Existing surveillance data on epilepsy do not provide current or complete information on how this disorder affects the U.S. population. This information is critical to guide prevention and intervention efforts, service delivery programs, quality improvement efforts, and health policy. Currently epilepsy-related data are not standardized across studies, which limits the accuracy of case ascertainment and coding and hinders monitoring of health services and quality of life. In addition, epilepsy is not routinely included in major population surveys, registries, and other databases. Actions needed to provide more timely information on a number of key attributes of the epilepsies—such as incidence, prevalence, comorbidities, services utilization, and costs—include the standardization of definitions and criteria for epilepsy surveillance and research as well as the continuation and expansion of epilepsy-related data collection from a variety of sources. The increasing use of electronic health records, which can be linked across providers and payers, may facilitate the gathering of surveillance data.
Data collection is the first step toward better classification and understanding of the problems individuals with epilepsy and their families face. These data are critical to position us to make informed decisions on deploying limited resources … [to] improve the life of individuals and their families…. We are dealing with a poorly addressed public health problem, and we urge you to help us better define its many dimensions and magnitude in order to begin to offer desperately needed solutions.
Public health surveillance systems provide public health agencies, health care providers, policy makers, and the general public with critically important information on the health of people in the United States. Data collected through these efforts provide better understanding of a health condition’s burden (e.g., frequency, severity, impact on functioning and quality of life, health care use, cost) and risk factors for its onset, comorbidities, and outcomes. This information facilitates priority setting, program development, and evaluation decisions (IOM, 2011a; Trevathan, 2011). Surveillance for public health is defined as “the ongoing, systematic collection, analysis, interpretation, and dissemination of data regarding a health-related event for use in public health action to reduce morbidity and mortality and to improve health” (German et al., 2001); surveillance may include the collection of data from a variety of sources, including registries and disease-specific reporting systems, surveys, and administrative and clinical data sets (CDC, 2011f).1 This chapter describes the need for more comprehensive, timely, and accurate epilepsy surveillance by
• discussing gaps in current data,
• assessing the measurement and methodological challenges of collecting data, and
• reviewing available data sources.
It also discusses how epilepsy surveillance might be improved by enhancing data collection and standardizing methods of measurement and case ascertainment.
Epilepsy surveillance data inform all of the other chapters of this report. However, current epilepsy surveillance resources and mechanisms are inadequate, and improvements are necessary to increase understanding of the epidemiologic aspects of epilepsy and to identify effective action in prevention, health care, and community services, as well as education and awareness. At present, public health researchers, policy makers, and advocates are “flying blind” due to the lack of adequate epilepsy surveillance data and infrastructure (Trevathan, 2011). While the focus of this chapter is on epilepsy surveillance and data collection in the United States, the assessment is informed by epilepsy surveillance efforts internationally as well as by surveillance systems for other health conditions in the United States.
The committee’s vision for effective epilepsy surveillance involves the development of active and passive data collection systems that follow stan-
1 As noted in the Data Collection section, data collected through these sources can also be used for epidemiologic research, including longitudinal cohort studies such as the Rochester Epidemiology Project (discussed below).
dardized methodologies to obtain valid measurement. Such systems need to be coordinated, comprehensive, accurate, and timely. In times of economic constraint, collaborative efforts may facilitate this surveillance, which will provide critical information to stakeholders at the local, state, regional, and national levels. Surveillance data can be used to achieve a range of goals, including
• guiding programs and policies aimed at prevention, treatment, and rehabilitation;
• detecting barriers in health care access and quality, such as delayed diagnosis, treatment gaps, and disparities;
• determining optimal service delivery models that are cost-effective; and
• providing a basis for further epidemiologic and health services research.
We need factual data. This would include the incidence and severity of refractory [epilepsy], disparities in access to care, comorbidities … and [epilepsy’s] impact financially and on quality of life for patients and providers.
At the heart of public health surveillance are data. The information presented in Chapter 1 and throughout this report about the significant burden of the epilepsies on health and quality of life is based on data collected through a variety of surveillance data sources, such as administrative and clinical records, population-based surveys, and registries (discussed later in the chapter). To meet the informational needs of the broad epilepsy community, data collected through epilepsy surveillance systems should be able to provide timely and accurate estimates of
• etiology (i.e., causes), risk factors, and comorbidities (Chapter 3);
• health disparities (Chapter 4);
• quality of care (Chapter 4); and
2Incidence is the number of new cases of a disease or disorder in a set period of time; prevalence is the number of existing cases of a disease or disorder at a given point in time.
For example, accurate and detailed surveillance data on the etiologies of, and risk factors for, epilepsy are needed in order to identify opportunities for public health efforts to prevent epilepsy from developing in the first place or to prevent a range of possible consequences. Furthermore, because the burden of the comorbidities often outweighs that of the epilepsy itself, surveillance of its comorbidities is also crucial to appropriate targeting of public health interventions. Currently, gaps in data collection prevent accurate and timely information to monitor and evaluate these basic public health dimensions of the epilepsies, one of the most common neurological disorders in the United States.
The data generated by the Rochester Epidemiology Project3 have formed the foundation of much of the current understanding about the epilepsies in the United States. This project’s contributions have been substantial, but many of the epidemiologic estimates it has generated are outdated and may not reflect the diversity of the current U.S. population. Up-to-date and representative data are needed on epilepsy trends and disparities in specific populations in order to generate actionable information that enables the public health community to target its resources for prevention and intervention in areas that will produce maximum benefit.
Obtaining a complete picture of epilepsy in the United States would require collecting many data elements (Box 2-1). Although all these elements are important—and in an ideal world would be available at the national, state, and local levels—some are more difficult to obtain than others and compromises will need to be made, given limited resources and technology. However, developing the capacity to gather many, if not all, of these data elements—using validated instruments and different data sources on representative populations and subgroups over time—will enable an informed public health response to promote health and well-being for people with epilepsy.
Improving epilepsy surveillance will involve overcoming several challenges in measurement and methodology. Many of the data currently
3The Rochester Epidemiology Project (http://www.rochesterproject.org/) is a collaborative effort by health care providers in Olmsted County, Minnesota, and the surrounding area. This project links medical records across practices that may see Olmsted County residents, making the linked records available to researchers. These records include inpatient, outpatient, and emergency room visits. Records linkage-based research is ongoing in Rochester for a variety of disorders. Data on epilepsy from 1935 to 1994 have been analyzed to provide estimates of epilepsy incidence, prevalence, and cost, as well as information on etiologies, risk factors, and outcomes (e.g., Annegers et al., 1996; Begley et al., 2001; Ficker et al., 1998; Hauser et al., 1991, 1993). Additional projects undertaken include studies on status epilepticus and the genetics of the epilepsies (e.g., Hesdorffer et al., 1998; Ottman et al., 1996).
- Age (including birth date when possible)
- Geographic location
- Personal and family demographics
- Relationship status
- Household composition
- Educational attainment
- Employment status
- Current health status
- General health status
- Epilepsy-specific status
- Current medical treatment status
- Disability status
- Mortality, including sudden unexpected death in epilepsy and other epilepsy-related deaths
- Age at onset
- Seizure type and frequency
- Epilepsy syndrome
– Stability of underlying condition
- Somatic disorders
- Neurological disorders
- Mental health conditions
- Cognitive disorders
- Infectious diseases
- Physical disabilities
- Nutritional problems
- Health insurance status
- Health care
- Source of care
- Type and frequency of use
- Quality of care
- Patient‘s perceptions of care quality
- Direct costs
- Use of informal and community services
- Type of caregiver
- Type of community service
- Quality of life
- Overall quality of life
- Seizure worry
- Emotional well-being
- Cognitive functioning
–Attention or concentration
- Medication effects
- Social functioning
- Role limitations
- Indirect costs
SOURCE: Adapted from Thurman et al., 2011.
collected cannot be validated, are not comparable, cannot be used to understand trends over time, are not representative of the U.S. population, and cannot be analyzed for important population subgroups. Many of these challenges are shared by clinical researchers as well, who are currently collaborating on the Common Data Elements project (described
below). The following are among the major measurement and methodological considerations that are barriers to epilepsy surveillance and research:
• variations in measurement of health service use, quality, access, and costs;
• heterogeneous approaches to assessing the impact of epilepsy on health status and quality of life; and
• challenges in identifying and recruiting health care providers and people with epilepsy to participate in surveillance and research projects.
Case Ascertainment and Diagnostic Accuracy
Unlike other disorders that have definable stages, … we have not defined epilepsy for epidemiologic [research] in a reproducible manner.
Determining timely and accurate incidence and prevalence estimates of epilepsy requires identifying individuals within a population who have epilepsy and determining when they developed the disorder. Although this sounds simple, it is unfortunately quite difficult. Case ascertainment and diagnostic accuracy depend on a number of factors, including standardization and validation6 of definitions and coding of the data, as well as the strengths and limitations of the source of the data (discussed later in this chapter).
Surveillance of the epilepsies strives toward complete ascertainment of people with epilepsy. For epidemiologic studies, this is particularly important to reduce the chance of artificially increasing or decreasing the proportion of the study population with epilepsy. Under- or overestimating the number of people with epilepsy in a population can occur for many reasons. For example, if data from health care facilities are used to identify who has epilepsy, some cases will be missed because some people with epilepsy never seek medical care for their seizures (Beran et al., 1985) or cannot access
4Case ascertainment is the identification and inclusion of people who meet the criteria being studied.
5An algorithm is the combination of codes and other criteria used to identify a case.
6 Validation involves testing and verifying the accuracy of a specific research method, such as the ability of a set of criteria to identify > 90 percent of the individuals in a population who have epilepsy.
health care because of socioeconomic or health system barriers (Szaflarski et al., 2006).
Variations in the Definition of Epilepsy
The use of varying definitions of epilepsy leads to some studies including cases that others would not, which increases the potential for under- or over-reporting epilepsy incidence and prevalence and prevents researchers from being able to compare data across sites and studies. Efforts are ongoing within the epilepsy research field to develop and use standardized definitions and algorithms for identifying epilepsy, epilepsy remission, refractory epilepsy, and active epilepsy, despite using different data sources. As described in Chapter 1, the occurrence of two or more unprovoked seizures separated by at least 24 hours is the broad operational definition of epilepsy, which was proposed by the International League Against Epilepsy (ILAE, 1993) and remains the most widely accepted. However, alternative definitions of general epilepsy and epilepsy subgroups continue to be discussed, and accurate and consistent case ascertainment depends on translating these definitions into standard data collection measures; for example, the length of time covered by the term “active” epilepsy may need to be shorter in surveys than in studies of medical records, in order to account for memory recall of survey respondents. The strengths and weaknesses of current methods of case ascertainment in a number of data sources used for epilepsy surveillance are considered later in this chapter.
On the clinical level, epilepsy can be difficult to diagnose (Chapters 1 and 4) because the health care provider rarely sees the seizure occur and accurately identifying the nature of the seizure or seizure-like event involves determining whether it was due to electrical disruptions in the brain (i.e., a seizure) or other reasons and whether it was provoked (e.g., by a fever). For example, seizures suffered during alcohol withdrawal or seizure-like events with a psychological basis may incorrectly be assumed to be epilepsy and may lead to over-reporting of epilepsy cases. On the other hand, underreporting of epilepsy may occur if the health professional does not recognize the symptoms as a seizure. Further, if seizure activity begins following a brain insult such as stroke, the focus may be on the primary diagnosis of cerebrovascular disease, and the seizures may not be diagnosed as epilepsy. Educating primary care providers and other health professionals regarding seizures and epilepsy can lead to more accurate diagnoses (Chapter 5), as can tools such as decision prompts in electronic health records (EHRs) to guide health professionals toward accurate diagnoses.
Diagnostic and Treatment Coding
Health care diagnoses and treatment decisions are coded in the patient’s medical record, generally with International Classification of Disease Clinical Modification (ICD-CM) codes, for billing and follow-up purposes. Researchers use these disease- or disorder-specific diagnostic and treatment codes to identify records for individuals with specific health conditions. Using codes for case ascertainment is more cost-effective than conducting reviews of each record by hand or interviewing each individual in the study population (Jetté et al., 2010). Furthermore, the current nationwide drive to implement EHRs (discussed later in the chapter) offers unprecedented opportunities to capture, share, and analyze coded data for surveillance purposes.
Nonetheless, epilepsy is challenging to diagnose and match to the appropriate code, and variations in coding practices can lead to over- or under-reporting of epilepsy. Both the ninth revision of the coding structure (ICD-9-CM), which is currently used in the United States, and the ICD-10-CM classification, which will be implemented in 2013 (HHS, 2009), have a number of codes for different types of seizures, signs, and symptoms and a limited number of codes for epilepsy (ICD-9-CM: 345.xx; ICD-10-CM: G40.x).
Currently there are several limitations to the use of codes for surveillance purposes. First, the ICD-9-CM and ICD-10-CM versions lack specificity in the epilepsy codes with respect to etiology, which limits researchers’ ability to elucidate risk factors for epilepsy and report outcomes by cause. Second, coding practices differ; for example, an epileptologist often provides more detailed and accurate information for coding as to type of epilepsy or seizure than an emergency department physician or general neurologist (Jetté et al., 2010). Third, few studies have been conducted to validate the algorithms used to identify epilepsy in different health care settings and across age groups; standardization is lacking in the codes used and in the period of “look back” to determine the incidence of epilepsy (see discussion of the data-gathering effort below). “Seizure, convulsion, epilepsy” were systematically reviewed as part of the Food and Drug Administration’s (FDA’s) Mini-Sentinel pilot project (discussed later in the chapter) to establish validated algorithms that can be applied in surveillance using administrative and claims data, and Kee and colleagues (2012) found that currently the validity of algorithms for identifying epilepsy in comparison to non-epilepsy seizures varies and further research is needed.
Validation studies have found that the presence of multiple occurrences of epilepsy codes—along with record of a prescribed seizure medication— improves accuracy in identifying someone with epilepsy (Holden et al., 2005a,b). Importantly, the algorithms used by Holden and colleagues required that multiple data types be linked (e.g., claims data, data from a
visit to a health professional, pharmacy data, membership data). Some studies have the time and resources to review the medical records in order to validate a subsample of the population (e.g., Parko and Thurman, 2009; Pugh et al., 2008); this enables understanding of the degree to which false positives and false negatives exist. Some studies combine information from the medical record with information from patient interviews and subject this information to review by experts to reach a consensus diagnosis (e.g., Benn et al., 2008; Berg et al., 1999; Olafsson et al., 2005). In these circumstances, cases are most often excluded due to syncope (i.e., fainting) or to seizure-like events with a psychological basis (Scheepers et al., 1998; Smith et al., 1999). For surveillance work, algorithms need to follow agreed-upon definitions and sets of codes so that searches of coded data sources will consistently retrieve cases with epilepsy.
Even given accurate coding, challenges remain for many studies because not all records are coded or complete with all required data, such as type of physician seen and race/ethnicity of the patient. Over time, as patients move from one health care provider or system to another, duplicate case counts can occur and attempts to measure incidence are compromised by the movement of patients within and between health care systems. Recently, researchers have begun using natural language processing to search the free text of the EHR in order to validate the ICD codes for specific conditions, such as pneumonia, pancreatic cancer, and psoriatic arthritis (Dublin et al., 2011; Friedlin et al., 2010; Love et al., 2011), and for other purposes, such as identifying patients who were due for recommended screening tests (Denny et al., 2012) or postoperative complications (Murff et al., 2011). One of the next steps in the validation of epilepsy codes is the use of natural language processing to determine their accuracy.
Self-Reporting Through Surveys
Researchers often use population-based surveys to collect health data. To identify individuals with epilepsy, an initial set of screening questions is generally asked, and these questions vary from survey to survey. These population surveys, such as the Behavioral Risk Factor Surveillance System (BRFSS) (discussed below), rely on self-reports of physician-diagnosed epilepsy and tend to generate considerably higher prevalence estimates than those from medical records or community-based studies. Following up on an initial identification of persons with epilepsy based on self-report, more in-depth questions and validation or review, such as medical examinations or review of medical records, help to reconcile these estimates. For example, a prevalence study in New York City produced initial rates of epilepsy similar to the BRFSS; additional information to aid case ascertainment and expert review of responses by a panel of epileptologists lowered prevalence
levels to those of other studies (Kelvin et al., 2007). Generating accurate incidence estimates from self-report population-based studies is not possible due to the difficulty of validating cases and faulty recall concerning the timing of epilepsy onset. In addition, some types of information may not be captured reliably through the self-reports of people with epilepsy and their families; for example, some studies have found that seizure frequency counts are underestimates because the majority of respondents are unaware of some seizures (Akman et al., 2009; Blum et al., 1996; Hoppe et al., 2007).
A focus is needed on identifying the screening questions that accurately determine the epilepsy status of individuals and contribute to information on overall prevalence. Recently, Brooks and colleagues (2012) validated the use of the five epilepsy-related screening questions developed by the Centers for Disease Control and Prevention (CDC) Epilepsy Group, which have been used by the BRFSS and other population-based surveys, in a sample of patients who receive care at a tertiary care center in Boston. Their findings suggest that prevalence estimates of lifetime and active epilepsy based on self-reports, while slightly higher than estimates based on medical review, are reasonably accurate and valuable for population-based studies. Further work is needed to determine whether their findings are generalizable to other populations. Because individuals may say they have a seizure disorder and not realize they have epilepsy and because communities may differ in the words used for seizures and epilepsy, as well as the extent and nature of the stigma associated with epilepsy, questions should not only follow standardized concepts and methods, but also be culturally adapted, designed using the principles of clear communication, and validated in the specific population being studied. Like other conditions with a similar prevalence, obtaining sufficient data for studying specific segments of the population (e.g., racial/ethnic minorities, age groups) is difficult because large sample sizes are needed. Further, surveys generally do not collect data on certain subpopulations such as homeless individuals or institutionalized individuals, and many do not include children.
Monitoring Health Care Quality, Access, and Direct Costs
In the last few decades, greater attention has been focused on the need to conduct surveillance of the quality, access, and value aspects of health care in order to maximize health outcomes and control costs (Chapter 4). Quality of care can be measured in several ways:
• by characteristics of health care structure (e.g., type of health care provider seen during visits by patients with epilepsy, type of health care facility where care was sought such as an epilepsy center),
• by elements of the process during a visit by the patient to a health professional (e.g., procedure or test ordered, such as video-electroencephalograph [EEG]), or
• by data on the individual’s outcomes or resulting health status (e.g., seizure frequency, disability status) (Brook et al., 1996).
