Critical determinants of the epilepsy treatment gap: A cross-national analysis in resource-limited settings

Authors


Address correspondence to Ana-Claire L. Meyer, San Francisco General Hospital, 1001 Potrero Avenue, Bldg 1, Room 101, Box 0870, San Francisco, CA 94110, U.S.A. E-mail: meyerac@sfgh.ucsf.edu

Summary

Purpose:  Epilepsy is one of the most common serious neurologic disorders worldwide. Our objective was to determine which economic, health care, neurology, and epilepsy-specific resources were associated with untreated epilepsy in resource-constrained settings.

Methods:  A systematic review of the literature identified community-based studies in resource-constrained settings that calculated the epilepsy treatment gap, the proportion with untreated epilepsy, from prevalent active epilepsy cases. Economic, health care, neurology, and epilepsy-specific resources were taken from existing datasets. Poisson regression models with jackknifed standard errors were used to create bivariate and multivariate models comparing the association between treatment status and economic and health resource indicators. Relative risks were reported.

Key Findings:  Forty-seven studies of 8,285 individuals from 24 countries met inclusion criteria. Bivariate analysis demonstrated that individuals residing in rural locations had significantly higher risks of untreated epilepsy (relative risk [RR] 1.63; 95% confidence interval [CI] 1.26–2.11). Significantly lower risks of untreated epilepsy were observed for higher physician density (RR 0.65, 95% CI 0.55–0.78), presence of a lay (RR 0.74, 95% CI 0.60–0.91) or professional association for epilepsy (RR 0.73, 95% CI 0.59–0.91), or postgraduate neurology training program (RR 0.67, 95% CI 0.55–0.82). In multivariate models, higher physician density maintained significant effects (RR 0.67; 95% CI 0.52–0.88).

Significance:  Even among resource-limited regions, people with epilepsy in countries with fewer economic, health care, neurology, and epilepsy-specific resources are more likely to have untreated epilepsy. Community-based epilepsy care programs have improved access to treatment, but in order to decrease the epilepsy-treatment gap, poverty and inequalities of health care, neurology, and epilepsy resources must be dealt with at the local, national, and global levels.

Epilepsy is one of the most common serious neurologic disorders; it affects 50 million people worldwide, 80% of whom live in the developing world (Leonardi & Ustun, 2002). Cost-effective epilepsy treatments are available and accurate diagnosis can be made without technological equipment, such as electroencephalograpy or neuro-imaging. Nonetheless, most individuals with epilepsy in many resource-poor regions do not receive treatment (Chisholm, 2005). Untreated epilepsy is a critical public health issue as people with untreated epilepsy face potentially devastating social consequences and poor health outcomes; particularly in resource-poor regions, persons with epilepsy contend with severe stigma, lower employment, and education levels, and with lower socioeconomic status (Jilek-Aall & Rwiza, 1992; Birbeck, 2000; Baskind & Birbeck, 2005; Jacoby et al., 2005; Amoroso et al., 2006; Ding et al., 2006; Birbeck et al., 2007; de Boer et al., 2008).

The epilepsy-treatment gap is defined as the proportion of people with epilepsy who require but do not receive treatment; it has been proposed as a useful parameter to compare access to and quality of care for epilepsy across populations (Kale, 2002; Begley, et al. 2007). Prior research suggests that countries with higher income, urban populations, and children have lower treatment gaps (Meyer et al., 2010). However, to our knowledge, few cross-national comparisons of resources for epilepsy care have been performed and none have explored the associations with the epilepsy-treatment gap (World Health Organization & World Federation of Neurology, 2004; The Global Campaign against Epilepsy, et al., 2005).

Cross-national comparisons have advanced our understanding of the influence of economic and health care resources on important health indicators such as vaccination coverage or maternal, infant, and under-five mortality rates (Anand & Barnighausen, 2004, 2007). Economic and health care resources were found to be critical determinants of vaccination coverage; specific indicators that were important determinants included the density of human resources for health and female adult literacy, but not per capita income (Anand & Barnighausen, 2007). Similarly, per capita income, female adult literacy, absolute income poverty, and the density of human resources for health were found to be important indicators of maternal mortality, infant mortality, and under-five mortality (Anand & Barnighausen, 2004).

