Early studies consisted primarily of case series, which were important in documenting the types of SE, their relative frequencies, and common etiologies. In order to determine the incidence of a condition such as SE, one must define a population, and then attempt to capture and document every case, over a defined time interval, usually at least a year, and often much longer. Such studies generally focus on a city, county, or defined community and then seek to capture every case that occurs. The active ascertainment process involves identifying cases using multiple strategic data sources such as ambulance reports, emergency department visits, hospital admissions, medical records, consultation requests, electroenchepalography (EEG) reports, telephone referrals, and so forth. Cases are then reviewed, confirming or excluding cases based on the study’s criteria for SE. By tabulating all cases, and using total population data from the defined region, incidence rate, mortality, and other population-based parameters can be calculated. Although time- and labor-intensive, such studies provide valuable information with a degree of accuracy that reflects the diligence of the surveillance.

Over the past decade, studies of large data sets have yielded a different type of information. In addition to demographic information, data sets from the Centers for Disease Control and Prevention, the Nationwide Inpatient Sample, and the National Hospital Discharge Survey contain coded diagnostic information, such as hospital discharges, complications and comorbidities, and death certificate diagnoses. One drawback is that the data included in these resources are subject to coding errors, changes in coding patterns, and limitations in the codes themselves. For example, the International Code of Diseases, Ninth Revision (ICD-9) contains three codes for SE: 345.3—“grand mal status,” 345.7—“epilepsia partialis continua,” and 345.2—“petit mal” SE. A case of intermittent complex partial seizures without recovery of awareness in between might not even be recognized as SE or assigned an appropriate code.

The definition of SE has important implications for accuracy of coding. A validation study of the 345.3 ICD-9 code for SE ascertained cases of SE (including convulsive and nonconvulsive cases) that received a discharge code of 345.3 over 1 year at a large urban hospital. Also identified were cases with discharge records suggesting SE that were not overtly documented or coded as SE. Many more SE diagnoses were missed when a longer time-dependent SE definition was used. Using seizure duration as the criterion for SE, diagnostic sensitivity was 100% for 5 min. For SE lasting at least 10 min, however, sensitivity fell to 55%, and for SE lasting at least 20 min, it was just 14% [36].

Another study found that ICD-9 and ICD-10 coding accurately identified epilepsy, although validity of coding for specific types of epilepsy was suboptimal. Positive predictive value (PPV) for grand mal SE was 84% for ICD-9 coding and 100% for ICD-10, while the PPV for complex partial SE was 83% (ICD-10) [37]. Epilepsy was commonly miscoded with a nonspecific code for “convulsions.” The authors concluded that accurate surveillance would require including the code for convulsions, with adjustment for the small number of cases labeled with this code who do not actually have epilepsy. These limitations in documenting and coding SE must be kept in mind when interpreting studies that rely on coding rather than careful case ascertainment through close examination of actual medical records.

What large data sets lack in terms of diagnostic accuracy and case ascertainment they make up for in size and breadth, encompassing an enormous number of cases from all over the United States. Data have been collected over decades, allowing a broad view of trends over time. In these formats, SE data can be correlated, relatively easily, with other measures, including length of stay, concurrent medical conditions, procedures, and other parameters.

The incidence of SE (all types) in the Richmond, Virginia metropolitan area was 41 per 100,000 individuals per year. The incidences for the pediatric, adult, and elderly populations were 38, 27, and 86 per 100,000 per year, respectively [20]. These figures did not include repeat episodes of SE in a single patient. With validation of the database, it was determined that approximately 90% of all SE cases at the university hospital had been identified, compared to only one-third of cases in the community hospitals. When underreporting was taken into account, the revised estimate of the incidence of SE in the Richmond area was 61 per 100,000. The overall mortality was 9 per 100,000, with a revised estimated mortality of 17 per 100,000. Extrapolating these figures to the United States population yielded an estimated annual national incidence of 152,000 cases of SE and 42,000 deaths associated with SE per year. These numbers underscore the broad scope of SE in the United States.

