Economic burden of migraine in Latvia and Lithuania: direct and indirect costs

The economic cost of migraine in Latvia and Lithuania is estimated in line with the best practice cost-of-illness approach. This approach requires the estimation of the prevalence rate, the proportion of people in a population who suffer from migraine in the base period, in 2017.

The prevalence rate is estimated by identifying systematic reviews on migraine prevalence from a systematic review of the literature in PubMed and Scopus.¹ From the 45 studies identified only three deemed relevant; these carefully executed, high-quality systematic reviews are listed in panel A of Table 1.² For current prevalence, for Europe, Stovner et al. [15] estimated an arithmetic average migraine prevalence of 15% in their meta-analysis. For lifetime prevalence, for Europe, the authors reported an arithmetic average migraine prevalence of 17% [15]. In an updated study, Stovner and Andrée [17] estimated that the mean prevalence of current migraine is slightly lower, 14.7% in Europe. In contrast, Woldeamanuel and Cowan [18] derived a migraine prevalence rate of 11.4% for Europe. The difference in migraine prevalence between the two studies (14.7% in [17] vs 11.4% in [18]) can be explained by the facts that Woldeamanuel and Cowan [18] employed weighted averages and included more up-to-date studies; both factors render their estimates more reliable.

In Europe, valuable prevalence data has been recently collected in the Eurolight project, an initiative supported by the European Commission Executive Agency for Health and Consumers [19]. Eurolight collected data on headache disorders in a cross-sectional questionnaire-based survey in ten European countries, which together represented over 60% of the adult population (18–65 years) of the European Union. The most important characteristics and the major findings of the Eurolight project is summarized in panel B of Table 1.

None of the systematic reviews shown in Panel A of Table 1 listed studies estimating prevalence rate of migraine in the Baltic countries. By undertaking a systematic search of the published literature in PubMed and Scopus, we concluded that no prevalence rate for migraine is available in Latvia, while three recent studies cover the other two Baltic countries.³ The Eurolight project assesses the prevalence rate of migraine in Lithuania [19, 20], while Toom et al. [21] report prevalence rate of migraine for Estonia. The most important characteristics and the major findings of studies covering the Baltic countries are summarized in Panel C of Table 1.

To conclude, it is reasonable to assume that the prevalence rate of migraine in the Baltics is close to those meta-analysis estimates which combine the findings of several European studies on migraine prevalence. The prevalence rate of 11.4% estimated by Woldeamanuel and Cowan [18] for Europe involving a sample size of 808,749 participants is used in further analysis as a conservative estimate. In sensitivity test the migraine prevalence rate of 12.71% from the Eurolight project for Europe is used as well [19]. The estimate from the Eurolight project is based on surveying 9247 individuals in ten EU countries and reflects migraine on more than 5 days/month.

For Lithuania, we have also considered using the gender-adjusted migraine prevalence rate of 18.8% from the Eurolight project [20]. This estimate is based on surveying 573 adults in Lithuania. Although the sample is population-based and has as reasonable sample size, the estimate classifies definite and probable migraine together as migraine; a drawback which let us exclude this estimate from further analysis. Similarly, we have also considered using the migraine prevalence rate of 17.7% reported for Estonia which is also a population-based estimate [21]. Although the authors distinguish between definite and probable migraine, the weighted prevalence rate of definite migraine is much smaller than in many other studies, only 7.3%, which again let us exclude this estimate from further analysis. In their study, the authors list several sources of possible underestimation [21].

For estimating the direct cost of migraine in Latvia and Lithuania, first migraine-related health care resource utilization shall be estimated. A targeted literature review produced a number of valuable studies; in this subsection these studies are reviewed briefly with the aim of identifying the most relevant source to estimate the migraine-related health care resource utilization in Latvia and Lithuania.

