Diagnostic Yield of Electroencephalography in a General Inpatient Population

ORIGINAL ARTICLE

Diagnostic Yield of Electroencephalography in a General Inpatient Population John P. Betjemann, MD; Ivy Nguyen, BS; Carlos Santos-Sanchez, MD; Vanja C. Douglas, MD; and S. Andrew Josephson, MD Abstract Objective: To determine the frequency and clinical predictors of seizures and markers of epileptiform activity in a nonecritically ill general inpatient population. Patients and Methods: We performed a retrospective cohort study of patients 18 years and older who underwent inpatient electroencephalography (EEG) between January 1, 2005, and December 31, 2010, for an indication of spells or altered mental status. The EEGs and reports were reviewed for ictal activity, interictal epileptiform abnormalities, and nonepileptiform abnormalities. Demographic and clinical data were gathered from the electronic medical record to determine seizure predictors. Results: Of 2235 patients screened, 1048 met the inclusion criteria, of which 825 (78.7%) had an abnormal EEG finding. Seizures occurred in 78 of 1048 patients (7.4%), and interictal epileptiform discharges were noted in 194 of 1048 patients (18.5%). An intracranial mass and spells as the indication for the EEG were independently associated with the group of patients experiencing seizures in a multivariate logistic regression model (adjusted for age, sex, EEG indication, intracranial mass, stroke, and history of epilepsy). Ninety-seven percent of patients (69 of 71) experienced their first seizure within 24 hours of monitoring, and the presence of seizures was associated with a lower likelihood of being discharged (odds ratio, 0.45; 95% CI, 0.27-0.76). Conclusion: Seizures occurred at a high frequency in hospitalized patients with spells and altered mental status. The EEG may be an underused investigative tool in the hospital with the potential to identify treatable causes of these common disorders. ª 2013 Mayo Foundation for Medical Education and Research

For editorial comment, see page 312 From the Department of Neurology, University of California, San Francisco.

326

A

ltered mental status (AMS) and paroxysmal spells of uncertain origin are common concerns among hospitalized patients. Although some cases of impaired consciousness will be found to have metabolic or cardiac etiologies, some may represent seizures or nonconvulsive status epilepticus, which can be detected only by electroencephalography (EEG). Although EEG is the key test to making these diagnoses, it is relatively underused in the inpatient setting owing to lack of availability and neurologic consultation at many hospitals in the United States.1 The yield of EEG in this setting has been explored only on a limited basis in previous studies that were small and included a large proportion of patients in the intensive care unit (ICU), a population in which it is well known that seizures are common.2-4 Varelas et al5 determined that emergency EEG had a 10.7% yield for status epilepticus, and a history of cardiac or respiratory arrest and

n

Mayo Clin Proc. 2013;88(4):326-331

“suspicious clinical activity” were the only variables that predicted the presence of status epilepticus on emergency EEGs. Another report found that remote risk factors for seizures and ocular movement abnormalities were more likely to be present in patients with nonconvulsive status epilepticus compared with other types of AMS.6 However, neither of these studies were limited to the non-ICU setting in a general medical population. The diagnostic yield of EEG in patients admitted to the general inpatient service is, therefore, unknown, and this information would be helpful to a broad range of physicians caring for patients with AMS who are considering additional testing. Although researchers have tried to develop predictive decision rules for selecting patients with AMS for EEG,7 no model has acceptable sensitivity or specificity.8 We aimed to determine the frequency of seizures and markers of epileptiform activity (interictal epileptiform abnormalities) detected

Mayo Clin Proc. n April 2013;88(4):326-331 n http://dx.doi.org/10.1016/j.mayocp.2012.12.013 www.mayoclinicproceedings.org n ª 2013 Mayo Foundation for Medical Education and Research

