The status-epilepticus sequence ended with durable, anesthetic-independent seizure control, but that endpoint can only be established if the brain is observed through treatment and withdrawal. Continuous EEG is therefore not simply a longer diagnostic EEG. It is neurotelemetry used to detect otherwise invisible seizures, quantify their burden, evaluate treatment, and follow a risk state that changes with sedation, temperature, physiology, medication withdrawal, and structural injury. Its value begins with a precise question: does persistent encephalopathy represent nonconvulsive seizures, is a suspicious event epileptic, does established status remain controlled, or is an evolving rhythmic or periodic pattern becoming sufficiently ictal to justify treatment? Monitoring without a decision pathway produces data; monitoring linked to predefined actions produces care.

Scalp EEG also has important biological limits. It primarily detects the summed, synchronous postsynaptic activity of superficial cortical populations. The skull attenuates and spatially blurs those signals, so small, deep, mesial temporal, orbitofrontal, insular, and interhemispheric seizures may be absent or poorly localized on scalp recordings. Craniectomy and skull defects can instead exaggerate amplitude and sharpness through a breach effect. A negative scalp EEG lowers the probability of an electrographic seizure but does not completely exclude a deep or spatially restricted ictal process when the clinical syndrome remains compelling. Full-montage video EEG is therefore preferred whenever focal seizures are plausible, while reduced-montage systems should be understood as triage instruments rather than equivalent replacements (Alkhachroum et al., 2022).

The highest-yield indication is failure to recover as expected after a clinically evident seizure. The American Clinical Neurophysiology Society recommends considering continuous EEG when alertness is not clearly improving within approximately 10 minutes after convulsive activity ends or when any impairment of consciousness persists beyond approximately 30 minutes. These intervals do not define the duration of a legitimate postictal state. They are triggers to stop assuming that postictal physiology explains everything. Monitoring is also appropriate after an acute supratentorial insult when consciousness is depressed, fluctuating, or worse than imaging would predict. Relevant conditions include subarachnoid and intracerebral hemorrhage, traumatic brain injury, large cortical ischemic stroke, encephalitis, central nervous system infection, hypoxic–ischemic injury, neurosurgical complications, and selected tumors. Paralysis, deep sedation, or therapeutic coma removes the motor examination that would otherwise reveal seizures and therefore lowers the threshold for surveillance (Herman et al., 2015a).

Continuous EEG also belongs in the evaluation of unexplained coma or fluctuating encephalopathy without an established primary brain lesion. Sepsis, hepatic or renal failure, medication toxicity, withdrawal, autoimmune disease, and severe electrolyte disturbance can simultaneously cause encephalopathy and lower the seizure threshold. Suspicious manifestations include gaze deviation, nystagmoid eye movement, eyelid flutter, facial or distal-limb twitching, episodic posturing, unexplained apnea, autonomic surges, abrupt aphasia, behavioral arrest, and stereotyped changes in responsiveness. None is specific. Shivering, rigors, clonus, tremor, dystonia, ventilator artifact, suctioning, chest percussion, ECG contamination, electrode movement, and periodic nursing care frequently imitate seizures. Synchronized video, bedside examination, and accurate event marking are therefore components of the diagnostic instrument rather than optional accessories.

The indication is strongest when finding or excluding a seizure will immediately alter management and weakest when neither result would change care. An awake patient who has returned to baseline after a single uncomplicated provoked seizure usually does not need days of monitoring. Neither does every patient with a stable toxic-metabolic encephalopathy and no epileptic features. Resource stewardship matters because prolonged low-yield studies can delay access for a patient with evolving nonconvulsive status. The relevant distinction is not simply illness severity. Recent seizures, cortical injury, coma, an unexplained fluctuating examination, and early epileptiform findings increase yield; a stable examination and an interpretable early recording without epileptiform abnormalities lower it.