Lantéri-Minet et al. [22] review systematically 34 studies of patients with chronic daily headache, occurring on at least 15 days per month. Although the authors focus on patients with chronic migraine to evaluate the evidence for quality of life impairment, disability, health care resource use and economic burden, they provide an excellent review of studies on resource utilization and economic impact covering several migraine types. The authors document that as of July 2009 resource utilization data was available only from two studies. The GRIM 2000 study is a large nationwide survey of headache characteristics and health care resource utilization in France [16], while the American Migraine Prevalence and Prevention (AMPP) study is a 5-year, longitudinal, national study of headache in the US [6]. Lantéri-Minet et al. [23] and Auray et al. [24] use data from the GRIM study, while Munakata et al. [6] builds upon the survey results of the AMPP study to access the total direct cost of migraine in France and in the US, respectively.

Lantéri-Minet et al. [22] review studies published before July 2009. Since that date, as identified by a targeted literature search, three large surveys have been carried out which can be considered as valuable source of information on migraine-related health care resource utilization: the first International Burden of Migraine Study (IBMS-I), the second International Burden of Migraine Study (IBMS-II), and the National Health and Wellness Survey. A detailed overview of the targeted, web-based methodology applied in IBMS-I is provided in [25]. In IBMS-I participants were recruited and surveyed in ten developed countries, the data collected covered sociodemographic and clinical characteristics, resource utilization, disability, health-related quality of life, anxiety, depression and productivity from over nine thousand participants. IBMS-II was also an international, web-based, cross-sectional survey; it investigated the burden of chronic and episodic migraine in six developed countries. IBMS-II targeted 600–600 patients with chronic and episodic migraine, respectively. In both IBMS-I and IBMS-II only participants meeting ICHD-2 criteria for migraine were included and they were classified as patients with chronic migraine (≥ 15 headache days per month) or episodic migraine (< 15 headache days per month) [25, 26]. The National Health and Wellness Survey was also a cross-sectional, web-based survey with over 80,000 participants from five EU countries [27]. Respondents were members of various online opt-in survey panels; in Italy and Spain online recruitment was supplemented by telephone recruitment among elderly. In that survey 16,340 respondents reported migraine headaches in the past 12 months from which 1680 randomly selected participants filled out the questionnaire on migraine.

The studies using these survey results and documenting migraine-related health care resource utilization are reviewed in Table 2.⁴ Three studies estimate health care resource utilization using the survey data from IBMS-I [28‐30]. Bloudek et al. [28] quantify the direct medical cost, excluding medication, in chronic and episodic migraine in five EU countries. Stokes et al. [29] assess health care resource use and related costs, including medication, in chronic and episodic migraine in the USA and Canada. In contrast, Blumenfeld et al. [30] compare episodic to chronic migraine sufferers and investigate whether the headache frequency determine headache-related disability, health-related quality of life and health care resource utilization pooling data from nine countries together. Blumenfeld et al. [30] report health care resource utilization data for the pooled sample and for visits only, whereas Bloudek et al. [28] and Stokes et al. [29] report results both for visits and diagnostic evaluations split by the five EU and two North-American countries, respectively.

Sanderson et al. [31] quantify migraine-related health care resource utilization using the survey data from IBMS-II. The authors quantify and compare health resource usage in chronic and episodic migraine across six countries. As compared to IBMS-I, the sample is smaller and covers less European countries. Another major drawback of the study is that it documents only the proportion of participants reporting one or more visits, the mean number of visits is not reported.

From the health care resource utilization data of Vo et al. [27] the incremental consumption of health care resources associated with migraine can be derived. Although the sample in [27] is small (n = 218), the estimates are very valuable as they show the incremental usage instead of the migraine-related usage, the approach followed by studies using the survey results from IBMS-I and IBMS-II. Vo et al. [27] report an incremental visit of 2.57, whereas Bloudek et al. [28] find that the total number of migraine-related health care provider visits is 3.17.⁵ This comparison signals that migraine-related health care resource utilization might be slightly overstated in studies without propensity matched controls.

In the recent study of Martelletti et al. [32], the authors describe the disease burden among individuals with migraine for whom preventive treatments failed. The authors administered an online survey in 31 countries worldwide using online bulletin boards. In their sample around 80% of the respondents had a history of prophylactic treatment failure. The authors report higher health care resource utilization (brain scan, emergency department visits, hospital stay) for patients who switched therapies at least two times.

