• Epidemiology;
  • Surveillance;
  • Epilepsy


  1. Top of page
  2. Summary
  3. Epidemiologic Methods for the Study of Epilepsy
  4. Conclusion
  5. Disclosures
  6. References
  7. Appendices

Worldwide, about 65 million people are estimated to have epilepsy. Epidemiologic studies are necessary to define the full public health burden of epilepsy; to set public health and health care priorities; to provide information needed for prevention, early detection, and treatment; to identify education and service needs; and to promote effective health care and support programs for people with epilepsy. However, different definitions and epidemiologic methods complicate the tasks of these studies and their interpretations and comparisons. The purpose of this document is to promote consistency in definitions and methods in an effort to enhance future population-based epidemiologic studies, facilitate comparison between populations, and encourage the collection of data useful for the promotion of public health. We discuss: (1) conceptual and operational definitions of epilepsy, (2) data resources and recommended data elements, and (3) methods and analyses appropriate for epidemiologic studies or the surveillance of epilepsy. Variations in these are considered, taking into account differing resource availability and needs among countries and differing purposes among studies.

Worldwide, there are an estimated at least 65 million people living with epilepsy (Ngugi et al., 2010). Reported estimates of epilepsy occurrence vary substantially among populations studied, but, in sum, indicate that in developed countries, the annual incidence of epilepsy is nearly 50 per 100,000 population, whereas the prevalence approximates 700 per 100,000 (Hirtz et al., 2007). In low and middle income countries, estimates of the corresponding rates are generally higher (Hauser, 1995; Kotsopoulos et al., 2002; Sander, 2003; Burneo et al., 2005b; Preux & Druet-Cabanac, 2005; Ngugi et al., 2010). Throughout the world, therefore, epilepsy imposes a substantial public health burden.

The term “epilepsy” encompasses many specific conditions in which unprovoked seizures occur that may have varying etiology, risk factors, and manifestations (Commission on Epidemiology and Prognosis of the International League Against Epilepsy, 1993). Although these are sometimes referred to as the “epilepsies,” we use the term epilepsy throughout this document. In epidemiologic studies of epilepsy conducted in both developed and low and middle income countries, different definitions and methods complicate the tasks of study interpretation and comparison. The purpose of this document is to promote consistency in methods and definitions in an effort to enhance future population-based epidemiologic studies, facilitate comparison between populations, and encourage the collection of data useful for the promotion of public health in diverse settings. At the same time, this guidance is intended to be flexible, recognizing that public health infrastructure, needs, and priorities vary among countries, especially when comparing developing and developed regions.

Epidemiologic Methods for the Study of Epilepsy

  1. Top of page
  2. Summary
  3. Epidemiologic Methods for the Study of Epilepsy
  4. Conclusion
  5. Disclosures
  6. References
  7. Appendices

Surveillance and epidemiologic studies

Public health surveillance is defined as “the ongoing systematic collection, analysis, and interpretation of health data necessary for designing, implementing, and evaluating public health prevention programs” (Guidelines Working Group, 2001). Examples of public health surveillance include department of health or ministry systems that monitor the incidence of diseases, including communicable diseases, as well as national mortality registries and systems relying on administrative data such as hospital discharge (separation) data. Only limited clinical details are usually collected by such systems, which in turn limits diagnostic precision and precludes more than rudimentary classification. As a standard public health practice, surveillance is usually distinguished from epidemiologic research, which typically involves collecting more extensive data over a limited period and often includes analyses to test novel hypotheses regarding specific questions of etiology or association. The distinctions between surveillance and research are not absolute.

A major purpose of epidemiologic studies and surveillance is to provide the information necessary for primary prevention, for early detection and treatment, for setting public health and health care priorities, and for identifying other education and service needs associated with health conditions. To assess the public health importance of epilepsy and to design and promote effective health care and service programs it is necessary to describe:

  •  the magnitude of the problem (e.g., total number of cases; incidence rate, i.e., the rate of new cases occurring in the population; mortality rate, i.e., the rate of deaths occurring with the condition; and prevalence, i.e., the proportion of the population with the condition)
  • populations at highest risk of epilepsy (e.g., demographic characteristics)
  • associations, risk factors, and causes
  • severity and outcome (e.g., seizure frequency and duration, symptoms, comorbidities, resulting disability, and cost of care)

Surveillance is also important to evaluate and monitor the effectiveness of health care programs, including trends in these measures over time. To serve all these purposes, it is important that epidemiologic and surveillance data be comparable over time and between locales. Therefore, inclusion criteria based on standard case definitions are important. Likewise, it is desirable to collect comparable data elements as described below.

Attributes of epidemiologic studies and surveillance

A number of attributes of epidemiologic studies and surveillance determine their success. The following are among the most important: (Guidelines Working Group 2001)

  • Economy. Economical methods reduce the costs and time required to collect and analyze data. Successful studies and systems avoid collecting unnecessary data and rely as much as possible on data from existing data collection systems, when of high quality and appropriate.
  • Acceptability. Data collection may require the acceptance and cooperation of many persons and organizations involved in reporting cases. Subjects may be asked to provide substantial time taking surveys and participating in other assessments. Successful studies and systems do not place costly, difficult, or unacceptable burdens on those who provide data.
  • Accuracy. The closeness of the measurement to the true value is critical. Accuracy of measurements of incidence or prevalence can be characterized by the qualities of sensitivity, specificity, and positive predictive value (PPV) of the tools to collect the data, which are described later in this paper.
  • Representativeness. Representative epidemiologic studies and surveillance systems include subjects whose characteristics and experience are similar to the population of interest. They accurately measure and describe the occurrence of epilepsy over time. Representativeness is critical if data are gathered in only a sample of epilepsy cases.

Data sources

The data needed for epidemiologic studies of epilepsy include items necessary for case ascertainment and verification, as well as items that help describe other characteristics of each case defined in later sections of this document. The methods chosen for such epidemiologic studies depend on the availability and content of different sources of data. Common sources include:

  •  Direct population surveys. Surveys conducting direct in-person interviews in households may attempt to contact: (1) all members of a smaller community (door-to-door surveys) (Haerer et al., 1986; Osuntokun et al., 1987; Aziz et al., 1994; Gourie-Devi et al., 1996; Karaagac et al., 1999; Nicoletti et al., 1999; Tran et al., 2006) or (2) a systematic representative sample of households (Centers for Disease Control and Prevention 1994; Wiebe et al., 1999). Such surveys are especially valuable in communities where many households lack telephones—as in some low and middle income countries—but are more time-consuming for interviewers. In other communities with nearly universal household telephone availability, it may be more efficient to conduct most surveys by telephone in a representative sample of households. With the recent proliferation of personal and mobile telephones, as well as increased screening of calls by potential respondents, rates of participation in telephone surveys have diminished substantially. Whether surveys are conducted in person or by telephone, limitations on the validity and clinical detail from self-reports and proxy reports must be considered (Pal et al., 1998).
    •   If resources permit, some limitations of self- or proxy-reported data may be overcome if the initial survey is used to screen for potential cases of epilepsy and if persons suspected of having epilepsy are subsequently evaluated in person by an epilepsy clinician for confirmation of the diagnosis and to collect additional information. Many population-based studies of epilepsy rely on such two-stage methods.
  •  Existing coded data. The International Classification of Diseases, Ninth Revision (ICD-9) (World Health Organization 1977) or Tenth Revision (ICD-10) (World Health Organization 2005), are used throughout the world to code death certificates for mortality registries. In addition, these classifications—or a variation like the U.S. International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) (U.S. Department of Health and Human Services 2003)—are widely used to code other medical records describing hospital admissions, or emergency department, clinic, or physician visits. In some areas, especially in developed countries, these coded data are collected for the population and maintained in computerized databases. Coding processes are prone to error, depending on the knowledge of the coders, accuracy of code transcriptions, and the accuracy and completeness of the original clinical records on which coding is based.
    •   Some countries with integrated universal health care systems have designed national or regional registers that can be used as sources for epidemiologic studies of epilepsy, for example, the Swedish Hospital Discharge Register and the Danish National Hospital Register (Nilsson et al., 1997, 1999; Adelow et al., 2006; Sun et al., 2006; Vestergaard et al., 2006). These systems can be used to identify study populations, and some may provide sufficient data to describe important characteristics of the population with epilepsy. In some countries or districts, it is possible to obtain population-based hospital discharge and other health care data that are being collected for administrative purposes. It should be noted that the validity and precision of these coded data may be limited, since a more general code, or the code that maximizes reimbursement, may be used in preference to the most accurate code.
    •   In some countries with universal health care systems, it is possible to link administrative data for an individual’s health care from hospitals, clinics, and physician offices. An example based on such data is the Canadian Chronic Disease Surveillance System (James et al., 2004; Dai et al., 2010). Although it lacks a national register, the U.S. National Center for Health Statistics does collect representative samples of coded health care data in its National Health Care Survey, which includes separate surveys of a national sample of hospitals, emergency departments, clinics, and physician offices. Individual patient care data across these surveys cannot be linked.

