Regional variation in prevalence and healthcare utilization due to epilepsy in Sweden




To estimate the regional differences in the prevalence of epilepsy and the associated costs due to inpatient and outpatient care and anti-epileptic drug (AED) utilization for the years 2005 and 2011 in Sweden.


Region-specific estimates of the prevalence of epilepsy were obtained using a method based on a linkage of the healthcare and pharmaceutical registries and the cause of death registry. Regional cost components were estimated using registry data by region on inpatient and outpatient care utilization, AED sales, and mortality. Per-patient utilization and monetary costs were calculated.


Estimated prevalence of epilepsy varied substantially across the regions in 2011, from 0.76% in Jämtland to 1.08% in Gotland. The national prevalence was 0.88%. The average number of hospitalizations per patient and year decreased at the national level between 2005 and 2011. At the national level, the per-patient specialized care (outpatient) utilization also decreased between 2005 and 2011. However, at the regional level, the decrease was not uniform, and in some counties, the per-patient utilization increased during the period studied. The per-patient utilization of AEDs increased in all counties, except Kronoberg, between 2005 and 2011. Moreover, between-region differences in healthcare and AED utilization, and significant differences between regions and national averages were revealed. Similarly, regional per-patient costs were shown to deviate from the national average in 13 of 21 regions.


There is significant variation in the prevalence of epilepsy and the provision of health care for patients with epilepsy across the different regions of Sweden.


Healthcare and pharmaceutical utilization may vary across different regions of a country for reasons pertaining to both population and healthcare-provider characteristics [1, 2]. Clearly, the larger the population, the greater the healthcare utilization, ceteris paribus. In addition, qualitative aspects of the disease in the patient population, such as disease severity, as well as other individual characteristics, such as age and socioeconomic status, are all likely to contribute to healthcare utilization [2, 3]. The supply of health care is restricted by organizational conditions and resource constraints. Also, treatment practices may vary among care providers. The Swedish National Board of Health and Welfare have developed and issued treatment guidelines for several conditions, but epilepsy is not among them. For epilepsy, only guidelines for the use of anti-epileptic drugs (AEDs), issued by the Swedish Medical Products Agency [4], exist. As treatment guidelines are advisory only, and treatment decisions are based on judgments by an individual, or a team, of physicians, variations in regional treatment practices would seem plausible.

Health care are provided by 21 different county councils in Sweden. The three main metropolitan areas – Stockholm, Västra Götaland (Gothenburg), and Region Skåne (Malmö) – together comprise about 50% of the Swedish population. Geographical distances to adequate specialized care centers are likely to be larger for people residing in less populated areas. Furthermore, metropolitan areas are typically populated with people who have higher socioeconomic status than residents of rural areas.

In this study, we determined the regional differences regarding the prevalence, healthcare utilization, and monetary costs of epilepsy in Sweden. Prevalence and cost of epilepsy by region was determined for 2 years, 2005 and 2011, for comparison purposes.

Material and methods

A detailed description of our methodology is provided in the online supplement and in Bolin et al. (this issue). Briefly, we applied a prevalence-based approach and a top-down methodology to estimate the different direct cost components, in- and outpatient care and pharmaceutical (AEDs) utilization [5-7]. Indirect costs – due to permanent disability, short-term illness and mortality – were not included. A previous study performed using the same method found that indirect costs accounted for more than 80% of total costs [5]. There are many aspects on how to measure prevalence of epilepsy, and obvious difficulties to make appropriate estimations, as the condition can be silent over long time [8]. In this article, the prevalence of epilepsy by region was estimated using a method based on a identifying and linking individual patients in the healthcare registries and the cause of death registry [5]. The prevalence of epilepsy in 2011 was estimated as the number of living individuals that utilized health care due to epilepsy at least once in the years 1998–2011 (the years for which the healthcare registries are coded using ICD-10). This provided a measure of the life-time prevalence of epilepsy. Region-specific estimates of the prevalence of epilepsy were obtained by counting the number of living individuals that have received in- or outpatient care due to epilepsy (ICD10: G40 and G41) separately for each region. Patients that have received care in more than one county were included in the county that provided the most recent specialized visit or hospital stay.