Health care resource utilization estimates in this study are derived from Bloudek et al. [28] for a number of reasons. First, in [28], similar to all other studies using the survey results of IBMS, migreneurs were selected carefully, patients had to meet the ICHD-II diagnostic criteria for migraine. Second, the methodology of IBMS was carefully designed and validated and followed the same approach across the sample countries [25]. Third, the international sample of patients with chronic and episodic migraine is large; as of now the largest and the most recent for Europe. Fourth, it covers five European countries, not just one. Fifth, it delivers estimates both about visiting health professionals and diagnostic evaluations. As a drawback, no patients were surveyed from Central and Eastern Europe and the authors did not document the usage of headache medication.

Health care resource utilization estimates used in this study are shown in Table 3. In the sample of 5655 patients in [28], 4.88% of patients were suffering from chronic migraine, while 95.12% from episodic migraine. We consider this proportion of chronic/episodic migraineurs as a valid estimate for Latvia and Lithuania as well.

Migraine may lead to reduced participation in the labour force through difficulties in obtaining and keeping full-time work. Migraine sufferers face barriers in finding secure, fulltime employment, and the ones being employed are constantly exposed to the fear of losing their jobs because of repeated absences or migraine-induced dysfunction. Generally, migraineurs are suggested to look for flexible jobs, such as writing, graphic design, programming, accounting and the ones that enable remote work [33]. Migraineurs thus may deliberately not enter the labour market as full-time employees or may willing to do so but unable to find and keep their jobs.

Labour force participation of migraine sufferers is lower than in the total working-age population. Empirical evidence shows that the unemployment rate for patients with less severe headache is very similar to the unemployment rate for the average population [4]. Migraineurs with high pain but moderate activity limitations had twice as high unemployment rate as the average population, while migraineurs with severe activity limitations had more than four times as high unemployment rate. Stang et al. [34] find that over the three-year study period, 12% of patients suffering from headache were unemployed, while 13% unable to obtain or keep their full time work due to their condition. The authors identified five factors increasing the likelihood of being unemployed, suffering from migraine being one of them, while being a female, young (aged 18–24), less educated, and having depressive symptoms being the others. Similarly, Stewart et al. [14] also document reduced workforce participation for migraineurs. The authors find that individuals with chronic and high frequency migraine were less likely to be actively working for pay compared with migraineurs with low frequency headache. In particular, the authors report that only 37% of individuals with chronic migraine were employed full time; the respective figure is 48% for migraineurs with less than ten headache days in the past 3 months.

In majority of studies, unemployment is not captured as a component of indirect cost [5, 6, 35]. Unemployed respondents are typically systematically excluded from the analysis; lost workdays are only estimated for the ones being at least part-time employed. Nevertheless, several studies show that workforce participation of migraine sufferers is lower than in the total working-age population [4, 14, 34]. In this study we assess productivity loss as a result of migraine sufferers being unemployed due to their condition. Although people with high frequency headache is less likely to be employed than the general population [4, 14], we conservatively assume that the prevalence rate of migraine among employed and unemployed migraineurs is the same. Moreover, we assume that only individuals with high frequency migraine are unemployed as a result of their migraine. For the remaining migraineurs the primary reason of their unemployment are conditions other than migraine—lower level of education, injury, poor physical or mental health, etc. These assumptions are in line with the findings reported in [4]. Based on Stewart et al. [14], 10.55% of migraineurs is considered as individuals with high frequency migraine.

Table 4 displays the most important findings of population-based studies documenting the number of days missed from work due to migraine identified by targeted literature search. Four inclusion criteria were defined. First, studies shall report the number of days missed from work for patients with migraine in general. Studies were excluded if they documented absenteeism for specific patient groups only, such as migraine with aura or chronic migraine. Second, respondents should be recruited from the general population. Studies were excluded if they were not representative of the entire population. For example, estimates on absenteeism for 723 headache sufferers at a large Swiss university hospital were excluded [38]. The third restriction was related to the geographic coverage of the studies; studies from North-America or Europe were included only. Fourth, studies should be published in the last 10 years (after 31 December 2008), for obtaining up-to-date information and arriving at a reliable, valid estimate. As a result of the targeted literature search, nine studies met the inclusion criteria, these studies are listed in Table 4 and summarized briefly in the following paragraphs.