If coded records for all individual patients in a population can be linked to describe all inpatient and outpatient medical encounters over an extended time, then it becomes possible to ascertain most cases of epilepsy in that population and to estimate prevalence and incidence (Holden et al., 2005b). Heretofore such methods have not been commonly used; their sensitivity, specificity, and positive predictive values may be limited, and until the quality and completeness of coded data have been widely assured, an evaluation of the validity of these linked data is advised in each population studied (Nilsson et al., 1997; Christensen et al., 2005; Holden et al., 2005b; Parko & Thurman, 2009; Jette et al., 2010).

Linking additional coded data that describe prescriptions for antiepileptic drugs (AEDs) can be used to strengthen the sensitivity and diagnostic accuracy of coded data (Hauser et al., 1993; Holden et al., 2005b). This method can also be used for ascertaining putative cases of active epilepsy in epidemiologic studies (Olafsson & Hauser, 1999; Artama et al., 2005, 2006), but it has limited application if the treatment gap is large.

  •  Existing uncoded data. Medical records from hospitals, emergency departments, clinics, and physician offices (and on occasion, coroner or medical examiner reports) are the source material for coded data described above and contain much more information. They are highly useful both to validate (confirm diagnosis) and to supplement (obtain more clinical and other details) the information obtained from coded data. The process of locating, reviewing, and abstracting such records can be time-consuming, and is sometimes attempted on only a representative (e.g., random) sample of medical records.

The sensitivity of case ascertainment in population-based studies of epilepsy can be improved by using multiple sources of information. For example, direct survey methods may be enhanced by also consulting key informants who are likely to be aware of persons with epilepsy in the community, for example, teachers, health workers, religious leaders, and traditional healers. Methods relying on coded data as a screening tool may be enhanced by also searching records from neurology and epilepsy clinics and practices, electroencephalography (EEG) laboratories, or pharmacy prescription records. Collecting information from multiple sources of course requires measures to consolidate duplicate reports and to confirm diagnoses from less dependable sources—a challenging task.

Concealment of epilepsy due to negative attitudes about epilepsy can affect ascertainment in some communities. Additional problems affecting case ascertainment include cognitive impairment. Several problems in case ascertainment are compounded in resource poor communities. These may include poor recall of age of onset of seizures, poor recall of date of last seizure due to lack of use of calendars, limited detection of nonconvulsive seizures, and sometimes the absence of a clear terminology for epilepsy and seizures. Each of these factors may interfere with ascertainment of epilepsy.

It is essential to understand the significance of seizures and epilepsy in a specific cultural setting before an epidemiologic study is initiated. In particular, the language and concepts used to describe seizures have a profound influence on the reliability of questionnaires and screening methods to detect seizures and epilepsy, recognizing that many cultures do not have a single term that describes epilepsy, seizures generally, or seizure types.

Definitions of epilepsy and epileptic seizures

The International League Against Epilepsy (ILAE) has proposed both conceptual and operational definitions of epilepsy. In 2005 the following definition of epilepsy was proposed: “a disorder characterized by an enduring predisposition to generate epileptic seizures and by neurobiologic, cognitive, psychological and social consequences of this condition. The definition requires the occurrence of at least one epileptic seizure” (Fisher et al., 2005). This is a conceptual definition, intended primarily for clinicians who are diagnosing epilepsy. Epidemiologic researchers require operational case definitions based on the conceptual definition. For the purpose of conducting most population-based studies of epilepsy epidemiology, we advise that epilepsy be defined in practice as two or more unprovoked seizures occurring at least 24 h apart. This operational definition is unchanged from that adopted in 1993 by the ILAE (Commission on Epidemiology and Prognosis of the International League Against Epilepsy (1993). This approach has the additional advantage of permitting comparison of epidemiologic studies across time periods.

Evidence of recurrence may be the only information available to most epidemiologic studies with which to identify the presence of an “enduring predisposition to seizures.” Therefore, there may be no alternative to this operational definition for most epidemiologic studies. Exceptions may be considered if other very strong predictors of unprovoked seizure occurrence are identified as, for example, in studies of genetic etiologies of epilepsy.

The definition of epilepsy requires further definition of an epileptic seizure. Based on current ILAE definitions we propose that an epileptic seizure be defined in principle as “a transient occurrence of signs and/or symptoms due to abnormal excessive or synchronous neuronal activity in the brain” (Fisher et al., 2005). Operationally, these signs or symptoms include sudden and transitory abnormal phenomena such as alterations of consciousness, or involuntary motor, sensory, autonomic, or psychic events perceived by the patient or an observer (Commission on Epidemiology and Prognosis of the International League Against Epilepsy 1993). Active epilepsy indicates a person who is either currently being treated for epilepsy or whose most recent seizure has occurred within a time interval usually defined as the past 2 or 5 years (Commission on Epidemiology and Prognosis of the International League Against Epilepsy 1993), but the time should be specified. In some localities, active epilepsy may be defined over a 1-year period, due to problems in recalling dates beyond that period (Longe & Osuntokun, 1989; Birbeck & Kalichi, 2004; Mung’ala-Odera et al., 2004; Edwards et al., 2008).

Epileptic seizures, even if recurrent, are not always synonymous with epilepsy per se. Traditionally, and based on strong epidemiologic data supporting the distinctions, certain conditions under which seizures occur are generally not considered to be epilepsy proper, although they are still considered within a broader spectrum of seizure-related disorders: (Commission on Classification and Terminology of the International League Against Epilepsy 1989, Engel, 2006a)

There are exceptional clinical circumstances in which these criteria may not necessarily exclude a diagnosis of epilepsy. With increased understanding of the underlying genetic causes of some of these disorders, the criteria for exclusion may change in the future. However, at the current time and for the purposes of most epidemiologic studies, febrile seizures, neonatal seizures, solitary unprovoked seizures (or an isolated episode of status epilepticus), and provoked seizures should be segregated from epilepsy.

These definitions of epilepsy and epileptic seizures often must be translated into more practical criteria that can be used in epidemiologic studies of epilepsy, given the types and quality of data that may be available. In this process, further consideration must be given to the quality of evidence for a diagnosis of epilepsy: (a) clear evidence of recurrent epileptic seizures, with evidence that these are unprovoked by any acute medical condition or transient brain disorder; and (b) documentation of diagnosis by someone with appropriate specialized training in the recognition of epilepsy. Because the sources of information available for epidemiologic studies of epilepsy may fall short of either of these criteria, varying degrees of certainty can be recognized in a case definition. These categories are:

  • Definite, with primary documentation that meets criterion (a) or (b) above.
  • Probable, with other sources of information indicating the likelihood that criterion (a) or (b) is met.
  • Suspect, where primary or other sources of information suggest a possibility of epilepsy but neither criterion (a) nor (b) is met. The information provided is inadequate to confirm or refute the diagnosis of epilepsy.

This classification needs validation. In the Appendix of this document, operational case definitions for various sources of information (e.g., medical records, coded health care data, and interview data) are proposed with varying levels of probability.

Classifications of seizure type and syndrome

Epilepsy encompasses many different conditions with varying manifestations. Depending on the type and quality of information available for an epidemiologic study, there are likely to be limitations in the extent to which a study can classify cases of epilepsy with regard to seizure types and other characteristics.

In 2010, a revised terminology and concepts for classifying seizures and epilepsies was proposed by the Commission on Classification and Terminology of the ILAE (Berg et al., 2010). This revision recommended that the classification be grouped within categories of “electroclinical syndromes,”“constellations,”“epilepsies associated with structural or metabolic conditions,” and “epilepsies of unknown cause” (Berg et al., 2010). The classification of individual types within these categories is based on characteristics including “age at onset, cognitive and developmental antecedents and consequences, motor and sensory examinations, EEG features, provoking or triggering factors, and patterns of seizure occurrence with respect to sleep” (Berg et al., 2010). If highly detailed information from epilepsy specialists is consistently available in an epidemiologic study, the ILAE classifications of seizures and epilepsy syndromes may be considered (Commission on Classification and Terminology of the International League Against Epilepsy 1981, 1989; Engel, 2006b; Berg et al., 2010).

Although accurate identification of the type of epilepsy is important for treatment and management, epidemiologic studies may not achieve such precision because: (1) investigators lack access to clinical records, (2) investigators lack the knowledge and training to make specific epilepsy diagnoses, or (3) patients lack access to the level of specialty care needed for specific diagnosis. In the absence of detailed, consistently collected classification information, an alternative is to collect more basic information elements. At a minimum, these include the age at onset, the typical manifestations (seizure types), and the underlying cause (etiology) as outlined below.

Age at onset

Defined as the age of occurrence of the first unprovoked seizure, age at onset is especially important to studies of epilepsy, as many specific syndromes have a characteristic age at onset useful to their recognition. The accuracy of pinpointing the age of onset varies by the patient’s seizure types, as some types (e.g., absence, myoclonic, complex partial) may recur over prolonged periods before being diagnosed as seizures (Jallon et al., 2001).