YIELD OF EEG IN A GENERAL INPATIENT POPULATION

by EEG in a large cohort of hospitalized nonintubated patients. We also sought to identify clinical variables that are associated with positive EEG findings to help physicians determine when this resource-intense test should be ordered and to guide health care systems regarding the need for EEG availability. PATIENTS AND METHODS We performed a retrospective cohort study examining the frequency of seizures in patients admitted to a tertiary care hospital. Our hospital functions as a major tertiary care referral center with a large catchment area in northern California and Nevada. The hospital is not a trauma center, but given its urban setting, a diverse patient population with a variety of conditions is served, and it includes an active organ transplant program. Similar to many urban hospitals, patients with drug and alcohol abuse histories are frequently encountered. Given the retrospective nature of the study, a waiver of informed consent was used, and the local Committee on Human Research approved the study. Using an administrative claims database, all adult admissions to the University of California San Francisco Medical Center between January 1, 2005, and December 31, 2010, were screened to identify patients for whom an inpatient EEG was ordered. We excluded all patients who were intubated at the time the EEG was ordered. We did, however, include nonintubated ICU patients as they were generally believed to have a reliable neurologic examination that could be followed and were often awaiting transfer to the a non-ICU ward. We excluded all outpatient and research EEGs, pediatric (<18 years of age) EEGs, and EEGs performed on prisoners. In addition, we excluded all EEGs obtained in the emergency department and the inpatient psychiatric hospital. Finally, we screened the database and medical records to exclude all patients with suspected seizures electively admitted to the video EEG monitoring unit for scalp and intracranial recordings (Figure 1). Clinical data were obtained via medical record review. We recorded baseline demographic data and screened the medical record for potential seizure risk factors, including history of a brain mass (eg, tumor, abscess, or subdural or epidural hematoma), stroke, immunocompromised state (eg, human immunodeficiency Mayo Clin Proc. n April 2013;88(4):326-331 www.mayoclinicproceedings.org

n

2235 Patients screened

197 Admitted for elective seizure monitoring

726 Intubated

201 Othera

1048 Patients includedb

FIGURE 1. Patient flow diagram. aOther includes research electroencephalograms, intracranial grid recordings, prisoners, boarded outpatients (psychiatry and emergency department), and those with missing electroencephalography reports. b Sixty-three patients were screened but not included because their location in the hospital was not clear.

virus/AIDS, organ/bone marrow transplant, or long-term corticosteroid use), and history of epilepsy. Hospital admission notes were reviewed to obtain an admission diagnosis, and discharge summaries were used to record a discharge diagnosis, presence of an antiepileptic medication at the time of discharge, and disposition. All the EEGs were digitally recorded, with nearly all standardized according to the conventional International 10-20 System. In rare instances, especially in 2005 and 2006, some of the EEGs were digitally recorded using doubledistance electrode placement in a modified status epilepticus montage. Between 2005 and 2010, the University of California San Francisco Medical Center had 3 epileptologists who interpreted all adult EEG recordings. In a previous study during the same period, we demonstrated excellent interrater reliability (k¼0.88) for diagnosing electrographic seizures when reviewing the raw EEG tracings compared with EEG reports.9 For the purpose of this study, 2 epileptologists (J.P.B. and C.S.) independently reviewed the original EEG reports, including the findings and the interpretation of those findings. Because some EEG reports can be ambiguous in their

http://dx.doi.org/10.1016/j.mayocp.2012.12.013

327

MAYO CLINIC PROCEEDINGS

wording and interpretation, the Cohen k was calculated to determine the interrater agreement of the results of the EEG reports. For each patient, we recorded the total duration of and the indication for EEG (spell vs AMS/encephalopathy). The EEG was meant to be placed for 24 to 48 hours. In some patients, owing to either patient factors (pending hospital discharge or called to undergo magnetic resonance imaging) or successful capturing of the spell in question, the EEG was terminated earlier. When uncertainty existed or no spell was captured, EEG could be continued for an extended period (>48 hours). If a patient was admitted to the hospital multiple times in a calendar year, we included only the first admission of the year in an effort to reduce the potential bias introduced by patients with medically refractory epilepsy who may be admitted multiple times. If a patient had multiple EEGs during hospitalization, we reviewed only the initial period of recording, defined as ending when the patient was disconnected from EEG for more than 24 hours. The EEG reports were reviewed to identify (1) ictal activity/seizures (including status epilepticus), (2) interictal epileptiform abnormalities TABLE 1. Baseline Demographic and EEG Characteristics of the 1048 Study Patientsa Characteristic