In this study, we build on a prior systematic review of the epilepsy-treatment gap (Meyer et al., 2010). In that work we found a dramatic global disparity in the care for epilepsy between high and low, lower middle, and upper middle income countries. In this study, our objective was to explore the impact of health and economic resources on the epilepsy-treatment gap among low, lower middle, and upper middle income countries. Specifically, our aim was to determine which health care resources and macroeconomic country characteristics critically affect the epilepsy-treatment gap in resource-constrained settings. In particular, we wanted to know which neurology and epilepsy-specific resources affected treatment gaps independent of general health care and economic resources. Greater understanding of the determinants of the epilepsy-treatment gap could inform future policy directions and intervention design and is an essential step toward decreasing the gap.

Methods

A systematic review of peer-reviewed literature in all languages was conducted using PubMed and EMBASE limited to January 1, 1987 through March 7, 2012. Please see Appendix S1 for detailed search strategy. This search represents an update of our prior systematic review, albeit with different inclusion criteria (Meyer et al., 2010). Two authors independently reviewed the studies to determine if they met the following inclusion criteria: (1) used a population-based sample; (2) used a standard definition of epilepsy; (3) determined the treatment gap on a population with active epilepsy; (4) calculated the treatment gap in a sample size of >20 people with epilepsy; (5) were performed in a low, lower middle, or upper middle income countries as per World Bank criteria (The World Bank, 2006).

A population-based sample was defined as a door-to-door or other probability sample of a regional or national population. Studies in which the sample was drawn from a medical care setting were excluded to avoid underestimating the treatment gap. School-based populations in countries where school attendance was low were also excluded. Finally, studies based on methods shown to produce unreliable community-based samples in epilepsy prevalence studies, such as the key informant method, were excluded as well.

The standard definition of active epilepsy had to be internally consistent and to differentiate epilepsy from provoked seizures, febrile seizures, and isolated seizures. Acceptable definitions of active epilepsy included a history of more than one unprovoked seizure and either recent seizures (within the previous 5 years) or current use of antiepilepsy medication. If the treatment gap or other information was missing from the report, we tried to contact the authors to obtain the information before excluding the study.

We excluded studies determining the treatment gap in a population with lifetime epilepsy as that could potentially overestimate the treatment gap. For example, some individuals captured in lifetime prevalence of epilepsy have a history of epilepsy but are currently in terminal remission off treatment. Including these individuals in estimating the treatment gap overestimates the gap because these individuals should not be currently treated with antiepilepsy medications; that is, by not receiving treatment, these individuals are receiving the recommended standard of care.

In addition, we did not include studies that calculated the treatment gap on a sample of ≤20 individuals with epilepsy. Finally, we excluded studies from World Bank–defined high-income countries. In our prior work we found that there was a dramatic global disparity in the treatment gap in which high-income countries had mean treatment gaps of <10%, whereas low, lower middle, and upper middle income countries had mean treatment gaps between 53% and 74%. In this study, our goal was to explore whether health and macroeconomic resources influence the epilepsy-treatment gap even among low, lower middle, and upper middle income countries.

In Table S1 we provide technical definitions and detailed source information for each indicator we used. Macroeconomic indicators were obtained for the prevalence year or publication year of the study and included the gross national income per capita (GNIpc), Atlas Method, in current US$, the amount of official development assistance and official aid in current US$ as a percentage of the gross national income (GNI) in current US$, the amount of external debt in current US$ as a percentage of the GNI in current US$ (Quick Query of selected World Development Indicators). Measures of poverty were obtained for the year nearest the prevalence or publication year of the study and included the percentage of the population below the poverty line of $38.00 per month ($1.25 per day) and the Gini Index, a measure of inequality of income distribution expressed as a percentage, where 0 would represent a situation where everyone had the same income, and 100 would represent a situation where the richest person in the nation had all the income (PovcalNet Poverty Measures).

General health resource indicators were obtained for the year nearest the prevalence or publication year of the study. Indicators included the following: the number of physicians per 1,000 population, the number of nurses and midwives per 1,000 population (Global Health Observatory Data Repository), the number of hospital beds per 10,000 population, and the total expenditure on health as a percentage of the gross domestic product. Proxy measures of individual health care resources and access to health care included the adult literacy rate (%), the proportion births attended by skilled health personnel, measles containing vaccine (MCV) coverage (or the percentage of 1 year old children immunized with a measles containing vaccine) and diptheria-tetanus-pertussis (DTP) coverage (the percentage of 1-year-old children immunized with three doses of DTP vaccine) (Global Health Observatory Data Repository).

