Detecting changes in financial protection: creating evidence for policy in Estonia

Abstract

Ninety-four per cent of the Estonian population is covered by public health insurance, but private expenditure has been increasing quickly both in real terms and as a percentage of total health expenditure. To date, little attention has been given to the impact this could have on the population's financial protection. Out-of-pocket payments, which account for the bulk of the private expenditure in many low- and middle-income countries, can push people into poverty and more generally represent too high a burden for some households. It is therefore very important that governments monitor the impact of out-of-pocket payments on health. Using an example from Estonia, this paper aims to illustrate that, if household budget survey data are available, monitoring a population's financial protection is not a complex undertaking. Further, by combining simple statistical analyses of these data with a good knowledge of a country's health system, it is possible to give a fairly detailed diagnostic of the nature of the population's coverage limitation. This allows for the presentation of easily interpretable results that can raise awareness among policy-makers and help to target adequate policy responses.

Using Estonian household budget surveys from 1995, 2001 and 2002, we show that the proportion of households who spend more than 20% of their capacity to pay on health increased from 3.4% in 1995 to 7.4% in 2002 and that in 2002, 1.3% of the population fell into poverty because of health payments. Logistic regression helps in identifying the population most at risk: elderly patients who belong to poor households and spend high amounts on medicines. This study, which can be replicated, did raise awareness among policy-makers about the changes in financial protection over the years in Estonia.

out-of-pocket payments, financial protection, Estonia, health system, health care services, equity

Introduction

The World Health Report 2000 proposed that one of the main objectives of any health system should be to protect households from the risk of becoming impoverished as a consequence of their health care expenditures. More generally, two related objectives of health financing policy are equity in the distribution of funding the health system (the financial dimension of solidarity) and protection against financial risk. Experience from many countries shows that out-of-pocket payments made at the time of utilization are harmful to this objective (Xu et al. 2003a). The need to make such payments can cause individuals to expend high amounts in relation to their available income, and some households are forced into debt and/or pushed into poverty as a result. These risks are greater for low-income families.

These problems are particularly acute in low- and middle-income countries, where public health expenditure tends to be low and people pay relatively more for care out of their pocket than in richer countries. As a consequence, the financial protection provided to the majority of the population is often inadequate. But universal financial protection is also a challenge for richer countries. For instance, in the United States, Waters et al. (2004) show that 3% of families spent more than 20% of their income on health through out-of-pocket payment in 1996. For families whose income is below the Federal Poverty Level, the out-of-pocket payment on health represented 13.5% of their income. A recent study in the US showed that about half (46.2%) of personal bankruptcies are due to medical expenditures (Himmelstein et al. 2005).

A growing body of literature (e.g. Wagstaff and van Doorslaer 2001; Xu et al. 2003b) presents various methods that can be used to estimate the burden that health expenditure represents and its impact on poverty, and provides illustrations from different countries.

Using an example from Estonia, this paper aims to illustrate that, if household budget survey data are available, monitoring a population's financial protection is not a complex undertaking. Further, by combining simple statistical analyses of these data with a good knowledge of a country's health system, it is possible to give a fairly detailed diagnostic of the nature of the population's coverage limitation. This allows for the presentation of easily interpretable results that can raise awareness among policy-makers and help to target adequate policy responses in other countries, and particularly in Estonia.

The Estonian context

Estonia, with a population of 1.35 million (Ministry of Social Affairs 2004b), is one of the Baltic countries which regained independence in 1991 and joined the European Union in 2004. During the transition, Estonia went through enormous economic development and undertook a series of sequenced health system reforms. The main changes in the health sector include the introduction of a social health insurance system funded by income-related contributions from 1991 on, the reshaping of health care provision around family physicians between 1997 and 2003 (Jesse et al. 2004), and a recent emphasis on restructuring the organization and financing of the hospital sector (Howe and Koppel 2001; Palu and Kadakmaa 2001). These changes have been complemented with increased access to modern medicines not available before the 1990s (Kiivet and Harro 2002; Jesse et al. 2004).

