SE was classically defined as a clinical event of 30 min duration or multiple events without return to baseline between over 30 min. In 2012, the Neurocritical Care Society (NCS) proposed modifying the above definition by reducing the duration to 5 min [8]. Clinical manifestations of SE are varied but typically include alteration of consciousness, focal motor movements, generalized convulsions, or some combination thereof. The variability of clinical manifestations often makes it difficult to determine the duration of a seizure. Increased availability of EEG monitoring has led to the recognition of subclinical seizures in many patients previously thought to have resolution of their seizure with cessation of motor movements [9–11].

Electrographic SE is defined by EEG characteristics alone. In neonates SE can be classified when the electrographic seizure lasts at least 30 min or when recurrent seizures last more than 50 % of total summed duration of an arbitrary 1-h epoch. In infants and older children, SE is diagnosed if electrographic seizure activity lasts at least 5 min or if recurrent seizures last at least the same duration without return to clinical baseline in between [8].

Increased use of cEEG over the last decade has led to recognition of a relatively high prevalence of electrographic seizures in neonates and children with critical illness. Multiple single-center studies and one large multicenter study report similar rates of electrographic seizures in critically ill neonates and children (~30 %) of which, up to a third are electrographic only [9–12, 14, 15]. Seizures often occur in the setting of an acute encephalopathy with and without a known central nervous system pathology [11]. Several risk factors for electrographic seizures have been identified and can be used as a guide to determine whom to monitor with EEG and for how long (Table 1). Neonates and infants seem to be at higher risk than older children [9, 11]. Following major neonatal cardiac surgery, the majority of seizures are electrographic only [17]. Neonates with acute encephalopathy and inborn errors of metabolism such as hyperammonemia or glycine encephalopathy or extensive dysgenesis also are at high risk of seizures [12].

The risk of electrographic seizures in those with convulsive SE is greater than 30 %, and up to a third of the patients with electrographic seizures have electrographic-only seizures [9, 12, 19]. Once a seizure medication is started, the majority of seizures are electrographic. Some types of neonatal clinical seizures are more likely to be associated with electrographic seizures: focal or multifocal clonic, focal tonic, generalized myoclonic or subtle seizure with gaze deviation in term infants, and subtle seizures like sustained eye opening with visual fixation in preterm infants [20]. There is a less consistent association between electrographic seizures and clinical subtle seizures such as blinking, behavior arrest, nystagmus, and motor automatisms.

There are EEG patterns that suggest a higher risk for seizure. These include interictal epileptiform discharges, periodic discharges and rhythmic patterns in children, and runs of spike and wave discharges in neonates (Fig. 6) [9, 12, 21]. Seizures are less likely with a normal EEG background but can still occur in neonates and children with clinical risk factors for seizures [6, 9, 12]. One study in children suggests there is a greater likelihood for seizures with initial EEG background patterns of burst suppression and attenuated/featureless, but seizures are also more likely to occur with other abnormal patterns such as discontinuous and slow/disorganized [9]. In neonates, an excessively discontinuous background without state changes, burst suppression, or inactive patterns are more likely to occur in seizure patients [23].

Until recently, there was little knowledge on how the presence of electrographic seizures impacted the outcome. Early studies revealed an association with seizures and poor outcome but were not designed to determine if the seizures themselves negatively impacted the outcome or were a marker of more severe disease. Several recent studies suggest worse outcomes in children and neonates with a larger electrographic seizure burden [22, 24–26]. This suggests that using cEEG for early identification and treatment of electrographic seizures may improve outcome.

CEEG is readily available in many tertiary care centers but is costly and labor intensive and thus not available at all centers. Nonconvulsive seizures were first described in several papers using serial routine EEGs, but since then multiple studies have demonstrated the advantages of cEEG over routine EEGs for seizure detection. Since seizures can occur more than an hour apart, routine EEGs may not detect all neonates with seizures [27]. In addition, seizures occur in the first hour of recording in 60 % or less of neonates and children [11, 12]. Finally, since the majority of seizures are electrographic, especially after starting seizure treatment, more seizures are detected on cEEG compared to routine EEG [11, 12, 19].

Optimal duration of cEEG to ensure adequate seizure detection may vary by clinical scenario. The literature is fairly consistent in reporting more than 85 % of electrographic seizures are captured in the first 24 h of cEEG [11, 12, 28]. If cEEG is started at or near the time of a neonatal insult, up to 36 hours of monitoring may be needed [29–32]. Most clinical guidelines recommend 24–48 h of monitoring for high-risk populations [12].

Duration of cEEG during therapeutic hypothermia for hypoxic-ischemic encephalopathy has not been well established. Neonates may need to be monitored for at least 78 h, as seizures can occur in neonates during the rewarming period [33, 34]. Children undergoing therapeutic hypothermia should also have cEEG until rewarmed since electrographic seizures present both during the 24 h of hypothermia and the rewarming period [35].