Neurology resource indicators were obtained from the Neurology Atlas (World Health Organization & World Federation of Neurology, 2004), compiled by the WHO and World Federation of Neurology, and included the following: the number of neurology beds per 100,000 population, the number of neurologists or neurology health care providers per 100,000 population, the presence of a professional association of neurologists, the presence of a postgraduate neurology training program, and a national budget line item for neurology care (World Health Organization & World Federation of Neurology, 2004). Epilepsy resource indicators were obtained from the Epilepsy Atlas, compiled by the International League Against Epilepsy (ILAE)/International Bureau for Epilepsy (IBE)/WHO Global Campaign against Epilepsy, and included the following: the number of health care providers who spend >50% of their time providing epilepsy care per 100,000 population, the presence of an epilepsy specialist, the presence of a professional association of epilepsy specialists, the presence of a patient or lay association for epilepsy, the presence of a postgraduate epilepsy training program, and a separate national budget for epilepsy care. Availability and cost of four antiepileptic drugs (AEDs) (phenobarbital, carbamazepine, phenytoin, and valproate) were obtained from the above survey and assessed using the following indicators: the number of those medications listed as essential medication and the annual cost of the medication as a proportion of the GNI per capita (The Global Campaign against Epilepsy, et al. 2005). Other indicators of neurology and epilepsy resources such as the availability of computed tomography (CT), magnetic resonance imaging (MRI), and electroencephalography (EEG) were drawn from a combination of the two surveys. However, because nearly all the included countries had CT and EEG available, these were not included for further analysis.

A quality score was developed for each economic and health resource indicator that corresponds to the mean difference between the prevalence or publication year of the study and the year the indicator was obtained. The quality score was calculated as follows for each data point: absolute value [(year economic or health care variable was estimated) − (year treatment gap was estimated)]. These were averaged for each health or economic variable. The lower the quality score, the closer the match between the study year and the indicator year, such that a quality score of 0 represents a situation where the year the indicator was measured matches the year of the study exactly. Therefore, the highest quality score was zero, and lower quality scores were larger in magnitude. The interpretation of the quality score varies between variables, as some economic and health care variables may be more stable (presence of lay association for epilepsy) or less stable (neurology hospital beds) over time. The quality score was not incorporated into multivariate models but was used as criteria to choose the highest quality measures for inclusion into multivariate models.

Because our goal was to look specifically at the independent impact of neurology and epilepsy-specific resources on rates of untreated epilepsy, we developed parsimonious multivariate models. Our prior systematic review demonstrated a strong influence of rural location; therefore, rural location was included in the models (Meyer et al., 2010). In addition, we controlled for only one macroeconomic indicator and health care resource indicator. We chose the macroeconomic and health care indicators according to the following criteria: (1) quality score near zero or stable estimates over time; (2) significant associations with the treatment gap in bivariate model; and (3) use in similar analyses as a predictor variable (Anand & Barnighausen, 2004, 2007). Therefore, we selected GNI per capita as a measure of macroeconomic resources and physician density as our measure of health care resources.

Poisson regression models with jackknifed standard errors were used to create bivariate and multivariate models that compared the association between treatment status and economic and health resource indicators, and relative risks were reported. Random effects were used to account for clustering at the study level, and country was used as a panel variable to account for clustering at a country level. Relative risks were reported instead of odds ratios because the prevalence of untreated epilepsy was high. Because we used aggregate data to explore features of individual risk, relative risks should be interpreted as relative rates of untreated epilepsy. Sensitivity analysis was performed for bivariate models using the following: (1) random effects Poisson models without jackknifed standard errors and (2) generalized estimating equations using robust standard errors. STATA (Version 11.0, StataCorp, College Station, TX, U.S.A.) was used for all analyses. A p-value of <0.05 was considered significant.

Results

The search generated 23,184 titles. Hand searching of 71 reviews of epilepsy prevalence generated an additional 30 unique titles. All titles were reviewed to identify potential epilepsy prevalence studies, then 571 abstracts and 296 full manuscripts were reviewed to identify 49 studies of 47 populations that met inclusion criteria (Fig. 1; Table S2).

