Cost-of-illness studies in nine Central and Eastern European countries

We conducted a systematic review following the PRISMA statement [14]. A literature search was performed using Medline, EMBASE, The Cochrane Library, CINAHL, and Web of Science databases to identify studies that report data on the cost of a disease. The search strategy was based on the keyword “cost of illness” and the name of the given CEE country (online Appendix 1). The search was limited to studies published in the past 10 years (1 January 2006—31 October 2016) and was updated on 30 June 2017 to shorten the time between the end of the search period and publication date. No language restrictions were applied. A complementary, non-systematic literature search was conducted in three countries. Three authors (SM, KT, and ZB) hand-searched for further papers in selected, peer-reviewed, non-indexed local journals in Austria, Bulgaria, and Hungary. The review protocol was not registered.

After removing duplicates, titles and abstracts of studies were reviewed independently by ZB, VB, and LG, and were retrieved if at least one of the reviewers considered the study to be relevant. First, abstracts (publication type) and reviews (publication type) were excluded. Full-text papers of the remaining studies were reviewed and included (ZB, VB, and LG). Any disagreement between reviewers was solved by discussions among the authors to reach consensus.

Studies were selected for further analysis if they met the following inclusion criteria: (i) COI data included for a specific disease without major restriction on the patient population, e.g., by age, co-morbidity, complication, or treatment, (ii) full-text paper, (iii) original research, and (iv) the study population was recruited in Austria, Bulgaria, the Czech Republic, Croatia, Hungary, Poland, Romania, Slovakia, or Slovenia. Studies were not selected for further analysis if they represented clinical trials, reviews, cost-effectiveness studies, budget impact analyses, treatment-related (drug) studies, costs of health programs (e.g., screening), or studies enrolling a patient population with co-morbidities (e.g., diabetic patient with depression).

A Microsoft Excel spreadsheet was developed to extract data from the identified studies, including general characteristics of the study (year of publication, geographical location, language, and funding source), methodological details of the study (disease, data collection method, study design, setting, costing year, currency, and perspective and costing methods), and results (direct costs, indirect costs, and total costs in euros). The list of extracted variables was created based on health economic checklists and adjusted by screening of six (10%) random articles [6, 15]. Costs reported in currencies other than euro were converted to euro at a mean annual exchange rate, and all costs were inflated to 2017 prices using the harmonised consumer price index extracted from Eurostat [16]. To facilitate cross-country comparisons, costs were also described as a percentage of 2017 GDP per capita. Diseases were categorised according to the International Statistical Classification of Diseases and Related Health Problems 10th Revision (ICD-10 Version:16) [17]. Data extraction was conducted by ZB and respective authors for national languages and double-checked.

As can be seen from Fig. S1 (online Appendix), after removing 246 duplicates, the search in the electronic databases resulted in 607 potentially relevant papers. Of these studies, 55 were not full-text papers and 98 were reviews. Furthermore, 282 papers did not report disease-related costs, 54 focused on costs of multiple diseases, and 67 focused on the cost of a certain treatment. Overall, 50 articles from the electronic search fulfilled the inclusion criteria. The supplementary local search resulted in another eight relevant articles in non-indexed, peer-reviewed journals (Austria: n = 2, Bulgaria: n = 5, and Hungary: n = 1).

Altogether, we included 58 articles (involving also multi-country studies) that reported results for Hungary (n = 24), Bulgaria (n = 16), Poland (n = 11), Czech Republic (n = 10), Austria (n = 9), Slovenia (n = 4), Croatia (n = 3), Slovakia (n = 3), and Romania (n = 3).

Thirteen additional COI studies did not meet to our eligibility criteria (e.g., involved samples restricted by age, co-morbidity, complication, or treatment), but we found their results worthy of attention, and hence, a summary of their characteristics and main results is presented in online Appendix 1.