Rochester, Minnesota. A retrospective study from the Mayo Clinic, looked at SE in Rochester, Minnesota, between 1965 and 1984 [23]. All cases of febrile seizures, acute symptomatic seizures, unprovoked seizures, or epilepsy were reviewed to identify and classify SE. The study identified 199 first episodes of SE during the 20-year period. The incidence of SE was 18.3 per 100,000. This is considerably lower than the incidence from the Richmond study, which may be due to different study methods (retrospective vs. prospective) and differing racial composition of the populations. The majority of the Minnesota study population (96%) was Caucasian, while the majority in Richmond (57%) was African–American. The incidence of SE in Richmond Caucasians was 20 per 100,000, significantly less than that in African–Americans (see section below on Race and Status Epilepticus). The incidence rates of SE in the Richmond and Rochester studies are comparable when racial factors are taken into account.

Italy. A study of incidence and short-term prognosis of SE used prospective surveillance of public general hospitals in Bologna, Italy, and reviewed all epilepsy discharge codes. An annual incidence of 13.1 per 100,000 was found, with the highest incidence in the elderly. The cause of the majority of cases of SE was acute symptomatic illness (48%), with stroke the most frequent etiology (41%). Over one-third (39%) of patients reported a history of seizures, and the 30-day case fatality was 39% [26].

A 2-year population-based study of SE in adults in a rural region of Northern Italy reported an adjusted annual SE incidence of 11.6/100,000. The crude incidence in adults over age 60 was 38.6/100,000—more than 10-fold the value for younger individuals. Acute symptomatic SE, primarily due to cerebrovascular disease, accounted for 30% of cases. Although the risk factors for SE in rural and urban Italy were similar, the rural 30-day case fatality of 7% was much lower than the Bologna rate of 39%. The authors inferred that short-term prognosis was influenced by differences in health service organization [27].

Hessen, Germany. A prospective population-based study in Germany identified 150 adult patients with SE over a 2-year period [25]. Patients were reported by neurologists and by intensive care unit and emergency department physicians and nurses. The calculated, corrected, age-adjusted incidence of SE was 17.1 per 100,000, higher in the elderly and in men. Seventy-four percent had a remote or acute brain insult as the etiology, with remote cerebrovascular disease the most frequent etiology, probably contributing to the increased incidence of SE in men and in the elderly. Fifty percent of the patients had a history of epilepsy, and the case fatality rate was 9.3%.

French-speaking Switzerland. A study of SE in Switzerland collected cases of SE prospectively in 60 hospitals in six French-speaking cantons over a 1-year period [24]. One hundred seventy-two cases were identified by physicians working in hospital emergency rooms, intensive care units, and EEG departments, and by neurologists and pediatricians. The standardized annual incidence rate was 10.3 per 100,000, higher among children under the age of one, and in the elderly, and higher among men than women. The case fatality rate was 7.6%.

La Reunion. La Reunion is a French overseas island territory east of Madagascar in the Indian Ocean. A population-based study, excluding patients with known epilepsy, found an incidence of 8.52 per 100,000 [28]. In this population, SE was most common in the elderly, and in men, and 60% of the SE was convulsive. Almost half of SE cases were provoked, by factors such as stroke, alcoholism, and infections. Mortality was 18.5%.

Table 3.2 [38–42] summarizes several studies of large data sets that rely on hospital coding.

California. A study of SE in California focused on generalized convulsive SE only and obtained data by reviewing a state-wide hospital discharge database covering hospitalizations between 1991 and 1998 [38]. It relied on ICD-9 coding of convulsive SE, which is subject to inaccuracies because SE is sometimes not recognized as such and may be coded as seizures or epilepsy rather than SE. Thus, the incidence rates in this study may be underestimates. The overall incidence was 6.2 per 100,000 population, and it declined significantly over the 1991–1998 study period, from 8.5 to 4.9 per 100,000. The case fatality rate for incident admissions was 10.7%.