Figure 1.


Literature search and papers selected for inclusion.

The median treatment gap among included studies was 76% (range 7–100%) and was calculated from 47 treatment-gap estimates representing 8,285 individuals from 24 countries (Table 1). A mean of one treatment-gap estimate was available per country (range 1–10), and studies were performed between 1982 and 2010. Study sizes ranged widely from 25 to 1,175 epilepsy cases, with a median of 90 cases. Treatment gap estimates originated primarily from low income countries (n = 29; 62%), although lower middle (n = 8; 17%) and upper middle (n = 10; 21%) countries were also represented.

Table 1.   Mean quality score and summary statistics for health and economic indicators
IndicatorMean quality scoreSummary statisticsa
  1. aThe number of studies included for analysis was 47 except as indicated.

Study population characteristics  
 Percent of studies performed in rural location [% (n)]055% (26)
 Percent of studies including only children [% (n)]09% (4)
Economic  
 Gross national income per capita in current US$ [median (range)]0420 (180–5,040)
 Aid as percentage of GNI in current US$ [median (range)]01% (0–24%)
 External debt as a percentage of GNI in current US$ [median (range)]044% (12–185%)
 Percent of population living below the poverty line of $1.25 per day [median (range)]1.843% (0.5–85%)
 Gini Index [median (range)]1.839 (29–67)
Health care resources  
Provider density   
 Number of physicians per 1,000 population [mean (SD)]5.00.75 (0.90)
 Number of nurses and midwives per 1,000 population [mean (SD)]5.11.4 (1.11)
 Number of hospital beds per 10,000 population [mean (SD)]7.614.6 (10.5)
Proxy measures of access to care   
 Total expenditure on health as a percentage of the gross domestic product [median (range)]1.94.3% (2.5–9%)
 Adult literacy rate (%) [median (range)]3.561% (19–100%)
 Percentage of births attended by skilled health personnel [median (range)]4.042% (3–100%)
 DTP coverage (the percentage of 1-year-old children immunized with three doses of diphtheria, tetanus toxoid, and pertussis [median (range)]0.0472% (9–99%)
 MCV coverage (the percentage of 1-year-old children immunized with a measles containing vaccine) [median (range)]0.0467% (1–99%)
Neurology and epilepsy-specific care  
Provider density and availability   
 Number of neurology beds per 100,000 population (n = 32) [mean (SD)]7.00.073 (0.12)
 Number of neurologists per 100,000 population (n = 40) [mean (SD)]7.00.80 (2.12)
 Number of neurology health care providers per 100,000 population (n = 40) [mean (SD)]7.01.66 (6.30)
 Presence of an epilepsy specialist (n = 40) [% (n)]8.080% (32)
 Number of health care providers who spend >50% of their time providing epilepsy care (n = 43) [mean (SD)]8.00.35 (1.9)
Organization of neurology and epilepsy care   
 Presence of a professional association of neurologists (n = 40) [% (n)]7.088% (35)
 Presence of a professional association of epilepsy specialists (n = 44) [% (n)]8.075% (33)
 Presence of a patient or lay association for epilepsy (n = 44) [% (n)]8.093% (41)
 Presence of a postgraduate neurology training program (n = 40) [% (n)]7.075% (30)
 Presence of a postgraduate epilepsy training program (n = 44) [% (n)]8.016% (7)
Other neurology and epilepsy resources   
 Presence of magnetic resonance imaging machine (n = 46) [% (n)]7.565% (30)
 Number of epilepsy drugs on the essential drug list (n = 36) [mean (SD)]8.03.0 (1.1)
 Annual cost of phenobarbital as a percentage of the GNI per capita (n = 31) [median (range)]8.01% (0–30%)
 Annual cost of carbamazepine as a percentage of the GNI per capita (n = 36) [median (range)]8.025% (0–277%)
 Annual cost of phenytoin as a percentage of the GNI per capita (n = 35) [median (range)]8.02% (0–43%)
 Annual cost of valproate as a percentage of the GNI per capita (n = 35) [median (range)]8.027% (0–569%)