Value

Age, mean (range) Male sex, No. (%) Race, No. (%) White Black Asian/Pacific Islander Native American Other Unknown Intensive care unit, No. (%) Reason for EEG, No. (%) Altered mental status Spells Duration of EEG, No.b 1 h >1-6 h >6-12 h >12-24 h >24-48 h >48 h

60 (41-79) 488 (46.6)

a

582 103 139 3 145 76 269

(55.5) (9.8) (13.3) (0.3) (13.8) (7.3) (25.7)

497 (47.4) 551 (52.6) 278 197 160 246 93 43

EEG ¼ electroencephalography. The EEG duration was unavailable for 31 patients and was not included in the analysis.

b

328

Mayo Clin Proc.

n

(periodic epileptiform discharges, focal spikes/ sharp waves, and temporal intermittent rhythmic delta activity [IRDA]), and (3) nonepileptiform interictal abnormalities (diffuse/focal slowing, triphasic waves, and other forms of IRDA). Ictal activity, which included any number of electrographic seizures, was defined as rhythmic activity or epileptiform discharges (spikes or sharp waves) that evolved in frequency, morphology, or location and lasted at least 10 seconds; this is a standard definition used in most epilepsy centers.10 Given the retrospective nature of this study, we did not attempt to determine which seizures were truly subclinical vs those that may have had subtle clinical manifestations that were unrecognized by physicians and staff. Widely accepted definitions were used for the following interictal epileptiform abnormalities: periodic lateralized epileptiform discharges, bilateral periodic lateralized epileptiform discharges, and generalized periodic epileptiform discharges.10 The IRDA was defined as relatively high-amplitude monorhythmic 1- to 3-Hz activity without clear evolution. The IRDA was further characterized by the location of the slowing, with particular attention given to temporal IRDA because it has been associated with seizures.11 We also recorded the time to first seizure gathered from the EEG reports and confirmed this by reviewing the original recordings when available (68 patients). Continuous data are summarized as mean  SD. Categorical data are summarized using frequency and percentage. The Cohen k was calculated to determine the interrater agreement of the results of the EEG reports. The c2 test was used to assess differences in categorical variables between patients having seizures and patients not having seizures. The t test was used to assess differences in continuous variables between patients having seizures and patients not having seizures. Multivariate logistic regression was used to determine independent predictors of having a seizure. The multivariate logistic regression model was adjusted for age, sex, spells as the indication for EEG, intracranial mass, stroke, and history of epilepsy. Statistical analysis was performed using a licensed copy of STATA, version 11.2.12 P<.05 was considered statistically significant. RESULTS We screened 2235 adults in whom inpatient EEGs were obtained between January 1,

April 2013;88(4):326-331

n

http://dx.doi.org/10.1016/j.mayocp.2012.12.013 www.mayoclinicproceedings.org

YIELD OF EEG IN A GENERAL INPATIENT POPULATION

Mayo Clin Proc. n April 2013;88(4):326-331 www.mayoclinicproceedings.org

n

223 (21.3%)

Normal EEG 78 (7.4%)

Seizures

41 (3.9%)

Periodic discharges

153 (14.6%)

Interictal discharges

358 (34.2%)

Focal slowing

706 (67.4%)