Collecting data on the process of care may provide the most sensitive estimates of high-quality care. Performance metrics derived from evidence-based practices can be used to assess and incentivize high-quality care. Importantly, to effect change, these metrics should be oriented to the health care provider’s direct role and responsibilities (Giuffrida et al., 1999). As discussed in Chapter 4, there has been significant progress recently in developing performance metrics specific to epilepsy, such as counseling about treatment side effects or referring a patient with refractory epilepsy for surgical evaluation (Fountain et al., 2011; Pugh et al., 2011). Much work remains to implement the metrics and establish a measurement framework and consistent mechanisms for monitoring the quality of different aspects of epilepsy care. EHRs are a possible source for the collection of relevant data for measuring quality. One goal of the implementation of EHRs is to improve the quality, safety, and efficiency of care by collecting structured data that will allow efficient information exchange (CMS, 2010b; HHS, 2010a).7
Assessing whether people with epilepsy have adequate access to care can be measured by examining potential (e.g., having a usual source of care), realized (e.g., visits to a physician), and outcome (e.g., health status) metrics (Andersen and Aday, 1978). The presence of significant differences in access metrics helps identify health care disparities between disadvantaged individuals or population groups that differ from the general population in demographic or socioeconomic status but have comparable needs. Factors to consider in measuring access to care include health system factors (e.g., availability of health care resources and providers, accessibility and acceptability of those resources to potential patients) and personal factors (e.g., gender, age, race/ethnicity, geographic location, education, income, type of insurance coverage).
Estimates of the frequency of service use and the related costs contribute to assessments of the value of health care (Chapter 4). Capacity is needed to accurately identify and track the frequency of service use by an individual and the costs directly related to those services (e.g., physician visits, diagnostic procedures, hospital stays, prescriptions). In measuring service use, data need to be collected about the type of provider seen and
7Other goals include reducing health disparities, engaging patients and families in their health care, improving care coordination, and improving public health (CMS, 2010b).
the health care setting. Critically, in order to capture complete data on an individual’s health care services and costs, databases need to be linked across relevant providers and health care settings. In epilepsy, this often requires obtaining data from multiple sources, since few data sources are comprehensive enough to include all relevant service types and settings.
The measurement of nonmedical direct costs, such as informal care by family members, and community service costs, such as education, training, and rehabilitation, is necessary to assess the full economic impact of the disorder. Because of the difficulty in obtaining nonmedical care cost data, some studies have not included these costs in their estimates (Begley et al., 2000), and variations in how studies have measured costs and made projections make it difficult to compare estimates across studies. In addition, epilepsy is known to be associated with mental health conditions and learning disabilities, and the costs associated with these comorbidities are not reflected in current estimates. To accurately assess the direct cost burden of the epilepsies on people with epilepsy and their families, additional work is needed to develop common methodologies that capture nonmedical direct costs in a more comprehensive, valid, and representative way.
Assessing Quality of Life and Indirect Costs
Increased emphasis on patients’ perspectives about their health and health care has led to the development of tools to measure quality of life. Quality of life is a multidimensional construct that includes components of emotional well-being, cognitive functioning, and social functioning (Chapter 6). Although a gold standard for assessing overall quality of life is not available currently, a number of validated generic and epilepsy-specific instruments can be used (Solans et al., 2008).
Generic instruments, such as the Medical Outcomes Study 36-Item Short Form health survey, focus on aspects of life that are widely applicable to all people (Brazier et al., 1992; Coons et al., 2000). Data collected with these types of surveys enable comparisons between people with epilepsy, those with other diseases and disorders, and the general public that identify the burden of disease attributable to epilepsy and how it compares to other conditions. However, generic instruments may not be able to identify more subtle aspects of epilepsy’s impact on quality of life (Sabaz et al., 2000). Validated epilepsy-specific instruments include the Liverpool Batteries (Baker, 1998), QOLIE-10 (Quality of Life in Epilepsy) (Cramer et al., 1996), QOLIE-31 for adults and QOLIE-AD-48 for adolescents (Cramer et al., 1998, 1999), QOLIE-89 (Devinsky et al., 1995), QOLCE (Quality of Life for Childhood Epilepsy) (Sabaz et al., 2000), the Seizure Severity Questionnaire (Cramer et al., 2002), and the Impact of Childhood Neurologic Disability Scale (Camfield et al., 2003).
Epilepsy-specific instruments assess the most important problems associated with the aspects of life directly affected by seizures and the side effects of medications taken to control them. For example, a patient-completed symptom checklist might be used to measure the impact of side effects from seizure medications. Quality-of-life instruments may also assess health state preferences (e.g., through scoring of various levels of functioning and well-being) to ensure that the perspective of the individual with epilepsy is captured (e.g., Stavem, 1998). Recent work has focused on development and validation of the Neuro-QOL instruments, which can be used for a number of neurological disorders (Nowinski et al., 2010), as well as qualitative interviews to understand the impact of epilepsy on various aspects of life (Kerr et al., 2011).
Each of these types of instruments has its own characteristics and requires careful consideration before being used to monitor quality of life in surveillance systems. Generally, surveys and questionnaires are the primary sources of this information since the data needed to assess quality of life must come from perceptions of people with epilepsy or from family members if the individual with epilepsy is a child or is intellectually impaired. With the variety of validated instruments that are available, standardizing the approach and frequent use of a common instrument will help generate comparable data on the impact of epilepsy on individuals.
The measurement of indirect costs associated with productivity losses reflects the full impact of epilepsy in economic terms. A few estimates of these costs have been calculated by estimating the lost productivity of people with epilepsy due to premature morbidity and mortality. Other dimensions of indirect costs, such as those associated with pain and suffering or those due to lost productivity of family members who care for an individual with epilepsy, have not been addressed. Studies that have examined the indirect costs of epilepsy find that they generally exceed direct costs by a significant margin (Begley et al., 2000; Strzelczyk et al., 2008). To accurately assess the full burden of epilepsy on people who have the disorder and their families and on the economy of the United States, additional work is needed to develop common methodologies that predict indirect costs in a comprehensive, valid, and representative way.
Participation in Surveillance and Research
Lack of participation by people with epilepsy and their health care providers in surveillance and research efforts can be a challenge to researchers. The low scientific and health literacy of the general U.S. population may lead to potential participants being unaware of the reasons why they should participate (IOM, 2011b; Macleish, 2011). Reporting information and responding to surveys can be time-consuming, and accurate, complete
reporting may be difficult for people with cognitive and memory impairments. Further, people with epilepsy may not want to openly discuss their condition out of fear of repercussions due to stigma (Jacoby, 2002).
To maximize participation in surveillance and research and to help ensure that research has valid results,
• the burden on participants should be minimized;
• participants should be informed of the value of their participation and the ways their data will be used;
• any relevant HIPAA8 or privacy considerations should be communicated to participants, who should also be informed that their data will be de-identified; and
• research instruments should follow the principles of clear commu-nication and be culturally appropriate.
Further, recruitment strategies should be evaluated to ensure that requests for participation are sufficiently disseminated to target audiences. Additional research questions include identifying specific subpopulations where response rates are low, determining the impact of this on the bias of the research, and assessing the degree to which improved recruitment of those populations eliminates this bias.
To overcome the paucity of surveillance data and use the data to improve the lives of people with epilepsy, expanded data collection efforts must use consistent methodologies. Currently the National Institute of Neurological Disorders and Stroke (NINDS) is leading a collaborative effort to encourage standardized data collection in clinical research across neurological conditions, including epilepsy (Loring et al., 2011). The Common Data Elements (CDE) project aims to establish common methodologies and terminologies to enable comparable datasets across studies. Public health researchers in epilepsy should look to the CDE project for guidance as the new standards are put into place and should apply its approach to surveillance of the epilepsies.
Demonstration projects that validate the use of specific definitions of epilepsy and criteria for case ascertainment, health care services use, quality of life, and cost measurement are needed to help standardize the current diversity of measures used in surveillance. To ensure validity for all people with epilepsy, these projects should be conducted in a range of
8HIPAA, the Health Insurance Portability and Accountability Act of 1996, established national privacy standards defining protected health information.
health care settings and among diverse population groups. During times of financial constraint, collaborations among federal agencies and advocacy and professional organizations could minimize the burden of conducting these projects.
Data elements for epilepsy surveillance (see Box 2-1 above) can come from many different sources. Each data source has strengths and limitations in providing insights into the disorder. The principal data sources for public health surveillance of the epilepsies include
• population surveys,
• registries and condition-specific reporting systems, and
• records from visits to health care providers (e.g., administrative and clinical records).
These data sources can be mined for broad, population-based surveillance purposes and can be used to inform a variety of population-based studies. Optimally, they could be linked within or across systems to generate a broad collection of data on large populations for use in improving prevention and treatment efforts. Specific research studies are included below to illustrate the types of analyses that could be conducted using these types of data if surveillance systems collected the data in a representative U.S. population.
CDC-Funded Population Health Surveys
General population health surveys are rich sources of data on a wide range of health-related topics. Population health surveys capture many aspects of health conditions and individual characteristics that are well suited for understanding the public health burden of the epilepsies. In the United States, the federal agency responsible for public health surveillance is the CDC. The CDC conducts two large general population surveys, the National Health Interview Survey (NHIS) and the BRFSS surveys (Box 2-2) and also provides support for some other state and local health surveys.
These population health surveys are an important part of epilepsy surveillance and provide representative data to estimate epilepsy prevalence as well as comparative data to understand the burden of the epilepsies. Further, they provide an evidence base to track trends over time in preva-
The Centers for Disease Control and Prevention’s National Health Interview Survey (NHIS) and the Behavioral Risk Factor Surveillance System (BRFSS) surveys operate in all 50 states, the District of Columbia, Puerto Rico, the U.S. Virgin Islands, and Guam. These large surveys use representative samples of the general population and typically interview civilian participants in person or by telephone. Survey responses are aggregated into data files and statistically weighted to represent the entire reference population (e.g., nation, state, county). Survey content changes from survey to survey and from year to year but generally includes detailed respondent demographics (e.g., age, sex, race/ethnicity), socioeconomic status (e.g., income, educational attainment), health conditions, and health behaviors. BRFSS surveys also include optional modules that states can administer. Some states conduct their own health surveys that include epilepsy content, such as the California Health Interview Survey (CHIS).
Analyses from these surveys provide information about the comorbidities of a disease or disorder and a population’s access to and utilization of health care services. Epidemiologic studies of epilepsy based on data collected from the NHIS, BRFSS surveys, and CHIS provide epilepsy prevalence estimates and have established a number of important and consistent relationships by comparing people with epilepsy to those in the general population without epilepsy (Elliott et al., 2008, 2009; Kobau et al., 2007, 2008; Strine et al., 2005). Analyses of some population health surveys have further differentiated people with a history of epilepsy into those with active epilepsy (one or more seizures in the past 3 months or taking medication for seizure control) and those with inactive epilepsy (no seizures in the past 3 months and not taking seizure medications) (Kobau et al., 2007, 2008). Studies based on these data have documented differences between people with and without epilepsy on numerous socioeconomic and health behavior dimensions, such as educational attainment, employment, income, quality of life, physical activity, and overweight or obesity, among others.
lence and treatment practices and in the relationship between epilepsy and a broad range of social and health-related outcomes. The sample size of these surveys tends to be large enough to compare people with and without epilepsy. As samples of the general population, results represent the entire population, including people who may not otherwise interact with the health care system, such as those without health insurance coverage.
These surveys, however, have several important limitations. First, participation is voluntary and declining, and some populations are not covered. Response rates to general population surveys, particularly those conducted by telephone, have declined significantly over the past several decades and may lead to nonresponse bias (Galea and Tracy, 2007). The increased use of cellular telephones has created challenges to adequately cover the general population with traditional landline random-digit dialing sampling methods. These surveys also generally omit other important segments of the population, such as people who are homeless or those living in institutions. This is of particular importance to epilepsy surveillance due to the grow-
ing population of older adults, who (along with children) have the highest incidence of epilepsy and who may live in nursing homes and assisted-living facilities.
Additionally, population survey data on children with epilepsy are insufficient. The CDC conducts the Youth Risk Behavior Surveillance System (YRBSS), but these surveys cover only high school students and are limited in scope as they focus on the six categories of high-risk behaviors that are leading causes of morbidity, mortality, and social problems in U.S. youth (Brener et al., 2004). As currently framed, the YRBSS is not a potential source of epilepsy data. The NHIS asks parents whether their child has had any seizures in the last 12 months (Boyle et al., 2011; CDC, 2011d) but does not ask whether these are epilepsy seizures, which limits its usefulness for epilepsy surveillance. The National Survey of Children’s Health did not include epilepsy in its 2003 version (Gurney et al., 2006), but the 2007 version asked whether the parent was ever told that the child had epilepsy. Thus far, studies based on these data have looked at epilepsy only as a comorbidity of another condition, such as attention deficit hyperactivity disorder (Larson et al., 2011), or as part of the comparison group for another condition, such as autism spectrum disorders (Schieve et al., 2011).
A second limitation of population surveys is that they cannot be used for data on specific populations with epilepsy, and epilepsy-related content thus far has not been regularly included. Although population health surveys have large samples, due to the relatively low prevalence rate of epilepsy, they produce samples that are too small to identify any rate differences across specific population groups, such as differences by race/ethnicity or by severity of epilepsy. The problem of sample size is exacerbated by the infrequent inclusion of epilepsy content in these surveys. BRFSS surveys have included content about epilepsy only in a few years and in a handful of states. In 2005, 19 participating states asked at least one question about epilepsy, and some asked additional questions about recent seizures and seizure frequency, use of seizure medications, and visits to a neurologist or epilepsy specialist in the previous year (Kobau et al., 2008).
Third, these surveys rely on self-reported data and are vulnerable to error (Kobau et al., 2008). For example, as discussed above, self-reported epilepsy may overestimate the presence of epilepsy within the population due to reports of seizures that are not epilepsy seizures (Ferguson et al., 2005; Kelvin et al., 2007), and, as discussed above, they may underestimate seizure count (Akman et al., 2009; Blum et al., 1996; Hoppe et al., 2007). Additionally, epilepsy-specific content has been limited to epilepsy diagnosis, frequency of seizures in the past year, use of medication, and visits to a neurologist in the past year; these surveys are unable to ascertain epilepsy syndrome or seizure type, severity, and etiology.
Epilepsy ruins many lives, and it is essential that we identify and address the enormous treatment gaps that still exist today.
Although incorporating questions about epilepsy into the BRFSS surveys has limitations, as described above, having a broader set of epilepsy-related questions asked in all participating states would generate more and improved surveillance data. One opportunity is for additional survey questions to explore the extent of treatment gaps in epilepsy. Although most research on treatment gaps in epilepsy is in developing countries (e.g., Meyer et al., 2010), BRFSS surveys and the California Health Interview Survey (CHIS) both show that a significant percentage of individuals who have had a seizure in the last 3 months report that they are not currently taking seizure medications (26 percent in CHIS in 2003 and 16 percent in 13 states from BRFSS in 2005) (Kobau et al., 2007, 2008). Additional survey questions on receiving medical care from epileptologists or at an epilepsy specialty center have been developed by CDC but not yet included in surveys. While there is speculation that people with epilepsy who receive specialty care have better outcomes than those who do not, there is currently no population-based evidence to test this hypothesis. Results from such studies could inform knowledge about the treatment gap for limited seizure medication usage in addition to the well-documented treatment gap in surgical treatment for refractory epilepsy (Engel, 2008; Haneef et al., 2010).
Additional questions on the BRFSS surveys would also increase their usefulness for epilepsy surveillance. Specifically, questions about memory and cognition problems would be useful, as would having the existing optional “anxiety and depression” module administered alongside the epilepsy questions to assess the frequency of mental health and cognitive comorbidities. This would permit an assessment of how depression may affect treatment outcomes, quality of life, and other health-related outcomes for people with epilepsy. Further, research is needed that focuses on epilepsy based on the results from the National Survey of Children’s Health.
The Medical Expenditure Panel Survey
While individuals can readily and accurately report many aspects of their health and health care during an interview, the complete cost of their medical treatment is not one of them. To measure and assess medical costs among the general population, the Agency for Healthcare Research and Quality (AHRQ) conducts the Medical Expenditure Panel Survey (MEPS), which is used to evaluate current and predict future health care costs and services use (Box 2-3).
MEPS can be used to specifically examine epilepsy-related data. Over
The Medical Expenditure Panel Survey (MEPS) is a series of household surveys of the U.S. civilian noninstitutionalized population using a sample of the previous year’s National Health Interview Survey, with supplemental information from a survey of medical providers and insurance providers (AHRQ, 2011a; Cohen, 2002). MEPS’s design is overlapping: each year a new panel begins whose cohort is followed for a period of 2 calendar years. It compiles data on patient demographics (including employment status), self-reported health status, use of health services, costs and payments by payer source, and health insurance status. Using a computer-assisted method, there are 5 personal interviews over 30 months, and the Agency for Healthcare Research and Quality (AHRQ) calculates costs for 2 years. De-identified data are provided by AHRQ for public use, including sufficient information about the survey’s methods and measurements to enable analyses of the results that are nationally representative.
multiple years, the collected data contain enough information about people with epilepsy to calculate estimates of the use of health services, costs, and informal care received. MEPS can also be used to monitor access to care and cost of employer-based health insurance as well as health status and well-being (Cohen, 2003). For example, Halpern and colleagues (2011) used MEPS data from 2002 to 2007 to analyze how insurance status affected health care utilization and out-of-pocket costs for people with epilepsy, and Yoon and colleagues (2009) used MEPS data from 1996 to 2004 to estimate the burden of direct health care costs for epilepsy in the United States. Importantly, the longitudinal nature of MEPS, although limited to 2 years, allows a rich source of data that describes service use and cost over time, while avoiding the need for lengthy recall periods by participants. Further, similar to the BRFSS surveys and NHIS, MEPS can be used for comparisons between people with and without epilepsy.
MEPS has similar limitations as the BRFSS and NHIS, such as non-response and too small a sample to allow for analysis of population subgroups of people with epilepsy; also, as MEPS participants are sampled from the NHIS, they do not include people who are institutionalized or who are homeless. Nor does MEPS capture data on indirect costs of epilepsy. However, MEPS has ways to at least partially compensate for some of its shortcomings. While data collected from households are self-reported and are thus subject to error, parallel surveys of the medical providers who care for participants help to improve the accuracy of these self-reports (Cohen, 2003). In addition, unpaid care services provided by family members are obtained (Yoon et al., 2009), so direct costs of nonmedical care are included to some extent.
Future revisions of MEPS could consider increasing the follow-up time to enable analysis of individuals’ patterns of care, health outcomes, and productivity. Given the chronic and recurring nature of epilepsy, this could help to identify trends in the progression of this condition over time. However, proposals to increase the time window should ensure that response rates and validity of data are not adversely affected (Cohen, 2003).
The Children’s Health Study
The Children’s Health Study is a new longitudinal study being planned and conducted by the National Institutes of Health (NIH), in partnership with the CDC and the Environmental Protection Agency. It aims to cover a representative, population-based sample of 100,000 children from birth until they are 21 years old and study the impact of the environment (e.g., water, diet, community influences) and genetics on their health, as well as their growth and development (NIH, 2011b,f). Data will be collected through in-person, telephone, and/or web-based interviews and questionnaires; additional data will include environmental and other samples, physical measurements, and neurological and other assessments (NIH, 2011e). The study plans to monitor the development of EHRs to determine the feasibility of including medical records in its data collection, but currently the primary mechanism will be surveys (NIH, 2011c). Epilepsy is one of the study’s outcomes of interest (NIH, 2011d), and several research projects have begun to develop and validate its questionnaires and other data collection mechanisms (NIH, 2011b). Given the scope in terms of the size and length of this study, it offers a valuable opportunity for prospective data collection in a representative group of U.S. children as part of broader surveillance efforts across age groups.