United States. An analysis of over 760,000 discharges with an ICD-0 diagnosis code of 345.3 (“grand mal status epilepticus”) found an increasing incidence of SE in the United States over time, while mortality rates remained relatively stable. Between 1979 and 2010, SE incidence rose from 3.5 to 12.5 per 100,000, with the largest increases in the first and last decade. In-hospital mortality was 9.2%. Factors that may have influenced the rising incidence of SE include increased recognition of post-anoxic myoclonic SE, increased EEG availability, the evolving definition of SE with a shorter duration, and an expanding elderly population, among whom SE is more likely to occur [39].

Taiwan. A 12-year cohort study in Taiwan identified patients using a database of discharge diagnoses [41]. This study found an incidence of 4.61 per 100,000 person-years, and like other studies, confirmed a J-shaped age distribution. In-hospital mortality was lower in males (7.4%) than in females (11%). The authors postulated that the lower incidence they found may have been due to ethnic factors or due to methodological differences.

Thailand. In Thailand, statistics from a database of reimbursement claims found 5.1 SE cases per 100,000 adults over 1 year, with 12% mortality [43]. A longitudinal study found that SE incidence in adults rose steadily from 1.29 to 5.2 per 100,000 between 2004 and 2012. The in-hospital mortality was 8.4% [42]. Risk factors for poor outcome included female sex, age greater than 60, and primary care hospital location.

There are few large-scale studies of SE in developing countries. Most of the information about SE in African countries comes from case series and cohort studies that suggest that SE is at least as common as in more developed countries. An 11-year study of SE in Senegal documented 697 cases, with a mortality rate of 24.8% [44]. The most common etiology was infection (67%), followed by epilepsy. In Nigeria, 41 cases were diagnosed over a 10-year period at University College Hospital in Ibadan, with the most common etiology being CNS infection (41%) [45]. The incidence of SE in a cohort study of Kenyan children was 35/100,000 [46].

Several studies address the occurrence of SE in people with epilepsy. A study at a university hospital in Benghazi, Libya, found that 55 of 568 adult patients had SE [47]. One study looked at infantile SE, and found 139 infants treated for SE at two Tunisian hospitals over a 7-year period. The mortality was 15.8% and the most frequent causes were fever and acute symptomatic diagnoses [48]. Several other studies of epilepsy suggest that SE is a common cause of death in epilepsy patients in Africa [49–51].

In most of the population-based studies discussed here, mortality is defined as death within 30 days of SE. The overall mortality of the Richmond study population was 22%, but there was a dramatic difference between pediatric and adult mortality. Pediatric mortality was only 3%, while adult SE mortality was 26%. The elderly had the highest mortality, 38% [21]. In the Rochester, Minn. population, 30-day mortality was 19% following a first episode of SE. Short-term mortality was associated with an underlying acute symptomatic etiology [22]. Mortalities and case fatality rates in other epidemiologic studies ranged from 7.6 to 39% (see Table 3.1). Clinical factors influencing mortality are discussed further in the section below, Determinants of Mortality in Status Epilepticus.

Large data set studies have found lower case fatality rates and lower SE incidence than older population-based studies (see Table 3.2). In-hospital mortality was 9.2% for generalized convulsive SE in the National Hospital Discharge Survey [39]. The calculated mortality rate based on the underlying cause of death from death certificate data was 2 per million in 2010 [40]. The discrepancy between this low mortality rate and the much higher rates of previous studies is due to different approaches in assessing mortality. This study counted deaths for which SE was listed on the death certificate as an underlying cause, while other studies included in their rates all subjects with SE who died, regardless of the cause.

In some of these studies, patients included those admitted to the hospital for SE or epilepsy-related problems, as well as those admitted for other medical or surgical reasons who subsequently developed SE. These two groups may represent distinct subpopulations of SE patients. Hospitalized patients who develop SE de novo have an exceptionally high mortality rate of 61% [52]. This mortality is not associated with SE duration and may be due to serious comorbid conditions, most commonly recent or remote stroke.