In a foundational cohort of 570 critically ill patients, electrographic seizures occurred in 19%, were exclusively nonconvulsive in 92% of affected patients, and first appeared within 24 hours in 88% and within 48 hours in 93%. Comatose patients were more likely to have delayed detection, demonstrating why a normal short recording cannot be treated as equally reassuring in every patient (Claassen et al., 2004). Randomized trials clarify the limits of this detection advantage. In CERTA, 364 adults with impaired consciousness but no recent seizure or status received either 30–48 hours of continuous EEG or two 20-minute routine EEGs. Continuous monitoring detected seizures more than three times as often and prompted more antiseizure-medication changes, but six-month mortality was unchanged. A later multicenter tele-EEG trial similarly found numerically more nonconvulsive seizures with 24–72 hours of monitoring than with a routine study, without a statistically significant improvement in detection, mortality, or function. These studies do not show that continuous monitoring is unnecessary for established status, anesthetic titration, or verification of seizure control; those questions were either excluded or insufficiently represented. They show that more detection and more treatment do not automatically establish outcome benefit in heterogeneous, lower-risk populations (Limotai et al., 2025; Rossetti et al., 2020).

Monitoring should begin early enough to answer the question before spontaneous evolution or empiric medication obscures the result. Rapid-response or point-of-care EEG can shorten a dangerous access delay. In the prospective DECIDE study, median time to rapid EEG acquisition was approximately five minutes, compared with a median delay of almost four hours for conventional EEG, and the additional EEG information improved diagnostic accuracy relative to clinical assessment alone. Reduced spatial coverage can nevertheless miss focal, parasagittal, or low-amplitude seizures. Rapid EEG is best used as a bridge: generalized or lateralized high-burden activity may justify immediate treatment, consultation, or transfer, but a negative reduced-montage study does not end the evaluation when suspicion remains high. Full-montage video EEG should follow as soon as feasible (Vespa et al., 2020).

Once an interpretable first hour has been obtained, 2HELPS2B provides the best-validated practical framework for estimating subsequent seizure risk in hospitalized adults. Two points are assigned for brief potentially ictal rhythmic discharges, or BIRDs. One point each is assigned for a prior seizure, sporadic epileptiform discharges, lateralized periodic discharges, lateralized rhythmic delta activity or bilateral independent periodic discharges, a periodic or rhythmic pattern faster than 2 Hz, and a plus modifier such as superimposed fast, rhythmic, or sharply contoured activity. Among patients without a seizure during the screening hour, scores of zero, one, and at least two corresponded to observed seizure risks of approximately 3%, 12%, and 27%. Monitoring for one hour, 12 hours, and 24 hours, respectively, reduced the estimated residual risk below 5% (Struck et al., 2020).

The score is frequently misused. It does not determine who should receive the initial EEG because most of its variables require EEG data. It is not a prognostic score, does not estimate neuronal injury, and was not validated after cardiac arrest. A score of zero during pharmacologic burst suppression cannot authorize discontinuation before the anesthetic is withdrawn. A low score also becomes stale when the patient deteriorates, develops a new cortical lesion, is rewarmed, undergoes surgery, has antiseizure therapy reduced, or emerges from deep sedation. Seizure risk is a state variable. A major transition in brain state resets the clinical question and may require renewed monitoring or rescoring.

Once a definite seizure has been captured, 2HELPS2B is no longer a stopping rule. Continuous EEG should ordinarily continue until seizures have remained controlled for at least 24 hours and throughout continuous intravenous anesthetic treatment and withdrawal. A 2026 single-center time-to-event analysis of 117 patients estimated that seizure-free monitoring for approximately 21 hours reduced recurrence risk below 5% in patients without status epilepticus, whereas approximately 37 hours was required after status. These estimates arose from only 187 seizure intervals and should not be converted into new biological thresholds, but they support extending monitoring beyond the first seizure-free day in selected status patients (Herman et al., 2015a; Krishnamurthy et al., 2026).