The majority of publications reported costs from one country (74%), but 15 studies presented results from multiple countries, and hence, altogether, 83 country-specific results were provided by 58 studies (Table 1). Three-quarters of the studies were published in English (n = 44), and except for five papers [18‐22], all non-English papers had an English abstract. Most of the publications (n = 45, 78%) presented costs in euro. In 37 studies, the national currency was converted to euro; of them, 17 (46%) studies stated explicitly exchange rate, 5 (14%) studies reported only the source of exchange rate, and 15 (40%) studies did not mention conversion at all. Among countries outside the euro zone, reporting costs in national currency was most common in Romania (67%). Overall, 47 (81%) studies stated the source of funding. The lack of a funding statement was most prevalent in Romania (n = 2, 67%) and in Bulgaria (n = 5; 31%). Only two studies received funds from two different sources, both of them were funded by the European Union (EU) and the local government. Regarding clinical areas, endocrine, nutritional, and metabolic diseases were the most common, in which costs were analysed (n = 15 country-specific results), followed by neoplasms (n = 12), and certain infectious and parasitic diseases (n = 10) (Fig. 1). Altogether 48 different diseases were analysed in the 58 included articles.

Eighty-three COI estimates were reported for 48 different diseases. Apart from rare diseases, multiple sclerosis caused the highest economic burden in terms of average total annual cost per patient in three countries (Austria €50,599, the Czech Republic €14,777, and Poland €12,343) [23‐25]. In Hungary, schizophrenia (€15,187), and in Bulgaria, gestational diabetes (€32,263) were the most costly diseases [22, 26].

Multi-country studies were conducted in nine diagnoses (rotavirus gastroenteritis, pneumonia, bladder cancer, hypoglycaemia, Duchenne muscular dystrophy, epidermolysis bullosa, Prader–Willi syndrome, cystic fibrosis, and haemophilia). One multi-country study (bladder cancer) was conducted in nine countries and another (hypoglycaemia) in six countries. Two studies were conducted (rotavirus gastroenteritis and pneumonia) in four countries and four studies (Duchenne muscular dystrophy, epidermolysis bullosa, Prader–Willi, and haemophilia) in two countries. The bladder cancer study involving nine countries resulted in mean total costs of €7421; however, costs differed significantly among countries, as the total cost was between €2320 (Bulgaria) and €16,479 (Slovenia). The direct medical cost ranged between €1090 (Bulgaria) and €8050 (Slovenia), and indirect cost varied between €912 (Bulgaria) and €6398 (Slovenia). The hypoglycaemia study was conducted in six countries, and the total overall societal cost per patient with diabetes was €11 and ranged between €5 (Bulgaria) and €18 (Slovenia) [27]. Rotavirus gastroenteritis and pneumonia studies were conducted in four countries and the average total costs were €541 and €764, respectively. Costs varied between €494 (Czech Republic) and €747 (Poland) in rotavirus gastroenteritis, and between €472 and €1111 in pneumonia. Duchenne muscular dystrophy, epidermolysis bullosa, Prader–Willi syndrome, cystic fibrosis, and haemophilia were studied in two countries (Hungary and Bulgaria) applying the same methodology in a European Commission founded rare disease study (BURQOL-RD project). Prader–Willi syndrome was the least costly (Bulgaria: €3842 Hungary: €12,532) and mucopolysaccharidosis was the most costly rare disease (Bulgaria: €77,414; Hungary: €25,326) [28, 29].

Unique studies in more than one country were conducted in eight diagnoses, namely multiple sclerosis, dementia, Parkinson’s disease, rheumatoid arthritis, osteoporosis, chronic obstructive pulmonary disease (COPD), systemic sclerosis, and diabetes. Multiple sclerosis and diabetes were studied most often (four studies each), while three unique studies in three different countries were conducted in Parkinson’s disease and two unique studies in three different countries were conducted in cystic fibrosis. Two unique studies on both dementia and COPD were conducted in two different countries. In multiple sclerosis, there was a 4.1 times difference in total costs between Austria (€50,599) and Poland (€12,343) [24, 30]. In diabetes, the highest direct cost was observed in Hungary (€1309) and the lowest total cost was observed in Bulgaria (€472) [31, 32]. In Parkinson’s disease, there was a 3.3 times difference in total costs between Austria (€22,984) and the Czech Republic (€6970) [33, 34]. In dementia, we found a 3.5 times difference in total costs between the Czech Republic (€2013) and Hungary (€671) [35, 36]. The costs of COPD were similar in Bulgaria (€1839) and Romania (€2103) [21, 37].

Adjusting costs for GDP per capita level, differences between countries decreased (Table 2). For instance, a 7.1-fold difference in bladder cancer and a 4.1-fold difference in multiple sclerosis were reduced to 2.4- and 1.5-fold, respectively. Comparing diseases with available cost estimates from more than one country (Fig. 2), a positive relationship was identified between costs and GDP per capita.