The first three studies shown in Table 4 used data from US patients. Munakata et al. [6] retrieve data from the 2006 follow-up survey part of the AMPP study. The authors document that patients with migraine missed 13.7 h per year from work, which is equivalent to 1.71 missed days per year. Kessler et al. [36] use data from the National Comorbidity Survey Replication. The authors investigate the predictive associations between migraines and workplace outcomes, and whether comorbidity can explain this association. The authors measure absenteeism over the past month with the WHO Health and Work Performance Questionnaire. The authors perform linear regression analysis to evaluate the predictive effects of migraines on absenteeism while controlling for socio-demographic characteristics. They find that migraine is significantly associated with absenteeism; it results in 10.7 excess sickness absence days per year. Stewart et al. [14] report estimates on absenteeism for migraine sufferers participating in the AMPP study in 2005 in the US. Respondents who reported active, severe headache in the screening survey received a second, self-administered headache questionnaire about employment and lost productive time. Out of 11,624 respondents meeting the criteria for migraine and included in the analysis, 6204 were employed actively for pay, either full-time or part-time. The authors document missed hours per actively employed worker per week for four patient groups: low, moderate and high frequency headache, and chronic migraine (< 10 days, 10–29 days, 30–44 days, or > 45 days of headache in 3 months, respectively). Among participants with active employment status, lost productive time was substantially higher for respondents with high frequency headache compared to those with low frequency headache. Migraineurs with 30–44 headache days in the last 3 months had the highest probability to report missed workdays due to headache in the last 2 weeks; they were followed by respondents suffering from chronic migraine. As Stewart et al. [14] report hours per worker per week missed from work, the estimates were converted into annual figures for comparison purposes. Moreover, the four patient groups were merged into two groups (patients with episodic migraine and patients with chronic migraine), for details see the footnote added to Table 4.

Steiner et al. [35] surveyed 8271 participants in 9 EU countries in the Eurolight project with the aim of measuring the personal impact of headache. The authors deliver estimates of lost workdays, housework days and social days due to migraine. The authors report that the impact of migraine is severe; 17.7% of males and 28.0% of females lose more than 10 days of activities (workdays, housework days and social days) in a 3-month period. Due to migraine, the total sample population lost 3.2 working days in preceding 3 months, which translates into losing almost 13 days per year. The figure reported in Table 4 can be considered as a combined absenteeism and presenteeism estimate; it includes both the workdays lost completely and the workdays with productivity reduced to 50% or more of the expected productivity.

Ayzenberg et al. [7] evaluate headache-attributed burden and its impact on productivity and quality of life in Russia. Face-to-face interviews were conducted with 2725 adults in urban and rural areas; the random sample is representative of the population. Headache-attributed lost time results were available for 1273 participants reporting headache. For all headaches, the authors estimate that the mean lost paid-work days in the preceding 3 months were 1.9 days. The authors provide detailed data for the following subsamples: participants with migraine, participants with tension-type headache and participants with headache on 15 days or more (either migraine or tension-type headache). In Table 4 data for respondents with migraine is shown; Ayzenberg et al. [7] report that they missed paid work only on 0.2 days in the preceding 3 months.

Vo et al. [27] provide absenteeism estimates for 218 patients with and without migraine from five European countries. The authors find that respondents with migraine when compared with non-migraine controls reported significantly higher absenteeism, the percentage of work time missed in the past 7 days was 14.43% vs 9.46%, respectively. Based on our calculations, migraine sufferers missed 12.55 days more than the non-migraine controls (see Table 4, footnote).

