Elsevier

Epilepsy Research

Volume 109, January 2015, Pages 48-56
Epilepsy Research

Non-expert use of quantitative EEG displays for seizure identification in the adult neuro-intensive care unit

https://doi.org/10.1016/j.eplepsyres.2014.10.013Get rights and content

Highlights

  • Non-convulsive seizures can only be identified by continuous EEG.

  • Lack of electrophysiologists in ICUs may delay recognition of these seizures.

  • Digital trend analysis methods can help identify ictal events.

  • Non-expert ICU staff can be trained to use these methods efficiently.

  • Alerting the electrophysiologist may allow earlier seizure identification.

Summary

Video-EEG monitoring is the ultimate way to diagnose non-convulsive status epilepticus (NCSE) in intensive care units (ICU). Usually EEG recordings are evaluated once a day by an electrophysiologist, which may lead to delay in diagnosis. Digital EEG trend analysis methods like amplitude integrated EEG (aEEG) and density spectral array (DSA) have been developed to facilitate recognition of seizures. In this study, we aimed to investigate the diagnostic utility of these methods by non-expert physicians and ICU nurses for NCSE identification in an adult neurological ICU. Ten patients with NCSE and ten control patients without seizures were included in the study. The raw EEG recordings of all subjects were converted to both aEEG and DSA and displayed simultaneously without conventional EEG. After training for seizure recognition with both methods, two physicians and two nurses analyzed the visual displays individually, and marked seizure timings. Their results were compared with those of a study epileptologist. Participants analyzed 615 h of EEG data with 700 seizures. Overall, 63% of the seizures were recognized by all, 15.6% by three, 11.6% by two, 8.3% by one rater and only 1.5% were missed by all of them (sensitivity was 88–99%, and specificity was 89–95% when the ratings were assessed as 1-h epochs). False positive rates were 1 per 2 h in the study and 1 per 6 h in the control groups. Interrater agreement was high (κ = 0.79–0.81). Bilateral independent seizures and ictal recordings with lower amplitude and shorter duration were more likely to be missed. There was no difference in performance between the rating of physicians and nurses. Our study demonstrates that bedside nurses, ICU fellows and residents can achieve acceptable level of accuracy for seizure identification using the digital EEG trend analysis methods following brief training. This may help earlier notification of the electrophysiologist who is not always available in ICUs.

Introduction

Status epilepticus (SE) is a neurological emergency that usually necessitates treatment in the intensive care unit (ICU). Besides, a considerable proportion of critically ill patients who are admitted to ICUs for various reasons suffer from seizures and SE, which might not uncommonly manifest as non-convulsive episodes in the intensive care setting (Towne et al., 2000). Long-term video-EEG monitoring (VEEGM) is the ultimate way to diagnose non-convulsive status epilepticus (NCSE) and monitor treatment of convulsive and non-convulsive SE especially when intravenous anesthetics are used. Despite these apparent advantages, the feasibility of VEEGM in the critical care settings is questioned due to its several disadvantages: the first is generation of a huge amount of EEG data requiring lengthy time and heavy effort for interpretation and analysis. Second, these EEG data are usually complicated with environmental and patient-originated artifacts, physiological events, pharmacological agents and pathological changes. Therefore, accurate and timely assessment of critical care EEG is not realistically achievable for ICU personnel, unless an around the clock available neurophysiologist with experience in epileptology is incorporated into the team. Considering the unquestionable and repeatedly demonstrated yield of critical care EEG, but unfortunately the paucity of such experienced professionals, a faster, easy to master and practical EEG methodology is therefore definitely needed to increase the utility of VEEGM in the intensive care setting.