System-wide cost containment and improvements of the health system's efficiency have always been driving forces of the health care reforms (Jesse et al. 2004). Equity in funding and access are also officially promoted. However, a growing body of evidence shows that all groups in society do not enjoy equal access to health care services (Kunst et al. 2002), even after their health status is taken into account (Habicht and Kunst 2005).

A possible explanation for this phenomenon could be an increase in financial barriers in access to care. To date, however, there have been no analyses conducted on the extent to which the need to make out-of-pocket expenditures is a barrier to utilization, nor of the impact that such payments are having on the population's financial protection. Using surveys from 1995, 2001 and 2002, we shed some light on the latter issue, showing that while the system does protect the vast majority of the population, coverage appears to be eroding over time, in particular for some vulnerable groups. The analysis of changes in financial protection also allows for some tentative conclusions about the impact on equity of access. Indeed, because our data set included expenditures but not utilization of services, our analysis understates inequities in population protection against high costs of care to the extent that poorer people are less likely to seek needed care.¹

Figure 1 gives an overview of the economic and public finance context for the period under consideration.² Between 1996 and 2002, the Estonian Gross Domestic Product (GDP) increased by around 40% in real terms, at a much faster pace than total expenditure on health (+10%). Over that period, total public expenditure grew almost as fast as the GDP, but public expenditure on health decreased slightly in real terms. In fact, the share of total public spending devoted to health fell from 16% in 1996 to 11% in 2002. The 10% rise in total health expenditure was therefore driven by private expenditure, which increased by nearly 80% in real terms. In 1996, 13.3% of the health expenditure was paid privately (mainly out-of-pocket); the proportion reached 19% in 2001 and 20% in 2002. Although the out-of-pocket share increased rapidly, Estonia's 2002 figure is still quite favourable compared with neighbouring Baltic countries, for example, Latvia at 47% in 2001 (World Health Organization 2004) and Lithuania at 27% in 2002 (World Health Organization 2005).

Figure 1.

Changes in the health expenditures in Estonia by category, Gross Domestic Product and total public expenditures

The Estonian Health Insurance Fund (EHIF) covers 94% of the population, including children under 18 and pensioners who are covered statutorily [see Couffinhal and Habicht (2005) for a detailed analysis of the Estonian health financing system]. The uninsured are typically non-working adults and they tend to be poorer than the rest of the population (Marksoo et al. 2001). Between the mid-1990s and 2002, their proportion in the population remained stable. The uninsured have access to publicly funded emergency care but have to pay for other types of care.

For the population with insurance, the benefit package provided by the EHIF is broad and covers the majority of the costs of health care services. At the end of 2002, these services included primary health care, specialist outpatient and inpatient care as well long-term care. A list of specific pharmaceuticals is covered by the health insurance fund. Between the mid-1990s and 2002, some statutory co-payments and co-insurance were introduced and increased, in particular for pharmaceuticals, but the overall structure of the covered package did not change fundamentally (see Table 1 for overview).

Table 1.