Registries and Condition-Specific Reporting Systems
In recent years, registries9 have become a common source of data that facilitate health condition-specific research. While registries vary from system to system, they share a common goal of collecting condition-specific, comprehensive incidence and related diagnostic data in a defined population. These condition-specific reporting systems may also be used to track health outcomes over time. Well-developed registries can be a valuable resource for conditions such as epilepsy that may yield relatively small samples in population surveys and other surveillance data sources.
9Registries are databases that contain information about people who have something in common, such as women with epilepsy who are pregnant and taking seizure medications.
Epilepsy Pregnancy Registries
I have been successful at responding to epilepsy treatment but I was not prepared for how my epilepsy would impact my son’s life the way it has. We need to know and address the full effects of antiepileptic drugs prescribed to patients with epilepsy. These drugs impact the mother, as well as her unborn children. This must be included when we talk about the true impact of epilepsy.
Since ethical considerations limit prospective clinical trials for studying pregnancy outcomes, pregnancy registries have become an important source of information about the impact of individual seizure medications on developing fetuses. Several types of pregnancy registries for women with epilepsy have been established, including national databases, independent academic registries, and registries sponsored by pharmaceutical companies (Box 2-4). At present, the only U.S. epilepsy-specific registry is for pregnant women with epilepsy (the North American AED10 Pregnancy Registry). Prior to the establishment of these registries, the only information available to patients and their physicians to guide decisions on epilepsy management during pregnancy came from studies based on case reports and anecdotal experience. These studies enabled the identification of potential risks to fetal development from exposure to seizure medications, including major congenital malformations, such as heart defects, spina bifida, and cleft lip and palate, and also minor malformations such as small digits, although to a lesser extent (Anderson, 1976; Arpino et al., 2000; Holmes et al., 2001; Kaneko et al., 1999; Koch et al., 1992; Lindhout et al., 1992; Olafsson et al., 1998; Omtzigt et al., 1992; Rosa, 1991; Samrén et al., 1997, 1999). However, these studies did not have sufficient statistical power to identify whether specific seizure medications differed in their teratogenic11 potential. The rapid increase of new seizure medications has brought urgency to the need for better understanding of the risks that these drugs pose to the developing fetus (Tomson et al., 2007).
Most of the epilepsy pregnancy registries are prospective, aiming to enroll large numbers of seizure medication-exposed pregnancies. In addition to pregnancy registries providing opportunities to study the effects of seizure medications on developing fetuses, they can also provide information on the impact of seizures during pregnancy and labor. Further, pregnancy registries can identify whether infants who are born with major congenital malformations had these malformations prenatally diagnosed or identified through prenatal screening, and they can also provide data on the number
11Teratogenic means relating to or causing malformations.
Two notable national databases exist for tracking pregnancy outcomes: the Swedish Medical Birth Register and the Finnish Prescription Drug and National Medical Birth Registry. The Swedish Medical Birth Register is population based and collects data from prenatal maternal health records as well as maternity department records. All pregnant Swedish women attending maternity health clinics are screened for chronic disease and medication history. This information is entered into a national database. It is believed that 98 percent of all pregnant women in Sweden attend these clinics. The Finnish Prescription Drug and National Medical Birth Registry identifies all women who are prescribed seizure medications during pregnancy and cross-references these data with the Finnish National Medical Birth Registry in an effort to identify all pregnant women who take seizure medications during pregnancy.
Independent Academic Registries
The North American AED Pregnancy Registry is a prospective voluntary registry where enrollment may be recommended by a physician, and pregnant women in the United States and Canada self-enroll. The primary goal is to determine the frequency of major malformations in infants who are exposed to as many as 34 different seizure medications during pregnancy. Since its inception in 1997 and as of September 2010, this registry has enrolled more than 7,700 self-reporting subjects from the United States and Canada.
The UK Epilepsy and Pregnancy Register, established in 1996, was one of the first pregnancy registries to follow patients prospectively through their pregnancies. To capture outcomes of seizure medication-exposed pregnancies in the United Kingdom (and, since 2007, in Ireland), pregnant women with epilepsy are self-referred or are recruited for participation by their general practitioners, midwives, or other
of pregnancies that were terminated due to prenatal diagnoses or screening results. A large registry such as EUROCAT (European Concerted Action on Congenital Anomalies and Twins) can use pooled data to identify rare malformations and their association with infrequent exposures.
Pregnancy registries have several limitations. A principal weakness is that they are observational studies, not randomized controlled trials. Women are not randomly assigned to receive different seizure medications, and the selection of a particular seizure medication and its dose depends on individual environmental and genetic variables that in themselves may influence the risk of a malformation. Further, if a registry does not actively recruit participants but relies on passive, voluntary participation, it has the potential to introduce bias. For example, in the North American AED Pregnancy Registry, the majority of participants are insured, white, and have a minimum of some college education, making the captured data not representative of the U.S. population (Tomson et al., 2007). Since some
health professionals. Entry into this study requires that the pregnancy outcome not be known at the time of enrollment. The health care provider is contacted after the birth for data collection.
The largest of epilepsy pregnancy registries, the International Registry of Anti-epileptic Drugs and Pregnancy (EURAP), has become an international collaboration representing 40 countries in Europe, Australia, Asia, and South America and is focused on the prospective observational study of pregnancies with seizure medications. EURAP also has a retrospective arm for those pregnancies that do not meet criteria for the prospective study. As of the end of 2011, EURAP had enrolled more than 16,900 pregnancies.
EUROCAT (European Concerted Action on Congenital Anomalies and Twins), a significantly more comprehensive but general (not epilepsy-specific) pregnancy outcome registry, gathers data from dozens of population-based registries to conduct surveillance of congenital malformations, including the impact of seizure medications taken during pregnancy. EUROCAT encompasses 43 registries from 23 countries, covering 29 percent of the birth population of Europe, amounting to 1.7 million births annually. It is a multisource registry collecting data on births as well as terminations of pregnancies following a prenatal diagnosis of congenital malformation.
Pharmaceutical Company Registries
The GlaxoSmithKline International Pregnancy Registry and the UCB, Inc., AED Pregnancy Registry have been used to monitor outcomes from lamotrigine- and levetiracetam-exposed pregnancies, respectively.
SOURCES: EURAP, 2012; EUROCAT, 2012; GlaxoSmithKline, 2012; Irish Epilepsy and Pregnancy Register, 2012; Morrow et al., 2006; North American AED Pregnancy Registry, 2012; Socialstyrelsen, 2012; Tomson et al., 2007, 2010; UCB, Inc., 2012; UK Epilepsy and Pregnancy Register, 2012.
participants have diagnoses other than epilepsy (e.g., migraine), conclusions may be confounded by the underlying maternal health condition, resulting in the impact of epilepsy and seizures during pregnancy and labor not being clearly isolated. Although control subjects are a problem in most registries, the North American AED Pregnancy Registry responds to this problem by recruiting friends and family members of enrolled women as unexposed controls (Tomson et al., 2007, 2010). Limitations of pharmaceutical company-driven seizure medication registries include small samples, lack of control groups, and the potential for bias and conflict of interest (real or perceived) in data interpretation. An important limitation is that existing registries vary in design, which makes systematic comparison of results between registries difficult. Recently, discussions have begun in an attempt to improve the standardization of data collected by several registries in order to enable pooled data comparisons (Tomson et al., 2010).
In moving forward, the North American AED Pregnancy Registry
would benefit from increasing the diversity of its participants through active recruitment and through standardization of its data elements with other major registries to allow analysis among a larger and more diverse sample. Accomplishing these goals might involve dissemination efforts to raise awareness and encourage the participation of women from demographic groups that are currently underrepresented as well as collaborative, international efforts to establish common methodologies. These are not small tasks, but pregnancy registries are currently the major source of data on the safety of seizure medications for the developing fetus. An alternative mechanism for this type of data collection may be the creation of EHR linkages of data on the mother’s seizure medication use with data on the child’s birth outcome, but the capacity to do this has not yet been developed.
The EpiNet Registry
A voluntary, international registry is being developed in New Zealand to collect data on people with epilepsy with the goal of using the database to help recruit participants and run large randomized clinical trials as well as prospective observational studies (Bergin and the EpiNet Study Group, 2011a; Bergin et al., 2007). The EpiNet registry, a secure web-based database, is accessible to approved investigators (i.e., neurologists with expertise and interest in epilepsy) who can input information on seizure type, epilepsy syndrome, etiology, and treatment. Bergin and colleagues (2010) conducted a pilot project in New Zealand and demonstrated that people with epilepsy can be recruited through the Internet for clinical trials. Currently a number of other countries, including Australia, Belgium, Canada, Italy, Pakistan, South Korea, and the United States, are participating in an international pilot project to evaluate the feasibility of the project’s website and database (Bergin and the EpiNet Study Group, 2011b). If privacy protections are put into place and the project is able to enroll sufficient numbers of participants whose data are reported in uniform ways, this registry could be a valuable source of longitudinal data on people with epilepsy around the world.
Registries for Other Conditions
Cancer registries Registries have played an important role in national-level cancer surveillance in the United States for nearly four decades. The Surveillance, Epidemiology, and End Results (SEER) program, which is operated by the NIH’s National Cancer Institute (NCI), began collecting cancer-related data in 1973 as a result of the National Cancer Act of 1971 (NCI, 2012). In 1992, the CDC’s National Program of Cancer Registries (NPCR) was created through the Cancer Registries Amendment Act of 1992 to develop a national system of state-based registries (CDC, 2010b) (Box 2-5).
The Centers for Disease Control and Prevention’s National Program of Cancer Registries (NPCR) supports state-based cancer registries in 45 states, the District of Columbia, and 3 U.S. territories. Combined, these registries cover approximately 96 percent of the U.S. population and collect data such as cancer occurrence, type, extent, and location (CDC, 1999). Data are reported to each state’s registry by health care facilities. Over the last decade, NPCR has worked with states to establish registries where they did not exist previously and to improve the completeness of the data collected (CDC, 2011e).
The National Cancer Institute’s Surveillance, Epidemiology, and End Results (SEER) program is a national database that links data from population-based cancer registries. SEER collects an array of information, including data on patient demographics, diagnosis, treatment, and outcomes. SEER includes data from 8 state registries and 12 city or regional registries within states (e.g., Los Angeles, Seattle, Puget Sound) and covers approximately 28 percent of the U.S. population (Cockburn, 2011; NCI, 2011; Warren et al., 2002). A database has also been formed that links both SEER and Medicare claims data (NCI, 2009).
In 2001, CDC’s NPCR and NCI’s SEER agreed to collaborate in order to form an integrated network and report national-level cancer statistics on incidence, type of cancer, stage of cancer at diagnosis, geographic location, demographics, and mortality (CDC, 2011e; Jemal et al., 2010; Wingo et al., 2003). The comparison and coordinated analysis of their data were possible, in part, because the data compiled through both programs use standards developed by the North American Association of Central Cancer Registries (NAACCR) for case ascertainment and measurement. NAACCR is a professional organization that develops and encourages the use of consensus data standards for the cancer registries’ data collection and categorization, including standard definitions and codes. NAACCR annually certifies registries in the United States and Canada to ensure standardization and availability of high-quality data; all state-based cancer registries were certified in either 2010 or 2011 (NAACCR, 2010a,b, 2011, n.d.).
Data from these registries have been used for a variety of valuable research and reporting purposes, including analysis of cancer risk and treatment disparities by social factors and cancer-related outcomes over time and by group. In addition to annual cancer statistics reports produced by the CDC and the American Cancer Society (CDC, 2011e,g; Jemal et al., 2010), these registries have been used to examine cancer comorbidities, screening and prevention opportunities, treatments, outcomes, quality of care, and costs (Cockburn, 2011; Klabunde et al., 2002; Warren et al., 2002). The data have also been used to evaluate prevention programs, such as sun-exposure awareness programs and the effectiveness of screening programs in reducing rates of late-stage cancer diagnosis (Cockburn, 2011).
Despite the wealth of research that has been conducted as a result of the availability of these registries, limitations do exist. For example, SEER data are limited to patient demographics, diagnosis, initial treatment, and mortality outcome; the SEER-Medicare database is needed for studies on comorbidities, long-term treatment, or health status over time (Warren et al., 2002). Other than broad categories of initial treatment, the SEER-Medicare database is not able to capture data on services that are not covered by Medicare (e.g., long-term care), and the Medicare claims data that SEER links to do not include individuals who receive care through health maintenance organizations (HMOs). The SEER-Medicare database also does not provide adequate data on cancers that occur primarily in younger populations (e.g., leukemia, testicular cancer) (Warren et al., 2002). At the state level, challenges for registries include data inaccuracies or misclassifications (e.g., race/ethnicity), duplicate reporting and multiple diagnoses in the same patient, and reporting delays (Izquierdo and Schoenbach, 2000).
Alzheimer’s disease South Carolina developed a comprehensive registry in 1988 to collect data on diagnosed cases of Alzheimer’s disease and related disorders. This registry links multiple data sources—including records from hospitals, emergency departments, long-term care settings, and memory clinics—with mental health and vital records as well as Medicaid data. These data are used to track and estimate prevalence, including by specific population groups, and to plan resource allocation (University of South Carolina Arnold School of Public Health, 2010, 2011). Other states, including West Virginia (West Virginia University, 2011) and New York (New York State Department of Health, 2004), have also developed registries for Alzheimer’s disease and other dementia disorders. Each of these registries has limitations. The South Carolina registry is voluntary, and there may be underreporting due to failure to capture a diagnosis of dementia in the coded data (e.g., because other health conditions were the focus of the health care visit) (Sanderson et al., 2003). The New York registry captures only data from inpatient hospital stays and nursing homes, and the quality and completeness of the coded data are unknown (New York State Department of Health, 2004, 2006). The West Virginia registry recently completed a pilot test (West Virginia University, 2011); analyses of its value and limitations should be conducted after it is implemented across the state.
Autism spectrum disorders A number of state-based registries devoted to autism spectrum disorders surveillance have been developed in recent years, including in Delaware, New Hampshire, and New Jersey. These states have passed legislation that requires reporting of autism spectrum
disorders by health professionals in order to better understand the incidence and geographic distribution of cases and to assist with planning for resource allocation (Delaware Health and Social Services, 2011; New Hampshire Department of Health and Human Services, 2011; New Jersey Department of Health and Senior Services, 2007). Evaluations of the completeness and quality of the data collected by these mandatory registries will be needed as they are developed. On a national level, the Kennedy Krieger Institute has developed the Interactive Autism Network (IAN), a voluntary online registry that includes more than 41,000 participants and collects data on family history, environment, and treatment, which may allow for exploration of potential causes and of diagnosis and treatment options (Kennedy Krieger Institute, 2011). Lee and colleagues (2010) reported that using IAN’s parent-reported data was a reliable method of case ascertainment; however, the web-based registration may introduce bias in the representativeness of the population covered. In April 2011, the Mental Health Research Network announced the development of a new autism spectrum disorders registry that will include 20,000 children and adolescents from 5 health care organizations in Boston, Northern and Southern California, Oregon, Washington, and Georgia (Kaiser Per-manente, 2011).
Summary The experiences of registries specific to other conditions offer some insights for surveillance of epilepsy. Standardization of data collection, including definitions and coding, is essential. To successfully achieve this goal, collaborations such as NAACCR are critical. Further, linkages across data sources, such as between registries and Medicare claims data, offer opportunities to understand cost and patterns of health service utilization, and centralized databases facilitate data compilation and processing. Successfully establishing and operating a number of these registries has depended on legislative support at the national and state levels, which provides funding and requires reporting. However, there are also limitations to these registries; for example, they may not be comprehensive, and the quality of their data may be hindered by inaccurate or incomplete coding. The existing infrastructure of registries focused on conditions such as cancer, Alzheimer’s, and autism spectrum disorders offers an opportunity to capture data on people who have one of these conditions and also have epilepsy. This could both expand available epilepsy data and offer a better understanding of the relationship between epilepsy and its comorbidities. In addition, further exploration is needed to determine the value and limitations of alternative ways to collect valid self-reported data, such as through online databases (e.g., IAN, PatientsLikeMe.com) (Wicks et al., 2012) and possibly through self-management tools (e.g., My Epilepsy Diary) (Le et al., 2011).
Data from Visits to Health Care Providers
Administrative datasets are collected from medical records of health care providers and claims files of insurance companies that were generated in the course of managing, paying for, or monitoring the provision of health care services. Health encounters create claims for payment, and public and private health care providers and insurance plans collect these claims data and include them in their own administrative databases. Additionally, birth and death records serve administrative purposes by creating legal records. Common administrative data sources include national and state hospital discharge data; Medicare, Medicaid, and private insurance claims data; and vital statistics (i.e., birth and death records).
Hospital discharge data Since the 1980s the federal government has required submission of uniform data on all acute hospital inpatient discharges paid through Medicare and Medicaid (Kanaan, 2000). In 2010, a total of 48 states had systems for reporting hospital discharge data, many of which included statewide all-payer, all-patient data on inpatient hospital stays (Love et al., 2010). Recently, trends toward increasing use of outpatient care has led 32 states to collect data from ambulatory treatment centers and 30 states to include data from emergency department visits. Hospital discharge data are population based and can be used for analyses that examine patient demographics, use of codes for diagnosis and treatment, hospital service use, and total costs (Love et al., 2010). The data typically contain diagnosis, treatment, and cause-of-injury codes for each admission or visit; unique personal identifiers can be used to link admissions and visits to specific individuals for determining admission type, length of stay, acute care charges, primary and secondary procedures, sources of payment, and discharge disposition (Iezzoni, 2003).
Hospital discharge data are relatively easy to obtain from the state agencies that maintain the database. The Nationwide Inpatient Sample (NIS), maintained by AHRQ, contains a 20-percent randomly stratified sample of all discharges from U.S. community, nonfederal hospitals (AHRQ, 2011b). As the United States’ largest all-payer hospital care database, the NIS collects data from about 1,000 hospitals, resulting in data on approximately 8 million hospital stays each year. Hospital discharge data can be used in combination with other data sources for a range of analyses, such as estimates of age- and race-specific hospital admission rates for people with epilepsy and of disparities in surgery (CDC, 1995; McClelland et al., 2010; Szaflarski et al., 2006).