Diffuse slowing Triphasics

37 (3.5%) 0

200

400

A

600

800

No. of patients

100 First seizures detected (%)

2005, and December 31, 2010. After exclusions, 1048 patients were included in the final analysis (Figure 1). The baseline demographic and EEG characteristics are given in Table 1. The median duration of an EEG was 7.5 hours (interquartile range, 0.75-17.75 hours). Two epileptologists (J.P.B. and C.S.) independently reviewed the EEG reports and demonstrated excellent agreement (k>0.95). Of the 1048 patients reviewed, 825 (78.7%) had an abnormal EEG finding, with the most common abnormality being diffuse slowing, seen in 706 of 1048 patients (67.4%). Seizures (ranging from a single seizure to status epilepticus) were noted in 78 of 1048 patients (7.4%); 970 of 1048 patients (92.6%) did not experience discrete seizures (Figure 2, A). In addition to the 78 patients with seizures, epileptiform discharges, markers of seizure potential, were found in 194 of 1048 patients (18.5%). Epileptiform discharges were found in 130 of the 970 patients (13.4%) in whom no electrographic seizures were recorded. There were no significant differences in age, race, or sex between patients who experienced a seizure and those who did not. Similarly, there was no significant difference in immunocompromised state between patients who had seizures and those who did not. In a univariate analysis, we found that a history of epilepsy, the presence of a brain mass, and spells as the indication for EEG (as opposed to AMS) were significantly associated with the group of patients who experienced seizures (Table 2). Stroke was significantly associated with the group of patients who did not experience seizures. In the multivariate analysis, only the presence of a brain mass and spells as the indication for EEG were independently associated with the group of patients experiencing seizures (Table 3). We were able to determine the timing of the first seizure in 71 of the 78 patients who experienced seizures. Forty-six of the 71 patients (64.8%) experienced their first seizure within the first hour of EEG recording and 63 of 71 (88.7%) by 6 hours (Figure 2, B). Nearly all the patients who experienced a seizure, 69 of 71 (97.2%), had their first electrographic seizure within 24 hours of monitoring. Hospital discharge disposition data were available for 1040 of the 1048 patients (99.2%). There was a trend toward increased

88.7

90

94.3

97.1

97.1

12

24

48

100

80 70

64.8

60 50 40 30 20 10 0 1

B

6

>48

Time of EEG monitoring (h)

FIGURE 2. Electroencephalography (EEG) findings and time to first seizure. A, Findings from EEG in 1048 patients. Periodic discharges include generalized periodic epileptiform discharges, bilateral periodic epileptiform discharges, and periodic lateralized epileptiform discharges. Interictal discharges include generalized spike waves, focal spikes/sharp waves, and temporal intermittent rhythmic delta activity. B, Time to first seizure. Data were available for 71 of the 78 patients who experienced seizures.

in-hospital mortality in patients experiencing seizures (11 of 78 [14.1%]) compared with those without (76 of 962 [7.9%]) (P¼.06). Patients experiencing seizures were significantly less likely to be discharged (29 of 78 [37.2%]) compared with patients without seizures (495 of 962 [51.5%]) (P¼.02). In a multivariate analysis adjusting for age, sex, indication for EEG (spells or AMS), history of epilepsy, stroke, and brain mass, the odds ratio for being discharged with a seizure was 0.45 (95% CI, 0.27-0.76; P¼.003). DISCUSSION We found that 78 of 1048 patients (7.4%) admitted to a tertiary care hospital and

http://dx.doi.org/10.1016/j.mayocp.2012.12.013

329

MAYO CLINIC PROCEEDINGS

TABLE 2. Univariate Analysisa Variable Age, mean (range) Male sex, No. (%) Race, No. (%)b White Black Asian/Pacific Islander American Indian Other Intensive care unit, No. (%) Spells as indication for EEG, No. (%) Intracranial mass, No. (%)c Stroke, No. (%)d Immunocompromised, No. (%) History of epilepsy, No. (%)

No seizures (n¼970)

Seizures (n¼78)

60 (41-79) 453 (46.7)

58 (40-76) 35 (44.9)

541 97 128 3 132 247 487 193 395 70 171

41 (57.7) 6 (8.5) 11 (15.5) 0 13 (18.3) 22 (28.2) 64 (82.1) 40 (51.3) 23 (29.5) 4 (5.1) 23 (29.5)

(60.0) (10.8) (14.2) (0.33) (14.7) (25.5) (50.2) (19.9) (40.7) (7.2) (17.6)