Cardiac arrest is a major exception to short risk-based protocols because EEG is used for seizure detection, longitudinal background assessment, and prognostication across sedation and temperature transitions. The joint American Heart Association and Neurocritical Care Society statement recommends starting EEG as early as possible, monitoring comatose patients through rewarming, and continuing for 72–120 hours when consciousness does not recover. Daily intermittent EEG is an alternative when continuous monitoring is unavailable. When seizures or status occur, monitoring should continue for at least 24 hours after electrographic control and then be individualized. These consensus recommendations belong within a multimodal prognostic timeline; they do not imply that every abnormal postanoxic pattern should be suppressed (Hirsch et al., 2024).

The next task is to classify what the EEG shows. Under ACNS terminology, an electrographic seizure consists of epileptiform activity averaging more than 2.5 Hz for at least 10 seconds or any pattern with definite evolution lasting at least 10 seconds. Electrographic status epilepticus is continuous electrographic seizure activity for at least 10 minutes or electrographic seizures occupying at least 20% of any 60-minute period. An electroclinical seizure may be shorter than 10 seconds when it has a definite time-locked clinical correlate. The 20% hourly threshold creates reproducible language and was informed partly by pediatric observations associating more than 12 minutes of seizure per hour with greater neurologic decline. It is not proof that 19% is safe, 21% is injurious, or that the same dose-response relationship applies identically to every adult brain (Hirsch et al., 2021; Payne et al., 2014).

The ictal–interictal continuum begins below those seizure thresholds. ACNS includes periodic discharges or spike-wave activity averaging more than 1 and no more than 2.5 Hz; slower periodic activity between 0.5 and 1 Hz when it fluctuates or carries a plus modifier; and lateralized rhythmic delta activity faster than 1 Hz when accompanied by fluctuation or a plus modifier. These patterns must persist for at least 10 seconds and must not already meet seizure criteria. Generalized rhythmic delta activity without additional qualifying ictal features is not an IIC pattern. The label is deliberately electrographic. It communicates that a pattern has ictal characteristics and may contribute to impaired consciousness or neuronal stress; it does not declare nonconvulsive status or mandate anesthetic treatment (Hirsch et al., 2021).

Seizure risk is not uniform across the continuum. In a multicenter cohort of 4,772 critically ill adults, lateralized periodic discharges were associated with seizures at every studied frequency, with greater risk when plus modifiers were present. The seizure associations of lateralized rhythmic delta activity and generalized periodic discharges increased with frequency, particularly at 1.5 Hz or faster and with plus features. At frequencies of at least 2 Hz, seizures occurred in approximately 40% of patients with lateralized rhythmic delta activity and 32% with generalized periodic discharges. Generalized rhythmic delta activity did not confer greater seizure risk than having no rhythmic or periodic pattern. Stimulus induction did not independently add risk beyond the underlying pattern, so a SIRPID is neither reassuring merely because stimulation caused it nor automatically ictal merely because it is reproducible (Rodriguez Ruiz et al., 2017).

Frequency, prevalence, evolution, spatial spread, burden, and modifiers are therefore more informative than the pattern name alone. Abundant 2-Hz lateralized periodic discharges with superimposed fast activity adjacent to an acute cortical lesion are much closer to the ictal end than occasional static 0.5-Hz generalized periodic discharges during severe systemic encephalopathy. “Triphasic” is a morphologic modifier rather than an etiologic diagnosis. Hepatic, uremic, septic, medication-related, postictal, and anoxic encephalopathies can produce generalized periodic discharges with triphasic morphology, but similar patterns can occur during nonconvulsive status. The examination, background organization, temporal evolution, prior EEG, imaging, physiology, and response to correction of the systemic disorder remain essential.