Vo et al. [37] use 28-day data captured through the Migraine Buddy© Smartphone application from the period of June 2015 – July 2016. Users were self-diagnosed adults from 17 European countries. Data were retrieved for 3900 individuals suffering from migraine; they were selected randomly from a population of 13,032 meeting the inclusion criteria. Of these, 3106 had a work; in Table 4 days missed from work is reported for this subsample. The authors find that migraine attack affected 8.3 days per month an average. In work absenteeism-related attack users most commonly noted body pain, mood and cognition, environmental handicap, depression and/or sleep alterations. Vo et al. [37] thus find that migraineurs miss more than twice as much days from work as the respondents in the Eurolight project [35]. The authors argue that this difference might be related to the study samples. In the sample in [37] more severe patients were included; patients suffered from at least four monthly migraine days and had headache in at least two consecutive weeks from the time of initial registration.

Finally, in a recent study Martelletti et al. [32] report both days missed from work and paid sick days in the last month for over six thousand migraineurs in employment from 31 countries. The authors find that employees suffering from migraine missed an average of 4.6 working days in the last month. The number of paid sick days was significantly smaller, 2.4 days per month. When considering the paid sick days, the findings of Martelletti et al. [32] is comparable to that of Vo et al. [37] . Their sample included more severe patients; they have oversampled patients for whom at least two preventive migraine treatment had failed.

For all migraine patients, the annual number of days being absent from work ranges from the conservative estimate of 0.8 days in [7] to the liberal estimate of 28.8 days in [32]; the latter figure is more than 30 times larger than the former. Had we taken not only the paid sick days as reported in [32] but all the days missed from work, we would arrive at an estimate of 55.2 days missed from work per year. The two highest estimates are derived from a sample of severe migraineurs [32, 37]; these figures are most probably not representative of all migraine sufferers. The remaining estimates are fairly polarized and can be divided into two separate groups. The first group contains estimates ranging from 0.8 to 2.91 days missed from work [6, 7, 14], while the second one consists of estimates ranging from 10.7 to 12.8 days missed from work [27, 35, 36]. Before selecting the absenteeism estimate to be used in this study, we review the relevant literature on presenteeism as well.

Measuring impairment is challenging; researchers must assess whether the disability to think clearly, lack of focus, and loss of concentration result in productivity loss at all, or the work can be performed as usual. If the work cannot be performed as expected, then the number of hours lost shall be calculated. Reduced productivity due to migraine is typically assessed by the Migraine Disability Assessment (MIDAS) questionnaire; participants are asked to report the number of days when their productivity is reduced by half or more [39]. Days when the productivity is reduced by less than half of the expected are ignored.

Table 5 displays the most important findings of population-based studies on productivity loss when working with migraine. The inclusion criteria were the same as for the studies reporting the number of days missed from work due to migraine. As a result of the targeted literature search, four studies met the inclusion criteria, these studies are listed in Table 5.

The first two studies shown in Table 5 used data from US patients, the latter two build upon survey results from Europe. Both studies from the US use data from the AMPP study. Munakata et al. [6] employ the results from the 2006 follow-up survey, whereas Stewart et al. [14] rely on the responses from the 2005 survey. Munakata et al. [6] find that patients with migraine lost 48.3 h per year due to presenteeism, which is equivalent to 6.04 lost workdays per year. Stewart et al. [14] report the hour-equivalent of headache-related reduced performance on days at work for 6204 respondents who were employed actively for pay. Similar to absenteeism, the authors report the hours per worker per week for four patient groups: low, moderate, high frequency headache, and chronic migraine. Patients with chronic migraine had the highest probability to report reduced performance due to headache in the past 2 weeks (3.8 h per week); they were followed by respondents suffering from very frequent headache (2.8 h per week). As Stewart et al. [14] report hour-equivalent of headache-related reduced performance per worker per week, the estimates were converted into annual figures for comparison purposes, for details see Table 5, footnote.