The regional prevalence of epilepsy in 2011 ranged from 0.76% in Jämtland to 1.08% in Gotland, while the nationwide prevalence was estimated at 0.88% (Table 1). The estimated prevalence of epilepsy increased by an average of 42% across the country between 2005 and 2011. The largest increase was observed in Värmland (78%) and the lowest in Stockholm (30%).

Table 1. Prevalence of epilepsy and relative epilepsy-related healthcare utilization in 2005 and 2011 by region. Share of total health care, total inpatient care, total outpatient care, and total anti-epileptic drug (AED) utilization is calculated as regional utilization divided by total utilization. Regions are presented in descending order according to the 2011 per 100,000 prevalence. National prevalence is reported in the row labelled SWEDEN. Notice, that the ordering according to prevalence does not carry over to the results reported in the other columns
RegionPrevalence per 100,000 inhabitantsShare of inpatient careShare of outpatient careShare of AED utilization
  1. a

    Change from 2005 to 2011.

  2. b

    Statistically significant difference from the national level (P ≤ 0.01, Student's t-test). The 1% confidence interval is 832–934.

Västra Götaland63388941%17.9%17.2%14.0%13.0%17.6%17.2%

We calculated epilepsy-related healthcare utilization and monetary costs in 2005 and 2011 by region, and, for comparison purposes, at the national level. Regional differences in per-patient healthcare and AED utilization are shown in Table 1 [comprehensive AED sells data are reported in a parallel study [9]]. The per-patient number of hospitalizations (inpatient visits) decreased in all counties – at the national level, the number was 0.15 in 2005 and 0.12 in 2011 (Table 2). The per-patient number of visits to hospital-based outpatient care also decreased at the national level, from 0.60 in 2005 to 0.49 in 2011. However, at the regional level, the change was not uniform: An increase was observed in five counties, namely Gotland, Skåne, Värmland, Norrbotten, and Jämtland. AED utilization, measured as number of daily defined doses, DDD (DDD is defined as the assumed average maintenance dose per day for a drug used for its main indication in adults), varied between 315 (Norrbotten) and 197 (Stockholm) – nationwide utilization was estimated at 233. Per-patient AED utilization increased between 2005 and 2011 in all regions with the exception of Kronoberg.

Table 2. Per-patient healthcare utilization by region in 2005 and 2011. Regions are presented in descending order according to the 2011 prevalence per 100,000 inhabitants (see Table 1)
County councilInpatient visitsOutpatient visitsDrug utilizationa (1000 DDDs)
  1. DDDs, daily defined dose (DDD is defined as the assumed average maintenance dose per day for a drug used for its main indication in adults).

  2. a

    The sum of the pharmaceutical utilization in the different regions is not exactly equal to the pharmaceutical utilization in Sweden in 2011. The discrepancy is due to missing data for age and county council for some observations.

  3. b

    Statistically significant difference from the national level (1% significance level; Student's t-test). The 1% confidence intervals are 0.106–0.134; 0.421–0.561; 212–254.

Västra götaland0.160.120.490.38b152238

When comparing regional and national healthcare and AED utilization in 2011, a number of statistically significant differences were observed (Table 2) (the 1% confidence interval is reported at the bottom of the Table). With regard to inpatient care, the per-patient number of visits was significantly lower than the national average in Blekinge, Gävleborg, Jämtland, Kalmar, Sörmland, and Västmanland, while it was higher in Uppsala and Västerbotten. The per-patient number of outpatient specialized visits was significantly higher than the national average in Gotland and Kalmar, while it was lower in Stockholm. The per-patient AED utilization (number of DDDs) was significantly lower than the national average (233) in Stockholm, while the utilization was significantly higher in eight regions.