Summary statistics for selected study population characteristics, economic and health care resources, as well as indicators of neurology and epilepsy-specific care resources are presented in Table 1. Of the studies included in our sample, 55% were performed in a rural location. In our sample, the median GNIpc was $420, and 43% of the population lived below the poverty line of $1.25 per day. Health care, neurology, and epilepsy-specific resources were similarly limited. On average, in our sample there were 0.8 neurologists per 100,000 population, and all reported at least one EEG in the country. One country reported that there were no CT scanners in the country, and only 65% reported at least one MRI in the country. Epilepsy drugs varied in relative cost. The median cost of an annual supply of phenobarbital was 1% of the GNIpc, whereas the median cost of an annual supply of valproate cost was 27% of the GNIpc. Most of the economic and some of the general health care resources had the highest quality scores (scores of 0 which represent that the indicator and study data were obtained in the same year). However, the neurology and epilepsy-specific indicators had lower quality scores (scores between 7 and 8 which represent that the indicator and study data were obtained 7–8 years apart). The discrepancy in quality scores likely because the neurology and epilepsy-specific data were obtained in a special one-time survey, whereas the other indicators are collected on a routine basis by large international organizations.

Bivariate analysis demonstrated that many of the economic, health care, neurology, and epilepsy-specific indicators had significant associations with the treatment gap (Table 2). Among the most striking, individuals from a rural location were 1.63 (95% CI 1.26–2.11) times more likely to have untreated epilepsy. Individuals from countries with one additional doctor per 1,000 population were 0.65 (95% CI 0.55–0.78) times less likely to have untreated epilepsy. Although residing in a country with greater GNIpc was not significantly associated with the gap, other measures of poverty such as the amount of aid received or the amount of external debt held by a country were associated with a higher risk of untreated epilepsy.

Table 2.   Bivariate and multivariate relationships between having untreated epilepsy and selected population characteristics, economic, health, and neurology care indicators
IndicatorBivariateMultivariatea
Relative risk
95% CI
Relative risk
95% CI
  1. aAll multivariate models included rural, GNI per capita, and the number of physicians per 1,000 population.

  2. bThe international dollar is a hypothetical unit of currency that has the value of the $US at a certain point in time. It is used to standardize values across countries and over time.

  3. cDistribution not appropriate for multivariate modeling due to multicollinearity.

  4. *p ≤ 0.05; †p ≤ 0.01; ‡p ≤ 0.001.

Study population characteristics  
 Studies performed in rural location1.63‡ [1.26–2.11]1.40† [1.10–1.78]
 Studies including only children0.70 [0.28–1.77] 
Economic  
 Gross national income per capita (GNI) in international $b0.85 [0.69–1.05]1.05 [0.84–1.30]
 Aid as percentage of GNI in current US$1.02† [1.01–1.04] 
 External debt as a percentage of GNI in current US$1.00* [1.00–1.00] 
 Percent of population living below the poverty line of $1.25 per day1.01† [1.00–1.01] 
 Gini Index0.99 [0.98–1.01] 
Health care resources  
Provider density   
 Number of physicians per 1,000 population0.65‡ [0.55–0.78]0.67† [0.52–0.88]
 Number of nurses and midwives per 1,000 population0.90 [0.79–1.03] 
 Number of hospital beds per 10,000 population0.98 [0.96–1.00] 
Proxy measures of access to care   
 Total expenditure on health as a percentage of the gross domestic product0.89 [0.78–1.02] 
 Adult literacy rate0.99 [0.99–1.00] 
 Percentage of births attended by skilled health personnel0.99* [0.99–1.00] 
 Percentage of 1-year-old children immunized with a measles containing vaccine1.00 [0.99–1.00] 
 Percentage of 1-year-old children immunized with three doses of diphtheria-tetanus-pertussis vaccine1.00 [0.99–1.00] 
Neurology and epilepsy-specific care  
Provider density and availability   
 Number of neurology beds per 100,000 populationc0.26 [0.01–7.74] 
 Number of neurologists per 100,000 population0.84 [0.66–1.07]0.98 [0.64–1.50]
 Number of neurology health care providers per 100,000 populationc0.95 [0.74–1.22] 
 Presence of an epilepsy specialist0.92 [0.66–1.29]0.81 [0.57–1.15]
 Number of neurology health care providers who spend >50% of their time providing epilepsy care per 100,000 populationc0.85 [0.45–1.59] 
Organization of neurology and epilepsy care   
 Presence of a professional association of neurologists0.86 [0.59–1.25]1.13 [0.82–1.58]
 Presence of a professional association of epilepsy specialists0.73† [0.59–0.91]1.05 [0.84–1.32]
 Presence of a patient or lay association for epilepsy0.74† [0.60–0.91]0.74* [0.55–0.98]
 Presence of a postgraduate neurology training program0.67‡ [0.55–0.82]0.91 [0.72–1.17]
 Presence of a postgraduate epilepsy training program0.82 [0.49–1.37]1.07 [0.73–1.56]
Other neurology and epilepsy resources   
 Presence of magnetic resonance imaging machine0.77* [0.62–0.96]1.16 [0.91–1.49]
 Number of epilepsy drugs on the essential drug list0.99 [0.86–1.13]1.05 [0.91–1.23]
 Annual cost of phenobarbital as a percentage of the GNI per capita1.01 [0.80–1.27]1.00 [0.82–1.21]
 Annual cost of carbamazepine as a percentage of the GNI per capita1.00† [1.00–1.00]1.00 [1.00–1.00]
 Annual cost of phenytoin as a percentage of the GNI per capita1.00 [0.99–1.02]0.99 [0.98–1.00]
 Annual cost of valproate as a percentage of the GNI per capita1.00‡ [1.00–1.00]1.00 [1.00–1.00]