By interviewing Russian adults, Ayzenberg et al. [7] document for participants with migraine that the numbers of days in which productivity was less than 50% of the expected productivity was 1.7 days in preceding 3 month translating into 6.8 days per year (Table 5). Vo et al. [27] provides presenteeism estimates for 218 patients with and without migraine from five European countries. The authors find that respondents with migraine when compared with non-migraine controls reported significantly higher presenteeism, the percentage of impairment while at work in the past 7 days was 35.52% vs 20.97%, respectively. To arrive at the number of days lost due to impairment for patients with migraine, the incremental difference in impairment between patients with migraine and non-migraine controls is multiplied by the number of days working with migraine (for details see Table 5, footnote).

In sum, in recent nation-wide studies the number of days lost due to impairment ranges from 6.04 to 9.27 days per year. This lower range – as compared to the range reported for missed workdays – might be explained by less severe consequences of being present with decreased functional capacity as compared to missing a workday completely; there is no need to submit a medical certificate and employees are less exposed to the fear of losing their jobs.

In further analysis, the figure from [35] is used for estimating productivity loss; each year individuals suffering from migraine lose 12.8 workdays due to headache. This estimate includes workdays lost both due to sick leave and impairment; the authors estimate total productive time lost at work as the sum of workdays lost completely due to absenteeism and workdays with productivity reduced to 50% or more of the expected productivity. Steiner and Lipton [40] argue that this approach counterbalances those working days when the productivity was reduced by less than half of the expected, which are ignored otherwise. This approach has already been introduced by MIDAS and was validated in [41].

The estimate of Steiner et al. [35] is considered as a reliable estimate for a number of reasons. First, the authors surveyed over 8000 participants, their sample is the largest. Second, the authors collected data from nine EU countries in a cross-sectional survey. Data collected in the US or Russia might not be valid for Latvia and Lithuania due to the difference in sick leave regulation, among others. Although the sick pay and sickness benefit schemes are not harmonized in the European Union, they share several common characteristics [42]. Steiner et al. [35] show that personal impact is terms of lost useful time was surprisingly uniform across the sample countries. Third, the authors employ an already validated method being widely used for assessing the impact of headache.

The 12.8 days productive time lost at work estimate from [35] for Europe is comparable to the findings from [14] for the US. As shown in Tables 4 and 5, Stewart et al. [14] document that migraine sufferers are absent from work on 2.9 days and lose additional 8.9 workdays due to reduced productivity, resulting in 11.8 lost workdays in total.

In order to estimate the economic cost of migraine in Latvia and Lithuania, the number of individuals suffering from migraine should be determined. Prevalence rates are typically derived for the adult population [18‐20]; prevalence rates for children and elderly are typically much smaller [3]. As a result, the number of people aged 18–65 is taken from both Latvia and Lithuania. As of 1 January 2017, in Latvia 1,228,779 people, whereas in Lithuania 1,819,685 people were aged 18–65 [43]. By assuming a prevalence rate of 11.4% as reported in [18], the number of individuals suffering from migraine is 140,081 in Latvia, and 207,444 in Lithuania.

Unit costs were derived from several national data sources in Latvia and Lithuania. For Latvia, unless indicated otherwise, information about publicly funded medical care was retrieved from [44]. For Lithuania, unless indicated otherwise, information about publicly funded medical care was retrieved from [45]. In Latvia, several medical services are provided by both publicly and privately funded providers. In the most conservative scenario (labelled as the base case), unit cost of publicly funded services is preferred over privately funded services. As we aim to access the total cost of migraine for the society regardless of the payer, patient co-payment, if present, is included in the unit cost. The unit cost of various health care services and procedures is shown in Table 6; details about estimating the unit cost of one visit in primary care, and the split between public funding and patient co-payment is reported in Additional file 3.

Estimating the medication cost for patients with migraine is challenging. Some medications are listed explicitly as migraine medications; they could be identified from administrative databases by drug or disease classification codes. Nevertheless, several over-the-counter analgesic drugs are used to treat migraine; these pain killers, such as paracetamol, aspirin, and ibuprofen have a wide array of applications; the ones used for migraine cannot be easily extracted from administrative databases.