Hospital discharge data have important limitations. Hospitalization datasets do not include actual payments to the health care facility, nor do
they collect data on the majority of pharmacy services or ambulatory care services provided outside of hospitals (Love et al., 2010), and these categories, when combined, represent a significant portion of expenditures for epilepsy care (Begley et al., 2000). Validating the data obtained from hospital discharge databases is rarely possible, and coding errors and diagnostic misclassification that result in over- or underdiagnosis are known to occur for epilepsy and other conditions (Andaluz and Zuccarello, 2009; Baaj et al., 2008; Huang et al., 2011). Even when accurately coded, the diagnoses available in such datasets provide limited clinical information and are not sufficient to determine the type of epilepsy or its severity (Kaiboriboon et al., 2011). Finally, costs of hospitalization can only be approximated by applying hospital cost-to-charge ratios to hospital charges obtained from discharge data or by applying Medicare payment rates to hospital stays (Drummond et al., 2005).
Claims data In the process of providing public (e.g., Medicare, Medicaid, Children’s Health Insurance Plan) and private (e.g., Blue Cross Blue Shield, United Health, CIGNA) health insurance coverage and paying providers, fiscal intermediaries collect large quantities of data. Many of the data elements that are included in hospital discharge data also are included in claims data for every covered visit or service, including demographic information, dates of service, service type, diagnosis and treatment codes, charges, and payments.
Claims data are particularly useful because they may include information on a comprehensive set of services, including hospital, physician, and medication use, which can be linked to de-identified individuals to track cases, service use patterns, and costs over time. Because these datasets are often large and cover many people and services (Iezzoni, 2003), they can be used for studies of people with epilepsy and even, in some instances, for studies comparing incident versus prevalent cases or subgroup analyses of different demographic groups or types of epilepsy. Claims data have been used recently to study the use and cost of care for people with epilepsy, medication adherence, and the impact of adherence on health care use and costs (Davis et al., 2008; Faught et al., 2009; Griffiths et al., 1999; Ivanova et al., 2010). HMO claims data have been used to study incidence and variation in the use and cost of care by seizure type and frequency (Begley et al., 2001). Medicare data are useful for studying specific populations, such as older adults with epilepsy, and studies have been conducted using this dataset to look at costs, disparities in care, and use of seizure medications (Bond and Raehl, 2006; Christian-Herman et al., 2004; Hope et al., 2009; Pugh et al., 2010). These and other claims-based studies have been useful in identifying the major medical services that contribute to the cost of epilepsy care and analyzing how the medical cost burden is distributed
The Health Care Cost Institute is a recently formed partnership among Aetna, Humana, Kaiser Permanente, and UnitedHealthcare to provide data for surveillance and research on health care costs and service use. Launched in September 2011, these health systems formed a database that covers claims from 5,000 hospitals and more than 1 million health care providers from 2000 to the present; it includes 5 billion claims and $1 trillion in costs. This database will be updated on a regular basis, and the institute will conduct research on its data to identify trends in costs as well as making the data available to independent researchers.
SOURCE: Health Care Cost Institute, 2011.
across individuals. Emerging efforts such as the Health Care Cost Institute stem from the cooperation of different health systems to share claims data for improved surveillance of cost and service use trends (Box 2-6).
As valuable as these data are for surveillance and research purposes, they have important limitations. Claims data provide no information on populations lacking health insurance coverage or those who avoid care because co-pays and deductibles are too expensive. Without all-payer claims data (Box 2-7), analyses of where patients receive health care if they change their type of insurance coverage are not possible (Love et al., 2010). As with hospital data, accurate case identification is difficult for several reasons: ICD-9-CM codes are not consistently applied or sufficiently detailed, and treatment codes are complex and may be prescribed for other conditions besides epilepsy. In addition, the various methods used to identify cases and services are infrequently validated and the representativeness of the population samples for which data have been obtained has not been confirmed. Shatin and colleagues (1998) found variations in service use patterns between children with epilepsy who have Medicaid and those who have employer-based insurance and emphasized the need to look at data
All-payer claims databases (APCDs) are state-based resources that aim to collect comprehensive claims data. Some states have mandated reporting while others are voluntary. One goal of the APCDs is to help standardize the reportable data elements to enable comparisons across payers. These databases provide data on a range of measures, including costs, quality of care, service use, access, and barriers to care (Love et al., 2010). Nearly two-thirds of states currently have APCDs or are evaluating their feasibility (APCD Council, 2011).
from multiple sites to ensure the representativeness of the study population. Studies to validate the identification of people with epilepsy and the services they receive are needed, as is a closer look at the representativeness of epilepsy populations for which claims data are available.
Vital statistics The U.S. Standard Certificate of Live Birth has a section for any “abnormal conditions in the newborn,” which includes a line item for “seizure or serious neurologic dysfunction” (CDC, 2003b). However, this line refers to neonatal seizures, which are not generally considered to be epilepsy (ILAE, 1993). Birth certificates are not a source for data collection on the epilepsies other than as the means to capture data on major congenital malformations, which could be linked to the mother’s use of seizure medications, as described in the pregnancy registries section above.
As a father, I had to tell the coroner what my son’s cause of death was. His response made it very clear that he was not familiar with SUDEP [sudden unexpected death in epilepsy], didn’t know what the term meant. When I explained what it was, he said, “Oh, we’ve had three or four similar cases in Boulder County in the past year.” The clear implication is that SUDEP is vastly under-reported.
Accurate death certificates that capture data on mortality in people with epilepsy are necessary to monitor trends in the overall mortality, identify risk factors, and estimate the incidence of cases where epilepsy may have contributed to, or caused, death, including instances of SUDEP. Epilepsy must be entered somewhere on the death certificate in order to accomplish these goals (Antoniuk et al., 2001). However, Bell and colleagues (2004) examined UK death certificates and found that epilepsy was recorded for only 7 percent of the people who had epilepsy, with more frequent recording among people who had frequent seizures. Currently in the United States, death certificates include cause of death (Part I) and “significant conditions contributing to death” (Part II in the United States), but not a full medical history (CDC, 2003a). The CDC provides national mortality data to researchers, and these data can be requested by underlying cause of death, which are categorized by ICD-9 or ICD-10 codes, depending on the time frame of the study (CDC, 2011b). In the United Kingdom, Goldacre and colleagues (2010) found that the underlying cause of mortality was listed as epilepsy in less than half of cases with epilepsy on the death certificate; thus, mortality rates for epilepsy that are based on one cause of death only, and not also on “significant conditions contributing to death” (Part II), are likely to be underestimates.
Current estimates of SUDEP incidence based on death certificates are inadequate for several reasons. First, there is no specific code for SUDEP in ICD-9 or ICD-10, which may contribute to underdiagnosis and a lack of
awareness of the problem (Hitiris et al., 2007; Lathers et al., 2011b). Second, in some cases, the cause of death may be inappropriately recorded; for example, a review of death certificates with cause of death listed as status epilepticus found that nearly half of the cases were actually SUDEP (Langan et al., 2002).12 Hanna and colleagues (2002) reported that 41 percent of autopsy reports, which are used to inform death certificates, inadequately documented epilepsy-related causes of death.
Third, there is a lack of awareness about SUDEP among coroners, medical examiners, and others who fill out death certificates (Devinsky, 2011; Lathers et al., 2011a). SUDEP may be under-reported due to the misconception that seizures do not have fatal consequences (Nashef and Sander, 1996; Schraeder et al., 2006). Coroners (who are not necessarily medically trained) are often unaware of SUDEP as a major cause of death in epilepsy (Leestma, 1997). Recognition of SUDEP as a valid diagnosis is more likely among trained pathologists compared to those without training in pathology or medicine (84 versus 63 versus 58 percent); seeing some epilepsy cases per year and having higher autopsy rates are also linked to greater recognition (Schraeder et al., 2006). However, Schraeder and colleagues (2006) found that SUDEP was used as a final diagnosis in few of the cases where it was appropriate, even among those who recognized SUDEP as a valid diagnosis. Instead, the cause of death was often attributed to status epilepticus, fatal seizure, respiratory failure, or cardiac arrhythmia. Educational efforts should focus on providing information on SUDEP to coroners and medical examiners (Schraeder et al., 2006) to improve the reliability of death certificate data. To inform these efforts, additional research is needed on how SUDEP is used as a diagnosis in the United States.
Surveillance of SUDEP is difficult because cases are ascertained using a variety of definitions, source populations, and data sources, including death certificates and autopsy records (Tomson et al., 2005, 2008). Complete ascertainment of the incidence of SUDEP can be achieved only through autopsies in order to exclude other definite causes of death (Antoniuk et al., 2001; Lathers et al., 2011a; Schraeder et al., 2006). Further, although detailed and accurate autopsies may improve understanding of SUDEP, currently there is no mandatory autopsy requirement (Schraeder et al., 2006). In addition, there is no national standard in the United States for documenting conditions at the time of death (e.g., where, body position) or for deciding whether to perform an autopsy (Schraeder et al., 2006).
12Status epilepticus is usually defined as an extended seizure or a series of seizures where consciousness is not regained in between, and it occurs in people with and without a diagnosis of epilepsy (Bazil and Pedley, 2009). In contrast, SUDEP is defined as a “sudden, unexpected, witnessed or unwitnessed, nontraumatic and nondrowning death, occurring in benign circumstances, in an individual with epilepsy, with or without evidence for a seizure and excluding documented status epilepticus” (Nashef et al., 2012).
To accurately count the number and distribution of SUDEP cases, to determine its cause, and—ultimately—to seek opportunities for prevention, more accurate forensic data are needed. Achieving these objectives will require standard criteria to define SUDEP and standard protocols for autopsies (So et al., 2009). Verbal autopsy—in the form of information from family and friends of the deceased about the circumstances of death—may add to understanding of SUDEP (Aspray, 2005). In 2008, the NINDS hosted a workshop on SUDEP and participants identified the need for standardized autopsy protocols (Hirsch et al., 2011), and in 2010, the NINDS solicited applications for collaborative research on SUDEP, including on approaches to “standardized procedures for collecting postmortem tissue and clinical data” (NINDS, 2010). If standardized reporting to a registry or other mechanism were required, coroners and medical examiners would be held accountable for knowing about and using SUDEP as a diagnosis.
Summary Though not created for surveillance and research purposes, administrative data—including data from hospital discharges, reimbursement claims, and vital statistics—may include sufficient details to provide information on the incidence and prevalence of epilepsy, the amount and cost of services that patients receive, the characteristics of people who receive services, and mortality patterns. Administrative data offer important advantages because they include large numbers of people, employ service and diagnostic coding that can be used to identify people with epilepsy, permit the tracking of people over time, and follow standardized federal and/or state regulations to ensure comparable content is collected among the health care systems. Administrative data provide information on patterns of care in real-world practice that may be more generalizable than those observed in clinical trials, where study subjects may not be typical of patients in actual practice settings. Since administrative data are collected for purposes other than research, they are relatively inexpensive to obtain and can be manipulated to examine various surveillance questions.
However, these datasets also have several limitations for surveillance and research purposes. For example, the use of service and diagnostic coding to identify cases is problematic. As discussed earlier, the accuracy of the coding often has not been verified and may not provide sufficient detail to determine the type and severity of epilepsy, the types of services received, the outcomes of care, or whether death was attributable to epilepsy. The validity of administrative data depends on the quality and consistency of record keeping among the many providers submitting the data. There may be difficulty in linking and comparing data across sites, populations (e.g., insured and uninsured), provider types, and systems of care.
Moving forward, the increasing use of EHRs (discussed below) and electronic systems for the capture of discharge and claims data will enable
more timely and efficient retrieval of data from each health care facility or insurance provider. Additional opportunities involve emerging collaborations, such as the Health Care Cost Institute and all-payer claims databases, as well as efforts to improve knowledge about and protocols for evaluating SUDEP and other epilepsy-related deaths. However, as noted throughout this chapter, standardized methods for recording information in these databases will be critical if these data are to be used for broad surveillance of epilepsy.
Retrospective use of clinical data In addition to the data that providers collect for billing and administrative purposes, researchers can also retrospectively review clinical data that are collected and recorded as part of the patient-provider interaction, through such methods as chart reviews, in an attempt to systematically glean information related to a condition or its treatment. Such data permit the identification of probable cases of epilepsy, and studies have used clinical records to investigate a variety of epilepsy-related topics, including incidence, prevalence, cause of death, health outcomes, and cost-effectiveness (Annegers et al., 1999; Knoester et al., 2005; Mohanraj et al., 2006; Ojemann et al., 1987; Parko and Thurman, 2009). Limitations of surveillance and research using clinical data include many that are similar to those discussed for administrative data, including a general absence of validation of various case ascertainment algorithms and service use and outcome measures. When using retrospective data, coding inaccuracies and missing data can make it hard to identify people with epilepsy and determine their characteristics. Additionally, the coding may not include seizure type and syndrome, particularly for records from visits to health care providers who do not specialize in epilepsy. Since patients may seek care from more than one provider, identification of incident epilepsy can be difficult if databases from different providers are not linked and multiple records for an individual reconciled.
Prospective use of clinical data Data from clinical settings can also be collected prospectively to investigate aspects of a particular condition or treatment plan. Prospective studies afford the opportunity to screen for possible cases of epilepsy and then validate the diagnosis using standardized or semi-structured interviews, which provide far greater detail about seizures than the typical medical record. Patients may be screened from hospitals, neurologists’ offices, primary care settings, long-term care facilities, and other care settings for studies of epilepsy incidence. These studies provide the opportunity to interview people with epilepsy and follow them for a discrete time period to monitor a range of outcomes, including health
status, quality of life, quality of care, and mortality. Although prospective studies are more expensive and take more time than retrospective ones, they have advantages in their ability to generate rich and comparable data on an array of questions about epilepsy, including incidence, comorbidities, pregnancy outcomes, refractory epilepsy, health outcomes, cause of death, and stigma (Benn et al., 2008, 2009; Berg et al., 2006; Danielsson et al., 2005; Leaffer et al., 2011; Meador et al., 2009; Perucca et al., 2011; Viinikainen et al., 2006). Prospective studies in epilepsy centers, such as Friedman and colleagues’ (2010) study of seizure-related injuries, may be especially useful for the collection of data on more severe or chronic epilepsy.
Prospective ascertainment of epilepsy data also faces challenges. Identification of subjects can be costly, involving active screening of several sources of care to make a preliminary identification of a sufficient number of potential cases, letters sent to potential cases inviting study participation, telephone calls to screen potential cases, lengthy interviews, and other data collection to confirm an epilepsy diagnosis. End points for follow-up must be carefully selected to maximize the information that can be obtained from medical records. Furthermore, losses to follow-up can limit the representativeness of the study population.
Electronic health records As repositories for both administrative and clinical data, the adoption and expanded use of linkable EHRs will enhance the utility of these data for public health surveillance of the epilepsies. A report of the President’s Council of Advisors on Science and Technology (PCAST, 2010) examined how health information technology, and specifically EHRs, could improve the quality of health care and reduce costs. The PCAST report concluded that information technology has the potential to facilitate surveillance of public health trends if a standardized infrastructure and language for health information are implemented. However, the council reported that, despite great promise, significant progress is needed to achieve integrated electronic health information and exchange. For example, only about one-third of office-based physicians have systems that meet the defined criteria for basic EHR capability, although this number is increasing (e.g., the number rose from 11 percent in 2006 to 34 percent in 2011, with about half of physicians using some form of EHR as of November 2011) (HHS, 2010b; Hsiao et al., 2011). Barriers to the use of EHRs for surveillance identified in the PCAST report include the following:
• EHRs are typically owned by vendors who have proprietary in-terests, which may lead to barriers in implementing standard data formats (in addition to the technical challenges) and participating in health information exchange.
• Health care organizations may view EHRs as internal resources and may be reluctant to enable external uses of the data, such as making them available in de-identified or aggregated formats for public health agencies and researchers.
• Concerns about privacy and data security may cause individuals to be uncomfortable with giving consent for their EHRs to be used in research (PCAST, 2010).
However, the council report also highlighted the successes of organizations such as Kaiser Permanente and the Veterans Health Administration (VHA) in implementing EHRs to improve care and emphasized the potential value of EHRs in providing large quantities of data in a timely manner for surveillance and research (PCAST, 2010).
The Health Information Technology for Economic and Clinical Health Act of 2009 was created to help overcome these and other barriers by authorizing $27 billion in funds for the Centers for Medicare and Medicaid Services to use as incentive payments to health care providers to promote the adoption and use of EHR technologies (Blumenthal, 2011). The incentive payments require “meaningful use” of EHRs, which means that health care providers must demonstrate that they are using certified EHRs that enable them to monitor data and use them for quality improvement (CMS, 2011). As part of the meaningful use process, one of the priority outcomes is ensuring that adequate privacy protections are in place for personal health information (CMS, 2010a). The value of this effort to implement compatible EHRs nationwide may be to enable much more accurate estimates of disease and disorder rates in the population, patterns of care and their outcomes, and treatment costs.
As noted by Tyler and colleagues (2011), EHRs are a cost-effective way to study a specific health condition, and they can enable improved monitoring of care for people with chronic health conditions (Baldwin, 2011). Charlton and colleagues (2011) reported that EHRs have some advantages over registries, including the potential for better follow-up and—since they do not rely on voluntary enrollment—greater representativeness. Additionally, VanWormer (2010) looked at the Heart of New Ulm Project, a possible model for EHR-based surveillance, and found that EHR-based estimates of coronary heart disease risk factors are in line with manually derived estimates. In that project, risk factors for coronary heart disease are derived from EHR data and reviewed annually over 10 years (VanWormer, 2010). Another model for EHR-based surveillance is the Department of Veterans Affairs’ (VAs) Cardiovascular Assessment, Reporting and Tracking (CART) System for tracking cardiovascular disease in real time (Box 2-8). Significantly, several ongoing and emerging collaborative efforts are focused on sharing EHR data to enhance surveillance and research opportunities (Box 2-9).
The Department of Veterans Affairs (VA) recognized the limitations of retrospective studies using administrative and clinical records—they had electronic health records (EHRs) and a registry for veterans with implantable defibrillators, but data were often in free text, and analysis required significant labor resources and time. In response, the VA established the CART (Cardiovascular Assessment, Reporting and Tracking) System, where data collection is integrated into the care process through the EHR, which allows for treatment and real-time surveillance of cardiovascular disease. The reports are standardized and completed at the time of care. To make this possible, collaborations between the relevant players (e.g., VA Offices on IT [Information Technology], Patient Care Services) were crucial. CART enables quality of care and patient safety reviews along with disease surveillance (Varosy, 2011).
The Health Maintenance Organization Research Network (HMORN) is a collaboration of 19 HMOs—all of which have electronic health records (EHRs)—that links hundreds of researchers and includes multicenter research projects. The HMORN holds an annual meeting and also convenes smaller committees and forums to discuss research and potential studies and methodologies, including data coordination, best practices, and operational strategies (HMORN, 2012a,b). One central feature of the HMORN is its Virtual Data Warehouse (VDW), where data remain at the original site but the VDW facilitates comparison of data between sites (HMORN, 2010).
Building on the successes of the HMORN, and through Common Fund support from the National Institutes of Health, a Health Care Systems Research Collabora-tory is being formed to facilitate collaborative research across U.S. health care systems (NIH, 2012; Van Den Eeden, 2011). Participating organizations represent integrated health care systems with EHRs and linked biospecimen repositories. The goal of the collaborator is to use the organizations’ data and operational infrastructure to facilitate longitudinal studies across multiple sites, including large-scale epidemiologic studies and prospective observational studies, as well as randomized clinical trials (NIH, 2011a). Planning for this work is still under way, but this program may offer valuable opportunities for future epilepsy surveillance.