The Salzburg criteria provide another structured approach when epileptiform activity remains below 2.5 Hz or rhythmic delta lacks definite evolution. Their original validation showed high diagnostic accuracy, but subsequent prospective testing demonstrates the danger of converting recognition criteria into a treatment mandate. In a 2024 cohort of 469 patients referred for possible nonconvulsive status, combining “certain” and “possible” Salzburg classifications produced 94% sensitivity and a 99% negative predictive value, but only 77% specificity and a 34% positive predictive value. False-positive classifications commonly involved postictal states, encephalopathies, serial discrete seizures, and SIRPIDs. The criteria are useful for systematic recognition and for avoiding false reassurance; they cannot replace clinical adjudication, longitudinal EEG, or a properly assessed treatment response (Leitinger et al., 2016; Ulvin et al., 2024).

Definite electrographic seizures and electrographic status should generally be treated when doing so remains consistent with the goals of care. Causal evidence is incomplete, but the rationale extends beyond simple association. In children, increasing seizure burden remained associated with neurologic decline after adjustment for diagnosis and illness severity. In 402 patients with subarachnoid hemorrhage, every additional hour of electrographic seizure burden was associated with approximately 10% higher odds of three-month disability or death and with worse cognitive performance. In a broader cohort of 1,967 hospitalized patients, greater peak epileptiform burden showed a dose-response association with poor discharge outcome. Residual confounding remains possible in every observational study: a more severely injured brain may generate more seizures and have a worse outcome for the same underlying reason. Nevertheless, convergent clinical, metabolic, and physiologic observations make prolonged electrographic seizures a reasonable treatment target (De Marchis et al., 2016; Payne et al., 2014; Zafar et al., 2021).

The argument for suppressing every IIC pattern is much weaker. In comatose patients with subarachnoid hemorrhage undergoing invasive multimodality monitoring, periodic discharges displayed frequency-dependent physiology. Cerebral blood flow initially increased as discharge frequency rose, but discharges above 2 Hz were associated with falling brain-tissue oxygen into the hypoxic range. This supplies biological plausibility that high-frequency periodic discharges can impose seizure-like metabolic demand. It does not prove that pharmacologic suppression improves function, that the finding generalizes beyond the monitored tissue of poor-grade subarachnoid hemorrhage, or that the harms of deeper sedation are justified (Witsch et al., 2017).

A treatment decision should integrate five questions into one judgment:

1. Is there a plausible clinical correlate such as aphasia, neglect, myoclonus, gaze deviation, or impaired responsiveness?
2. How close is the pattern to seizure in frequency, evolution, plus features, spatial spread, prevalence, and burden?
3. Is the underlying substrate capable of recovery and particularly vulnerable to secondary metabolic stress?
4. Is there corroborating evidence of injury, such as a new diffusion abnormality, hyperperfusion, focal hypermetabolism, rising intracranial pressure, falling brain-tissue oxygen, or metabolic crisis?
5. What will treatment cost in hypotension, mechanical ventilation, infection, delirium, weakness, and loss of the neurologic examination?

A 2.3-Hz evolving lateralized pattern over acutely inflamed cortex deserves a different response from static 0.5-Hz generalized periodic discharges in profound multisystem failure.

When uncertainty is clinically important, an antiseizure treatment trial should be designed as a diagnostic experiment rather than an open-ended escalation. The pattern, examination, proposed intervention, expected time course, and response criteria should be documented before treatment. A parenteral benzodiazepine may rapidly suppress an ictal pattern, but it can also suppress nonictal periodic discharges and worsen alertness. EEG suppression alone is therefore insufficient. The most convincing response combines disappearance or substantial reduction of the pattern with temporally related clinical improvement. Improvement in background organization may support biological relevance when immediate behavioral recovery is impossible, but an EEG-only response does not formally establish an electroclinical seizure. A relatively nonsedating parenteral antiseizure medication may permit a longer, less confounded assessment.

In a retrospective series of 64 patients with generalized triphasic-wave patterns of uncertain significance, 34% had a definite electroclinical response to either a benzodiazepine or a nonsedating antiseizure medication. Clinical responses occurred after benzodiazepines in 19%, whereas immediate or delayed responses followed nonsedating medications in 42%. Selection bias, nonrandom treatment, delayed responses, and simultaneous correction of systemic illness prevent those percentages from defining an optimal strategy. They do demonstrate that some apparently metabolic patterns are treatment responsive and that a negative benzodiazepine challenge does not reliably exclude an ictal contribution (O’Rourke et al., 2016).