As a conservative approach, we estimate medication costs by calculating the annual cost of typical acute migraine therapies and choosing the cheapest therapy. Additional file 4 lists the annual cost of acute migraine therapies in Latvia and Lithuania by assuming monthly 4 headache days. Annual cost of therapies with paracetamol, nonsteroidal anti-inflammatory drugs (NSAIDs) such as ibuprofen and aspirin, and triptans are calculated; these typical therapies were identified from [46]. We assume that paracetamol, ibuprofen and aspirin are taken on each day with headache, while triptans are taken once per attack, assuming monthly one attack. As shown in Additional file 4, in general, therapies with ibuprofen are the cheapest, followed by paracetamol and aspirin; treatments with triptans are much more expensive. In the absence of detailed data on medication usage by patients, we assume that patients with high frequency headache take the cheapest available triptans, whereas all other patients use the cheapest over-the-counter analgesics, both in the defined daily dose. Based on Stewart et al. [14], 10.55% of migraineurs is considered as individuals with high frequency migraine—migraine on more than 10 days per month, including chronic migraine. The assumption that 10.55% of patients use triptans once per month can be considered as a conservative assumption; in North-America 16.8% of patients with migraine use triptans on average on 4.75 days in the past month [29].

As shown in the last row of Table 6, for Latvia we arrive at an annual mean per patient medication cost of €12.26, derived as (89.45% × 10.21) + (10.55% x €29.67), where 89.45% is the proportion of patients with low and moderate headache frequency as reported in [14], €10.21 is the annual cost of the cheapest therapy with analgesics (Additional file 4), 10.55% is the proportion of patients with high headache frequency as reported in [14], and €29.67 is the annual cost of the cheapest therapy with triptans (Additional file 4). For Lithuania, we arrive at an annual mean per patient medication cost of €9.39, derived as (89.45% x €9.21) + (10.55% x €10.83), following the same logic as for Latvia. A sensitivity test with an alternative medication cost estimate is then performed. We also discuss the validity of our medication cost estimate and we compare it with estimates from other European countries.

In sensitivity test unit cost of privately funded services in Latvia is used (Table 6). Latvian patients frequently visit private providers due to significantly shorter waiting times. The fee of privately funded health care services varies greatly, see Additional file 3. In sensitivity test the lowest reported fee of the first visit is inserted for private consultations. Additional details about the unit costs used in the sensitivity test is reported in Additional file 3. For example, the fee of nurse practitioner/physician assistant visit and hospitalizations remain unchanged, whereas the fee of emergency room visit almost triples. Fees also remain unchanged if the entire fee is paid out of pocket (e.g., transcutaneous nerve stimulator procedure, acupuncture). In contrast, visiting a primary care physician in private setting is six times more expensive than a publicly funded visit. Similarly, fees of consultations with specialists and several diagnostic evaluations are much higher in private setting as compared to public setting (e.g., neurologist visit, pain specialist visit, brain CT). For medications we insert €21.30; related assumptions are reported when performing the sensitivity tests.

We have also carried out a systematic search in Pubmed to identify any previous unit cost estimates, for comparison purposes. As unit prices differ highly across countries, we only consider the estimates reported for Latvia and Lithuania as relevant.⁶ The systematic search yielded one relevant study, which provides unit prices for eight European countries, including Lithuania [5]. Their estimates for Lithuania are shown in the last column of Table 6; the estimates have not been adjusted by inflation and thus allow only rough comparison.

Several employment related data is retrieved from Latvian and Lithuanian databases; from the databases of the Central Statistical Bureau of Latvia and Statistics Lithuania, principal government institutions in charge of statistics and census data. The number of registered unemployed persons is required to estimate the productivity loss related to reduced workforce participation. As for unemployed migraineurs we consider that the entire annual gross income is lost, information about the average annual gross wages and salaries of employees in Latvia and Lithuania is also retrieved. Finally, for estimating the loss incurred through absenteeism and presenteeism, information about the average daily income and number of workdays in 2017 in Latvia and Lithuania is collected.