Seizure types

The ILAE seizure classifications (Commission on Classification and Terminology of the International League Against Epilepsy 1981, 1989; Engel, 2006b; Berg et al., 2010) require clinical expertise and evaluations that may not be available to people with epilepsy in some localities. Furthermore, if done, the results of such evaluations are often not available to epidemiologic researchers. Under these circumstances, the only available information may be descriptions from the person with epilepsy or from a witness. The seizure classification must then be simplified, based on limited characteristics of the ictus such as the presence or absence of:

  •  focal sensory or motor symptoms or signs,
  •  major convulsive (e.g., generalized tonic–clonic) activity,
  •  other types of motor activity, and
  •  impaired responsiveness or consciousness during the seizure.

To enable comparisons among past and future studies it is important that this simplified classification relate as closely as possible to more detailed categories and concepts of current ILAE classifications and terminologies (Berg et al., 2010; Blume et al., 2001; Commission on Classification and Terminology of the International League Against Epilepsy 1981; Engel, 2006b; International League Against Epilepsy (ILAE) (2008). To this end, we suggest that most seizures can be classified within the following matrix (Table 1). The matrix classifies seizures both according to onset (generalized or focal) and predominant ictal features (motor vs. nonmotor), and its categories allow for varying levels of available information and certainty regarding seizure type. EEG is not necessarily required for this classification—recognizing that it is often unavailable in epidemiological studies—although it can be critical in correct diagnosis and may be an important indicator of underlying focal abnormalities, particularly in low and middle income countries (Munyoki et al., 2010).

Table 1.   Simplified clinical classification of seizure type
  1. aSeizure onset is manifested by generalized tonic and/or clonic (convulsive) motor activity and unconsciousness. Focal features may occur.

  2. bSeizure onset has focal manifestations that evolve to generalized convulsive activity.

  3. cFocal or generalized nature of seizure onset is undetermined, but seizures manifest generalized convulsive activity.

  4. dInclude myoclonic seizures, eyelid myoclonus, epileptic spasms, atonic seizures, other, and unspecified generalized motor seizures with or without impairment of consciousness.

  5. eSeizure has focal manifestations (including myoclonic, inhibitory, Jacksonian march, focal asymmetric tonic, hemiclonic, hyperkinetic, and other focal motor seizures) that do not evolve to generalized convulsive activity.

  6. fUnspecified motor seizures; includes neonatal and other seizures.

  7. gStaring spells, unresponsiveness, or other alteration of consciousness.

  8. hIncludes typical and atypical absence seizures.

  9. iFocal seizure associated with impairment of consciousness (formerly termed “complex partial”) without secondary generalization (Commission on Classification and Terminology of the International League Against Epilepsy, 1989).

  10. jSeizure manifested by transient decreased responsiveness or “staring,” undetermined if absence or dyscognitive (“complex partial”) in type.

  11. kIncludes auras without alteration of consciousness or secondary generalization (including somatosensory and experiential seizures), autonomic, and other nonmotor seizures.

Predominantly motor
 ConvulsiveGeneralized convulsiveaFocal onset with secondary generalizationbConvulsive undeterminedc
 Other motorGeneralized other motordFocal motoreOther motor undeterminedf
Predominantly nonmotor
 Impaired responsivenessgGeneralized absencehDyscognitive focal seizures (formerly complex partial)iImpaired responsiveness, undeterminedj
 Other nonmotorNASensory, psychic, and other, including autonomickNA
 Generalized seizure, unspecifiedFocal seizure, unspecifiedSeizure, unspecified
Etiology (proximate cause)

New terminology and concepts advanced by the ILAE divide the causes of epilepsy into three broad categories: genetic, structural/metabolic, and unknown (Berg et al., 2010). Although these categories correspond roughly to the former categories of idiopathic, symptomatic, and cryptogenic, there are important conceptual distinctions.

  •  The cause of epilepsy is considered genetic when epilepsy is a direct result—and the core manifestation—of a known or presumed genetic defect.
  •  The cause is considered structural/metabolic when a structural lesion (either static or progressive) or metabolic condition (e.g., inborn errors of metabolism) is present and is known to be associated with an increased risk of epilepsy. When such a lesion or condition arises due to a genetic defect, that is, when “there is a separate disorder interposed between the genetic defect and the epilepsy,” the cause of epilepsy should be classified as structural/metabolic. Therefore, the cause is attributed to the condition that is most directly linked and proximate to the development of epilepsy.
  •  If the nature of the cause is not known, then such cause is classified as unknown.

Within each of these three broad categories are many heterogeneous specific causes. We suggest that the subcategories in Table 2 may provide a useful basis for epidemiologic studies of epilepsy etiologies. Previous epidemiologic studies have classified etiology in the broad categories of idiopathic/cryptogenic and symptomatic, with separate subcategories assigned to the latter. Under the new terminology, an analogous categorization that would permit comparisons with previous studies combines genetic, presumed genetic, and unknown and separately classifies structural/metabolic causes. Although no single classification scheme of causes will satisfy all the purposes among epidemiologic studies, this broadly organized structure may provide flexibility to allow modifications (and more detail) according to the focus of the study as well as to accommodate advancements in understanding causes. The differentiation between these categories will depend upon the available facilities [e.g., EEG and magnetic resonance imaging (MRI) may be available, but not access to selected metabolic tests and genotyping]. Validation of this categorization is needed.

Table 2.   Classification of epilepsy causes
Direct etiology
Genetic/Presumed geneticStructural/MetabolicUnknown or Undetermined
  1. aWithout known etiology despite adequate evaluation (e.g., history, examination, EEG, and other testing determined to be relevant such as neuroimaging or genetic testing).

  2. bWhere evidence is lacking that the structural pathology precedes the onset of epilepsy, it is not assumed that such pathology causes epilepsy.

  3. cWith epilepsy as a late effect. For distinction with acute symptomatic seizures, see page 6.

  4. dIncludes conditions where underlying etiology is undocumented or available information is limited to terms such as “intellectual disability”/”mental retardation” or “cerebral palsy” when these preceded the onset of seizures.

  5. eWithout adequate evaluation to determine etiology as defined by investigators.

Specific genetic epilepsy syndromesInfectionsEpilepsy of unknowna etiology
Genetic and chromosomal developmental encephalopathiesaTraumatic brain injuryEpilepsy of undeterminede etiology
Mesial temporal sclerosisb
Degenerative neurologic diseases
Metabolic or toxic insults to brainc
Perinatal insults
 Intraventricular hemorrhage
 Hypoxic–ischemic encephalopathy
Malformations of cortical or other brain development
Neurocutaneous syndromes
Inborn errors of metabolism


In population studies, comorbidity—the co-occurrence of two or more separate medical conditions in the same individual—is of greatest interest when it appears at above-chance levels, that is, when it is associated. Several comorbid conditions associated with epilepsy have been described (Sunder, 1997; Hermann et al., 2000; Boro & Haut, 2003; Trinka, 2003; Gaitatzis et al., 2004; McLachlan, 2006; De Simone et al., 2007) and are listed in Appendix, Table A2. Associations between epilepsy and other medical and psychiatric conditions may exist because epilepsy and the comorbid condition share an underlying etiology or because epilepsy or epilepsy treatment lead to a higher than expected occurrence of the comorbid condition. It is also possible that epilepsy comorbidities present at the time of the incident diagnosis may influence prognosis of epilepsy.

Although not validated in specific studies of epilepsy, comorbidity indices (Linn et al., 1968; Charlson et al., 1987; Deyo et al., 1992; Miller et al., 1992; Dodds et al., 1993; Elixhauser et al., 1998; Selim et al., 2004) can be used to assess the impact of comorbid conditions on the prognosis of epilepsy. Such indices may assess the number, type, and severity of other conditions, regardless of whether they are associated with epilepsy. These indices are most often coded according to information from medical records and have been most often used in studies of mortality (Kwon et al., 2011). Other assessments of comorbidity burden have relied on patient self-report, and several studies have shown that these questionnaires are reliable and valid in adult populations (Commission on Classification and Terminology of the International League Against Epilepsy, 1989; Harlow & Linet, 1989; Linet et al., 1989; Katz et al., 1996; Bergmann et al., 1998a,b; Selim et al., 2004). In some circumstances (e.g., psychiatric disorders, migraine), it may be useful to use validated interviews.

Epilepsy care

Access to care

To describe fully the determinants of access to care it is useful to employ a behavioral model of health service use that identifies predisposing factors (including demographics and health beliefs), enabling factors (such as income and health care availability), and need factors (such as disease severity) (Andersen, 1995). Based on the behavioral framework, access to care is measured by examining the association between patterns of healthcare use and predisposing and enabling factors while controlling for comparable need. At a more basic level, a number of studies, particularly in low and middle income countries (LMIC), address access to care with regard to the treatment gap, the difference between the number of people with active epilepsy and the number being appropriately treated (Meinardi, et al. 2001; Mbuba et al., 2008). The treatment gap measures alone do not provide information about underlying factors that limit access to care (Table 3). The treatment gap has traditionally referred to the use and provision of antiepileptic drugs (AEDs) where use of AEDs is usually based upon self-reporting. This may have low sensitivity and specificity when compared to the detection of AED levels in the blood (Edwards et al., 2008).