Most neurology and epilepsy-specific resources were associated with reduced risks of untreated epilepsy as well. The presence of a postgraduate neurology training program was associated with 0.67 (95% CI 0.55–0.82) reduced relative risk of untreated epilepsy, the presence of a professional epilepsy association with a 0.73 (95% CI 0.59–0.91) reduced relative risk, and the presence of an MRI with a 0.77 (95% CI 0.62–0.96) reduced relative risk. A lower cost of carbamazepine or valproate was significantly associated with reduced risk of untreated epilepsy, although the magnitude of the risk reduction was very small. A lower cost of phenobarbital and phenytoin was not significantly associated with reduced risk of epilepsy, but the median annual cost of phenobarbital (1%) and phenytoin (2%) as a percentage of GNIpc was substantially lower than carbamazepine (25%) or valproate (27%). Multivariate models demonstrated that after controlling for rural location, GNIpc, and physician density, only the presence of a patient or lay association for epilepsy was associated with a 0.74 (95% CI 0.55–0.98) lower risk of untreated epilepsy.

Discussion

Our study demonstrates that although estimates of treatment gap among resource-constrained countries are limited, among the world’s poorest countries, the economic gradient and availability of health care, neurology, and epilepsy-specific resources matter. Persons living in rural locations, in countries with high debt burden and dependence on foreign aid are more likely to have untreated epilepsy. Persons living in countries with few doctors, high proportion of home births, and without specialized care for neurology or epilepsy are more likely to have untreated epilepsy.

As in other disease processes, the distribution of health care resources between rural and urban areas is likely a critical factor in determining the epilepsy-treatment gap (Hobcraft et al., 1984; Bender et al., 1993; Brockerhoff, 1995; Sastry, 1997; Senior et al., 2000; Fotso, 2007). Because rural areas face the heaviest burden of untreated epilepsy, it is essential to ensure that the distribution of health care and economic resources extends to the rural areas.

Finally, the presence of specialty care in the country is essential to maintain a continuum of care. Specialists not only have a role to play in managing and treating complex cases, but they are especially needed to provide education and training, and ongoing supervision and support to nonspecialists working at the primary health care level. Although multivariate modeling did not demonstrate significant associations between the treatment gap and most neurology and epilepsy specific health care resources, our analysis reflects availability of these resources at a national level and may not be representative of the barriers faced by individuals attempting to obtain care or medications for their epilepsy. For example, in India, 70% of medical practitioners reside in urban locations, whereas 70% of the population lives in rural areas (Mani & Subbakrishna, 2003). Furthermore, although AEDs may be listed as essential drugs by a particular country, this may not correspond to a reliable supply of AEDs to rural areas. Although phenobarbital may be affordable in many countries, an annual supply cost up to 30% of the GNIpc in some countries. Other AEDs were similarly expensive; an annual supply of phenytoin cost up to 43% of the GNIpc, carbamazepine up to 277%, and valproate up to 569%. In addition, resources such as MRI may be available only in the capital city and only in the private sector, rendering such resources inaccessible to individuals who reside in rural areas or without the financial resources to pay for the study. Further research efforts to collect individual level demographic and socioeconomic indicators as well as region-specific information about health care resources would help further elucidate the social and economic determinants of the treatment gap on an individual and regional scale.