The trial also requires a stopping rule. If the pattern disappears without clinical or meaningful background improvement, repeated sedating challenges may merely erase the biomarker while worsening the patient. If a nonsedating medication reduces frequency and burden and the examination improves, continued treatment is reasonable. If the pattern evolves into definite seizures or status, management should follow the established status pathway. Anesthetic coma should generally be reserved for definite status, recurrent high-burden seizures, or an exceptionally compelling continuum pattern accompanied by convergent evidence of ongoing injury. It should not be the reflexive response to every low-frequency periodic discharge.

TELSTAR provides the strongest warning against equating EEG suppressibility with clinical benefit. The trial randomized 172 comatose cardiac-arrest survivors with rhythmic or periodic activity to intensive suppression for at least 48 hours or standard care. Generalized periodic discharges between 0.5 and 2.5 Hz predominated. Sustained suppression was achieved in 56% of treated patients and 2% of controls, yet poor three-month neurologic outcomes occurred in 90% and 92%, respectively; mortality was 80% and 82%, and intensive treatment modestly prolonged ventilation and ICU care. TELSTAR does not prove that definite post-arrest seizures should be ignored and does not establish futility for lateralized patterns, evolving seizures, encephalitis, or other recoverable disorders. It demonstrates that suppressibility is not evidence of causal pathogenicity and that indiscriminate suppression of predominantly generalized postanoxic rhythmic and periodic patterns does not improve outcome (Ruijter et al., 2022).

Current post-arrest guidance consequently supports treating definite clinical or electrographic seizures and status according to usual standards, considering treatment of IIC patterns selectively or when longitudinal worsening occurs, and not treating isolated sporadic epileptiform discharges. Sporadic spikes communicate cortical irritability and influence seizure-risk estimation; they are not themselves a seizure burden that must be abolished. Periodic patterns communicate greater network organization and risk, but treatment intensity must remain proportional to their electroclinical and physiologic significance (Hirsch et al., 2024).

Quantitative trends and automated seizure-burden displays can make days of EEG searchable, but they remain screening instruments. Spectrograms, rhythmicity displays, amplitude-integrated trends, suppression ratios, and automated alerts can identify time windows requiring raw EEG review. They are vulnerable to rhythmic artifact, periodic movements, muscle activity, electrode failure, and patterns outside their training distribution. In the 2026 AccuRASE II validation of one point-of-care system’s updated algorithm, a 50% seizure-burden alert threshold produced 86% sensitivity and 94% specificity for definite electrographic status, but the positive predictive value was only 30%. Performance changed with the selected threshold, whether possible status was included, and the software version. An automated alert should accelerate expert review; it should never independently trigger anesthetic therapy (Sheikh et al., 2026).

The same discipline applies to human monitoring. “Continuous EEG” usually means continuous acquisition with intermittent review, not an expert watching every second. Review frequency must match risk. Active status, anesthetic titration, rapidly increasing seizure burden, or a pattern near the seizure threshold requires near-real-time surveillance and a predefined communication pathway. A stable, low-risk recording may be reviewed less frequently. Electrode impedances, video position, event-button use, sedative and antiseizure-medication timestamps, temperature, stimulation, and bedside procedures must be maintained so that apparent electrographic evolution is not merely technical or physiologic artifact. Quantitative displays must always be linked back to raw EEG and synchronized video (Herman et al., 2015b).

An actionable report should state the monitoring indication, technical limitations, relevant sedation and temperature, background continuity and reactivity, seizure number and burden, time of the last seizure, and whether seizures have clinical correlates. Any continuum pattern should be characterized by localization, frequency, prevalence, duration, plus modifiers, fluctuation, evolution, stimulus relationship, and change from the previous epoch. The report should distinguish electrographic seizure, electroclinical seizure, electrographic status, possible status, IIC activity, and sporadic epileptiform discharges. Most importantly, it should communicate whether risk is rising, falling, or unchanged and identify the finding that requires immediate notification. A label without trajectory is less useful than a trajectory without drama.