In this respect, incorporation of quantitative EEG techniques or digital trend analysis (DTA) methods into routine monitoring might be a step in the right direction (Kurtz et al., 2009, Scheuer and Wilson, 2004, Stidham et al., 1980). Ictal events are typically characterized by prominent changes in the frequency and amplitude of EEG activity as compared to baseline, and therefore can be detected on the spectral display of quantitative EEG methods such as amplitude integrated EEG (aEEG) and density spectral array (DSA) or compressed spectral array (CSA). These techniques provide continuous on-line or real-time spectrographic display of trends in cerebral electrical activity at the bedside. Of note, in aEEG, the information from one channel is rectified, filtered, time compressed and displayed on a semi-logarithmic scale (El-Dib et al., 2009, Prior et al., 1971, Stidham et al., 1980, Toet and Lemmers, 2009), whereas in CSA or DSA fast-Fourier transformation is applied to convert raw EEG into time-compressed and color-coded display (Bickford et al., 1973, Salinsky et al., 1987, Scheuer and Wilson, 2004). Spectrogram screening has been shown to significantly reduce EEG review time, with minimal loss of sensitivity (Moura et al., 2014). Despite this potential advantage, quantitative EEG techniques have not attracted widespread attention in adult ICUs in contrast to their escalating use in neonatal ICU settings for various reasons (Hellström-Westas et al., 2008). Further, these techniques may allow ICU staff, who expectedly lack extensive EEG training and experience, to recognize significant EEG changes satisfactorily and timely. Two previous studies evaluating the utility of DTA methods for seizure detection in adult ICUs by inexperienced physicians have revealed contradictory results (Nitzschke et al., 2011, Williamson et al., 2014). The potential success rate of ICU nurses, on the other hand, is still unknown. Therefore, in this study, we aimed to revisit the maximum yield of DTA methods for seizure recognition in the hands of non-expert physicians, and more importantly nurses working in the critical care units.

Section snippets

Materials and methods

The study cohort was selected among consecutive patients who underwent continuous VEEGM in the adult neurological ICU (NICU) of our department between November 2009 and January 2013. The admission diagnoses of patients were mostly ischemic or hemorrhagic stroke, toxic–metabolic encephalopathy, central nervous system infection and hypoxic–ischemic encephalopathy. We excluded patients with convulsive SE. When the recording contained electrographic seizures without any accompanying convulsion, the

Results

The participants evaluated 289 h of EEG with 700 seizures in the study group and 326 h of EEG without seizures in the control group. Seizure count per recording varied from 10 to 182 in the study group. Interictally five patients had periodic lateralized epileptiform discharges (PLEDs); one patient had generalized periodic discharges (GPD) and another one had both GPDs and bilateral independent PLEDs. Seizures were generalized in one, focal unilateral in six and bilateral independent in three

Discussion

Our findings suggest that visual EEG display methods can be used effectively by the critical care nurses, residents and fellows for electrographic seizure monitoring, following a brief training period. Compared to the gold standard raw EEG reading of an epileptologist, these qualitative techniques provided high sensitivity (overall 93%), specificity (91–95%) and accuracy (AUC: 0.93) for seizure detection in the hands of these ICU personnel with no experience in critical care EEG and epilepsy.

Conclusions

Our study highlights that critical care nurses can identify non-convulsive seizures by DTA methods as good as ICU fellows and residents. A high accuracy among physician and non-physician ICU personnel is achievable by a brief training session; an additional basic education on raw EEG might even optimize the accuracy further by decreasing false positive readings. Application of these techniques in ICUs may lead to earlier diagnosis and thus treatment initiation in patients with NCSE. These

Conflict of interest

The authors declare that they have no conflict of interest.

Acknowledgement

The research was supported by Hacettepe University Research Fund Grant No: 1-801 105 001 (Dr. Nese Dericioglu).

References (25)

  • L. Hellström-Westas et al.

    Atlas of Amplitude Integrated EEG in the Newborns

    (2008)
  • P. Kurtz et al.

    Continuous EEG monitoring: is it ready for prime time?

    Curr. Opin. Crit. Care

    (2009)
  • Cited by (0)

    View full text