P value .15 .76 .86

.59 <.001 <.001 .05 .49 .009

EEG ¼ electroencephalography. Those with unknown race were excluded from the race analysis. c Intracranial mass includes tumors, abscesses, vascular malformations, subdural and epidural hematomas, and cysts. d Stroke includes ischemic and hemorrhagic stroke. a

b

monitored with inpatient EEG were experiencing electrographic seizures. Additionally, epileptiform discharges were found in 130 of 970 patients (13.4%) in whom no electrographic seizures were recorded, a finding that indicates that these patients are at risk for seizures. Overall, most patients who experienced a seizure did so within 24 hours of EEG recording, and many were even identified with a brief 1-hour EEG. Overall, there is a paucity of work addressing the frequency of seizures in hospitalized patients. In studies mainly focusing on mixed populations including those who are intubated, the frequency of seizures identified by continuous EEG ranged from 6% to 37%.10,13-15 One particular study found a seizure rate of 19% TABLE 3. Likelihood for Detecting Seizures: Multivariate Analysis Variable

Odds ratio (95% CI)

P value

Age 1.00 (0.99-1.01) .87 Female sex 1.05 (0.65-1.71) .83 Spells as indication for EEG 4.42 (2.41-8.1) <.001 History of epilepsy 1.42 (0.82-2.48) .21 Intracranial mass 4.23 (2.59-6.9) <.001 Stroke 0.71 (0.42-1.21) .21 EEG ¼ electroencephalography.

330

Mayo Clin Proc.

n

in a mixed pediatric and adult population mainly in the neurologic ICU.10 This same study found that coma, age younger than 18 years, convulsion immediately before EEG, and a history of epilepsy were independent predictors of seizures. In another study of medical ICU patients, the rate of seizures was found to be 10%.13 The present findings underscore the relatively high frequency of seizures in hospitalized patients with spells or AMS, a finding that has seemingly been underappreciated by neurologists and nonneurologists alike. Because ictal disorders are treatable, having a relatively low threshold to obtain an EEG may be critical. Interictal epileptiform discharges were also frequently identified, a finding that often changes treatment decisions regarding initiation of antiepileptic medications. Electroencephalography is a scarce resource, and many hospitals do not have access to the technology. One major barrier involves a lack of trained technologists and epileptologists to perform and interpret extended (24-hour) EEG studies. This study demonstrates that hospitals with limited resources can perform relatively brief EEGs (1-6 hours) and still identify most seizures in these nonecritically ill patients. It is difficult to accurately calculate the cost of an EEG; however, Medicare reimbursement can serve as a crude surrogate. Standard national Medicare reimbursements for Current Procedural Terminology codes 95816 (video EEG of <1 hour) and 95953 (video EEG for each 24-hour segment) inclusive of technical and professional fees are $368.14 and $469.29, respectively.16 Telemedicine may serve as an ideal intervention in instances in which access to trained epileptologists is the primary issue, as EEG studies are becoming easier to read remotely with current technology. If a hospital system cannot obtain this study, some consideration should be given to transfer to a hospital with this important resource. This study has several noteworthy limitations. Similar to previous studies examining the yield of EEG, this study is retrospective and, therefore, the yield of seizures and epileptiform activity is potentially influenced by a selection bias. In addition, the retrospective design limits the ability to analyze the relationship between seizure frequency and hospital admission and discharge diagnoses. The University of California San Francisco Medical