Co-payment policies in Estonia for selected years

		1995	2001/2002	2004
Primary care		• Co-payment for visits (€0.32).	• Co-payment for visits (€0.32).	• No co-payment for office visits.
		Retired, disabled persons, and children exempted.	Retired, disabled persons, and children exempted.	• Home visit fee (€3.2). Children under 2 years and pregnant women are exempted.
Outpatient specialist care	Outpatient specialists [contracted by health insurance (HI), the distinction important for 2004]	• Co-payment for visits (€0.32) if public hospital. Retired, disabled persons, and children exempted. If hospital status was private, then patient charged according to self-established pricelist.	• In addition to co-payment established by HI rules, some providers established additional fees.	• Co-payment of up to €3.2. Children under 2 years and pregnant women are exempted.
	Outpatient specialists (not contracted by HI)	• Patient charged according to provider-established pricelist.	• All patients charged according to provider-established pricelist.	• All patients charged according to provider-established pricelist, but up to a ‘reasonable' cost.
	Dental care	• Partially covered by HI, but additional charges applied by providers who were private entities.	• Partially covered by HI, but additional fees established and charged by providers who were private entities.	• No co-payment for children's dental care covered by HI.
				• Adult dental care not covered by HI, and yearly reimbursement rate from €9.8–28.8 applied for different population groups. Dentures for persons over 63 covered at €128 per 3 years.
Inpatient care		• No co-payment for hospital stays.	• No co-payment for hospital stays.	• Co-payment of up to €1.6 per day, for up to 10 days per episode of illness. Children, pregnant women and patients in intensive care units are exempted.
		• Co-payment established by providers for above-standard accommodation.	• Co-payment established by providers for above-standard accommodation.	• Co-payment established by providers for above-standard accommodation.
		• Co-insurance for a few specific services (such as voluntary termination of pregnancy).	• Co-insurance for specific services (such as IVF, rehabilitation, voluntary termination of pregnancy) set out by HI.	• Co-insurance for specific services (such as IVF, rehabilitation, voluntary termination of pregnancy) set out by HI.
Medicines (only outpatient prescription medicines as inpatient medicines are covered by HI)		• Prescription medicines for chronic and serious diseases (by conditions and for certain population groups such as disabled, retired and children) – co-payment of €0.32, plus co-insurance 0% or 10% (in May 1995 the co-payment increased to €0.64 and co-insurance rules remained).	• Prescription medicines for chronic diseases (by conditions and for certain population groups such as disabled or retired) – co-payment of €1.30, plus co-insurance 0% or 10%.	• Prescription medicines for chronic diseases – co-payment of €1.30 plus co-insurance of 0% or 25% of the drug price (or 10% for those aged 4–16, receiving disability or old age pensions, or older than 63).
		• General prescription medicines – co-payment of €1.91 per prescription, plus co-insurance of 50%, with no upper ceiling (until May 1995, when co-payment increased to €2.56 and medicines were not reimbursed by HI more than €12).	• General prescription medicines – co-payment of €3.20 per prescription, plus co-insurance of 50%, where HI will not reimburse more than €12 per prescription.	• Prescription medicines for those younger than 4 years – only co-payment of €1.3.
				• General prescription medicines – co-payment of €3.20 per prescription, plus co-insurance of at least 50% of the drug price, where HI will not reimburse more than €12 per prescription.
				• Annual spending on outpatient prescription medicines is eligible for additional reimbursement as: 50% (from yearly expenditures between €383–639); 75% (€639–1278); 0% (above €1278).

Overall, out-of-pocket payments consist of statutory co-payments but also include direct payments for goods and services excluded from the benefits package, in particular dental care, coverage of which was always limited. In 2002, according to the National Health Accounts (Ministry of Social Affairs 2003), pharmaceuticals represented 51% of the private expenditure on health, and dental care about 24%. Other types of private expenditure include outpatient services provided by non-EHIF contracted providers and some limited informal payments. Available information suggests that informal payments are not very common in Estonia. A survey by the Organisation for Economic Co-operation and Development (OECD) found that unofficial payments are rare and that such payments are mainly driven by patients’ own initiative (CIET International 2002). In a more recent study conducted by the Estonian Institute of Market Research in 2004, 3.4% of respondents reported that health care personnel had asked for some kind of payoff (Josin 2004).

The increase in private expenditure over the period under consideration probably resulted from a combination of different factors, predominantly changes in the user charges policy on pharmaceuticals, increases in prices of pharmaceuticals, and changes in the list covered when more effective medicines become available. To illustrate the latter, for example, the share allocated to pharmaceuticals from the health insurance budget grew from 8% in 1995 and to a high of 26% in 2002. In addition, according to the medicines sales statistics from pharmacies, the consumption of over-the-counter drugs (which are not reimbursed by insurance in Estonia) has increased over time, but the increase has been slower than that for prescription medicines (Jesse et al. 2004).