Regional health information organizations (RHIOs) aim to support health information exchange, one of the eligibility requirements for Centers for Medicare and Medicaid Services incentive payments for EHR meaningful use. RHIOs are organizations that coordinate the exchange of data in a region (e.g., city, state). The number of RHIOs has increased over the past few years, but health care provider participation rates vary as do the RHIOs’ ability to facilitate robust health information exchange. Infrastructure is still being developed to allow interoperability (Adler-Milstein et al., 2011).
All of these efforts point to opportunities for epilepsy surveillance through the use of EHRs. Considerable improvements must be made to standardize EHRs for valid health information exchange across providers, but the federal government’s investment in this process is helping to move these efforts forward. To determine the usefulness of EHRs for epilepsy surveillance, pilot projects that validate methods for case ascertainment, including look-back periods for incident cases, and service use will be necessary. Furthermore, strategies should be explored to determine the appropriate balance of coded versus free-text data collected in EHRs—searchable by code or natural language processing—to maximize both efficiency and the data available for surveillance and research. As noted elsewhere in this chapter, collaborations will be important to minimize costs and ensure interoperability.
Surveillance That Includes Linked Data Sources
The concept of records linkage was first formulated by Dunn (1946) to describe the combination of multiple sources of health information into a single file for each individual in a population from birth to death. As described below, in some populations it has been possible to link clinical records and administrative data across hospitals, practitioners, and payers, permitting ascertainment of epilepsy and reasonable follow-up for end points, such as number and type of contacts with the health care system or death. Linkage is not always perfect, particularly when a patient has more than one medical record number at the same facility or when date of birth, gender, or ZIP Code are missing (Bradley et al., 2010). These problems may lead to a high false-negative rate in the records linkage system. Also, records linkage systems may suffer from a high false-positive rate if records linked together do not belong to the same patient (Bohensky et al., 2010). Although there are a number of challenges to establishing EHRs systematically, they can help to link multiple kinds of data for individuals within and across health care systems moving forward.
The major example of records linkage in epilepsy is the Rochester Epidemiology Project, where records for Minnesota residents of Rochester, Olmsted County, and the region around Olmstead County have been centralized. The system includes medical records from private physician offices, hospitals, and nursing homes, as well as death records. Numerous studies have been conducted on epilepsy using the Rochester data (e.g., Annegers et al., 1995, 1996; Begley et al., 2001; Ficker et al., 1998; Hauser et al., 1991, 1993; Hesdorffer et al., 1996a,b, 2011). An advantage of Rochester’s records linkage system was that it enabled studies of epilepsy incidence and other attributes over a number of decades to allow analysis of trends. Another advantage of records linkage is that the cost of ascertaining and
following cases is much less than that of prospective studies that must examine multiple unlinked data sources (Bradley et al., 2010). One challenge in this type of study is identifying a population that reflects the diversity of the U.S. population.
Results of a Data-Gathering Effort
To better understand the opportunities and barriers to broad surveillance of the epilepsies, the Institute of Medicine committee requested that several health care systems (Henry Ford Health System, Geisinger Health System, and the VHA) and one state’s records linkage system (South Carolina Epilepsy Surveillance System [SCESS]) explore a list of surveillance questions for their populations and analyze the strengths and limitations of their systems to generate information about epilepsy (Appendix B). Researchers in each system generously responded to the committee’s request and provided candid evaluations of their system’s ability to capture data on epilepsy. Unfortunately, variability in the methods of these systems meant that the data were not comparable, but looking at each system individually is informative about the current state of surveillance capabilities in the United States and highlights some important lessons for future surveillance. While these systems have limitations, they offer a preview of the wealth of opportunities that records linkages and EHRs could offer for epilepsy surveillance in the future.
Michigan’s Henry Ford Health System is a large health system that includes 6 hospitals, more than 30 ambulatory care centers, and more than 2,000 physicians. Its managed care plan has approximately a half-million members. Henry Ford also has a Comprehensive Epilepsy Program that provides specialty care for people with epilepsy from the metropolitan Detroit area and the surrounding regions. Using administrative data and its EHR, Henry Ford researchers were able to estimate incidence and prevalence of epilepsy and comorbidities in their population using ICD-9-CM codes; service use, patterns of care, and care settings were identified as well. Strengths of the Henry Ford Health System to inform surveillance of epilepsy include that it has a comprehensive record of all paid claims for individuals in its Health Alliance Plan. Analysis of this cohort can identify incident cases, cases with comorbidities, and the comprehensive set of services used by an individual. However, Henry Ford’s population is not representative of the U.S. population, further validation is needed to ensure accurate estimation of incidence and prevalence, and validation of the algorithms used to identify comorbidities and use of health care services is necessary.
Pennsylvania’s Geisinger Health System includes 37 community practice sites and more than 1,800 clinical staff serving approximately 2.6
million people (Geisinger Health System, 2011). Geisinger also has an epilepsy center that provides specialty epilepsy care (Geisinger Health System, 2012). For the last decade, Geisinger has used a central electronic repository that integrates data from all clinical visits, laboratory reports, and claims. These data were used to estimate incidence and prevalence of epilepsy, comorbidities, and health service use using ICD-9-CM codes. Strengths of the Geisinger Health System to inform surveillance of epilepsy include that its EHR is comprehensive and contains a multiple-year period of look-back data to establish incidence, and it provides data on a largely rural population. However, the Geisinger population is not representative of the U.S. population, its algorithms for case ascertainment and service use have not been validated, and its incidence and prevalence estimates are likely overestimates due to the inclusion of ICD-9-CM code 780.09.13
The VHA runs the nation’s largest integrated health care system, with more than 53,000 health professionals at 152 medical centers and almost 1,400 clinics, community centers, and other settings providing care to more than 8.3 million veterans (VA, 2011). The VHA’s EHR encompasses care provided by VA hospitals, outpatient clinics, nursing homes, and other facilities, as well as services rendered by non-VA providers if VA funds are used for payment. These care data are sent to a central repository and linked with a patient identifier. Previous VA studies successfully linked epilepsy data from multiple VA databases, including an investigation of the impact of epilepsy on health status (Pugh et al., 2005) and an analysis of trends in seizure medication prescriptions among older adults with newly diagnosed epilepsy (Pugh et al., 2008). In the current data-gathering effort, diagnosis codes, dates and location of care visits, and data on prescribed medications were analyzed to provide estimates of incidence, prevalence, comorbidities, and service use for two populations: veterans 65 years old and older and veterans from Afghanistan and Iraq. Strengths of this system for surveillance of the epilepsies are the comprehensive, linked nature of the data repository and that many of the algorithms have been validated for comorbidities and service use, as well as for incidence and prevalence estimates in the older veterans cohort. However, a look-back period of more than a year would help rule out the possibility of overestimation of incidence. A limitation is that incidence and prevalence in the Afghanistan and Iraq cohort may be overestimated due to the high prevalence of posttraumatic stress disorder in this population, which is strongly associated with seizure-like events with a psychological basis that may be misdiagnosed as epilepsy. Also, there may be care received outside the VA that is not included in these estimates; the extent to which this would affect the results is unknown.
The SCESS, funded by the CDC, was formed in 2002. Collaboration has been critical to its successful acquisition of data from a variety of sources. The SCESS collects and links claims data on privately insured individuals, those insured through the State Employee Insurance Program, and Medicaid and Medicare beneficiaries; incorporates hospital admissions data, including emergency room visits; and has access to medical chart data in some hospitals and doctors’ offices. The data are collected and housed by the South Carolina Budget and Control Board’s Office of Research and Statistics, which assigns a unique identification number to each individual to allow linkage across data sources. A review of the clinical records was conducted in the initial funding cycle to validate the information obtained from the data sources. In the current data-gathering effort, incidence and prevalence were estimated, as well as comorbidities and services use. Strengths of the SCESS include its use of unique identifiers that enable accurate linkages, its ability to analyze cost and effects of services through its collection of costs by type of service and procedure, and that it is a passive surveillance system, which minimizes cost. However, while the SCESS is representative of the state’s civilian population, it does not include people in the military or veterans, and the accuracy of codes for specific types of epilepsy is undetermined.
These multisource surveillance systems permit reasonably complete case ascertainment in their populations and identification of fairly comprehensive service use, and they allow longitudinal follow-up of individuals and trend analysis. Problems with records linkage arise if individuals are counted twice or not at all due to incorrect matching of records in case ascertainment; further, the diagnostic and treatment codes used may not be accurate. The verification of the codes through cross-checking with other data sources makes the multisource approach very powerful for surveillance and research. The expanded use and adoption of linkable EHRs will enhance the opportunity for linked data sources in the future, and validation studies can confirm the methodologies and results. Linked surveillance systems have the potential to be invaluable resources for policy making, allocation of service resources, and prevention efforts.
The FDA’s Sentinel Initiative
In 2008 the FDA announced its Sentinel Initiative, which includes the creation of an electronic system that will conduct national surveillance to monitor the safety of FDA-regulated medical products (e.g., drugs, biologics, medical devices). A year later, the Mini-Sentinel, a 5-year pilot project, was started to develop and evaluate methods to capture these data across a variety of electronic sources (e.g., claims data, EHRs, registries) (Behrman et al., 2011). Challenges and barriers encountered during the pilot project
will inform the implementation of the full-scale Sentinel System. One key aspect of the Mini-Sentinel is the importance of collaboration between the FDA and other institutions, which provide access to health data and also contribute expertise in the development of the system (FDA, 2012); after 2 years, the Mini-Sentinel project includes participation from more than 30 academic and private institutions (Platt et al., 2012). Another critical feature of the Mini-Sentinel is its use of a “distributed data system,” where each collaborating institution has control of its own data, which may ease some privacy and proprietary concerns. As the Mini-Sentinel project, and the broader Sentinel Initiative, continue to evolve, relevant experiences and lessons learned should inform epilepsy surveillance efforts. Additionally, opportunities to collect epilepsy-related data (e.g., seizure medications, adverse events) should be explored when the full system is established.
The Autism and Developmental Disabilities Monitoring Network
An example of records linkage for surveillance of autism spectrum disorders, a comorbidity of epilepsy, is the Autism and Developmental Disabilities Monitoring (ADDM) Network. The ADDM Network evolved from previous CDC developmental disabilities surveillance efforts when the Children’s Health Act of 2000 was enacted, which provided the CDC with the authority to fund autism spectrum disorders surveillance across the country (CDC, 2011c; Yeargin-Allsopp, 2011). The ADDM Network uses standard methodologies to examine prevalence trends over time, prevalence across geographic regions, and characteristics of children with autism spectrum disorders (CDC, 2011a; Yeargin-Allsopp, 2011). To study the peak prevalence of autism spectrum disorders, the ADDM Network focuses on children who are 8 years of age. Cases are identified through a retrospective review of records from a variety of health and education sources, such as pediatric hospitals and clinics, diagnostic centers and other clinical settings, and schools. The review collects testing, developmental, and behavioral data, and identified cases are validated through clinician review (Yeargin-Allsopp, 2011). A number of studies have been published using data from the ADDM Network (CDC, 2010a). However, the ADDM Network does not include data on children who are home-schooled or who attend private or charter schools, and, like many other surveillance efforts, there are concerns about quality and completeness of the collected data (Yeargin-Allsopp, 2011).
Although a similar effort in surveillance of epilepsy would likely focus on a wider age range and collect different records (e.g., EEG results), the ADDM Network offers an example of the value of legislation in enabling a standardized surveillance mechanism. In particular, the use of educational records as a source of data to identify children with autism spectrum dis-
orders highlights a possible additional resource for epilepsy surveillance to capture data on children with both epilepsy and cognitive dysfunction (e.g., learning disorders) (Chapters 3 and 6). Critical to this effort has been the memorandum of understanding between the participating state’s Departments of Education and Human Resources to access the records. The ADDM Network then screens potential cases by looking in the educational records for a diagnosis, for Department of Education eligibility criteria, or for behavioral triggers that have been noted in the child’s clinical record (Yeargin-Allsopp, 2011). Thus, for epilepsy-related data to be collected from educational records, collaborations with the Department of Education would be critical, and clear criteria for screening these records to identify children with epilepsy and cognitive comorbidities would be necessary. Research is needed to develop appropriate criteria and screening methods and to assess the value of these records to further understanding about epilepsy when it is accompanied by cognitive comorbidities.
International Surveillance with Records Linkage
Several other countries, notably Denmark, Sweden, and Canada, have or are in the process of linking medical records across providers and administrative data from providers and payers for studies of epilepsy (Box 2-10). Significantly, these three countries have health care systems that are largely or entirely nationalized, which minimizes the variability among data sources and maximizes the representativeness of the results. Despite their different health care systems, these countries can offer lessons for epilepsy surveillance in the United States, including the importance of unique identifiers and of the collaboration needed for linking information. Further, they illustrate the common challenges faced in epilepsy surveillance, including the accuracy of codes for analyzing specific epilepsy types, syndromes, and etiologies.
EHRs, health information exchanges, and linked datasets have considerable promise for improved and cost-effective surveillance as they evolve in the years ahead. As described earlier in this chapter, the currently limited experience in obtaining comparable surveillance information from several electronic data systems demonstrates some of the challenges these systems present for epilepsy surveillance. In particular, efforts must be made to ensure that case ascertainment is complete and accurate, the length of look-back periods for determining incidence is adequate, patient mobility in and out of systems is accounted for, population representativeness is ensured, and comprehensive health care use and cost information are available and
Denmark has a number of longitudinal registries, including the National Patient Registry and the National Hospital Register, which contain more than three decades of health information, including diagnoses, treatments, and surgeries, from all patient contacts with the health care system. The Civil Registration System gives a unique identifying number to each individual, and the registries link their data using this number, enabling records linkage that avoids the false negatives and positives experienced in the United States. Based on these epilepsy-related data, studies have been conducted on a range of topics, including estimates of the effect of breastfeeding on risk for epilepsy, costs and impacts of epilepsy, risk for comorbid schizophrenia and psychosis, and risk for health outcomes including myocardial infarction, stroke, and death (Bredkjaer et al., 1998; Jennum et al., 2011; Olesen et al., 2011; Qin et al., 2005; Sun et al., 2011).
The Stockholm Incidence Registry of Epilepsy (SIRE), established in 2001, aims to identify cases with new-onset, unprovoked seizures among residents in a defined geographical area in Northern Stockholm, Sweden. This prospective registry uses multiple sources for case identification, including neurologists (both private and public), pediatricians, geriatricians, and nurses in nursing homes. Recently, SIRE data have been linked with other national registries, including the Swedish Hospital Discharge Register and the Population and Housing Census (Adelöw et al., 2011). Additional methods that help to ensure complete case ascertainment include review of all electroencephalographs (EEGs) at the central EEG lab, review of medical records for all new neuro-oncology referrals and all neurology and pediatric patients who receive their first epilepsy diagnostic code, and review of records from pediatric emergency rooms. Thus, the registry uses administrative data supplemented by more time-consuming review of records. Once all available information is obtained for a case, identified by the assigned identification number, a panel classifies the case. Studies conducted using the SIRE data have reported incident cases of unprovoked seizures and epilepsy as well as relevant risk factors (Adelöw et al., 2009, 2011).
Canada has recently undertaken a National Population Health Study of Neurological Conditions (NPSNC) to improve understanding of the epidemiology and impact of 14 neurological conditions, including epilepsy. Its aims include using linked administrative, electronic health record, and survey data to study incidence and prevalence; comorbidities; the impact of epilepsy on affected people, families, and society; health care services; and risk factors for the development of poor outcomes and other conditions. This work is made possible by the collaboration of many different federal agencies, Neurological Health Charities Canada (a collaborative effort of more than two dozen health organizations), researchers, and other stakeholders, including provincial health ministry managers. Work undertaken before the development of the NPSNC validated epilepsy coding in Canada when patients are seen in the emergency room or are hospitalized (Jetté, 2011). However, the validity of primary care data was not been previously examined and is being assessed as part of the NPSNC.
accurate. There is much to learn from the development and experiences of surveillance systems established for other purposes and conditions and in other countries, and pilot studies conducted in the near future could attempt to overcome these limitations. As part of these efforts, all privacy
concerns that may arise must receive an adequate response, and partnerships will be needed in order to improve sustainability.
A variety of data sources are currently used for epilepsy surveillance in the United States. These data sources can provide only partial estimates of many basic surveillance indicators, including epilepsy incidence, prevalence, etiologies, risk factors, comorbidities, health status, quality of life, access to care, quality of care, and cost of care. Demographic information is often inadequate, and sample sizes are generally too small to examine disparities in population subgroups. This patchwork of surveillance activity nevertheless has been mined to conduct important research on epilepsy; however, in terms of both completeness and timeliness, current data fall short of providing the information that would be most useful for understanding, planning, and guiding health care provision and policy for people with epilepsy.
Throughout this chapter, the committee has provided the basis for the research priorities and recommendations regarding improvements needed in the collection of epilepsy-related data that are detailed in Chapter 9. Improved surveillance of epilepsy will require linked electronic databases that cover large, representative populations. A crucial prerequisite for accurate and meaningful surveillance will be the validation of algorithms and methods for different age groups and settings. Standardized definitions and methods will allow surveillance data to be compared and actionable. Several opportunities may offer improved surveillance of the epilepsies, and existing examples such as those described throughout this chapter can provide useful lessons. The nationwide move to EHRs offers an unprecedented chance to capture data on epilepsy. Also, collection of epilepsy-specific data in population health surveys, registries for related conditions, and longitudinal studies will increase the amount of information about epilepsy, and the creation of a registry on epilepsy-related deaths would provide a valuable new information resource.
None of these efforts alone will accomplish comprehensive surveillance of the epilepsies, close current knowledge gaps, or adequately inform policy makers, public health agencies, health care providers, and the general public. Instead, coordinated action on multiple fronts is needed to ensure the collection of epilepsy-related data from a range of data sources. Collaboration with emerging data-sharing efforts across health care providers and with projects collecting data on related diseases and disorders will maximize resources, enable improved data collection, and, potentially,
increase momentum in advocacy efforts to fund and develop national-level surveillance, such as the National Neurological Diseases Surveillance System.14 Currently there is unparalleled change occurring within public health surveillance in terms of capability and innovation, and the epilepsy field should capitalize on the opportunity to transform knowledge about epilepsy and its burden in the United States.
Adelöw, C, E. Andell, P. Amark, T. Andersson, E. Hellebro, A. Ahlbom, and T. Tomson. 2009. Newly diagnosed single unprovoked seizures and epilepsy in Stockholm, Sweden: First report from the Stockholm Incidence Registry of Epilepsy (SIRE). Epilepsia 50(5):1094-1101.
Adelöw, C, T. Andersson, A. Ahlbom, and T. Tomson. 2011. Prior hospitalization for stroke, diabetes, myocardial infarction, and subsequent risk of unprovoked seizures. Epilepsia 52(2):301-307.