Monitoring can stop when the clinical question has been answered, residual seizure risk has fallen to an acceptable level, an adequate seizure-free interval has passed, and no impending transition is expected to raise risk again. It should continue while seizures remain active, treatment is being titrated, anesthetics are being withdrawn, the examination remains unexplained in the presence of epileptiform activity, or a high-risk pattern is evolving. The best stopping decision is not “24 hours because that is customary.” It is “the residual risk is now low enough, the brain state is stable enough, and additional recording is unlikely to change management.”

The next installment will remain in Category 08 and move beyond seizure detection to the rest of ICU EEG: background continuity, reactivity, symmetry, state changes, sleep architecture, suppression and burst suppression, quantitative trends for delayed cerebral ischemia, and the boundaries between diagnostic information and neuroprognostication.

References

Alkhachroum, A., Appavu, B., Egawa, S., Foreman, B., Gaspard, N., Gilmore, E. J., Hirsch, L. J., Kurtz, P., Lambrecq, V., Kromm, J., Vespa, P., Zafar, S. F., Rohaut, B., & Claassen, J. (2022). Electroencephalogram in the intensive care unit: A focused look at acute brain injury. Intensive Care Medicine, 48(10), 1443–1462. https://doi.org/10.1007/s00134-022-06854-3

Claassen, J., Mayer, S. A., Kowalski, R. G., Emerson, R. G., & Hirsch, L. J. (2004). Detection of electrographic seizures with continuous EEG monitoring in critically ill patients. Neurology, 62(10), 1743–1748. https://doi.org/10.1212/01.WNL.0000125184.88797.62

De Marchis, G. M., Pugin, D., Meyers, E., Velasquez, A., Suwatcharangkoon, S., Park, S., Falo, M. C., Agarwal, S., Mayer, S., Schmidt, J. M., Connolly, E. S., & Claassen, J. (2016). Seizure burden in subarachnoid hemorrhage associated with functional and cognitive outcome. Neurology, 86(3), 253–260. https://doi.org/10.1212/WNL.0000000000002281

Herman, S. T., Abend, N. S., Bleck, T. P., Chapman, K. E., Drislane, F. W., Emerson, R. G., Gerard, E. E., Hahn, C. D., Husain, A. M., Kaplan, P. W., LaRoche, S. M., Nuwer, M. R., Quigg, M., Riviello, J. J., Schmitt, S. E., Simmons, L. A., Tsuchida, T. N., & Hirsch, L. J. (2015a). Consensus statement on continuous EEG in critically ill adults and children, part I: Indications. Journal of Clinical Neurophysiology, 32(2), 87–95. https://doi.org/10.1097/WNP.0000000000000166

Herman, S. T., Abend, N. S., Bleck, T. P., Chapman, K. E., Drislane, F. W., Emerson, R. G., Gerard, E. E., Hahn, C. D., Husain, A. M., Kaplan, P. W., LaRoche, S. M., Nuwer, M. R., Quigg, M., Riviello, J. J., Schmitt, S. E., Simmons, L. A., Tsuchida, T. N., & Hirsch, L. J. (2015b). Consensus statement on continuous EEG in critically ill adults and children, part II: Personnel, technical specifications, and clinical practice. Journal of Clinical Neurophysiology, 32(2), 96–108. https://doi.org/10.1097/WNP.0000000000000165

Hirsch, K. G., Abella, B. S., Amorim, E., Bader, M. K., Barletta, J. F., Berg, K., Callaway, C. W., Friberg, H., Gilmore, E. J., Greer, D. M., Kern, K. B., Livesay, S., May, T. L., Neumar, R. W., Nolan, J. P., Oddo, M., Peberdy, M. A., Poloyac, S. M., Seder, D., … Geocadin, R. G. (2024). Critical care management of patients after cardiac arrest: A scientific statement from the American Heart Association and Neurocritical Care Society. Circulation, 149(2), e168–e200. https://doi.org/10.1161/CIR.0000000000001163