Another reason for the increase in out-of-pocket spending could be the development of the private sector (as providers such as hospitals or dentists have started to work as private entities), where fees are unregulated and providers are allowed to establish their own charge rates. This is particularly true for dental services, but also for some outpatient services where the provider has no contract with the EHIF (though the magnitude of this is not well documented).

Interacting with the first and possibly the second of these factors is, as noted above, the decline in the real level of government health spending. This decline, combined with changes in co-payment policy,³ has shifted a greater responsibility for funding directly to patients. In addition, the growth of demand for and spending on privately provided care may have also been stimulated by the decline in public spending. And of course, another likely explanation for the rise in real private health spending is the growth in real incomes of the Estonian population, on the assumption that the demand for health care is income elastic.

In terms of policy objectives, however, we are not concerned with the rise in private expenditure per se (nor its causes), but rather whether this increase led to an inequitable redistribution of the burden of funding the system, and an erosion of financial protection for the population as a whole as well as across different groups within the population.

Methods

Data are from nationally representative household budget surveys conducted in 1995, 2001 and 2002 (EMOR 1996; Statistical Office of Estonia 2002, 2003a,b). These are the ‘routine’ annual surveys conducted by the statistical office rather than special surveys focused on health expenditure patterns.

The focus of our analysis is out-of-pocket payments on health, i.e. payments made by households at the point of receiving health services net of any insurance reimbursement. The 1995 survey does not provide a break-down per type of health expenditure, but in 2001 and 2002 four broad categories are identified: outpatient services (including, and mostly consisting of, dental care), inpatient services, medicines (both prescription and over-the-counter medicines), and medicinal products, appliances and devices.

The household survey includes all household-reported expenditures, i.e. both formal and informal. Distinguishing between them is not possible, but other studies undertaken in Estonia (CIET International 2002; Josin 2004) suggest that the share of informal payments is among the lowest in Central and Eastern Europe (CEE) and the Baltic countries (see Lewis 2002 and Ensor 2004 for regional reviews of informal payments).

The distribution of total out-of-pocket payments in absolute and relative terms across income groups over time is analysed, as well as, when possible, its composition by income group.

Household income is measured by total consumption expenditure. Reported consumption expenditure is used in the analysis in preference to reported income for two reasons. First, the variance of current expenditure is smaller than that of current income over time. Secondly, in most household surveys, expenditure data are more reliable than income data. This is particularly true in developing countries (Deaton 1992; Bouis 1994).

In order to assess the financial protection provided to the population, we use a measure of the financial burden proposed by the World Health Organization (Xu et al. 2003b), and in addition we determine the proportion of households that fall below the poverty line due to health care payments. These two measures are fairly easy to compute, simple to explain and, to an extent, complementary.

The ‘household financial burden’ is measured by the ratio of total out-of-pocket payments to households’ capacity to pay or non-subsistence spending. The methodology developed by the World Health Organization was applied in estimating a household's capacity to pay, which is the total household expenditure minus the poverty line. The poverty line is estimated from the food expenditure of the household with the median food share of total household expenditure adjusted for household size (Xu et al. 2003b). The financial burden can therefore be roughly interpreted as the proportion of their income that households spend on health once they have paid for their food.

Once each household's burden is computed, the resulting information can be summarized by choosing a level that is considered ‘high’ and determining the proportion of the population which falls in that category and is therefore paying an ‘excessive’ burden. Obviously, the choice of the threshold will have an impact on the perceived intensity of the problem. The choice should probably in part be driven by the economic context and the capacity the country may or not have to address the problem, given its level of development. For instance, Waters et al. (2004) look at a 10% threshold in the US context; Xu et al. (2003a), who focus on low-income countries, label ‘catastrophic’ a 40% level.⁴ But the choice also undeniably reflects the author's value judgment of what an excessive burden might be. In the present context, we concentrated our analysis on a 20% threshold but present other results as well.