Adler-Milstein, J., D. W. Bates, and A. K. Jha. 2011. A survey of health information exchange organizations in the United States: Implications for meaningful use. Annals of Internal Medicine 154(10):666-671.
AHRQ (Agency for Healthcare Research and Quality). 2011a. Medical Expenditure Panel Survey. http://www.meps.ahrq.gov/mepsweb/ (accessed July 14, 2011).
———. 2011b. Overview of the Nationwide Inpatient Sample (NIS). http://www.hcup-us.ahrq.gov/nisoverview.jsp (accessed February 7, 2012).
Akman, C. I., M. A. Montenegro, S. Jacob, K. Eck, C. Chiriboga, and F Gilliam. 2009. Seizure frequency in children with epilepsy: Factors influencing accuracy and parental awareness. Seizure 18(7):524-529.
Andaluz, N., and M. Zuccarello. 2009. Recent trends in the treatment of spontaneous intracerebral hemorrhage: Analysis of a nationwide inpatient database. Journal of Neurosurgery 110(3):403-410.
Andersen, R., and L. A. Aday. 1978. Access to medical care in the U.S.: Realized and potential. Medical Care 16(7):533-546.
Anderson, R. C. 1976. Cardiac defects in children of mothers receiving anticonvulsant therapy during pregnancy. Journal of Pediatrics 89(2):318-319.
Annegers, J. F, W. A. Hauser, J. R. Lee, and W. A. Rocca. 1995. Secular trends and birth cohort effects in unprovoked seizures: Rochester, Minnesota 1935-1984. Epilepsia 36(6): 575-579.
Annegers, J. F, W. A. Rocca, and W. A. Hauser. 1996. Causes of epilepsy: Contributions of the Rochester Epidemiology Project. Mayo Clinic Proceedings 71(6):570-575.
Annegers, J. F, S. Dubinsky, S. P. Coan, M. E. Newmark, and L. Roht. 1999. The incidence of epilepsy and unprovoked seizures in multiethnic, urban health maintenance organizations. Epilepsia 40(4):502-506. Antoniuk, S. A., L. V. Oliva, I. Bruck, M. Malucelli, S. Yabumoto, and J. L. Castellano. 2001. Sudden unexpected, unexplained death in epilepsy autopsied patients. Arquivos de Neuro-Psiquiatria 59(1):40-45.
APCD (All-Payer Claims Database) Council. 2011. Interactive state report map. http://apcd-council.org/state/map (accessed November 21, 2011).
14 A bill to support the development of this surveillance system was passed in the U.S. House of Representatives in 2010 and reintroduced in the U.S. Senate in March 2011.
Arpino, C., S. Brescianini, E. Robert, E. E. Castilla, G. Cocchi, M. C. Cornel, C. de Vigan, P. A. Lancaster, P. Merlob, Y. Sumiyoshi, G. Zampino, C. Renzi, A. Rosano, and P. Mastroiacovo. 2000. Teratogenic effects of antiepileptic drugs: Use of an international Database on Malformations and Drug Exposure (MADRE). Epilepsia 41(11):1436-1443.
Aspray, T. J. 2005. The use of verbal autopsy in attributing cause of death from epilepsy. Epilepsia 46:15-17.
Baaj, A. A., S. R. Benbadis, W. O. Tatum, and F. L. Vale. 2008. Trends in the use of vagus nerve stimulation for epilepsy: Analysis of a nationwide database. Neurosurgical Focus 25(3):E10.
Baker, G. A. 1998. Quality of life and epilepsy: The Liverpool experience. Clinical Therapeutics 20(Suppl. A):A2-A12.
Baldwin, G. 2011. Chronic care, chronic I.T. Problems. Health Data Management 19(8): 26-33.
Bazil, C. W., and T. A. Pedley. 2009. Epilepsy. In Merritt’s neurology. 12th ed., edited by L. P. Rowland and T. P. Pedley. Philadelphia, PA: Lippincott Williams and Wilkins. Pp. 927-248.
Begley, C. E., M. Famulari, J. F. Annegers, D. R. Lairson, T. F. Reynolds, S. Coan, S. Dubinsky, M. E. Newmark, C. Leibson, E. L. So, and W. A. Rocca. 2000. The cost of epilepsy in the United States: An estimate from population-based clinical and survey data. Epilepsia 41(3):342-351.
Begley, C. E., D. R. Lairson, T. F. Reynolds, and S. Coan. 2001. Early treatment cost in epilepsy and how it varies with seizure type and frequency. Epilepsy Research 47(3):205-215.
Behrman, R. E., J. S. Benner, J. S. Brown, M. McClellan, J. Woodcock, and R. Platt. 2011. Developing the Sentinel System: A national resource for evidence development. New England Journal of Medicine 364(6):498-499.
Bell, G. S., A. Gaitatzis, A. L. Johnson, and J. W. Sander. 2004. Predictive value of death certification in the case ascertainment of epilepsy. Journal of Neurology, Neurosurgery and Psychiatry 75(12):1756-1758.
Benn, E. K., W. A. Hauser, T. Shih, L. Leary, E. Bagiella, P. Dayan, R. Green, H. Andrews, D. J. Thurman, and D. C. Hesdorffer. 2008. Estimating the incidence of first unprovoked seizure and newly diagnosed epilepsy in the low-income urban community of northern Manhattan, New York City. Epilepsia 49(8):1431-1439.
———. 2009. Underlying cause of death in incident unprovoked seizures in the urban community of northern Manhattan, New York City. Epilepsia 50(10):2296-2300.
Beran, R. G., J. Michelazzi, L. Hall, P. Tsimnadis, and S. Loh. 1985. False-negative response rate in epidemiologic studies to define prevalence ratios of epilepsy. Neuroepidemiology 4(2):82-85.
Berg, A. T., S. R. Levy, F. M. Testa, and S. Shinnar. 1999. Classification of childhood epilepsy syndromes in newly diagnosed epilepsy: Interrater agreement and reasons for disagreement. Epilepsia 40(4):439-444.
Berg, A. T., B. G. Vickrey, F. M. Testa, S. R. Levy, S. Shinnar, F. DiMario, and S. Smith. 2006. How long does it take for epilepsy to become intractable? A prospective investigation. Annals of Neurology 60(1):73-79.
Bergin, P., and the EpiNet Study Group. 2011a. The EpiNet Study Group. http://www.epinet.co.nz/ (accessed February 22, 2011).
———. 2011b. Studies currently underway. http://www.epinet.co.nz/index.cfm?PageID=6 (accessed November 22, 2011).
Bergin, P., R. Frith, E. Walker, and P. Timmings. 2007. How to get the answer to nearly everything: Using the Internet for epilepsy research. Epilepsia 48(7):1415-1417.
Bergin, P. S., T. Ip, R. Sheehan, R. W. Frith, L. G. Sadleir, N. McGrath, A. Ranta, and E. B. Walker. 2010. Using the Internet to recruit patients for epilepsy trials: Results of a New Zealand pilot study. Epilepsia 51(5):868-873.
Blum, D. E., J. Eskola, J. J. Bortz, and R. S. Fisher. 1996. Patient awareness of seizures. Neurology 47(1):260-264.
Blumenthal, D. 2011. The age of meaningful use. http://healthit.hhs.gov/portal/server.pt?open=512&mode=2&objID=3541 (accessed February 8, 2012).
Bohensky, M. A., D. Jolley, V. Sundararajan, S. Evans, D. V. Pilcher, I. Scott, and C. A. Brand. 2010. Data linkage: A powerful research tool with potential problems. BMC Health Services Research 10:346.
Bond, C. A., and C. L. Raehl. 2006. Clinical and economic outcomes of pharmacist-managed antiepileptic drug therapy. Pharmacotherapy 26(10):1369-1378.
Boyle, C. A., S. Boulet, L. A. Schieve, R. A. Cohen, S. J. Blumberg, M. Yeargin-Allsopp, S. Visser, and M. D. Kogan. 2011. Trends in the prevalence of developmental disabilities in U.S. children, 1997-2008. Pediatrics 127(6):1034-1042.
Bradley, C. J., L. Penberthy, K. J. Devers, and D. J. Holden. 2010. Health services research and data linkages: Issues, methods, and directions for the future. Health Services Research 45(5 Pt. 2):1468-1488.
Brazier, J. E., R. Harper, N. M. Jones, A. O’Cathain, K. J. Thomas, T. Usherwood, and L. Westlake. 1992. Validating the SF-36 health survey questionnaire: New outcome measure for primary care. BMJ 305(6846):160-164.
Bredkjaer, S. R., P. B. Mortensen, and J. Parnas. 1998. Epilepsy and non-organic non-affective psychosis. National epidemiologic study. British Journal of Psychiatry 172:235-238.
Brener, N. D., L. Kann, S. A. Kinchen, J. A. Grunbaum, L. Whalen, D. Eaton, J. Hawkins, and J. G. Ross. 2004. Methodology of the Youth Risk Behavior Surveillance System. Morbidity and Mortality Weekly Report Recommendations and Reports 53(RR-12):1-13.
Brook, R. H., E. A. McGlynn, and P. D. Cleary. 1996. Quality of health care. Part 2: Measuring quality of care. New England Journal of Medicine 335(13):966-970.
Brooks, D. R., R. Avetisyan, K. M. Jarrett, A. Hanchate, G. D. Shapiro, M. J. Pugh, D. R. Berlowitz, D. Thurman, G. Montouris, and L. E. Kazis. 2012. Validation of self-reported epilepsy for purposes of community surveillance. Epilepsy and Behavior 23(1):57-63.
Camfield, C., L. Breau, and P. Camfield. 2003. Assessing the impact of pediatric epilepsy and concomitant behavioral, cognitive, and physical/neurologic disability: Impact of Childhood Neurologic Disability Scale. Developmental Medicine and Child Neurology 45(3):152-159.
CDC (Centers for Disease Control and Prevention). 1995. Hospitalization for epilepsy— United States, 1988-1992. Morbidity and Mortality Weekly Report 44(43):818-821.
———. 1999. National Program of Cancer Registries cancer surveillance system rationale and approach. http://www.cdc.gov/cancer/npcr/pdf/npcr_css.pdf (accessed February 7, 2012).
———. 2003a. U.S. standard certificate of death. http://www.cdc.gov/nchs/data/dvs/death11-03final-acc.pdf (accessed September 21, 2011).
———. 2003b. U.S. standard certificate of live birth. http://www.cdc.gov/nchs/data/dvs/birth11-03final-ACC.pdf (accessed September 21, 2011).
———. 2010a. Autism and Developmental Disabilities Monitoring (ADDM) Network publications. http://www.cdc.gov/ncbddd/autism/addm-articles.html (accessed November 22, 2011).
———. 2010b. Cancer Registries Amendment Act. http://www.cdc.gov/cancer/npcr/amendmentact.htm (accessed February 7, 2012).
———. 2011a. Autism and Developmental Disabilities Monitoring (ADDM) Network. http://www.cdc.gov/ncbddd/autism/addm.html (accessed November 22, 2011).
———. 2011b. CDC WONDER: Compressed Mortality File 1979-1998 and 1999-2007. http://wonder.cdc.gov/wonder/help/cmf.html#ICD (accessed December 21, 2011).
———. 2011c. CDC’s Autism and Developmental Disabilities Monitoring (ADDM) Network. http://www.cdc.gov/ncbddd/autism/states/ADDM-network-fact-sheet.pdf (accessed November 22, 2011).
———. 2011d. Draft 2011 NHIS questionnaire: Sample child. ftp://ftp.cdc.gov/pub/Health_Statistics/NCHS/Survey_Questionnaires/NHIS/2011/english/qchild.pdf (accessed November 22, 2011).
———. 2011e. National Program of Cancer Registries: About the program. http://www.cdc.gov/cancer/npcr/about.htm (accessed November 18, 2011).
———. 2011f. Overview of public health surveillance. http://www.cdc.gov/osels/ph_surveillance/nndss/phs/files/overview.ppt (accessed February 23, 2012).
———. 2011g. United States cancer statistics. http://apps.nccd.cdc.gov/uscs/ (accessed November 18, 2011).
Charlton, R. A., J. G. Weil, M. C. Cunnington, S. Ray, and C. S. de Vries. 2011. Comparing the General Practice Research Database and the UK Epilepsy and Pregnancy Register as tools for postmarketing teratogen surveillance: Anticonvulsants and the risk of major congenital malformations. Drug Safety 34(2):157-171.
Christian-Herman, J., M. Emons, and D. George. 2004. Effects of generic-only drug coverage in a Medicare HMO. Health Affairs (Millwood) Suppl. Web Exclusives:W4-455-468.
CMS (Centers for Medicare and Medicaid Services). 2010a. 42 CFR parts 412, 413, 422 et al. Medicare and Medicaid programs; electronic health record incentive program; final rule. Federal Register 75(144):44314-44588.
———. 2010b. Medicare and Medicaid EHR incentive program: Meaningful use stage 1 requirements summary. http://www.cms.gov/EHRIncentivePrograms/Downloads/MU_Stage1_ReqSummary.pdf (accessed February 6, 2012).
———. 2011. CMS EHR meaningful use overview. http://www.cms.gov/ehrincentiveprograms/30_Meaningful_Use.asp (accessed November 21, 2011).
Cockburn, M. 2011. Lessons from cancer surveillance relevant to improving epilepsy surveillance. PowerPoint presented at the IOM Workshop on Public Health Surveillance, Population Health Research, and Data Collections for the Epilepsies, March 21. http://iom.edu/~/media/Files/Activity%20Files/Disease/Epilepsy/Panel%203-1%20Cockburn.pdf (accessed November 18, 2011).
Cohen, S. B. 2002. The Medical Expenditure Panel Survey: An overview. Effective Clinical Practice 5(3 Suppl.):E1.
———. 2003. Design strategies and innovations in the Medical Expenditure Panel Survey. Medical Care 41(7 Suppl.):III5-III12.
Coons, S. J., S. Rao, D. L. Keininger, and R. D. Hays. 2000. A comparative review of generic quality-of-life instruments. Pharmacoeconomics 17(1):13-35.
Cramer, J. A., K. Perrine, O. Devinsky, and K. Meador. 1996. A brief questionnaire to screen for quality of life in epilepsy: The QOLIE-10. Epilepsia 37(6):577-582.
Cramer, J. A., K. Perrine, O. Devinsky, L. Bryant-Comstock, K. Meador, and B. Hermann. 1998. Development and cross-cultural translations of a 31-item quality of life in epilepsy inventory. Epilepsia 39(1):81-88.
Cramer, J. A., L. E. Westbrook, O. Devinsky, K. Perrine, M. B. Glassman, and C. Camfield. 1999. Development of the quality of life in epilepsy inventory for adolescents: The QOLIE-AD-48. Epilepsia 40(8):1114-1121.
Cramer, J. A., G. A. Baker, and A. Jacoby. 2002. Development of a new seizure severity questionnaire: Initial reliability and validity testing. Epilepsy Research 48(3):187-197. Danielsson, S., I. C. Gillberg, E. Billstedt, C. Gillberg, and I. Olsson. 2005. Epilepsy in young adults with autism: A prospective population-based follow-up study of 120 individuals diagnosed in childhood. Epilepsia 46(6):918-923.
Davis, K. L., S. D. Candrilli, and H. M. Edin. 2008. Prevalence and cost of nonadherence with antiepileptic drugs in an adult managed care population. Epilepsia 49(3):446-454.
Delaware Health and Social Services. 2011. Autism surveillance and registration program. http://dhss.delaware.gov/dph/chca/dphnsp7.html (accessed November 19, 2011).
Denny, J. C., N. N. Choma, J. F. Peterson, R. A. Miller, L. Bastarache, M. Li, and N. B. Peterson. 2012. Natural language processing improves identification of colorectal cancer testing in the electronic medical record. Medical Decision Making 32(1):188-197.
Devinsky, O. 2011. Sudden, unexpected death in epilepsy. New England Journal of Medicine 365(19):1801-1811.
Devinsky, O., B. G. Vickrey, J. Cramer, K. Perrine, B. Hermann, K. Meador, and R. D. Hays. 1995. Development of the quality of life in epilepsy inventory. Epilepsia 36(11): 1089-1104.
Drummond, M. F., M. J. Sculpher, and G. W. Torrance. 2005. Methods for the economic evaluation of health care programmes. 3rd ed. New York: Oxford University Press.
Dublin, S., E. Baldwin, R. Walker, P. Haug, D. Carrell, and W. Chapman. 2011. Plenary III-02: Accuracy of natural language processing to identify pneumonia from electronic radiology reports. Clinical Medicine and Research 9(3-4):169-170.
Dunn, H. L. 1946. Record linkage. American Journal of Public Health and the Nation’s Health 36(12):1412-1416.
Elliott, J. O., B. Lu, J. L. Moore, J. W. McAuley, and L. Long. 2008. Exercise, diet, health behaviors, and risk factors among persons with epilepsy based on the California Health Interview Survey, 2005. Epilepsy and Behavior 13(2):307-315.
Elliott, J. O., B. Lu, B. F. Shneker, J. L. Moore, and J. W. McAuley. 2009. The impact of “social determinants of health” on epilepsy prevalence and reported medication use. Epilepsy Research 84(2-3):135-145.
Engel, J., Jr. 2008. Surgical treatment for epilepsy: Too little, too late? JAMA 300(21): 2548-2550.
EURAP (International Registry of Antiepileptic Drugs and Pregnancy). 2012. EURAP: An International Registry of Antiepileptic Drugs and Pregnancy. www.eurapinternational.org/index (accessed February 7, 2012).
EUROCAT (European Concerted Action on Congenital Anomalies and Twins). 2012. European surveillance of congenital anomalies. http://www.eurocat-network.eu/ (accessed February 7, 2012).
Faught, R. E., J. R. Weiner, A. Guerin, M. C. Cunnington, and M. S. Duh. 2009. Impact of nonadherence to antiepileptic drugs on health care utilization and costs: Findings from the RANSOM study. Epilepsia 50(3):501-509.
FDA (Food and Drug Administration). 2012. Mini-Sentinel program: Background. http://mini-sentinel.org/about_us/ (accessed February 23, 2012).
Ferguson, P. L., A. W. Selassie, B. B. Wannamaker, B. Dong, R. Kobau, and D. J. Thurman. 2005. Prevalence of epilepsy and health-related quality of life and disability among adults with epilepsy—South Carolina, 2003 and 2004. Morbidity and Mortality Weekly Report 54(42):1080-1082.
Ficker, D. M., E. L. So, W. K. Shen, J. F. Annegers, P. C. O’Brien, G. D. Cascino, and P. G. Belau. 1998. Population-based study of the incidence of sudden unexplained death in epilepsy. Neurology 51(5):1270-1274.
Fountain, N. B., P. C. V. Ness, R. Swain-Eng, S. Tonn, and C. T. Bever. 2011. Quality improvement in neurology: AAN epilepsy quality measures. Report of the Quality Measurement and Reporting Subcommittee of the American Academy of Neurology. Journal of Neurology 76(1):94-99.
