of focal sign , motor deficits [2, 3] or hemiparesis [19–21]. Some studies have described speech and visual field deficits  but sensory deficits, cranial nerve or cerebellar signs have only been reported in one study . In contrast seizures at presentation have been more consistently reported in the pediatric literature, occurring in 11-28% of children at the time of initial presentation which is consistent with our findings.
Paramedics and emergency physicians are usually the first health professionals to assess patients with possible stroke. Both play important roles in expediting triage and assessment of patients and facilitating appropriate diagnostic imaging. Pre-hospital tools are used by ambulance paramedics to recognise potential stroke. The Face Arm Speech test (FAST) [6, 10], has been shown to have high positive predictive value when used by paramedics (78%), primary care physicians (77%) and emergency physicians (71%) and is widely used in some countries including in Australia. It has a diagnostic sensitivity of 79% when used by paramedics. Our data suggests that the FAST has similar sensitivity to adults when applied to a group of children with confirmed childhood ischemic stroke.
Clinical diagnostic accuracy of ED physicians varies from 22-96% . Therefore clinical tools have been developed for adults presenting with brain attack symptoms to assist clinicians in distinguishing between stroke and non stroke mimics. The Recognition of Stroke in the Emergency Room (ROSIER) scale has been developed for use by emergency physicians . In order of discriminatory value acute onset of symptoms, arm weakness, leg weakness, speech disturbance or facial weakness predict stroke and seizures, confusion or loss of consciousness predict non stroke diagnosis. Sensory symptoms, vertigo, dizziness and HA are non discriminatory between stroke and mimic. ROSIER has a sensitivity of 91%, better than the prehospital tools such as FAST, CPSS and LAPSS [6–10] and a specificity of 92%, similar to or better than the prehospital tools. The ROSIER had better sensitivity than the FAST in our pediatric patients with a positive result seen in 81% of cases but it is still less than adults.
Signs and symptoms of stroke may be more difficult to identify in very young children but the proportion of children under age two with a positive ROSIER was similar to that of the group as a whole. Seizures are a predictor of non stroke diagnosis in adults with the ROSIER tool but they are a relatively common occurrence in childhood stroke, reported in 11-28% of cases [2, 4, 19]. The ROSIER remained positive in more than half of the children with seizures in our study, due to presence of other neurological signs predictive of stroke. However, firm conclusions cannot be drawn due to the small numbers in these subgroups and evaluation in a larger population of children is warranted.
Stroke topography has been shown to affect applicability of these tools in adults. None of the tools assess visual field defects, disorders of perception or balance so they are insensitive to posterior circulation. For example the CPSS has a sensitivity of 88% for anterior circulation strokes but only 29% for posterior circulation strokes . The ROSIER incorrectly diagnosed stroke or mimic in 11% of cases and the majority of false negative cases were posterior circulation events . It is possible that similar problems may be encountered in the pediatric population because approximately one quarter of our pediatric strokes involved the posterior circulation, at a similar frequency to adults .
There is a need to develop and validate appropriate pediatric bedside tools in the emergency department to improve the diagnostic accuracy in detection of childhood stroke. Our data indicate that adult stroke tools are a reasonable starting point but high false negative rates of 24% for the FAST and 19% for the ROSIER tool suggest they require further development and modification if they are to be useful in the pediatric emergency department.
The limitations of this retrospective study were that some symptoms or signs may not have been documented by the assessing physician or that those recorded may not have reflected the findings at initial presentation as the medical records included notes of the senior neurologist under whom all patients were admitted. Ideally the sensitivity of the scales would have been assessed for use by ED staff and neurologists separately to determine the utility of the scales when used by front line ED staff. These differences, however, could not be determined due to the retrospective methodology of this study. Again, due to the retrospective nature of the study, we accepted broadly interpreted descriptions in the medical records for items such as "sudden onset" or "woke from sleep" which were adapted from, though not further defined in an adult study by Hand et al . Adult stroke scales were also designed to be applied prospectively. The study was conducted in a tertiary paediatric hospital so the results may not be generalisable to the broader paediatric population.