Reporting cost results in euros was dominant over national currencies, suggesting that researchers in the CEE region find it important to make their results available for the international scientific community and allow for comparability with other studies. To assess study quality, we selected some quality indicators, such as those are used in health economics checklists. Reporting study perspective, reference year, costing method (top–down vs. bottom–up), source of resource use, valuation of informal care, valuation of productivity loss, and funding source were considered as quality indicators. We find it noteworthy to mention that whilst the source of data on resource utilization and reference year of costing were stated in nearly every paper (98% and 95%, respectively), other important quality indicators were less often reported. The study perspective was reported in 78%, the approach to valuing indirect costs in 77%, costing method in 64%, at least one unit cost in 42%, and method for valuing informal care in 31% of the studies. A recent review of economic evaluations in Austria found that the study perspective and reference year were not reported by 60% and 25% of the studies, respectively [13]. Differences may be explained by inclusion of non-peer-reviewed or grey literature (e.g., economic evaluation reports from national health technology assessment agencies) and of other forms of economic evaluations in the study by Mayer et al. The review by Mayer et al. included 93 (partial and full) economic evaluations, 14 of which were cost-of-illness analyses. Out of the 93 included studies, 23 were not indexed according to the Journal Citation Reports (Social) Sciences Edition and 12 were non-peer-reviewed reports [13].

A large variety of diseases was covered by the studies, and most of them occurred in a one study. Each disease was studied by, on average, 1.3 papers. Considering country-specific results by ICD categories, endocrine, nutritional, and metabolic diseases (18%), neoplasms (14%), infectious (12%), neurologic (11%), and musculoskeletal diseases (11%) represented the five main fields of COI research in CEE. It is difficult to judge the drivers of the selection of clinical fields. The public health importance of a disease might be an important factor as, for instance, all the studies in the ‘Endocrine, nutritional and metabolic diseases’ ICD category were related to diabetes, and among neoplasms studies, the most prevalent malignancies (breast, colorectal, lung, and prostate cancer) were present (Table 2). According to the Global Burden of Disease study, the leading three causes of total Disability-Adjusted Life Years (DALY) included ischaemic heart disease, cerebrovascular disease, and lower respiratory infection, comprising 16% of all DALYs [38]. Leading causes of DALYs were represented only in six (10%) studies (cerebrovascular disease: n = 1, ischaemic heart disease: n = 2, and lower respiratory infection: n = 3) in our review, questioning public health importance as a driver of topic selection in COI studies. The need for COI data to support decision-making on reimbursement of highly effective but costly new drugs seems to be another relevant issue, and this hypothesis is supported by the relatively high rate of studies in inflammatory rheumatic diseases, where biological drugs were introduced in the CEE countries in the observed period. Multiple sclerosis is another disorder where biologicals revolutionized the treatment that partly explains the relatively high rate of neurological studies in the region. Moreover, when counting papers, neurologic diseases were most frequently studied (19%). A possible explanation could be that neurologic conditions in the CEE region were priorities for state-funded or EU-funded research. Eight out of the ten COI studies focusing on neurologic diseases received funding from the local governments or EU organisations. It is interesting that neurologic diseases were found also the most frequently studied clinical area according to a recently published systematic review of EQ-5D studies in the CEE region [39]. These results suggest that neurologic diseases have a high priority in health economics research in the CEE.

With respect to diseases for which cost estimates were present in multiple countries, costs varied substantially across countries. However, there are apparent differences in the level of comparability between studies. There were multi-country studies following a standardized methodology in which more than one CEE country together with Western European countries was participated. We also identified single-country studies in various diseases using very different methods. Both multi-country and single-country studies reported significant cost differences in diseases across countries.

For the interpretation of data, it is important to take into consideration that the number of patients, sample characteristics (e.g., age, gender, disease duration, and disease severity), and the availability of costly treatments at the time of the study (e.g., biological drugs for inflammatory diseases) varied a great deal across studies that may strongly influence the COI results and their comparability. Large differences in unit costs can also cause significant variations in costs. In bladder cancer, for example, the cost of an inpatient day was seven times higher in Austria (€495) than in Romania (€67). Methodological differences, such as prevalence- and incidence-based costing, form an obstacle for the comparison of costs. Therefore, the incidence-based prostate cancer study by Brodszky et al. cannot be compared with the prevalence-based prostate cancer study by Inotai et al., although both studies were conducted in Hungary [40, 41]. It should also be noted that differences in health care systems (private/public, financing, etc.) might have a significant impact on costs; for instance, global budget, fee-for-service or DRG financing mechanisms, the presence of co-payments, minor or major share of private services, and many more aspects might influence the actual costs, access to health care, and, finally, the COI figure [42].