Friedlin, J., M. Overhage, M. A. Al-Haddad, J. A. Waters, J. J. Aguilar-Saavedra, J. Kesterson, and M. Schmidt. 2010. Comparing methods for identifying pancreatic cancer patients using electronic data sources. AMIA Annual Symposium Proceedings 2010:237-241.
Friedman, D. E., R. S. Tobias, C. I. Akman, E. O. Smith, and H. S. Levin. 2010. Recurrent seizure-related injuries in people with epilepsy at a tertiary epilepsy center: A 2-year longitudinal study. Epilepsy and Behavior 19(3):400-404.
Galea, S., and M. Tracy. 2007. Participation rates in epidemiologic studies. Annals of Epidemiology 17(9):643-653.
Geisinger Health System. 2011. Statistics. http://www.geisinger.org/about/stats.html (accessed February 9, 2012).
———. 2012. Epilepsy center. http://www.geisinger.org/professionals/services/neuro/epilepsy.html (accessed February 9, 2012).
German, R. R., L. M. Lee, J. M. Horan, R. L. Milstein, C. A. Pertowski, and M. N. Waller. 2001. Updated guidelines for evaluating public health surveillance systems: Recommendations from the Guidelines Working Group. Morbidity and Mortality Weekly Report Recommendations and Reports 50(RR-13):1-35.
Giuffrida, A., H. Gravelle, and M. Roland. 1999. Measuring quality of care with routine data: Avoiding confusion between performance indicators and health outcomes. BMJ 319(7202):94-98.
GlaxoSmithKline. 2012. GlaxoSmithKline pregnancy registries. http://pregnancyregistry.gsk.com/ (accessed February 7, 2012).
Goldacre, M. J., M. Duncan, M. Griffith, and M. R. Turner. 2010. Trends in death certification for multiple sclerosis, motor neuron disease, Parkinson’s disease and epilepsy in English populations 1979-2006. Journal of Neurology 257(5):706-715.
Griffiths, R. I., P. N. Schrammel, G. L. Morris, S. H. Wills, D. M. Labiner, and M. J. Strauss. 1999. Payer costs of patients diagnosed with epilepsy. Epilepsia 40(3):351-358.
Gurney, J. G., M. L. McPheeters, and M. M. Davis. 2006. Parental report of health conditions and health care use among children with and without autism: National Survey of Children’s Health. Archives of Pediatrics and Adolescent Medicine 160(8):825-830.
Halpern, M. T., J. M. Renaud, and B. G. Vickrey. 2011. Impact of insurance status on access to care and out-of-pocket costs for U.S. individuals with epilepsy. Epilepsy and Behavior 22(3):483-489.
Haneef, Z., J. Stern, S. Dewar, and J. Engel, Jr. 2010. Referral pattern for epilepsy surgery after evidence-based recommendations: A retrospective study. Neurology 75(8):699-704.
Hanna, N., M. Black, J. Sander, W. H. Smithson, R. Appleton, S. Brown, and D. Fish. 2002. The national sentinel clinical audit of epilepsy-related death: Epilepsy—Death in the shadows. London, UK: The Stationery Office.
Hauser, W. A., J. F. Annegers, and L. T. Kurland. 1991. Prevalence of epilepsy in Rochester, Minnesota: 1940-1980. Epilepsia 32(4):429-445.
———. 1993. Incidence of epilepsy and unprovoked seizures in Rochester, Minnesota: 1935-1984. Epilepsia 34(3):453-468.
Health Care Cost Institute. 2011. About HCCI. http://healthcostinstitute.org/about (accessed November 21, 2011).
Hesdorffer, D. C., W. A. Hauser, J. F. Annegers, E. Kokmen, and W. A. Rocca. 1996a. Dementia and adult-onset unprovoked seizures. Neurology 46(3):727-730.
Hesdorffer, D. C., W. A. Hauser, J. F. Annegers, and W. A. Rocca. 1996b. Severe, uncontrolled hypertension and adult-onset seizures: A case-control study in Rochester, Minnesota. Epilepsia 37(8):736-741.
Hesdorffer, D. C., G. Logroscino, G. Cascino, J. F. Annegers, and W. A. Hauser. 1998. Incidence of status epilepticus in Rochester, Minnesota, 1965-1984. Neurology 50(3):735-741.
Hesdorffer, D. C., G. Logroscino, E. K. T. Benn, N. Katri, G. Cascino, and W. A. Hauser. 2011. Estimating risk for developing epilepsy: A population-based study in Rochester, Minnesota. Neurology 76(1):23-27.
HHS (Department of Health and Human Services). 2009. HIPAA administrative simplification: Modifications to medical data code set standards to adopt ICD-10-CM and ICD-10-PCS. Federal Register 74(11):3328-3362.
———. 2010a. Health information technology: Initial set of standards, implementation specifications, and certification criteria for electronic health record technology; final rule. Federal Register 75(144):44590-44654.
———. 2010b. Measuring health IT adoption. http://healthit.hhs.gov/portal/server.pt/community/healthit_hhs_gov_adoption_and_meaningful_use/1152 (accessed November 21, 2011).
Hirsch, L. J., E. J. Donner, E. L. So, M. Jacobs, L. Nashef, J. L. Noebels, and J. R. Buchhalter. 2011. Abbreviated report of the NIH/NINDS workshop on sudden unexpected death in epilepsy. Neurology 76(22):1932-1938.
Hitiris, N., S. Suratman, K. Kelly, L. J. Stephen, G. J. Sills, and M. J. Brodie. 2007. Sudden unexpected death in epilepsy: A search for risk factors. Epilepsy and Behavior 10(1):138-141.
HMORN (Health Maintenance Organization Research Network). 2010. HMO Research Network. http://www.hmoresearchnetwork.org/resources/tools/HMORN_Precis.pdf (accessed November 21, 2011).
———. 2012a. Annual conference. http://www.hmoresearchnetwork.org/confpgs/conf.htm (accessed February 9, 2012).
———. 2012b. Governing and operational committees of the HMO Research Network. http://www.hmoresearchnetwork.org/resources/toolkit/HMORN_Committees-and-Contacts.doc (accessed February 9, 2012).
Holden, E. W., E. Grossman, H. T. Nguyen, M. J. Gunter, B. Grebosky, A. Von Worley, L. Nelson, S. Robinson, and D. J. Thurman. 2005a. Developing a computer algorithm to identify epilepsy cases in managed care organizations. Disease Management 8(1):1-14.
Holden, E. W., H. Thanh Nguyen, E. Grossman, S. Robinson, L. S. Nelson, M. J. Gunter, A. Von Worley, and D. J. Thurman. 2005b. Estimating prevalence, incidence, and disease-related mortality for patients with epilepsy in managed care organizations. Epilepsia 46(2):311-319.
Holmes, L. B., E. A. Harvey, B. A. Coull, K. B. Huntington, S. Khoshbin, A. M. Hayes, and L. M. Ryan. 2001. The teratogenicity of anticonvulsant drugs. New England Journal of Medicine 344(15):1132-1138.
Hope, O. A., J. E. Zeber, N. R. Kressin, B. G. Bokhour, A. C. Vancott, J. A. Cramer, M. E. Amuan, J. E. Knoefel, and M. J. Pugh. 2009. New-onset geriatric epilepsy care: Race, setting of diagnosis, and choice of antiepileptic drug. Epilepsia 50(5):1085-1093.
Hoppe, C., A. Poepel, and C. E. Elger. 2007. Epilepsy: Accuracy of patient seizure counts. Archives of Neurology 64(11):1595-1599.
Hsiao, C.-J., E. Hing, T. C. Socey, and B. Cai. 2011. Electronic health record systems and intent to apply for meaningful use incentives among office-based physician practices: United States, 2001-2011. NCHS Data Brief 79(November). http://www.cdc.gov/nchs/data/databriefs/DB79.pdf (accessed February 9, 2012).
Huang, M. C., A. A. Baaj, K. Downes, A. S. Youssef, E. Sauvageau, H. R. van Loveren, and S. Agazzi. 2011. Paradoxical trends in the management of unruptured cerebral aneurysms in the United States: Analysis of nationwide database over a 10-year period. Stroke 42(6):1730-1735.
ICD9data.com. 2011. ICD-9 and ICD-9-cm data. http://www.icd9data.com/ (accessed September 22, 2011).
Iezzoni, L. I. 2003. Risk adjustment for measuring health care outcomes. 3rd ed. Chicago, IL: Health Administration Press.
ILAE (International League Against Epilepsy). 1993. Guidelines for epidemiologic studies on epilepsy. Commission on Epidemiology and Prognosis, International League Against Epilepsy. Epilepsia 34(4):592-596.
IOM (Institute of Medicine). 2011a. A nationwide framework for surveillance of cardiovascular and chronic lung diseases. Washington, DC: The National Academies Press.
———. 2011b. Public engagement and clinical trials: New models and disruptive technologies. Workshop summary. Washington, DC: The National Academies Press.
Irish Epilepsy and Pregnancy Register. 2012. The Irish Epilepsy and Pregnancy Register. http://www.epilepsypregnancyregister.ie/about%20the%20register.html (accessed February 7, 2012).
Ivanova, J. I., H. G. Birnbaum, Y. Kidolezi, Y. Qiu, D. Mallett, and S. Caleo. 2010. Direct and indirect costs associated with epileptic partial onset seizures among the privately insured in the United States. Epilepsia 51(5):838-844.
Izquierdo, J. N., and V. J. Schoenbach. 2000. The potential and limitations of data from population-based state cancer registries. American Journal of Public Health 90(5): 695-698.
Jacoby, A. 2002. Stigma, epilepsy, and quality of life. Epilepsy and Behavior 3(6S2):10-20.
Jemal, A., R. Siegel, J. Xu, and E. Ward. 2010. Cancer statistics, 2010. CA: A Cancer Journal for Clinicians 60(5):277-300.
Jennum, P., J. Gyllenborg, and J. Kjellberg. 2011. The social and economic consequences of epilepsy: A controlled national study. Epilepsia 52(5):949-956.
Jetté, N. 2011. Lessons from the development of a Canadian national system of surveillance. PowerPoint presented at the IOM Workshop on Public Health Surveillance, Population Health Research, and Data Collections for the Epilepsies, March 21. http://iom.edu/~/media/Files/Activity%20Files/Disease/Epilepsy/Panel%202-3%20Jette.pdf (accessed February 9, 2012).
Jetté, N., A. Y. Reid, H. Quan, M. D. Hill, and S. Wiebe. 2010. How accurate is ICD coding for epilepsy? Epilepsia 51(1):62-69.
Kaiboriboon, K., N. Schiltz, S. M. Koroukian, S. D. Lhatoo, and M. Z. Koubeissi. 2011. Limitations of NIS database in evaluation of epilepsy surgery morbidity and mortality. Archives of Neurology 68(11):1483.
Kaiser Permanente. 2011. Autism registry is first division project funded by new national Mental Health Research Network. http://researchnews.kaiser.org/?p=1002 (accessed November 19, 2011).
Kanaan, S. B. 2000. The National Committee on Vital and Health Statistics, 1949-1999: A history. http://ncvhs.hhs.gov/50history.htm (accessed February 7, 2012).
Kaneko, S., D. Battino, E. Andermann, K. Wada, R. Kan, A. Takeda, Y. Nakane, Y. Ogawa, G. Avanzini, C. Fumarola, T. Granata, F. Molteni, G. Pardi, L. Minotti, R. Canger, L. Dansky, M. Oguni, I. Lopes-Cendas, A. Sherwin, F. Andermann, M. H. Seni, M. Okada, and T. Teranishi. 1999. Congenital malformations due to antiepileptic drugs. Epilepsy Research 33(2-3):145-158.
Kee, V. R., B. Gilchrist, M. A. Granner, N. R. Sarrazin, and R. M. Carnahan. 2012. A systematic review of validated methods for identifying seizures, convulsions, or epilepsy using administrative and claims data. Pharmacoepidemiology and Drug Safety 21(Suppl. 1):183-193.
Kelvin, E. A., D. C. Hesdorffer, E. Bagiella, H. Andrews, T. A. Pedley, T. T. Shih, L. Leary, D. J. Thurman, and W. A. Hauser. 2007. Prevalence of self-reported epilepsy in a multiracial and multiethnic community in New York City. Epilepsy Research 77(2-3):141-150.
Kennedy Krieger Institute. 2011. Interactive Autism Network. http://www.ianproject.org/ (accessed July 13, 2011).
Kerr, C., A. Nixon, and M. Angalakuditi. 2011. The impact of epilepsy on children and adult patients’ lives: Development of a conceptual model from qualitative literature. Seizure 20(10):764-774.
Klabunde, C. N., J. L. Warren, and J. M. Legler. 2002. Assessing comorbidity using claims data: An overview. Medical Care 40(8 Suppl.):IV-26-IV-35.
Knoester, P. D., A. J. Boendermaker, A. C. Egberts, Y. A. Hekster, A. Keyser, J. L. Severens, W. O. Renier, and C. L. Deckers. 2005. Cost-effectiveness of add-on lamotrigine therapy in clinical practice. Epilepsy Research 67(3):143-151.
Kobau, R., H. Zahran, D. Grant, D. J. Thurman, P. H. Price, and M. M. Zack. 2007. Prevalence of active epilepsy and health-related quality of life among adults with self-reported epilepsy in California: California Health Interview Survey, 2003. Epilepsia 48(10):1904-1913.
Kobau, R., H. Zahran, D. J. Thurman, M. M. Zack, T. R. Henry, S. C. Schachter, and P. H. Price. 2008. Epilepsy surveillance among adults—19 states, Behavioral Risk Factor Surveillance System, 2005. Morbidity and Mortality Weekly Report Surveillance Summaries 57(6):1-20.
Koch, S., G. Losche, E. Jager-Roman, S. Jakob, D. Rating, A. Deichl, and H. Helge. 1992. Major and minor birth malformations and antiepileptic drugs. Neurology 42(4 Suppl. 5):83-88.
Langan, Y., L. Nashef, and J. W. Sander. 2002. Certification of deaths attributable to epilepsy. Journal of Neurology, Neurosurgery and Psychiatry 73(6):751-752.
Larson, K., S. A. Russ, R. S. Kahn, and N. Halfon. 2011. Patterns of comorbidity, functioning, and service use for U.S. children with ADHD, 2007. Pediatrics 127(3):462-470.
Lathers, C. M., S. A. Koehler, C. H. Wecht, and P. L. Schraeder. 2011a. Forensic antiepileptic drug levels in autopsy cases of epilepsy. Epilepsy and Behavior 22:778-785.
Lathers, C. M., P. L. Schraeder, M. W. Bungo, and J. E. Leestma. 2011b. Sudden death in epilepsy: Forensic and clinical issues. Boca Raton, FL: CRC Press.
Le, S., P. O. Shafer, E. Bartfeld, and R. S. Fisher. 2011. An online diary for tracking epilepsy. Epilepsy and Behavior 22(4):705-709.
Leaffer, E. B., A. Jacoby, E. Benn, W. A. Hauser, T. Shih, P. Dayan, R. Green, H. Andrews, D. J. Thurman, and D. Hesdorffer. 2011. Associates of stigma in an incident epilepsy population from northern Manhattan, New York City. Epilepsy and Behavior 21(1):60-64.
Lee, H., A. R. Marvin, T. Watson, J. Piggot, J. K. Law, P. A. Law, J. N. Constantino, and S. F. Nelson. 2010. Accuracy of phenotyping of autistic children based on Internet implemented parent report. American Journal of Medical Genetics. Part B, Neuropsychiatric Genetics 153B(6):1119-1126.
Leestma, J. E. 1997. Forensic considerations in sudden unexpected death in epilepsy. Epilepsia 38(11 Suppl.):S63-S66.
Lindhout, D., J. G. Omtzigt, and M. C. Cornel. 1992. Spectrum of neural-tube defects in 34 infants prenatally exposed to antiepileptic drugs. Neurology 42(4 Suppl. 5):111-118.
Loring, D. W., D. H. Lowenstein, N. M. Barbaro, B. E. Fureman, J. Odenkirchen, M. P. Jacobs, J. K. Austin, D. J. Dlugos, J. A. French, W. D. Gaillard, B. P. Hermann, D. C. Hesdorffer, S. N. Roper, A. C. Van Cott, S. Grinnon, and A. Stout. 2011. Common data elements in epilepsy research: Development and implementation of the NINDS epilepsy CDE project. Epilepsia 52(6):1186-1191.
Love, D., W. Custer, and P. Miller. 2010. All-payer claims databases: State initiatives to improve health care transparency—Issue brief. http://apcdcouncil.org/sites/apcdcouncil.org/files/All-Payer%20Claims%20Databases%20State%20Initiatives%20to%20Improve%20Health%20Care%20Transparency.pdf (accessed September 27, 2011).
Love, T. J., T. Cai, and E. W. Karlson. 2011. Validation of psoriatic arthritis diagnoses in electronic medical records using natural language processing. Seminars in Arthritis and Rheumatism 40(5):413-420.
Macleish, M. 2011. Challenges and opportunities for surveillance: The patient and family perspective. PowerPoint presented at the IOM Workshop on Public Health Surveillance, Population Health Research, and Data Collection for the Epilepsies, March 21. http://iom.edu/~/media/Files/ActivityFiles/Disease/Epilepsy/Panel2-4Macleish.pdf (accessed September 19, 2011).
McClelland, S., 3rd, H. Guo, and K. S. Okuyemi. 2010. Racial disparities in the surgical management of intractable temporal lobe epilepsy in the United States: A population-based analysis. Archives of Neurology 67(5):577-583.
Meador, K. J., G. A. Baker, N. Browning, J. Clayton-Smith, D. T. Combs-Cantrell, M. Cohen, L. A. Kalayjian, A. Kanner, J. D. Liporace, P. B. Pennell, M. Privitera, and D. W. Loring. 2009. Cognitive function at 3 years of age after fetal exposure to antiepileptic drugs. New England Journal of Medicine 360(16):1597-1605.
Meyer, A.-C., T. Dua, J. Ma, S. Saxena, and G. Birbeck. 2010. Global disparities in the epilepsy treatment gap: A systematic review. Bulletin of the World Health Organization 88(4):260-266.
Mohanraj, R., J. Norrie, L. J. Stephen, K. Kelly, N. Hitiris, and M. J. Brodie. 2006. Mortality in adults with newly diagnosed and chronic epilepsy: A retrospective comparative study. Lancet Neurology 5(6):481-487.
Morrow, J., A. Russell, E. Guthrie, L. Parsons, I. Robertson, R. Waddell, B. Irwin, R. C. McGivern, P. J. Morrison, and J. Craig. 2006. Malformation risks of antiepileptic drugs in pregnancy: A prospective study from the UK Epilepsy and Pregnancy Register. Journal of Neurology, Neurosurgery, and Psychiatry 77(2):193-198.