Longitudinal data demonstrate an increase in SE incidence, while case fatality rates have remained stable. In the Rochester, Minn. population, the age-adjusted incidence of a first episode of SE increased over time from 14.1 per 100,000 between 1945 and 1954, to 18.1 per 100,000 between 1975 and 1984. The increase in incidence was due to the increasingly frequent occurrence of myoclonic SE after cardiac arrest, an uncommon condition in the earlier decades. Before 1965 there were no cases of myoclonic SE in this study. By 1975–1984, approximately 16% of SE was myoclonic SE, usually in the setting of anoxic encephalopathy following cardiac arrest in the elderly [53]. SE etiologies remained similar over the two decades spanning 1970–1989 [54].

Another potential contributing factor to the increasing incidence of SE may be better recognition of subtle forms of SE. The incidence of “nonmotor SE” increased fivefold between 1935–1944 and 1945–1954 and increased modestly thereafter to 2.7 per 100,000 for 1975–1984 [53].

Despite rising incidence rates, mortality rates increased only slightly between 1955 and 1984, from 3.6 to 4.0 per 100,000, and the 30-day case fatality rate was unchanged [53]. When cases of myoclonic SE were excluded from analysis, SE survival improved during 1975–1984, compared with previous decades [23, 53].

Large data set studies also report a growing incidence of SE without a corresponding bump in mortality. A study of 32 years of U.S. National Hospital Discharge Survey data found that SE incidence increased nearly fourfold, while mortality remained relatively unchanged [39]. Another study of SE-related hospitalization over 12 years found that SE incidence increased 56% between 1999 and 2010, while SE-related mortality rose just 5.6% [40]. The largest increase in SE incidence occurred in intubated patients for whom SE was not the principal diagnosis, suggesting better recognition of SE in ICU patients, possibly due to EEG monitoring.

SE occurs in 1.1–1.4% of first-time stroke patients [59, 60]. In most studies, stroke is a major etiology for SE in older adults. When ischemic stroke and cerebral hemorrhages, both acute and remote, are considered together, cerebrovascular disease is associated with 41% of adult SE cases [20]. Stroke was also the most common etiology of SE in European studies [25, 26]. In the California study of convulsive SE, the most common etiologies were “late effects of stroke/brain injury” (10.8%), developmental delay (9.9%), sodium imbalance (8.7%), alcoholism (8.1%), and anoxia (8%) [38]. A study at a large urban hospital in the 1980s found that ASD withdrawal, rather than stroke, was the most common cause of SE in adults, with alcohol-related causes the second most common etiology [54].

The etiology associated with the highest mortality in the Richmond study was anoxia [21, 39]. In the Rochester population, cerebrovascular disease and anoxic encephalopathy following cardiac arrest were the most frequent causes of SE followed by death within 30 days [22]. Low ASD levels had a mortality of only 4% [43]. Other studies reporting low overall SE mortality included a high percentage of patients with ASD withdrawal and alcohol-related etiologies [54]. Low mortality has been associated with unknown or remote symptomatic etiology [38, 61].

Causes of SE differ significantly in the pediatric and adult populations. In children, the most common etiology is infection with fever, present in slightly over half of cases. This was the only pediatric etiology that had any associated mortality—5%. Remote symptomatic etiologies occurred in 38% and low ASD levels in 21% of children with SE [20].

When the association between seizure type and mortality was examined, the mortality rate for partial seizures was surprisingly high, at 30.5%. Those with generalized tonic-clonic seizures, including secondarily generalized seizures, had a mortality rate of 20.7% [21]. Mortality rates for secondarily generalized SE ranged from 22 to 47% [22, 26]. Absence SE was not associated with any significant mortality, while mortality following myoclonic SE was as high as 68% [22, 26]. Seizure type was not a significant independent risk factor for mortality [22, 56].