11 Critical care EEG in children

Continuous EEG in a pediatric intensive care unit finds seizures that nobody at the bedside can see. In a cohort of 550 children at 11 centers, electrographic seizures occurred in 30%, and in 36% of those children every seizure was exclusively electrographic, with no clinical sign at all. Electrographic status epilepticus occurred in 38% of the children with seizures and was associated with in-hospital death (Abend 2013). A sedated, paralysed or comatose child cannot show you a seizure, so the monitor has to.

The second reason the ICU needs its own vocabulary is that the patterns it produces do not fit the routine-EEG categories. A comatose child after cardiac arrest, a child with encephalitis, or a child in the hours after a long convulsion produces rhythmic and periodic activity that is neither clearly a seizure nor clearly interictal. Two readers looking at the same page will disagree unless they share a set of defined terms. That is what the American Clinical Neurophysiology Society's 2021 critical care EEG terminology provides: a standardized way to say where a pattern is, what it is, how much of the record it occupies, and whether it meets criteria for a seizure or sits on the continuum below one (Hirsch LJ 2021).

Who gets monitored

Indications for continuous EEG in critically ill children follow the ACNS consensus statement on continuous EEG in critically ill adults and children (Herman 2015). The common thread is a child whose mental status cannot be explained, or whose clinical events cannot be trusted: after convulsive status epilepticus, with encephalopathy of uncertain cause, after cardiac arrest, with acute brain injury, or when paralysis or sedation removes the clinical signs.

This module builds on two earlier ones. The seizure definitions in Module 10 are the ceiling of the ICU vocabulary; everything here describes activity that either meets them or falls short. The neonatal definitions in Module 7 differ, and the numbers must not be mixed: a neonatal seizure needs only evolution and 10 seconds, and neonatal status is 50% of an hour, whereas the pediatric and adult threshold for electrographic status is 10 continuous minutes or 20% of any hour.

Every rhythmic or periodic ICU pattern gets a two-part name, and the parts always come in the same order (Hirsch LJ 2021).

Main term 1: where

Main term 2: what

Combined, they give GPDs, LPDs, BIPDs, GRDA, LRDA, GSW and so on. The distinction between periodic and rhythmic is the distinction between discharges separated by a measurable gap and waves that run into each other without one. It matters because the two carry different weight on the continuum below, and because a periodic discharge with a stereotyped morphology is often the residue of an injury, whereas rhythmic delta can be a nonspecific sign of dysfunction.

Plus modifiers

A plus modifier says the pattern carries something extra that makes it look more ictal:

The modifiers are not decoration. As the ictal-interictal continuum section shows, a plus modifier can move a slow periodic pattern onto the continuum, and it is part of what a reader must record before deciding whether a pattern deserves a treatment trial.

Read the name aloud

Build the name in order every time: location, pattern, modifier, then the frequency and prevalence qualifiers from the next section. "Left-sided rhythmic delta with superimposed sharp waves" becomes LRDA+S. "Periodic discharges over both hemispheres, arising independently on each side, with fast activity riding on them" becomes BIPDs+F. A name built this way tells a colleague on the phone exactly what the page looks like, and lets the ictal-interictal continuum criteria be applied mechanically rather than by impression.

Once a pattern has a name, the reader states how much of the record it occupies and how fast it runs. Both qualifiers come from the same ACNS 2021 standard (Hirsch LJ 2021), and both feed the decisions in the next section.

Prevalence

Prevalence is a fraction of the epoch being described, so state which epoch: the whole day, the last hour, or the page. A pattern that is rare over 24 hours but continuous in the last hour is telling you something has changed.

Frequency

Frequency is stated in hertz for periodic discharges and rhythmic activity, and it is the single number that decides where a pattern sits on the ictal-interictal continuum. The thresholds in the next section are all frequency thresholds, so measure the frequency rather than estimate it: count discharges against the time marks over a full 10-second page, and check the page duration first, as Module 1 warns.

BIRDs

Brief potentially ictal rhythmic discharges are rhythmic activity at more than 4 Hz, lasting 0.5 seconds to under 10 seconds, with no clinical correlate. They are too short to meet the seizure definition in Module 10, which requires 10 seconds, but they look like the beginning of one. A BIRD is more convincing when its location matches the patient's known epileptiform discharges or seizures, and a reader who finds BIRDs should look harder at the rest of the record for longer runs that do meet the seizure threshold.