Table 3.   Potential causes of the treatment gap for epilepsy
Primary (diagnostic gap)Secondary (therapeutic gap)
Lack of adequate paraclinical services (Diop et al., 2003)Lack of treatment availability (Odermatt et al., 2007)
Lack of qualified medical personnel (Diop et al., 2003)Inability to pay for drugs (Mac et al., 2007)
Lack of access to health care (distance or cost) (Tran et al., 2008)Low quality of drugs (Mac et al., 2008)
Error in diagnosis (Kanner, 2008)Error in treatment prescribed (Feely, 1999); Rejection of treatment by patient (Broadley, 2004)
Patient and family rejection of diagnosis due to stigma (Muela Ribera et al., 2009; Rafael et al., 2010)Cultural beliefs (Meinardi, et al. 2001)
Patient misconceptions about the nature of condition (Shibre et al., 2009)Uninformed choices or poor understanding of the nature of treatment by patient or family (Stores, 1987; Jacoby, 2002)
 Poor compliance of alternative medicine (Tandon et al., 2002; Ricotti & Delanty, 2006)

Some studies of access to care among people with epilepsy have been conducted in populations of developed countries, emphasizing access to specialty care (Bhatt et al., 2005; Burneo et al., 2005a; Gaitatzis et al., 2002). Patterns of health service use and associated costs can be addressed using both patient-based and population-based approaches (Halpern et al., 2000). Direct population surveys (or representative samples)—including standardized questions regarding type and quantity of treatments—are preferred methods of measuring treatment gaps. Indirect methods can, however, be employed when data are scarce, for example, comparing estimates of epilepsy prevalence from other sources to the number of people known (or estimated) to be in treatment. Such estimates are subject to numerous biases and should only be used as broad indicators of access to care.

Cost studies

Cost-of-illness studies are important to determine the burden of epilepsy on individuals and society (Begley & Beghi, 2002). Epilepsy-specific costs are defined as the economic value of services consumed in the prevention, treatment, or rehabilitation of people with the disorder (direct costs), and the estimated economic value of work and leisure activity lost from associated morbidity and mortality (indirect costs). The cost of services is usually estimated according to the average payment made to service providers. Indirect costs are typically estimated in terms of lost earnings and the imputed value of lost household work associated with morbidity and mortality. Two broad approaches (the bottom-up and the top-down) are used in generating estimates of the direct costs of epilepsy. In the former, cost estimates of health care, social services, and family member resources used by patients with epilepsy are derived from studies (Gaitatzis et al., 2002) of real cases or, alternatively, based on hypothetical information from expert panels and related literature sources. In the latter, estimates are made of the cost of services for all illnesses (defined by disease-specific codes) based on nationally representative provider surveys, and a portion is attributed to epilepsy. In general, the top-down approach cannot be used in resource-poor countries due to inadequate information and differing measures of cost.

Cost-effectiveness studies involve the comparative assessment of alternative courses of action in terms of their relative cost and effectiveness (Heaney & Begley, 2002). A cost-effectiveness ratio is calculated for each treatment or service being assessed, where the denominator reflects the incremental gain in health (e.g., reduction in seizure frequency or severity) and the numerator reflects the additional cost of achieving that health gain. When the gain in health is expressed in terms of health-related quality of life, such as Quality Adjusted Life Years (QALYs) or Healthy Year Equivalents (HYEs), the economic assessment study is described as cost-utility analysis (Langfitt et al., 2006).

In cost-benefit analyses the denominator (e.g., resultant health state) is expressed in monetary terms, and interventions are recommended if the monetary value of the resultant health state exceeds that of the intervention. Because these analyses entail substantial complexity and potential ethical concerns, they are performed infrequently (Johannesson & Jonsson, 1991). An alternative form of economic evaluation is cost-minimization analysis in which assumptions are made that treatments produce identical health benefits. The aim here is to determine the relative cost of each treatment (Heaney & Begley, 2002).

Because precise data on resource utilization and treatment effects are generally lacking, most studies are conducted for hypothetical cohorts of patients based on data derived from clinical trials and expert panels (Heaney & Begley, 2002).


Disparities can occur in epilepsy incidence, prevalence, access to care, treatment choices and response, complications, and comorbid conditions (Gakidou et al., 2000; Braveman, 2006). The topic is of concern to advocates and policy makers seeking to eliminate inequalities and improve health in high-risk populations. Disparities may exist in several dimensions: socioeconomic status (SES); education; race or ethnicity; geography; physical and social environment; and others. These factors, particularly SES, may covary with the other factors. Some categories, for example, race or ethnicity, represent social constructs that are difficult or impossible to measure objectively or precisely.

Socioeconomic disparities in epilepsy

The concept of socioeconomic status is complex, relating to material and social assets and resources as well as social prestige (Krieger et al., 1997). SES can be assessed for individuals, households, or communities (Krieger et al., 1997). Several standardized indices of SES that incorporate elements such as educational attainment, occupation, and income have been employed and validated in developed countries (Duncan, 1961; Hollingshead, 1975; Blishen et al., 1987; Nakao & Treas, 1992; Cirino et al., 2002). Simpler scales emphasizing resource access and item ownership may be of greater use in low and middle income countries (Patel et al., 2007). Where studies of epilepsy obtain information from subject interviews, standardized SES assessment tools may be considered; however, many may be too lengthy to be practical. Alternatives in developed countries are to estimate SES categories employing community-based methods, such as the use of census tract or postal codes, when the average SES of people in the district is known, recognizing that some misclassification of individuals and households is likely with these methods (Krieger et al., 2002). Briefer individualized methods for assigning SES categories include using limited information about occupation, education, income, or property ownership to the extent that these can be obtained from medical charts or from interviews. The relationship of SES to epilepsy incidence, prevalence, healthcare use, and outcomes deserves further study in both developed and LMIC; among the latter, average SES may be considerably lower, but within-population variation greater.

Ethnic/racial disparities

A number of studies have identified disparities in epilepsy occurrence, healthcare use, and outcomes among categories of race or ethnicity (Szaflarski et al., 2006a). Assigning study subjects to racial and ethnic categories confronts ambiguities, especially with frequent intergroup marriage and affiliation. In the absence of objective methods to assign categories, studies may use self-designations, rely on designations recorded in medical records, or even impute values based on family name, language, or geographic information (Fiscella & Fremont, 2006). Limitations of such methods and their potential for misclassification must be recognized, and these may be superseded by more detailed genetic studies.

Other dimensions and manifestations of disparity

Age, sex, size of community, cultural beliefs and practices, religion, type of health care system, and variations in health knowledge and attitudes toward epilepsy are other factors that may influence disparities in epilepsy occurrence, care, or outcome and, therefore, merit study.

Numerous studies have identified disparities in epilepsy in North America (Szaflarski, et al. 2006b; Theodore et al., 2006; Burneo, et al. 2009) and elsewhere. With some, there is a large potential for confounding among the factors they examine. In order to isolate the underlying factors of most importance, great care is required in the selection of diverse study populations for comparison, the collection of data, and their analysis.

Epilepsy severity and outcomes

The overall impact of epilepsy on health and well being depends on many factors, beginning with the types of seizures experienced, seizure frequency, underlying neurologic conditions, other comorbidities, treatment side effects, and the extent to which these are mitigated or resolved with medical treatment over time. Clinical parameters are not sufficient to evaluate the effect of epilepsy on the health of the single patient. Other factors also contribute to the outcomes of epilepsy: in particular, whether familial and cultural attitudes toward epilepsy are stigmatizing or supportive, and whether families and society provide educational, vocational, and other resources that enable people with epilepsy to circumvent potential disabilities they might encounter. It is of value for epidemiologic studies of epilepsy to address these factors, to collect data regarding seizure type and frequency, care received, and basic indicators of outcome such as social integration, educational and vocational attainment, employment status, and perceived quality of life.


Standardized measures of seizure severity have been developed for clinical trials (Table 4). These, in general, take account of variables such as the type, duration, and frequency of seizures; other ictal phenomena such as falls; and duration of postictal recovery. They require the collection of clinically detailed information to a degree that may be impractical for many population-based epidemiologic studies. In emphasizing the characteristics of seizures, they do not measure the total impact of epilepsy. A full assessment of the severity of epilepsy may require items based on patient perception of seizure severity, since observer evaluation does not capture the physical and emotional burden experienced by the patient (Baker et al., 1998a). However, the inclusion of a subjective evaluation may decrease the reliability of the scale. To varying degrees, both kinds of seizure severity scales have problems regarding ease of use, sensitivity to changes, and subjectivity of assessments, and no single scale is accepted generally as a standard. No scale has been developed for low and middle income countries where other factors such as physical consequences (e.g., burns) are important. Their characteristics are compared in detail elsewhere (Cramer & French, 2001).