April 2013;88(4):326-331

n

http://dx.doi.org/10.1016/j.mayocp.2012.12.013 www.mayoclinicproceedings.org

YIELD OF EEG IN A GENERAL INPATIENT POPULATION

Center is a tertiary care center, and follow-up data may be limited because many patients are, ultimately, transferred back to their referring hospitals to complete their hospitalization. Future endeavors should involve prospective investigations that test decision rules to determine who is at highest risk for seizures and examine the ideal duration of monitoring. CONCLUSION We demonstrated a high rate of electrographic seizures in a general inpatient population with spells or AMS as the indication for EEG. Many of these seizures can be identified in a relatively short period of EEG monitoring. This finding is important for physicians evaluating and treating patients with spells or AMS. ACKNOWLEDGMENTS We thank the entire EEG laboratory at the University of California, San Francisco, especially the EEG technologists, for their tireless efforts. Dr Betjemann and Ms Nguyen contributed equally to this work. Abbreviations and Acronyms: AMS = altered mental status; EEG = electroencephalography; ICU = intensive care unit; IRDA = intermittent rhythmic delta activity Potential Competing Interests: Dr Douglas receives compensation as editor of The Neurohospitalist. He received honorarium for speaking about neurohospitalists at a Grifols Inc national sales meeting. Ms Nguyen was awarded a Quarterly Research Fellowship by the University of California, San Francisco, Dean’s Office Medical Student Research Program to fund an 8-week summer project. Data Previously Presented: These data were presented in part at the 2012 American Academy of Neurology annual meeting in New Orleans, LA. Correspondence: Address to John Betjemann, MD, University of California, San Francisco, San Francisco General Hos-

Mayo Clin Proc. n April 2013;88(4):326-331 www.mayoclinicproceedings.org

n

pital, 1001 Potrero Ave, Bldg 1, Room 101, San Francisco, CA 94110 ([email protected]).

REFERENCES 1. Quigg M, Shneker B, Donner P. Current practice in administration and clinical criteria of emergent EEG. J Clin Neurophysiol. 2001;18(2):162-165. 2. Mirski MA, Varelas PN. Seizures and status epilepticus in the critically ill. Crit Care Clin. 2008;24(1):115-147. 3. Jordan KG. Continuous EEG and evoked potential monitoring in the neuroscience intensive care unit. J Clin Neurophysiol. 1993;10(4):445-475. 4. Privitera M, Hoffman M, Moore JL, Jester D. EEG detection of nontonic-clonic status epilepticus in patients with altered consciousness. Epilepsy Res. 1994;18(2):155-166. 5. Varelas PN, Spanaki MV, Hacein-Bey L, Hether T, Terranova B. Emergent EEG: indications and diagnostic yield. Neurology. 2003;61(5):702-704. 6. Husain AM, Horn GJ, Jacobson MP. Non-convulsive status epilepticus: usefulness of clinical features in selecting patients for urgent EEG. J Neurol Neurosurg Psychiatry. 2003;74(2): 189-191. 7. Khan RB, Yerremsetty PK, Lindstrom D, McGill LJ. Emergency EEG and factors associated with nonconvulsive status epilepticus. J Natl Med Assoc. 2001;93(10):359-362. 8. Chang AK, Shinnar S. Nonconvulsive status epilepticus. Emerg Med Clin North Am. 2011;29(1):65-72. 9. Kamel H, Betjemann JP, Navi BB, et al. Diagnostic yield of electroencephalography in the medical and surgical intensive care unit [published online July 21, 2012]. Neurocrit Care. http://dx.doi. org/10.1007/s12028-012-9736-7. Accessed December 1, 2012. 10. Claassen J, Mayer SA, Kowalski RG, Emerson RG, Hirsch LJ. Detection of electrographic seizures with continuous EEG monitoring in critically ill patients. Neurology. 2004;62(10):1743-1748. 11. Normand MM, Wszolek ZK, Klass DW. Temporal intermittent rhythmic delta activity in electroencephalograms. J Clin Neurophysiol. 1995;12(3):280-284. 12. STATA [computer program]. Release 11. College Station, TX: StataCorp LP; 2009. 13. Oddo M, Carrera E, Claassen J, Mayer SA, Hirsch LJ. Continuous electroencephalography in the medical intensive care unit. Crit Care Med. 2009;37(6):2051-2056. 14. Pandian JD, Cascino GD, So EL, Manno E, Fulgham JR. Digital video-electroencephalographic monitoring in the neurologicalneurosurgical intensive care unit: clinical features and outcome. Arch Neurol. 2004;61(7):1090-1094. 15. Vespa PM, O’Phelan K, Shah M, et al. Acute seizures after intracerebral hemorrhage: a factor in progressive midline shift and outcome. Neurology. 2003;60(9):1441-1446. 16. US Department of Health and Human Services. Centers for Medicare & Medicaid Services website. http://www.cms.gov. Accessed December 2, 2012.

http://dx.doi.org/10.1016/j.mayocp.2012.12.013

331