According to the literature, one might expect a higher COI in a country with a higher GDP [43‐45]. In many diseases (multiple sclerosis, bladder cancer, Parkinson’s disease, rheumatoid arthritis, Prader–Willi syndrome, haemophilia, diabetes, and hypoglycaemia), there was a clear positive association between total costs and GDP per capita. As opposed to this, cost estimates, sometimes, inversely correlated with the per capita GDP. For instance, GDP per capita in Bulgaria is almost half of that in Hungary; nevertheless, costs of mucopolysaccharidosis were threefold higher in Bulgaria. Thus, in some cases, adjusting costs for the GDP further increased the inter-country differences. On the other hand, the 3.5-fold higher GDP per capita in Austria decreased the cross-country differences (from 4- to 1.3-fold) in costs of multiple sclerosis. In spite of the considerable heterogeneity observed in the studies included in this review, some trends could be identified. The magnitude of costs increased with the level of per capita GDP. In other words, cross-country differences decreased or even vanished when the costs were adjusted. In contrast, higher costs with lower GDP per capita could be observed only in some rare diseases (cystic fibrosis, epidermolysis bullosa, and mucopolysaccharidosis) and rotavirus gastroenteritis. Moreover, methodological differences did not seem to affect this relationship. Comparing multi-country studies in a disease applied the same methodology for more than one country and single-country studies analysed costs in the same disease, the relationship between cost-of-illness and GDP per capita showed similar pattern in these two groups of studies (see Fig. 2).

Cost-of-illness studies varied considerably both in methods and in cost estimates, and serve many purposes. Methodological deficiencies, such as the lack of reporting either on the three distinct phases of costing (identifying the relevant cost items, measuring the use of the identified resources, and placing a value on these cost items) [46], or other important characteristics such as the perspective of the study, related to the production function (direct and indirect costs) were the leading causes of shortcomings in comparability. However, no specific costing guidelines for health care interventions are available in these countries, and except in Austria, there is no national cost database available, providing some kind of unit cost data in a collected form [13, 47, 48]. Another important difficulty in costing relates to the different Managed Entry Agreements (MEA), such as price volume agreements, discounts, outcome guarantees, and many more, in the reimbursement of the health technologies in the different countries [49, 50]. Due to the MEAs, for instance, the real purchasing price of the medicinal products is not publicly available.

Several papers were published about transferability in the past 2 decades [51‐56]. At the moment, health economics and health technology assessment guidelines in CEE countries either include very limited advice or provide no guidance on the transferability or adaptation of clinical and economic data from other jurisdictions. Thus, establishing better guidelines for COI studies on transferability would be valuable for robust decision-making in the CEE countries [56]. As Gao et al. stated, confirming the transferability of COI estimates across jurisdictions would contribute significantly to resolving the issue of transferability of cost-effectiveness results [45]. Transferability is a very important issue around the world and especially in Central or Eastern Europe with limited resources to provide COI studies [53‐55]. Data transferability and transferability of the results are not discussed in these COI studies. Both should be improved using Drummond’s check list for evaluating economic evaluations [57]. Transferability might be an important alternative to conduct local COIs. However, due to the methodological, data, and publication heterogeneity, the usefulness of the COI results in other jurisdictions is limited.

There are a few limitations to note. A systematic approach was taken to identify studies that have considered the costs of diseases; however, the possibility that relevant studies were not identified and included in this systematic literature review remains. Some COI results might have been missed due to excluding grey literature (i.e., conference abstracts and project reports) from our search. Other limitationis that the local search in non-indexed journals was conducted only in three of the nine countries. On the other hand, no language restriction was applied in the systematic search. Adopting a Medical Subject Heading (MeSH)-based search strategy may have led to missing some studies using keywords improperly. At the same time, the PubMed search engine uses a broad range of entry terms which may minimize the number of excluded studies. Further limitation is that no comprehensive checklist was applied, because, according to our best knowledge, there is no COI study-specific checklist in English. This might bias our conclusions on study quality, but we believe that the presented study characteristics could give a good overall description of the included studies.