Table 4.   Examples of epilepsy-related severity assessment instruments
 No. items/(dimensions)References
Seizure severity measures
 Seizure Frequency Scoring System(5)Engel et al. (1993)
 VA Seizure Frequency and Severity Scale6–21Cramer et al. (1983)
 National Hospital (Chalfont) Seizure Severity Scale8O’Donoghue et al. (1996)
 Occupational Hazard Scale(6)Janz (1989)
 Liverpool Seizure Severity Scale16, 20Baker et al. (1991, 1998b)
 Hague Seizure Severity Scale (children)13Carpay et al. (1997)
Syndrome Severity Measures
 Syndrome Severity ScaleDunn et al. (2004)
Epilepsy Severity Measures
 Global Severity of Epilepsy Scale (children)1Speechley et al. (2008)

A classification of epilepsy syndrome severity for children has also been proposed (Dunn et al., 2004). This classification is based on the diagnosis of syndrome and does not take into account individual differences in seizure manifestations (e.g., type and frequency) among persons with a given syndrome. Its usefulness is limited by the availability of complete and accurate syndrome diagnosis information. Finally, a simple, physician-rated, single-item scale to assess global severity of pediatric epilepsy severity has demonstrated reliability and validity (Speechley et al., 2008) and is being used in multicenter studies of quality of life in pediatric epilepsy.

Quality of life

Health-related quality of life (HRQoL) can be regarded as the broadest and most important outcome of any chronic health condition. The relationship between chronic health conditions, HRQoL, and disability is interrelated and complex, and can sometimes pose measurement challenges (Krahn et al., 2009). Comprehensive measures of HRQoL take into account social, vocational, cognitive, and mood states, in addition to overall perceived physical and mental health.

Some also address processes of health care or quality of care (Donabedian, 2005). Preference-based or “utility-based” measures of HRQoL are used by some to assess clinical outcomes and for cost-utility studies [e.g., EuroQOL (EQ)-5D] (The EuroQol Group 1990).

Assessments of HRQoL can be generic, intended for use in general populations, and not restricted to assessing outcomes of a specific disease or disorder. The use of generic instruments in studies of epilepsy allows comparisons with other conditions and with the general population (Langfitt et al., 2006). Other HRQoL measures are disease-specific, designed to address the physical, psychological, and social aspects of a particular condition. Recently, recommendations to avoid conflating function and HRQoL were provided (Krahn et al., 2009). For epilepsy, various measurement instruments have been developed or adapted for which reliability, validity, and sensitivity to change have been demonstrated (Table 5) (Leone et al., 2005). These epilepsy-specific measures can provide insight on the impact of epilepsy care and may be more likely to identify variations in outcomes of epilepsy care.

Table 5.   Examples of validated health-related quality of life assessment instruments
 No. itemsReferences
Generic measures
 SF-3636Brazier et al. (1992), Ware & Sherbourne (1992), Weinberger et al. (1991)
 SF-1212Jenkinson & Layte (1997), Kazis et al. (2006), Ware et al. (1996)
 HRQOL-1414Centers for Disease Control and Prevention (2010); Moriarty et al. (2003)
 Healthy Days/HRQOL-44Centers for Disease Control and Prevention (2007); Moriarty et al. (2003)
 Satisfaction with Life Scale (SWLS)5Diener (1984), Larsen et al. (1985), Pavot et al. (1991)
Epilepsy-specific measures
 Liverpool Batteries Jacoby et al. (1992)
 Quality of Life in Epilepsy Questionnaires
  QOLIE-8989Devinsky et al. (1995)
  QOLIE-3131Cramer et al. (1998)
  QOLIE-1010Cramer et al. (2000)
  QOLIE-AD-48 (adolescents)48Cramer et al. (1999)
  QOLCE (children)91Sabaz et al. (2000)

Important considerations in selecting measures of HRQoL for use in surveys of epilepsy include:

  •  The appropriateness of the instrument to local language and cultures. Most of these scales have been validated primarily in a few developed countries. Their appropriateness in other regions and the comparability of their results across cultures has not been fully assessed. At the same time, however, changes in such scales to adapt them to new cultures are best minimized in order to maintain their equivalence across cultures.
  •  The feasibility of implementing the assessment in a survey or epidemiologic study is determined in part by its length and mode of administration. Most instruments addressing quality of life require extensive information that is not routinely documented in medical records; thus, they may not be suited for many population-based studies of epilepsy. Lengthy HRQoL instruments that are self-administered may be cognitively burdensome for people with epilepsy, thereby limiting the quality of the data.
  •  Validation of the instrument with respect to its reliability (consistency of results internally and with repeated administration), construct validity, and criterion validity.
  •  Sensitivity of the instrument’s measures of effect to changes in treatment and other determinant circumstances.

In countries with sufficient technical means, HRQoL may be more efficiently assessed through new resources available from the Patient Reported Outcome Measurement Information System (PROMIS) supported by the U.S. National Institutes of Health (NIH) (National Institutes of Health, 2010a,b). By integrating item-response theory and computer adaptive testing, PROMIS has developed on-line resources at no cost for researchers to use to measure patient-reported symptoms and HRQoL across a wide variety of chronic diseases and conditions. A distinct, but closely related NIH project, Neuro-QOL (National Institutes of Health, 2010c), will validate item banks to measure quality of life specifically in adult and pediatric populations with various neurologic diseases, including epilepsy.

Other specific outcomes related to quality of life

It should be noted that seizure frequency alone, an important element of measuring severity, is also an important predictor of HRQoL. In general, those whose seizures fully remit (with or without ongoing treatment) show little reduction in HRQoL, whereas those with refractory epilepsy demonstrate substantial decrements in quality-of-life scores (Devinsky et al., 1999; O’Donoghue et al., 1999; Wiebe et al., 1999; Leidy et al., 2001; Strine et al., 2005; Kobau et al., 2007).

Other important assessments related to quality of life that may be considered in population-based studies of epilepsy address specific psychosocial dimensions such as depression, anxiety, perceived stigma, cultural beliefs and attitudes about epilepsy, as well as positive aspects of functioning (e.g., positive affect and satisfaction with life domains).

Terminal remission

Epilepsy, characterized by recurrent seizures, may continue for a limited time in some individuals, whereas it can be a lifetime condition in others. Terminal remission of epilepsy is defined as a period of 2 or 5 years of seizure freedom off AEDs (Hauser & Kurland, 1975).

Epilepsy burden on the families

The burden of epilepsy reaches well beyond those with the condition, affecting other persons of close relationship defined by family, household, or friendship, as well as broader relationships in the community. The effects are strongest on the family or household and will vary depending on:

  •  family composition and structure and the role of the person with epilepsy (e.g., parent or caregiver, minor child, or adult child),
  •  the severity of epilepsy and presence and severity of comorbid conditions,
  •  the family’s socioeconomic and health insurance status,
  •  the family’s and community’s perceptions of epilepsy, and relevant social norms and expectations, and
  •  the support systems available.

Parents with epilepsy may have restricted opportunities for gainful employment outside the home, or restricted abilities to perform work as homemakers and primary caregivers for their children. In general, these limitations may disproportionately affect women with epilepsy. Parents of children with epilepsy may be taxed by additional caregiver burdens, especially if seizures are intractable or accompanied by cognitive, behavioral, or developmental disabilities. Siblings of children with epilepsy may be affected by relatively reduced parental attention or by needing to assume some caregiving responsibilities themselves. Families with members who have epilepsy may be isolated from the neighborhood and community because of stigma and discrimination due to culture and folk traditions. Stigma can be measured with the Epilepsy Stigma Scale (Baker et al., 2000; Baker, 2002).

Although the overall impact of epilepsy on the family can be profound, there are currently no specific instruments that measure this burden. The Impact of Pediatric Epilepsy Scale (IPES) and structured questions about the burden on siblings have been used to evaluate the influence of epilepsy on several aspects of family functioning (Mims, 1997; Camfield et al., 2001; Tsuchie et al., 2006). The quality of life and psychosocial dimensions (e.g., depression, anxiety, stigma, etc.) of family members, especially care providers, can also reflect the epilepsy burden on the families. Further research is needed in this area.

Analysis in population studies

Just as consistent definitions, variables, and data collection methods are important to enable comparisons among studies, so are consistent analyses and reported measures important. The common standard measures of the frequency of epilepsy in a population are point prevalence, incidence, and risk:

Point prevalence is the proportion of individuals in the population who are affected by a health condition at a single point in time, usually designated as a specific day. With epilepsy, the point prevalence of active epilepsy is typically considered of most interest.

A case of active epilepsy indicates a person who is either currently being treated for epilepsy or whose most recent seizure has occurred (usually) within the past 2–5 years (Commission on Epidemiology and Prognosis of the International League Against Epilepsy, 1993), although the past year has been used in low and middle income countries (Longe & Osuntokun, 1989; Birbeck & Kalichi, 2004; Mung’ala-Odera et al., 2004; Edwards et al., 2008). Variations on this definition of active epilepsy can be used for the purposes of some studies lacking long-term retrospective data on seizure occurrence; however, to enhance comparability across studies, variations in this definition should be minimized, and the time period must be stated in the report.

Point prevalence is useful for indicating the degree of disease burden in a population. Prevalence studies are especially relevant when assessing healthcare and other service needs. They are less useful for etiologic investigations.