Hirsch, L. J., Fong, M. W. K., Leitinger, M., LaRoche, S. M., Beniczky, S., Abend, N. S., Lee, J. W., Wusthoff, C. J., Hahn, C. D., Westover, M. B., Gerard, E. E., Herman, S. T., Haider, H. A., Osman, G., Rodriguez-Ruiz, A., Maciel, C. B., Gilmore, E. J., Fernandez, A., Rosenthal, E. S., … Gaspard, N. (2021). American Clinical Neurophysiology Society’s standardized critical care EEG terminology: 2021 version. Journal of Clinical Neurophysiology, 38(1), 1–29. https://doi.org/10.1097/WNP.0000000000000806

Krishnamurthy, P. V., Philibert-Rosas, S., Rivkin, M., Brace, C. J., Haworth, S. E., Fatima, S., Selte, A., Kalkach Aparicio, M., & Struck, A. F. (2026). Duration of EEG monitoring needed to ensure a low risk of seizure recurrence in hospitalized patients. Journal of Neurology, 273, Article 70. https://doi.org/10.1007/s00415-025-13596-x

Leitinger, M., Trinka, E., Gardella, E., Rohracher, A., Kalss, G., Qerama, E., Höfler, J., Hess, A., Zimmermann, G., Kuchukhidze, G., Dobesberger, J., Langthaler, P. B., & Beniczky, S. (2016). Diagnostic accuracy of the Salzburg EEG criteria for non-convulsive status epilepticus: A retrospective study. The Lancet Neurology, 15(10), 1054–1062. https://doi.org/10.1016/S1474-4422(16)30137-5

Limotai, C., Jirasakuldej, S., Wongwiangiunt, S., Tumnark, T., Suwanpakdee, P., Wangponpattanasiri, K., Rakchue, P., Tungkasereerak, C., Pleumpanupatand, P., Tansuhaj, P., Ekkachon, P., Kittipanprayoon, S., Kerddonfag, A., Pobsuk, T., Pattanateepapon, A., Phanthumchinda, K., Suwanwela, N. C., Thaipisuttikul, I., Boonyapisit, K., … Rattanayuvakorn, P. (2025). Efficacy of delivery of care with tele-continuous EEG in critically ill patients: A multicenter randomized controlled trial (Tele-cRCT study). Critical Care, 29(1), Article 15. https://doi.org/10.1186/s13054-024-05246-x

O’Rourke, D., Chen, P. M., Gaspard, N., Foreman, B., McClain, L., Karakis, I., Mahulikar, A., & Westover, M. B. (2016). Response rates to anticonvulsant trials in patients with triphasic-wave EEG patterns of uncertain significance. Neurocritical Care, 24(2), 233–239. https://doi.org/10.1007/s12028-015-0151-8

Payne, E. T., Zhao, X. Y., Frndova, H., McBain, K., Sharma, R., Hutchison, J. S., & Hahn, C. D. (2014). Seizure burden is independently associated with short-term outcome in critically ill children. Brain, 137(5), 1429–1438. https://doi.org/10.1093/brain/awu042

Rodriguez Ruiz, A., Vlachy, J., Lee, J. W., Gilmore, E. J., Ayer, T., Haider, H. A., Gaspard, N., Ehrenberg, J. A., Tolchin, B., Fantaneanu, T. A., Fernandez, A., Hirsch, L. J., & LaRoche, S. M. (2017). Association of periodic and rhythmic electroencephalographic patterns with seizures in critically ill patients. JAMA Neurology, 74(2), 181–188. https://doi.org/10.1001/jamaneurol.2016.4990