The practical value of these qualifiers is that they force the report to say what a colleague needs: not "periodic discharges on the left" but "abundant LPDs at 1.5 Hz in the last hour, with occasional BIRDs in the same location." The first phrase is an impression; the second can be acted on.

The ictal-interictal continuum is the set of patterns that are not seizures by the definitions in Module 10 but are too fast, too sharp or too unstable to be dismissed as interictal. ACNS 2021 defines it by frequency, pattern and modifier (Hirsch LJ 2021). A pattern is on the continuum if it is any one of the following:

Above 2.5 Hz for at least 10 seconds, or with definite evolution for at least 10 seconds, the pattern is an electrographic seizure, not a continuum pattern. Below these thresholds without a modifier or fluctuation, it is interictal.

Fluctuation versus evolution

The continuum criteria hinge on a distinction that residents often blur. Evolution is the seizure criterion from Module 10: at least 2 unequivocal, sequential changes in frequency (by at least 0.5 Hz), morphology or location, each lasting at least 3 cycles. Fluctuation is a pattern that changes without meeting that standard: 3 or more changes, no more than 1 minute apart, in frequency, morphology or location, that do not satisfy the evolution criteria. Fluctuation moves a slow pattern onto the continuum; evolution moves any pattern across the line into seizure. Ask first whether the changes are sequential and sustained enough to be evolution; if not, ask whether there are enough of them to be fluctuation.

What to do with a continuum pattern

A continuum pattern is a description, not a diagnosis, and the response is a set of actions rather than a reflex. Name it fully, with frequency, prevalence and modifiers, so that the next reader sees what you saw. Correlate it with the child's examination, imaging and course, because the same pattern means different things after cardiac arrest and in encephalitis. Watch it over time: a pattern that becomes faster, more prevalent or more organized is heading toward the seizure threshold, and one that slows and fragments is heading away. Where the clinical picture supports it, the electroclinical seizure definition from Module 10 offers a test: EEG and clinical improvement after a parenteral antiseizure medication makes the pattern electroclinical. Whether to run that trial is a decision for the treating team with the epilepsy service, and it should be documented as a trial, not as treatment of a proven seizure.

Whether ICU seizures matter is an empirical question, and three pediatric cohorts answer it in a consistent direction.

Read together, the studies say that burden, not the mere presence of seizures, tracks outcome. In the Topjian cohort, seizures that did not reach status did not independently predict a worse outcome, whereas status did, and strongly. In the Payne cohort the relationship was continuous, with each 1% rise in maximum hourly burden carrying an odds ratio of 1.13, and a sharp increase in neurologic decline above a burden of 20%, or 12 minutes in an hour. That 20% figure is the same threshold ACNS uses to define electrographic status epilepticus in children and adults, which is one reason the definition is worth applying exactly. These are associations in observational cohorts, and none of them proves that treating the seizures changes the outcome; they justify finding and quantifying seizures, and they set the level at which a reader should be alarmed.

Quantitative EEG

A day of continuous EEG is thousands of pages, and no one reads them all at the same attention. Quantitative trends compress the record so that events stand out: spectrograms show a rise in power at seizure frequencies, amplitude-integrated EEG shows a rise in the amplitude envelope, and asymmetry trends show one hemisphere departing from the other. Use them to find candidate events and to see the rhythm of a day, including whether a treatment changed anything.

The rule that goes with them has no exceptions: always confirm a trend change on the raw EEG before calling it. Trends cannot tell rhythmic artifact from a seizure, a change in state from a change in the brain, or a disconnected electrode from suppression. Chest physiotherapy, patting and ventilator rhythms, the artifacts of Module 3, all produce convincing trend changes. The trend points; the raw page decides.

Sign-off

The sign-off for this module has two parts. The resident passes the ACNS Critical Care EEG Monitoring Research Consortium certification test, which follows free training slides with a self-assessment of 50 questions and 25 EEG examples and awards a certificate at 80% or above. Then the resident reviews two pediatric continuous EEG days with faculty, naming every pattern in full ACNS terms.