The incidence rate is the rate with which new cases occur in a population. It is expressed as a frequency per standard population (e.g., 100,000) per time period (usually per year). Incidence rates are informative about the rates of new cases regardless of the prognosis or cause of the disorder. For etiologic and prognostic investigations, incidence-based studies are superior to prevalence-based studies.

New onset versus newly diagnosed epilepsy incidence

In practice, incidence studies are often limited to studying the incidence of new diagnosis, since many persons with newly diagnosed epilepsy have had a long, uncertain, or undocumented period of seizure occurrence preceding their diagnosis. The time from onset to diagnosis varies substantially with the type of epilepsy and its predominant seizure type (Anonymous, 2000).

The magnitude of the incidence of new onset epilepsy and the incidence of newly diagnosed epilepsy will differ, because these measures have different numerators. For new onset epilepsy, the numerator includes people identified at their second unprovoked seizure. In contrast, the numerator for newly diagnosed epilepsy includes both new onset epilepsy and people with more than two unprovoked seizures who are first diagnosed with epilepsy during the study period. The incidence of new onset epilepsy has never been compared to that for newly diagnosed epilepsy, but it is expected that the incidence of new onset epilepsy will be less than that for newly diagnosed epilepsy.

A variant expression of incidence is incidence density, which uses a different denominator, population-time (person-time), to describe the accrual of events (new epilepsy diagnosis) over an interval of time when population is observed under the assumption of stability. The advantage of incidence density is it can easily be converted into risk and is more suitable for rare events such as epilepsy when numerator is stabilized as an accrual of cases over a specified time interval. Furthermore, it is the best approximation of cumulative incidence for rare diseases (Morgenstern et al., 1980).

The risk of epilepsy is the probability that a person will develop the disorder. It represents the cumulative effect of incidence over a longer period of time or span of age, for example, a 5-year risk, the cumulative risk in childhood, or a lifetime risk. The cumulative risk over the lifetime is sometimes called “lifetime prevalence.” In practice, the last measure can seldom be determined without relying on retrospective data sources and is thus more prone to error.

Using administrative health datasets to estimate incidence or prevalence

Linked hospital, emergency department, clinic, and physician office data that describe all medical encounters for all individuals in a population may be used to estimate epilepsy incidence or prevalence. The specificity and positive predictive value of diagnostic codes for epilepsy and seizures must be considered (see Appendix). When individuals have multiple medical encounters described by epilepsy or seizure codes, the likelihood of identifying a true case of epilepsy is high; however, these codes do not distinguish new-onset from long established cases. Because some people with established epilepsy seek follow-up medical care infrequently, for example, less than once yearly, the first appearance of a code for epilepsy with an individual enrolled in a dataset cannot be assumed to represent a new diagnosis of epilepsy unless there is a long preceding period of enrollment with no record of seizure or epilepsy. Ideally for determination of incidence, there should be no prior code for epilepsy in the administrative data, particularly when individuals are represented over all or most of the lifetime. However, in other situations, this may be impractical. When administrative data do not cover all or most of the lifetime, the minimum length of enrollment before the first epilepsy-related code that may represent a new epilepsy diagnosis is uncertain: 1 or 2 years appears to be inadequate, and at least 3 (or preferably more) years are advisable. For prevalence studies, similar periods of follow-up appear necessary to ascertain nearly all existing cases.


Mortality rates are higher in people with most types of epilepsy, and this may be attributed both to consequences of seizure occurrence as well as to direct effects of some underlying diseases that give rise to epilepsy (Gaitatzis & Sander, 2004). Rates of sudden unexpected death in people with epilepsy (SUDEP) are also greater than rates of sudden unexpected death in the general population (Tomson et al., 2005). Determination of the cause of death may be useful in identifying measures to prevent deaths in this population.

Different measures are used to estimate mortality, depending upon the study design and available information on deaths. The measures include mortality rate, case fatality, standardized morality ratio, and proportionate mortality. A standardized mortality ratio (SMR) is the most common measure used to compare rates of death between a population of people with epilepsy and a referent population. The magnitude of one SMR cannot be compared to the magnitude of another SMR, because the SMR is calculated using indirect standardization. As a result, the referent for each SMR has a different age distribution that alters the expected deaths. Mortality rates can also be used when information on the number of people in the general population is available. The proportion of deaths in the epilepsy population can be reported using case fatality. Sometimes only death data that list cause of death are available. In this case, it is possible to calculate the proportionate mortality, which is the ratio of the number of deaths due to a specific cause in a population to the total number of deaths in the same period (Logroscino & Hesdorffer, 2005) describes the proportion of deaths in a community that are due to epilepsy compared to deaths due to other causes.

Evaluation of measurement validity

The validity of basic measures of incidence, prevalence, and mortality is described by these proportions:


The proportion of cases of disease detected by the case ascertainment method. Sensitivity is low if many true cases of disease are unreported (or misclassified as noncases), resulting in too-low estimates of disease occurrence.


The proportion of the study population without disease that is correctly classified by the case ascertainment method represents its specificity. It is reduced if true noncases of disease are misclassified as cases (false positives), which can inflate estimates of disease occurrence.

Positive predictive value (PPV)

The proportion of ascertained cases that actually have the disease. As with specificity, PPV is also reduced if true noncases of disease are misclassified as cases, leading to excessive estimates of occurrence. The PPV varies directly with the prevalence of disease in the population under study, that is, diminishing as the prevalence of the condition decreases. For conditions of lesser prevalence, however, a high PPV provides greater assurance of accuracy than a specificity measurement of the same value.

Unlike PPV, sensitivity and specificity are conditional probabilities predicated on the availability of a definitive test that should be predetermined and fixed as antecedents. In other words, unless otherwise the marginal totals corresponding to the definitive test, “gold standard,” are fixed ahead of time, one can find a desirable level of sensitivity and specificity by manipulating the cell values (Fleiss et al., 2003).

Measures of association

Measures of frequency are typically compared to each other to produce measures of association or measures of effect. Associations are typically quantified as differences or ratios.

Absolute (difference) measures are calculated as differences in prevalence, risk, or rate between one group and another (e.g., men vs. women, one country vs. another, persons exposed to a risk factor of interest vs. persons not exposed). Absolute difference measures express the effect of a risk factor in terms of the actual proportion of people affected. This proportion, when applied to a specific population, can be used to estimate the number of people affected.

Relative (ratio) measures are calculated as the rate (risk, prevalence) in one group divided by that in a referent group. This is a preferred method for portraying the impact of a factor on the individual and quantifying risk factor–disease associations in etiologic research, because it provides information on the degree to which the risk of disease is increased in groups with a factor of interest compared to those without the factor.

It is often helpful to use absolute and relative measures of association together, as they provide somewhat different and complementary information. Relative measures, in particular, have to be interpreted cautiously as they are heavily dependent on the frequency of disease in the referent population (incidence rate in the reference population). The same absolute difference can result in very different relative differences depending on the frequency in the referent group. Point estimates of measures of associations are an index of magnitude of effect and are the best measure for studying causality in epidemiologic research under biologic hypotheses (Rothman et al., 2008). It is important to underline that in population-based studies is not unusual to find small effects with values of relative risk or odds ratio between 1 and 1.5, whereas in clinical studies larger effects are found more frequently.

Measures of impact

These are based on absolute and relative measures of association and provide estimates of the amount of disease (absolute or relative) that may be caused by a particular factor.

The population attributable risk is the number of new cases in a defined period that are due to (attributable to) a particular causative factor. The population attributable risk (percent or fraction) is the reduction in the incidence of disease that would be expected in a population if a specific factor presumed to be causal is removed from the population.

These approaches are extremely useful for putting in perspective the value of a specific prevention program as it quantifies the maximum impact such a program might have. For example, one can ask questions such as, how many cases of epilepsy could be prevented in a specific country if effective measures were put into place to eradicate malaria? Or “What proportion of epilepsy is due to endemic neurocysticercosis (Medina et al., 2005)?”

Related calculations can yield estimates such as the attributable risk and attributable risk percent among the exposed. These are more relevant to making a statement about the likelihood that a given factor caused a specific individual’s illness.

Formulae for the calculation of these measures and related statistics can be found in standard textbooks of epidemiology (Rothman et al., 2008), neuroepidemiology (Nelson et al., 2004), and biostatistics (Fleiss et al., 2003).


  1. Top of page
  2. Summary
  3. Epidemiologic Methods for the Study of Epilepsy
  4. Conclusion
  5. Disclosures
  6. References
  7. Appendices

This document proposes definitions, methods, and resources for investigators of the epidemiology of epilepsy that are intended to promote future population-based studies of high validity and to improve the comparability of such studies. Although the emphasis of this article is on population-based studies, they are useful also in clinical epidemiologic studies of epilepsy. This guidance is further intended to be flexible, allowing for differences of purpose among particular studies, as well as variations in the available methods and information resources on which they depend.