The major limitation of this study is that we analyzed country level determinants of an individual’s risk for untreated epilepsy; therefore, a major limitation of this study is that interpretation of the results is at risk of the ecological fallacy. Although this is important preliminary research, it is not a substitute for more detailed study of individual-level determinants of risk. However, few individual or regional level data about economic or health care resources were available, so aggregate national data were used in its place. Only one study assessed individual level determinants of the treatment gap; in this study from Brazil, individual socioeconomic status was not correlated with treatment gaps (Noronha et al., 2007).

Another limitation is related to the method used to ascertain the neurology and epilepsy-specific variables; these data were derived from a survey of key informants, and its representativeness of the actual resources available in a country has not been verified. In addition, several of the measures of neurology and epilepsy-specific care demonstrated high multicollinearity with the more general measures of economic and health care resources. Furthermore, we were unable to perfectly match health and economic indicators to the year the treatment gap data was collected, so estimates may not reflect the true availability of these resources at the time the treatment gap data was collected. Finally, because treatment gap estimates are limited, our sample size was relatively small, which limited our ability to analyze the effects of many of our indicators; only about half of eligible studies that we reviewed collected data on the treatment gap (Meyer et al., 2010). In addition, most of these studies were done in populations that are not representative of the nation as a whole, which may have affected our overall treatment-gap estimates. Nationally representative population-based data for the epilepsy treatment gap and more data characterizing neurology and epilepsy-specific resources available for care in resource-poor regions would improve future cross-country comparisons.

In summary, even among resource-limited regions, people with epilepsy who live in countries with fewer economic, health care, neurology, and epilepsy-specific resources are more likely to have untreated epilepsy. A critical area of future research will be to ascertain individual level determinants of the risk for untreated epilepsy. The consequences of untreated epilepsy include high morbidity and mortality, social consequences including stigma and discrimination, and high economic costs. Large community-based trials in China and Brazil conducted by the ILAE/IBE/WHO Global Campaign against Epilepsy have demonstrated that epilepsy can be effectively treated at a community level with inexpensive drugs by health care workers with basic training (Wang et al., 2006; Li et al., 2007). In addition to community-based epilepsy care, our study suggests that the epilepsy treatment gap can only be reduced if poverty and inequalities of health care, neurology, and epilepsy resources are dealt with at the local, national, and global levels.

Acknowledgments

This study was supported by the Veterans Affairs/Robert Wood Johnson Clinical Scholars Program and the American Academy of Neurology Practice Research Training Fellowship. Dr. Birbeck was supported by the Global Burden of Diseases, Injuries and Risk Factors Study as well as the U.S. National Institutes of health (NIH). (R01NS061693) We would like to gratefully acknowledge the assistance of three individuals who helped translate articles from Chinese, Russian, and Croatian, respectively: Dr. Ding Ding, Ms. Marina Marcus, and Prof. Igor Rudan. Dr. Meyer translated from Spanish, Portuguese, French, and Italian.

Disclosure

Dr. Meyer was funded by the Veterans Affairs/Robert Wood Johnson Clinical Scholars Program and the American Academy of Neurology Practice Research Training Fellowship. She is also funded by the National Institutes of Health and the Hellman Family Foundation. She has served as a paid and unpaid consultant to the World Health Organization. She has nothing further to disclose. Dr. Dua has nothing to disclose. Dr. Boscardin receives support from NIH and Department of Veteran’s Affairs funded research projects and serves on a Data Monitoring Committee for a Pfizer study. Dr. Escarce receives funding for research projects from the National Institutes of Health, the Robert Wood Johnson Foundation, and the Agency for Healthcare Research and Quality. Dr. Saxena has nothing to disclose. Dr. Birbeck was supported by the Global Burden of Diseases, Injuries and Risk Factors Study as well as NIH-funded research on epilepsy and HIV. She has served as an unpaid advisor to the World Health Organization. She has nothing further to disclose. We confirm that we have read the Journal’s position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.

Ancillary