Rossetti, A. O., Schindler, K., Sutter, R., Rüegg, S., Zubler, F., Novy, J., Oddo, M., Warpelin-Decrausaz, L., & Alvarez, V. (2020). Continuous vs routine electroencephalogram in critically ill adults with altered consciousness and no recent seizure: A multicenter randomized clinical trial. JAMA Neurology, 77(10), 1225–1232. https://doi.org/10.1001/jamaneurol.2020.2264

Ruijter, B. J., Keijzer, H. M., Tjepkema-Cloostermans, M. C., Blans, M. J., Beishuizen, A., Tromp, S. C., Scholten, E., Horn, J., van Rootselaar, A. F., Admiraal, M. M., van den Bergh, W. M., Elting, J. W. J., Foudraine, N. A., Kornips, F. H. M., van Kranen-Mastenbroek, V. H. J. M., Rouhl, R. P. W., Thomeer, E. C., Moudrous, W., Nijhuis, F. A. P., … Hofmeijer, J. (2022). Treating rhythmic and periodic EEG patterns in comatose survivors of cardiac arrest. The New England Journal of Medicine, 386(8), 724–734. https://doi.org/10.1056/NEJMoa2115998

Sheikh, Z. B., Fong, M. W. K., Dhakar, M. B., Fang, W., Ayub, N., Molino, J., Haider, H. A., Foreman, B., Gilmore, E. J., Mizrahi, M., Karakis, I., Schmitt, S., Osman, G., Yoo, J. Y., & Hirsch, L. J. (2026). Accuracy of the rapid-response electroencephalography’s Automated Seizure Burden Estimator: A follow-up validation study of version 8 (AccuRASE II). Epilepsia. Advance online publication. https://doi.org/10.1002/epi.70352

Struck, A. F., Tabaeizadeh, M., Schmitt, S. E., Rodriguez Ruiz, A., Swisher, C. B., Subramaniam, T., Hernandez, C., Kaleem, S., Haider, H. A., Cissé, A. F., Dhakar, M. B., Hirsch, L. J., Rosenthal, E. S., Zafar, S. F., Gaspard, N., & Westover, M. B. (2020). Assessment of the validity of the 2HELPS2B score for inpatient seizure risk prediction. JAMA Neurology, 77(4), 500–507. https://doi.org/10.1001/jamaneurol.2019.4656

Ulvin, L. B., Nilsen, K. B., Taubøll, E., Etholm, L., & Heuser, K. (2024). Sensitivity and specificity of the Salzburg EEG criteria for nonconvulsive status epilepticus. Annals of Clinical and Translational Neurology, 11(10), 2685–2695. https://doi.org/10.1002/acn3.52184

Vespa, P. M., Olson, D. W. M., John, S., Hobbs, K. S., Gururangan, K., Nie, K., Desai, M. J., Markert, M., Parvizi, J., Bleck, T. P., Hirsch, L. J., & Westover, M. B. (2020). Evaluating the clinical impact of rapid response electroencephalography: The DECIDE multicenter prospective observational clinical study. Critical Care Medicine, 48(9), 1249–1257. https://doi.org/10.1097/CCM.0000000000004428

Witsch, J., Frey, H. P., Schmidt, J. M., Velazquez, A., Falo, C. M., Reznik, M., Roh, D., Agarwal, S., Park, S., Connolly, E. S., & Claassen, J. (2017). Electroencephalographic periodic discharges and frequency-dependent brain tissue hypoxia in acute brain injury. JAMA Neurology, 74(3), 301–309. https://doi.org/10.1001/jamaneurol.2016.5325

Zafar, S. F., Rosenthal, E. S., Jing, J., Ge, W., Tabaeizadeh, M., Aboul Nour, H., Shoukat, M., Sun, H., Javed, F., Kassa, S., Edhi, M., Bordbar, E., Gallagher, J., Moura, V., Jr., Ghanta, M., Shao, Y.-P., An, S., Sun, J., Cole, A. J., & Westover, M. B. (2021). Automated annotation of epileptiform burden and its association with outcomes. Annals of Neurology, 90(2), 300–311. https://doi.org/10.1002/ana.26161