The estimated 2011 per-patient total cost ranged between €913 (Gävleborg) and € 1221 (Värmland) (Table 3). The nationwide per-patient cost was estimated at €1047 for the same year. The estimated per-patient total cost increased from 2005 to 2011 in all regions. The results are reported in Tables 1-3, below. For comparison purposes, each table also reports national-level figures. The national-level results were calculated using aggregate data, while regional estimates employed data that excluded observations without valid information about county council. This accounts for the discrepancy between aggregate estimates and sum of the corresponding regional estimates. Further, in Tables 1 and 2, the regions were ordered in descending order from top to bottom according to prevalence per 100,000 inhabitants. In Table 3, the ordering was based on per-patient total cost. In each table, the national-level figures have been inserted for comparison purposes.

Table 3. Monetary costs of epilepsy by region in the years 2005 and 2011. Regions are presented in descending order according to the 2011 per-patient total cost
County councilPer-patient total cost (€)Per-patient costs (€)Total cost per 100,000 inhabitants (million €)
Inpatient careOutpatient carePharmaceuticals
  1. a

    Statistically significant difference from the national level (1% significance level; Student's t-test). The 1% confidence interval is 991–1103.

Västra Götaland883983a4423401501652924780.560.87


In this study, we calculated regional-specific estimates of the life-time prevalence of epilepsy, and the healthcare and AED utilization due to epilepsy (Sweden, 2005 and 2011). The use of national registers and disease-specific quality registries has increased the last years in Sweden. So far, no quality register has been established for epilepsy. This means that potential improvements in quality and cost-effectiveness of epilepsy treatments may not be realized. This study demonstrates that there are significant regional differences in the utilization of epilepsy-induced healthcare and AED treatment. The data employed in this study does not allow for explanatory analyses of these differences, though. This would require a collection of more comprehensive data from available registries. The results obtained in this study may serve as a guide for the construction of such a database.

We found a large increase in the estimated prevalence of epilepsy between 2005 and 2011. To some extent, this is likely to be an artifact, introduced by the longitudinal content of the healthcare registries. Patients with epilepsy with a satisfactory controlled disease may have had only one visit to a specialist – at the onset of the disease – and may be followed by their general practitioners only [10]. As outpatient visits have only been systematically collected from, and including, the year 2001, such patients may not be included in our data (primary care is not included in the national healthcare registries). In a parallel study, we derived an approximation of the extent to which the 2005 prevalence was underestimated. We found that, over time, the national healthcare registries will comprise an increasing share of the total life-time epilepsy population [9].

The estimated prevalence of epilepsy in nine of 21 regions in 2011 differed significantly (P ≤ 0.01) from the national prevalence. The national epilepsy prevalence was estimated at 883 per 100,000 in 2011; the highest 2011 prevalence was found for Gotland (1082 per 100,000) and the lowest for Jämtland (762 per 100,000). Closeness to one of the three main metropolitan areas in Sweden could be expected to decrease the likelihood of epilepsy staying undiagnosed. Then, without any true regional differences in prevalence, we would observe the highest prevalence numbers for those regions. However, closeness to one of the metropolitan areas does not appear to explain significant differences between regional and national prevalence. Thus, the observed variance seems to be accounted for by true regional differences in the prevalence of epilepsy, which, in turn, may be generated by demographic and socioeconomic differences and/or institutional differences between regions. Empirical studies exploring these relationships necessitate more detailed information regarding socioeconomic, demographic and institutional regional differences.

Regional per-patient healthcare and pharmaceutical utilization showed large variation between regions. For 2011, the national per-patient number of inpatient visits was estimated at 0.12. The corresponding regional number of inpatient visits was estimated in the range 0.06–0.15. Similarly, the national per-patient number of outpatient visits was estimated at 0.49, and the regional number of visits was estimated in the range 0.34–0.68. Thus, the average patient visits his or her neurologist once every other year. Prescription renewals are frequently handled by GPs, or by neurologists by telephone, in which case no visit to the clinic is registered. In addition, there may be structural factors pertaining to the organization of the healthcare sector that favors the prioritization of incident cases. The national-level per-patient utilization of AEDs was estimated at 233 DDDs, and the regional utilizations were estimated in the range 197–315. Several of the comparisons between regional utilization and the corresponding national mean in 2011 resulted in statistically significant differences (P ≤ 0.01). As regards inpatient care utilization, five regions differed significantly from the national mean in 2011; the corresponding numbers of significantly different regions for outpatient care and pharmaceutical utilization, respectively, were 3 and 4.