Murff, H. J., F. FitzHenry, M. E. Matheny, N. Gentry, K. L. Kotter, K. Crimin, R. S. Dittus, A. K. Rosen, P. L. Elkin, S. H. Brown, and T. Speroff. 2011. Automated identification of postoperative complications within an electronic medical record using natural language processing. JAMA 306(8):848-855.
NAACCR (North American Association of Central Cancer Registries). 2010a. About NAACCR, Inc. http://www.naaccr.org/AboutNAACCR/NAACCRMission.aspx (accessed November 18, 2011).
———. 2010b. U.S. registries certified in 2010 for 2007 incidence data. http://www.naaccr.org/Certification/USCert2007.aspx (accessed November 18, 2011).
———. 2011. U.S. registries certified in 2011 for 2008 incidence data. http://www.naaccr.org/Certification/USCert2008.aspx (accessed November 18, 2011).
———. n.d. NAACCR brochure. http://www.naaccr.org/LinkClick.aspx?fileticket=waBuX8mPz-M%3d&tabid=73&mid=413 (accessed November 18, 2011).
Nashef, L., and J. W. Sander. 1996. Sudden unexpected deaths in epilepsy—Where are we now? Seizure 5(3):235-238.
Nashef, L., E. L. So, P. Ryvlin, and T. Tomson. 2012. Unifying the definitions of sudden unexpected death in epilepsy. Epilepsia 53(2):227-233.
NCI (National Cancer Institute). 2009. SEER-Medicare: Brief description of the SEER-Medicare database. http://healthservices.cancer.gov/seermedicare/overview/ (accessed February 7, 2012).
———. 2011. Overview of the SEER program. http://seer.cancer.gov/about/overview.html (accessed November 18, 2011).
———. 2012. The Surveillance, Epidemiology and End Results (SEER) program. http://training.seer.cancer.gov/operations/standards/setters/seer.html (accessed February 7, 2012).
New Hampshire Department of Health and Human Services. 2011. Autism spectrum registry. http://www.dhhs.nh.gov/dcbcs/bds/registry.htm (accessed November 19, 2011).
New Jersey Department of Health and Senior Services. 2007. Special child health services registry. http://www.state.nj.us/health/fhs/sch/schr.shtml (accessed November 19, 2011).
New York State Department of Health. 2004. Dementias reported in hospitalizations among New York State residents. http://www.health.ny.gov/diseases/conditions/dementia/alzheimer/dementia_registry.htm (accessed November 19, 2011).
———. 2006. About the New York State Dementias Registry. http://www.health.ny.gov/diseases/conditions/dementia/alzheimer/dementia_registry_fact_sheet.htm (accessed February 28, 2012).
NIH (National Institutes of Health). 2011a. Funding opportunity announcement RFA-RM-11-003: NIH-HMO collaboratory coordinating center limited competition. http://grants.nih.gov/grants/guide/rfa-files/RFA-RM-11-003.html (accessed November 22, 2011).
———. 2011b. The National Children’s Study. http://www.nationalchildrensstudy.gov/about/overview/Pages/NCS-General-Presentation-12-11.pdf (accessed February 6, 2012).
———. 2011c. National Children’s Study response to the National Academy of Sciences review of the National Children’s Study research plan. http://www.nationalchildrens-study.gov/newsandevents/announcements/Pages/ncs_response_NAS_review_082608.pdf (accessed November 22, 2011).
———. 2011d. National Children’s Study update. http://www.nationalchildrensstudy.gov/research/Pages/NCS-update-presentation.pdf (accessed November 22, 2011).
———. 2011e. The National Children’s Study: An evolving system concept of operations. http://www.nationalchildrensstudy.gov/about/overview/Pages/NCS_concept_of_operations_04_28_11.pdf (accessed November 22, 2011).
———. 2011f. Study overview. http://www.nationalchildrensstudy.gov/about/overview/Pages/default.aspx (accessed November 22, 2011).
———. 2012. Health Care Systems (HCS) Research Collaboratory: Overview. http://commonfund.nih.gov/hcscollaboratory/overview.aspx (accessed February 7, 2012).
NINDS (National Institute of Neurological Disorders and Stroke). 2010. Exploratory grant program in sudden unexpected death in epilepsy (SUDEP) (P20). http://grants.nih.gov/grants/guide/rfa-files/RFA-NS-11-006.html (accessed October 18, 2011).
North American AED Pregnancy Registry. 2012. The North American AED pregnancy registry. http://aedpregnancyregistry.org/ (accessed February 7, 2012).
Nowinski, C. J., D. Victorson, J. E. Cavazos, R. Gershon, and D. Cella. 2010. Neuro-QOL and the NIH Toolbox: Implications for epilepsy. Therapy 7(5):533-540.
Ojemann, L. M., C. Baugh-Bookman, and D. L. Dudley. 1987. Effect of psychotropic medications on seizure control in patients with epilepsy. Neurology 37(9):1525-1527.
Olafsson, E., J. T. Hallgrimsson, W. A. Hauser, P. Ludvigsson, and G. Gudmundsson. 1998. Pregnancies of women with epilepsy: A population-based study in Iceland. Epilepsia 39(8):887-892.
Olafsson, E., P. Ludvigsson, G. Gudmundsson, D. Hesdorffer, O. Kjartansson, and W. A. Hauser. 2005. Incidence of unprovoked seizures and epilepsy in Iceland and assessment of the epilepsy syndrome classification: A prospective study. Lancet Neurology 4(10):627-634.
Olesen, J. B., S. Z. Abildstrom, J. Erdal, G. H. Gislason, P. Weeke, C. Andersson, C. Torp-Pedersen, and P. R. Hansen. 2011. Effects of epilepsy and selected antiepileptic drugs on risk of myocardial infarction, stroke, and death in patients with or without previous stroke: A nationwide cohort study. Pharmacoepidemiology and Drug Safety 20(9):964-971.
Omtzigt, J. G., F. J. Los, D. E. Grobbee, L. Pijpers, M. G. Jahoda, H. Brandenburg, P. A. Stewart, H. L. Gaillard, E. S. Sachs, and J. W. Wladimiroff. 1992. The risk of spina bifida aperta after first-trimester exposure to valproate in a prenatal cohort. Neurology 42(4 Suppl. 5):119-125.
Ottman, R., J. H. Lee, N. Risch, W. A. Hauser, and M. Susser. 1996. Clinical indicators of genetic susceptibility to epilepsy. Epilepsia 37(4):353-361.
Parko, K., and D. J. Thurman. 2009. Prevalence of epilepsy and seizures in the Navajo nation 1998-2002. Epilepsia 50(10):2180-2185.
PCAST (President’s Council of Advisors on Science and Technology). 2010. Report to the President realizing the full potential of health information technology to improve healthcare for Americans: The path forward. http://www.whitehouse.gov/sites/default/files/microsites/ostp/pcast-health-it-report.pdf (accessed November 21, 2011).
Perucca, P., D. C. Hesdorffer, and F. G. Gilliam. 2011. Response to first antiepileptic drug trial predicts health outcome in epilepsy. Epilepsia 52(12):2209-2215.
Platt, R., R. M. Carnahan, J. S. Brown, E. Chrischilles, L. H. Curtis, S. Hennessy, J. C. Nelson, J. A. Racoosin, M. Robb, S. Schneeweiss, S. Toh, and M. G. Weiner. 2012. The U.S. Food and Drug Administration’s Mini-Sentinel program: Status and direction. Pharmacoepide-miology and Drug Safety 21(Suppl. 1):1-8.
Pugh, M. J. V., L. A. Copeland, J. E. Zeber, J. A. Cramer, M. E. Amuan, J. E. Cavazos, and L. E. Kazis. 2005. The impact of epilepsy on health status among younger and older adults. Epilepsia 46(11):1820-1827.
Pugh, M. J. V., A. C. Van Cott, J. A. Cramer, J. E. Knoefel, M. E. Amuan, J. Tabares, R. E. Ramsay, D. R. Berlowitz, and the Treatment in Geriatric Epilepsy Research Team. 2008. Trends in antiepileptic drug prescribing for older patients with new-onset epilepsy: 2000-2004. Neurology 70(22 Pt. 2):2171-2178.
Pugh, M. J. V., A. C. Vancott, M. A. Steinman, E. M. Mortensen, M. E. Amuan, C.-P. Wang, J. E. Knoefel, and D. R. Berlowitz. 2010. Choice of initial antiepileptic drug for older veterans: Possible pharmacokinetic drug interactions with existing medications. Journal of the American Geriatrics Society 58(3):465-471.
Pugh, M. J., D. R. Berlowitz, J. K. Rao, G. Shapiro, R. Avetisyan, A. Hanchate, K. Jarrett, J. Tabares, and L. E. Kazis. 2011. The quality of care for adults with epilepsy: An initial glimpse using the QUIET measure. BMC Health Services Research 11:1.
Qin, P., H. Xu, T. M. Laursen, M. Vestergaard, and P. B. Mortensen. 2005. Risk for schizophrenia and schizophrenia-like psychosis among patients with epilepsy: Population based cohort study. BMJ 331(7507):23-25.
Rosa, F. W. 1991. Spina bifida in infants of women treated with carbamazepine during pregnancy. New England Journal of Medicine 324(10):674-677.
Sabaz, M., D. R. Cairns, J. A. Lawson, N. Nheu, A. F. Bleasel, and A. M. Bye. 2000. Validation of a new quality of life measure for children with epilepsy. Epilepsia 41(6):765-774.
Samrén, E. B., C. M. van Duijn, S. Koch, V. K. Hiilesmaa, H. Klepel, A. H. Bardy, G. B. Mannagetta, A. W. Deichl, E. Gaily, I. L. Granström, H. Meinardi, D. E. Grobbee, A. Hofman, D. Janz, and D. Lindhout. 1997. Maternal use of antiepileptic drugs and the risk of major congenital malformations: A joint European prospective study of human teratogenesis associated with maternal epilepsy. Epilepsia 38(9):981-990.
Samrén, E. B., C. M. van Duijn, G. C. Christiaens, A. Hofman, and D. Lindhout. 1999. An-tiepileptic drug regimens and major congenital abnormalities in the offspring. Annals of Neurology 46(5):739-746.
Sanderson, M., J. T. Benjamin, M. J. Lane, C. B. Cornman, and D. R. Davis. 2003. Application of capture-recapture methodology to estimate the prevalence of dementia in South Carolina. Annals of Epidemiology 13(7):518-524.
Scheepers, B., P. Clough, and C. Pickles. 1998. The misdiagnosis of epilepsy: Findings of a population study. Seizure 7(5):403-406.
Schieve, L. A., S. L. Boulet, M. D. Kogan, M. Yeargin-Allsopp, C. A. Boyle, S. N. Visser, S. J. Blumberg, and C. Rice. 2011. Parenting aggravation and autism spectrum disorders: 2007 National Survey of Children’s Health. Disability and Health Journal 4(3):143-152.
Schraeder, P. L., K. Delin, R. L. McClelland, and E. L. So. 2006. Coroner and medical examiner documentation of sudden unexplained deaths in epilepsy. Epilepsy Research 68(2):137-143.
Shatin, D., R. Levin, H. T. Ireys, and V. Haller. 1998. Health care utilization by children with chronic illnesses: A comparison of Medicaid and employer-insured managed care. Pediatrics 102(4):E44.
Smith, D., B. A. Defalla, and D. W. Chadwick. 1999. The misdiagnosis of epilepsy and the management of refractory epilepsy in a specialist clinic. QJM: An International Journal of Medicine 92(1):15-23.
So, E. L., J. Bainbridge, J. R. Buchhalter, J. Donalty, E. J. Donner, A. Finucane, N. M. Graves, L. J. Hirsch, G. D. Montouris, N. R. Temkin, S. Wiebe, and T. L. Sierzant. 2009. Report of the American Epilepsy Society and the Epilepsy Foundation Joint Task Force on Sudden Unexplained Death in Epilepsy. Epilepsia 50(4):917-922.
Socialstyrelsen. 2012. The Swedish Medical Birth Register. http://www.socialstyrelsen.se/register/halsodataregister/medicinskafodelseregistret/inenglish (accessed February 7, 2012).
Solans, M., S. Pane, M. D. Estrada, V. Serra-Sutton, S. Berra, M. Herdman, J. Alonso, and L. Rajmil. 2008. Health-related quality of life measurement in children and adolescents: A systematic review of generic and disease-specific instruments. Value in Health 11(4):742-764.
Stavem, K. 1998. Quality of life in epilepsy: Comparison of four preference measures. Epilepsy Research 29(3):201-209.
Strine, T. W., R. Kobau, D. P. Chapman, D. J. Thurman, P. Price, and L. S. Balluz. 2005. Psychological distress, comorbidities, and health behaviors among U.S. adults with seizures: Results from the 2002 National Health Interview Survey. Epilepsia 46(7):1133-1139.
Strzelczyk, A., J. P. Reese, R. Dodel, and H. M. Hamer. 2008. Cost of epilepsy: A systematic review. Pharmacoeconomics 26(6):463-476.
Sun, Y., M. Vestergaard, J. Christensen, and J. Olsen. 2011. Breastfeeding and risk of epilepsy in childhood: A birth cohort study. Journal of Pediatrics 158(6):924-929.
Szaflarski, M., J. P. Szaflarski, M. D. Privitera, D. M. Ficker, and R. D. Horner. 2006. Racial/ ethnic disparities in the treatment of epilepsy: What do we know? What do we need to know? Epilepsy and Behavior 9(2):243-264.
Thurman, D. J., E. Beghi, C. E. Begley, A. T. Berg, J. R. Buchhalter, D. Ding, D. C. Hesdorffer, W. A. Hauser, L. Kazis, R. Kobau, B. Kroner, D. Labiner, K. Liow, G. Logroscino, M. T. Medina, C. R. Newton, K. Parko, A. Paschal, P.-M. Preux, J. W. Sander, A. Selassie, W. Theodore, T. Tomson, and S. Wiebe. 2011. Standards for epidemiologic studies and surveillance of epilepsy. Epilepsia 52(Suppl. 7):2-26.
Tomson, T., T. Walczak, M. Sillanpää, and J. W. A. S. Sander. 2005. Sudden unexpected death in epilepsy: A review of incidence and risk factors. Epilepsia 46(Suppl. 11):54-61.
Tomson, T., D. Battino, J. French, C. Harden, L. Holmes, J. Morrow, E. Robert-Gnansia, A. Scheuerle, F. Vajda, K. Wide, and J. Gordon. 2007. Antiepileptic drug exposure and major congenital malformations: The role of pregnancy registries. Epilepsy and Behavior 11(3):277-282.
Tomson, T., L. Nashef, and P. Ryvlin. 2008. Sudden unexpected death in epilepsy: Current knowledge and future directions. Lancet Neurology 7(11):1021-1031.
Tomson, T., D. Battino, J. Craig, S. Hernandez-Diaz, L. B. Holmes, D. Lindhout, J. Morrow, and J. French. 2010. Pregnancy registries: Differences, similarities, and possible harmonization. Epilepsia 51(5):909-915.
Trevathan, E. 2011. “Flying blind” without epilepsy surveillance data. Neurology 76(1):10-11.
Tyler, C. V., S. C. Schramm, M. Karafa, A. S. Tang, and A. K. Jain. 2011. Chronic disease risks in young adults with autism spectrum disorder: Forewarned is forearmed. American Journal on Intellectual and Developmental Disabilities 116(5):371-380.
UCB, Inc. 2012. UCB resources. http://www.keppraxr.com/resources/ucb.aspx (accessed February 7, 2012).
UK Epilepsy and Pregnancy Register. 2012. The UK Epilepsy and Pregnancy Register. http://www.epilepsyandpregnancy.co.uk/ (accessed February 17, 2012).
University of South Carolina Arnold School of Public Health. 2010. Annual report: South Carolina Alzeimer’s disease registry. http://www.sph.sc.edu/osa/documents/ANNUAL%20REPORT%202010.pdf (accessed November 19, 2011).
———. 2011. Alzeimer’s disease registry. http://www.sph.sc.edu/osa/alzheimers_registry.html (accessed November 19, 2011).
VA (Department of Veterans Affairs). 2011. About VHA. http://www.va.gov/health/aboutVHA.asp (accessed February 9, 2012).
Van Den Eeden, S. K. 2011. Surveillance of neurologic disorders within integrated system and HMORN Collaboratory. PowerPoint presented at the IOM Workshop on Public Health Surveillance, Population Health Research, and Data Collections for the Epilepsies, March 21. http://iom.edu/~/media/Files/Activity%20Files/Disease/Epilepsy/Panel%203-4%20Van%20Den%20Eeden.pdf (accessed November 22, 2011).
VanWormer, J. J. 2010. Methods of using electronic health records for population-level surveillance of coronary heart disease risk in the heart of New Ulm Project. Diabetes Spectrum 23(3):161-165.
Varosy, P. D. 2011. Veterans Health Administration: Surveillance of cardiovascular disease. PowerPoint presented at the IOM Workshop on Public Health Surveillance, Population Health Research, and Data Collections for the Epilepsies, March 21. http://iom.edu/~/media/Files/Activity%20Files/Disease/Epilepsy/Panel%203-3%20Varosy.pdf (accessed November 22, 2011).
Viinikainen, K., S. Heinonen, K. Eriksson, and R. Kalviainen. 2006. Community-based, prospective, controlled study of obstetric and neonatal outcome of 179 pregnancies in women with epilepsy. Epilepsia 47(1):186-192.
Warren, J. L., C. N. Klabunde, D. Schrag, P. B. Bach, and G. F. Riley. 2002. Overview of the SEER-Medicare data: Content, research applications, and generalizability to the United States elderly population. Medical Care 40(8 Suppl.):IV-3-18.
West Virginia University. 2011. West Virginia Alzheimer’s disease registry. http://www.wvadr.hsc.wvu.edu/Pages/ (accessed November 19, 2011).
Wicks, P., D. L. Keininger, M. P. Massagli, C. D. la Loge, C. Brownstein, J. Isojarvi, and J. Heywood. 2012. Perceived benefits of sharing health data between people with epilepsy on an online platform. Epilepsy and Behavior 23(1):16-23.
Wingo, P. A., P. M. Jamison, R. A. Hiatt, H. K. Weir, P. M. Gargiullo, M. Hutton, N. C. Lee, and H. I. Hall. 2003. Building the infrastructure for nationwide cancer surveillance and control—A comparison between the National Program of Cancer Registries (NPCR) and the Surveillance, Epidemiology, and End Results (SEER) program (United States). Cancer Causes and Control 14(2):175-193.
Yeargin-Allsopp, M. 2011. Lessons from population-based surveillance for ASD. PowerPoint presented at the IOM Workshop on Public Health Surveillance, Population Health Research, and Data Collections for the Epilepsies, March 21. http://iom.edu/~/media/Files/Activity%20Files/Disease/Epilepsy/Panel%203-2%20Yeargin-Allsopp.pdf (accessed November 22, 2011).
Yoon, D., K. D. Frick, D. A. Carr, and J. K. Austin. 2009. Economic impact of epilepsy in the United States. Epilepsia 50(10):2186-2191.