  1. Top of page
  2. Summary
  3. Epidemiologic Methods for the Study of Epilepsy
  4. Conclusion
  5. Disclosures
  6. References
  7. Appendices

The authors declare the following associations as potential conflicts of interest: David J. Thurman – None; Ettore Beghi – Honoraria/research grants from Kedrion Pharma Company, Eisai, Sanofi-Aventis, UCB Pharma, GlaxoSmithKline; Charles E. Begley – None; Anne T. Berg – Travel funding/honoraria/consulting fees from Eisai, UCB, Dow Agro Science; Jeffrey R. Buchhalter – Research grants from Ovation Pharmaceuticals, Pfizer; Ding Ding – None; Dale C. Hesdorffer – Honoraria/travel funds/stock Pfizer, GlaxoSmithKline, General Electric; W. Allen Hauser – Consultant Neuropace; Lewis Kazis – Research grants Amgen Inc., Genzyme, Eli Lilly and Company, Bristol Meyers Squibb, Sanofi-Aventis, Boerhringer-Ingelheim, and Astra Zeneca; Rosemarie Kobau – None; Barbara Kroner – None; David Labiner – Consultant/Speaker/Research grants from Cyberonics, Esai and Ortho McNeil; Kore Liow – Honoraria from UCB Pharma; Giancarlo Logroscino – None; Marco T. Medina – None; Charles R. Newton – None; Karen Parko – None; Angelia Paschal – None; Pierre-Marie Preux – None; Josemir W. Sander – Honoraria/grants/travel grants from UCB, Janssen-Gilag, Eisai, and GlaxoSmithKline; Anbesaw Selassie – None; William Theodore – Honoraria/stock Elseveir, General Electric; Torbjörn Tomson – Research grants/honoraria from Eisai, GlaxoSmithKline, Janssen-Cilag, Novartis, Pfizer, Sanofi-Aventis, UCB Pharma; Samuel Wiebe – None.

We confirm that we have read the Journal’s position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.


  1. Top of page
  2. Summary
  3. Epidemiologic Methods for the Study of Epilepsy
  4. Conclusion
  5. Disclosures
  6. References
  7. Appendices


  1. Top of page
  2. Summary
  3. Epidemiologic Methods for the Study of Epilepsy
  4. Conclusion
  5. Disclosures
  6. References
  7. Appendices


Table A1.   Examples of acute symptomatic seizures in association with disruption of the structural or functional integrity of the brain
Cause (Annegers et al., 1995; Beghi et al., 2010)Period of occurrenceNotes/Exceptions
Cerebrovascular disease (Labovitz et al., 2001; Camilo & Goldstein, 2004; Hesdorffer et al., 2009)First 7 days 
Traumatic brain injury (Jennett, 1973a,b; Jennett et al., 1973; Frey, 2003; Hesdorffer et al., 2009)First 7 daysIncludes intracranial surgery. Longer intervals are acceptable for subdural hematoma in the absence of known trauma or at first identification of new hematoma. Subsequent seizures are unprovoked.
CNS infection (Hesdorffer et al., 2009; Beghi et al., 2010) bacterial/viralFirst 7 daysIncludes seizures occurring after 7 days in patients with persistent clinical and/or laboratory signs of infection
NeurocysticercosisPresence of parasites in transitional or degenerative phase by imagingIncludes seizures occurring in the presence of mixed forms. Seizures occurring with viable parasites (acute phase) or calcified granuloma (calcified phase) are unprovoked
MalariaPresence of fever and malaria parasitemia 
Postmalaria neurologic syndromeClearance of parasitemia, associated with fever and psychosis 
Cerebral tuberculomaDuring treatmentSeizures occurring after successful treatment are unprovoked
Brain abscessDuring treatmentSeizures occurring after successful treatment are remote symptomatic
HIV infectionAcute infection or with severe metabolic disturbancesSeizures occurring in the absence of opportunistic CNS infections or severe metabolic disturbances are unprovoked
Arterovenous malformationsIn the presence of acute hemorrhageAll other seizures are unprovoked
Multiple sclerosisFirst presenting symptom within 7 days of relapse
Autoimmune diseasesSigns or symptoms of activation
Table A2.   Comorbid conditions associated with epilepsy
Psychiatric disorders (Hesdorffer et al., 2000; Gaitatzis et al., 2004; Hesdorffer et al., 2004; Qin & Nordentoft, 2005; Qin et al., 2005; Hesdorffer et al., 2006, 2007)Mood disorders, anxiety disorders, alcohol related disorders, somatoform disorders, attention deficit hyperactivity disorders, schizophrenia and psychotic disorders, personality disorders, suicidality
Somatic disorders (Gaitatzis et al., 2004)Stroke, cardiovascular disease, diabetes mellitus, migraine, asthma and other pulmonary conditions, celiac disease and other gastrointestinal disorders, osteoporosis and osteopenia
Infectious disease (Carpio et al., 1998)Neurocysticercosis
Infestations (Kabore et al., 1996; Pion et al., 2009)Possibly onchocerciasis and toxocara
Cognitive disorders (Elger et al., 2004; Sillanpaa, 2004; Hermann & Seidenberg, 2007)Cognitive impairment, learning disability
Disabilities (Gaitatzis et al., 2004)Hearing and vision loss
Accidents (Tomson et al., 2004)Accidents and injuries
Nutritional problems (Crepin et al., 2007)Malnutrition

Appendix: Standard Basic Data Elements for Epidemiologic Studies of Epilepsy

The following section presents standard criteria and coding elements for classifying cases by level of certainty, a minimum (“core”) dataset essential for most population-based studies of epilepsy, and an expanded (“optional”) dataset for studies with resources to obtain more detailed information.

Case definition criteria

Using data obtained by trained health care provider (interview or medical records)

  • Definite

  •  clear evidence of two or more unprovoked epileptic seizures that have occurred over interval(s) exceeding 24 h, OR
  •  confirmed diagnosis of epilepsy by a health care provider with appropriate specialized training in the recognition and treatment of epilepsy.
  • Probable

  •  documentation of a diagnosis of epilepsy by a trained nonspecialist health care provider without specific documentation of definite criteria above.
  • Suspect

  •  data suggest a possibility of epilepsy but criteria for definite or probable epilepsy are not met. The information provided is inadequate to confirm or refute the diagnosis of epilepsy.

Using population survey data collected by nonclinician interviewers

  • Probable

  •  respondent (subject or proxy) reports that a physician or trained health care provider has diagnosed epilepsy (probable).
  • Suspect

  •  information provided suggests a possibility of epilepsy but is inadequate to confirm or refute the diagnosis of epilepsy.

Using data provided by traditional healers or nonclinician observers

  • Probable

  •  description is available with adequate clinical detail that is judged by an expert in epilepsy to indicate a high probability of epilepsy, OR
  •  information is provided by traditional healers whose abilities to correctly identify cases have been evaluated and found to be adequate.
  • Suspect

  •  description suggests possibility of epilepsy, but criteria for probable epilepsy are not met.

Using existing coded health data (International Classification of Diseases)

Note: (1) The specificity and positive predictive values of ICD-coded medical encounter data have been shown to vary among studies of epilepsy in different localities (Holden et al., 2005a; Jette et al., 2010). The following scheme is suggested as rough guidance where only coded data are available. An evaluation of the specificity and predictive values of the following codes and code combinations in each study locality is advised if possible, with appropriate modifications of the following scheme as needed. (2) In most localities, adequate sensitivity may be expected only when complete data can be linked for both inpatient and outpatient medical encounters in order to rule out acute symptomatic seizures.

  • Probable

  •  a single medical encounter assigned an ICD-9-CM diagnostic code 345.xx or ICD-10 code G40.x, OR
  •  two or more medical encounters on separate days each assigned ICD-9-CM diagnostic codes 780.39 or ICD-10 codes G41.x or R56.8, OR
  •  a single medical encounter assigned ICD-9-CM diagnostic codes 780.39 or ICD-10 code R56.8 AND an AED is prescribed for outpatient use for 3 or more months.
  • Suspect

  •  a single medical encounter is assigned ICD-9-CM code 780.39, or ICD-10 codes R56.8 or G41.x

Core Data Variables

Case identifier number

Description: A unique identifying code assigned to each case.

Note: Used for linking data from multiple sources or data maintained in separate database files.

Birth date

Description: Include month, day, and year of birth, where practical.

Note: An approximation can be used (e.g., year only, or month and year) if more precise data are unavailable or not collected to preserve anonymity of subjects. Missing values for month or day may be estimated. A precise birth date, in addition to name and sex, may be necessary to link case records from multiple sources if these lack common unique identifiers for each case.


Description: Age at time of case ascertainment by the study or, in prevalence studies, age on “prevalence day.”

Note: Not essential if birth date information is complete and accurate. If not known precisely, may be estimated, if necessary, using historical milestones. See also the separate variable, Age of Onset.


Description: Case classification as female or male.

Locality of residence

Description: May refer to a locality or district, postal code, city, county, or province or state.

Note: In studies of general populations, may be adapted to local circumstances, if possible, corresponding to recognized geopolitical units for which census data are available.

Race, ethnicity or nationality

Note: Schema are suggested consistent with standard government reporting practices in the jurisdiction under study.