When compared to the findings reported by a German research group, Strzelczyk et al. [11], for the years 2003 and 2008, respectively, our results differ quite much. This pertains both to trends and to point estimates. We start by commenting on trends. First, the Strzelczyk et al.'s study found that the cost of hospitalizations increased between 2003 and 2008, while we found a sharp decrease in per-patient hospitalization costs. Second, the Strzelczyk et al.'s finding that the (mean) AED cost is lower for the 2008 population than for the 2003 population stands in sharp contrast to our findings regarding the utilization of AEDs. Our finding that the cost of specialized outpatient care has increased between 2005 and 2011 is similar to their findings, though. In a subsequent study [12], the healthcare utilization of adult epilepsy patients was estimated for the same geographical region (performed by what appears to be the same research group). In this latter study, the per-patient cost was found to be lower than what have been reported in other German studies, suggesting that the population used in [10] may not be representative for the nation-wide German population.

As regards point estimates, we found that the nation-wide per-patient direct cost in 2011 was € 1047. This is considerably lower than recent direct per-patient costs reported for the German population (€ 2406) and for a European population (Swedish cost: € 2461) [3, 11]. So, how can these differences be understood? One plausible explanation is methodological differences. Our study used total-population register data (top-down method) to estimate per-patient costs, while the other studies that we used for comparison utilized patient-level information for subpopulations to estimate population-level costs. The patient-level costs that were reported in [3] and [11, 12] were collected from subpopulations – the German studies enrolled 366 and 110 + 151 patients, respectively. Both methods of estimating per-patient costs have their advantages and disadvantages. The top-down method that we used minimizes the risk of double-counting costs, while the bottom-up method provides more detailed information about different cost components that are not systematically collected in registries.

In this study, we found a small increase in per-patient direct costs between 2005 and 2011. In the study parallel to this (see above), we found a significant increase in total costs. However, when comparing the per-patient costs between 2005 and 2011, the previous discussion about the estimated prevalence in 2005 and 2011, respectively, should be kept in mind: Had the prevalence rate used for calculating the 2005 per-patient figures been adjusted for prevalent epilepsy patient that was not observed, we would have seen a significant increase in per-patient cost between 2005 and 2011, for all regions. This is consistent with the findings in the German study discussed above [12].

Our findings reveal significant regional differences in both epilepsy prevalence and health care and AEDs utilized by patients with epilepsy. Moreover, the change in per-patient healthcare and drug utilization between 2005 and 2011 also differs between regions. As a consequence, the regional per-patient costs differed significantly from the national mean in more than half of the cases (13 of 21) in 2011. Analysis of the inter-region variance of the three cost components – in- and outpatient care and AEDs – suggested that differences in AED costs are the single largest contributing factor for differences in total per-patient costs. Geographical distances to advanced healthcare services do not seem to explain these results. This suggests that the observed differences originate from regional demographic and socioeconomic disparities, differences in disease severity, and diverse clinical practices. The findings in a recent study of the importance of treatability for direct healthcare costs suggest that per-patient direct costs are highly sensitive to average drug responsiveness in a population [13]. Thus, a small difference in average drug responsiveness between regions would produce significant differences in per-patient costs. Further, studies performed for Danish and Swedish populations suggest that socioeconomic differences are related to the incidence of epilepsy [14], and to the utilization of AEDs [2]. Thus, a plausible hypothesis for future research seems to be that such differences are also related to disease severity, which would offer another explanation for the regional per-patient cost differences that we found in this study. A first check of whether this may be the case would amount to simply comparing measures of average socioeconomic measures between regions.


We gratefully acknowledge the financial contribution from UCB Nordic, Denmark. The authors also thank Azita Tofighy, PhD, UCB Pharma for reviewing the manuscript and for editorial support. 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.

Conflict of interest

Authors Kristian Bolin and Anne-Marie Landtblom have received financial support from UCB Nordic A/S. Fredrik Berggren is an employee of UCB Nordic A/S.