Current medical treatment status (epilepsy under treatment, not under treatment)

Description: Whether epilepsy is currently under medical treatment or not.


  • Epilepsy currently under treatment

  • Epilepsy not currently under treatment

  • Unknown treatment status

Note: See also Type of Treatment under optional variables.

Age of onset

Description: The age at which seizures first began.

Note: Alternatively, the date (month and year) can be used. This is a core variable for studies of incidence, but is optional for studies of prevalence.

Date of last seizure

Description: Preferably, month, day, and year of most recent seizure.

Note: Day (and month, if necessary) can be imputed if precise date not known. Alternative classification may be most recent occurrence within intervals, for example, within the past month, or 3 or 6 months, or 1, 2, or 5 years of the date of data collection.

Seizure frequency

Description: For prevalence studies, number of seizures in the past month, 3 months, and year.

Values: For each interval, record number of reported seizures.

Source(s) of data

Description: sources providing data for each case.


  • Self-report

  • Proxy report

  • Key informants (e.g., teachers, traditional healers, community leaders)

  • Prospective clinical history and examination

  • Retrospective medical record data from clinic, emergency department (ED), or hospital

  • Administrative data (ICD-coded hospital, ED, or clinic data)

  • Administrative data (pharmacy records)

  • Vital records (ICD-coded)

  • Vital records (not coded)

Note: In cases where multiple sources have contributed data, each should be noted.

Additional Data Variables

The usefulness of the following variables will depend on the purpose of the study.

Other sociodemographics
Relationship status

Description: current marital (or equivalent) status


  • Married

  • Domestic partner (long-term relationship with shared living arrangements)

  • Divorced/separated

  • Widowed

  • Never married/never a domestic partner

  • Not applicable (child)

  • Unknown

Household composition

Description: Number of persons living in household and relationships to person with epilepsy. Values: The numbers of parents, other adults, and dependent children should be noted with ages and relationships described.

Educational attainment

Description: Highest level of formal education achieved.

Values: Categories vary greatly among countries. Standard categories used in reports issued by the relevant national government should be considered. In some resource-poor countries, it may be necessary to add whether or not a person can read and write as a separate category.

Note: For children with epilepsy, use information pertaining to the parent of highest attainment with whom they are living.

Employment status

Description: Whether gainfully or otherwise employed.

Values: Categories vary greatly among countries. Standard categories used in reports issued by the relevant national government should be considered.

Notes: For children with epilepsy, use information pertaining to their parent(s) with whom they are living. Account should be taken of employment in informal economies, especially in LMIC. If possible, determine if unemployment is attributed to epilepsy.


Description: Broad vocational categories

Values: Categories vary greatly among countries. Standard categories used in reports issued by the relevant national government—or categories used by the International Labor Organization (International Labor Organization, 1990) or the United Nations (Statistical Commission on International Economic and Social Classifications, 2008)—should be considered.

Note: For children with epilepsy, use information pertaining to their parents with whom they are living. The occupations of homemaker and parental childcare provider should be included, although these are omitted from some standard classifications.

Personal income

Description: Annual income reported by person with epilepsy.

Values: Relevant income strata vary among countries. Standard categories used in reports issued by the relevant national government may be considered.

Note: Source of income (e.g., earned income or disability allowance) may be considered. Not applicable for children.

Household income

Description: Annual income reported for household in which person with epilepsy resides.

Values: Relevant income strata vary among countries. Standard categories used in reports issued by the relevant national government may be considered. In the United States, relevant categories might be defined in terms of the percent of the Federal Poverty Level.

Note: Applicable for children

Health insurance

Description: Type of health insurance, if any, covering person with epilepsy.

Values: Relevant categories vary among countries and at a minimum, private insurance should be separated from public insurance coverage. In the United States, relevant categories might include:

  • Private insurance

  • Medicare

  • Medicaid (SCHIP for children)

  • Other government-sponsored insurance

  • “Self-insured”

  • None

  • Unknown

Seizure type and syndrome
Seizure type

Description: Ideally, the determination of seizure type is based on detailed clinical descriptions supplemented with information from EEG, neuroimaging studies, and examination findings. In populations where such information is routinely available, the ILAE classification (Engel, 2006b; Berg et al., 2010) may be considered for epidemiologic studies. In the absence of such detailed information simplifications to that classification may be useful. More simplified levels of classification rely to a greater extent on the predominant ictal semiology. However, they still allow information from studies with more classification detail to be collapsed onto the same meaningful categories, enabling comparisons across disparate settings. Where cases experience more than one seizure type; each type should be described.


  • Based on clinical data—see Appendix, Table A1 for simplified classification or see ILAE classification (Engel, 2006b).

  • Based on ICD-coded data—see table below. Note that accuracy of ICD-9-CM and ICD-10 coding to the level of the fourth and fifth digits has not been extensively validated.

Table A3.   Simplified Seizure Classification by mode of onset, based on ICD-coded data
Convulsive, nonconvulsive not differentiated G40.3, G40.4, G40.6
 Nonconvulsive345.0, 345.2;G40.7, G41.1
 Convulsive345.1, 345.3 
Simple, dyscognitive (complex) not differentiated G40.0
 Dyscognitive (complex)345.4G40.2, G41.2
  Unclassified345.6–345.9, 780.39G40.5, G40.8, G40.9, G41.0, G41.8, G41.9, R56.
Epilepsy syndrome

The ILAE provides a list of epilepsy syndromes that is periodically revised (Commission on Classification and Terminology of the International League Against Epilepsy 1989; Engel, 2006a; Berg et al., 2010). The classification requires detailed clinical and electroencephalographic information that may not be available for epidemiologic studies.

Note: Where adequate data are available, the full ILAE classification may be used.

Etiologies and comorbid conditions

Description: Direct cause of epilepsy.

Values: See Appendix, Table A3.

Stability of underlying condition


  • Static

  • Progressive

Note: This is primarily applicable to causes classified as “structural/metabolic” or “unknown.”

Neurological status

Description: broad assessment of overall neurological function and subcategories of cognitive, motor, and sensory function. The purpose is to identify persons who, in addition to epilepsy, also have substantial motor, cognitive, or sensory impairments sufficient to interfere with normal education, employment, or independent living.

Values (indicate each as normal or impaired):

  • General Neurologic Status

    • If impaired, specify:

    • Cognitive function

    • Motor function

    • Sensory function (vision and hearing)

Note: If cognitive or motor impairments are congenital or acquired early in life, that is, cerebral palsy (Christine et al., 2007) or intellectual disability (mental retardation) (Shevell, 2008), this should be specified.


Description: Defined as either (1) any medical or psychiatric condition that may occur or coexist with epilepsy or (2) those conditions that occur or coexist with epilepsy at rates that are greater than expected in general populations. For the former, classifications of general morbidity indices are advised (Linn et al., 1968; Charlson et al., 1987; Deyo et al., 1992; Miller et al., 1992; Dodds et al., 1993; Elixhauser et al., 1998; Selim et al., 2004). For the latter, describe specific conditions of interest to the study. These may be grouped under broad categories of psychiatric disorders, neurologic and special sensory disorders, and general medical conditions.

Current medical treatment status (epilepsy under treatment, not under treatment)

Description: Whether epilepsy is currently under medical treatment or not.


  • Epilepsy currently under treatment

  • Epilepsy not currently under treatment

  • Unknown treatment status

Note: See also Type of Treatment under optional variables.

Medical therapy

Description: Type of treatment received for epilepsy.


  • AED(s) (record and specify all AEDs used)

  • Diet therapy (e.g., ketogenic diet)

  • Stimulator/device (vagus nerve or brain)

  • Corticosteroids

  • Epilepsy surgery (e.g., past cortical resection, hemispherectomy, callosotomy, multiple subpial transection)

  • Traditional medicine or treatment (specify)

  • Other (specify)

  • No current or past medical therapy for epilepsy

Note: Record all that apply. Record whether each type of therapy is current (ongoing) or past (discontinued or completed). If all AEDs discontinued, record date last AED was stopped.

Most recent AED use

Description: Interval from time of assessment to time the last dose of AED was taken.

Values: Record the number of hours since last AED dose.

Note: Use longer unit(s) of time, if applicable.

Other care

Description: Other supportive care received for epilepsy or its consequences.


  • Occupational or physical therapy

  • Vocational rehabilitation

  • Individual Educational Plan (children)

  • Other special educational support

  • Other health service support (e.g., children’s health plan)

  • No current or past services for epilepsy recorded

Note: Record all that apply. Record whether each type of service is current (ongoing) or past (discontinued or completed).

Emergency treatment

Description: Number of occasions of emergency/acute medical treatment (e.g., intravenous, rectal, or buccal administration of benzodiazepine) for seizures in the past month, 3 months, and year.

Values: For each interval, record number of times emergency treatment was received, specify type of emergency treatment, and whether administered in hospital or not.

Other variables

The National Institutes of Health has developed common data elements for several diseases. These may provide supplemental information on data collection for researchers worldwide and can be viewed at Included are individual variables as well as recommended standardized instruments for the measurement of different factors such as quality of life, cognition, and comorbidity.