Another powerful way to raise awareness about out-of-pocket payments is to measure their impact on the poverty headcount in the population, in other words to estimate the proportion of the population that falls under the poverty line due to health payments. This can easily be done by comparing each household's total expenditure with its expenditure net of health care.⁵ If the former is below the poverty line and the latter not, it can be said that the household ‘became poor’ because of out-of-pockets payments. This measure indirectly focuses on the bottom of the income distribution (the closer you are to the poverty line, the higher your chances of falling below). The extent to which this is an asset or a weakness in terms of advocacy depends on the values of the society and particularly the emphasis given to helping the poor.

Having identified households that face excessive health care payments or fall into poverty because of them, the next step is to try to better identify who they are. Cross-tabulations of their high payer status with other characteristics (e.g. income level, health status) can shed some light on the reasons why they might have been put at risk, but regression analyses can provide additional insights. The number and type of variables that can be used in the regression will obviously depend on the quality of the data and adjustments usually have to be made. For instance, in the Estonian household budget survey, basic socio-economic characteristics were reported, but the household members’ health or insurance status were not available for us when the study was carried out. So finally, in order to characterize the factors associated with a ‘high’ financial burden, we applied a logistic regression to the 2002 data.

The dependent variable, high health payment, is set to 1 for a household whose out-of-pocket expenditure as a percentage of its capacity to pay is equal to or greater than 20%, and 0 otherwise. The independent variables include household structure, i.e. presence of members above 65 years old and of children under 16 years old, household income level (income quintile), and characteristics of the household head, such as sex and employment status (see Table 2). 5499 observations were included in regression analysis. An odds ratio is reported for each independent variable together with a coefficient. An odds ratio smaller than 1 for an independent variable indicates that this factor protects a household from facing high health payments, while the opposite is true if the odds ratio is greater than 1.

Table 2.

Variables used in the regression (data from 2002)

Variable	Label	Note	Mean	Std. dev.
Quintile	Income quintiles	The 1st quintile is the base	3.00	1.41
Urban	Household location	1, yes. 0, no	0.79	0.41
Old65	With members above 65 years old	1, yes. 0, no	0.23	0.42
Child	With members under 16 years old	1, yes. 0, no	0.30	0.46
Hhsize	Household size		2.41	1.30
Male	Household head	1, yes. 0, no	0.51	0.50
Primary	Complete primary but not secondary school	1, yes. 0, no	0.76	0.43
Secondary	Above secondary school	1, yes. 0, no	0.15	0.36
Employed	Household head	1, yes. 0, no	0.60	0.49

Results

The trend in out-of-pocket expenditure from 1995 to 2002

According to the surveys, in 1995, households spent on average 58 Kroon (3.7 Euro) per month out-of-pocket for health services, 139 Kroon (8.9 Euro) in 2001 and 152 Kroon (9.7 Euro) in 2002. Health payments also increased in relative terms: in 1995, they represented on average 2.2% of total household monthly expenditure and 4.1% of household capacity to pay (non-subsistence spending). These figures increased respectively to 3.2% and 5.0% in 2002, which implies that health spending is increasing faster than households’ other consumptions. Conversely, as shown above, health spending by government increased more slowly than other types of public expenditures.

Distribution and structure of out-of-pocket health payments across income groups

In absolute terms, nominal out-of-pocket health payments vary dramatically across income groups for all 3 years (Figure 2a). In fact, the level of out-of-pocket expenditure increases when income rises, as richer people spend more on health. However, comparatively small expenditures for health can have a great impact on a poorer household's situation. Indeed, comparing out-of-pocket expenditure with total expenditure, Figure 2b shows a very different picture. In 1995, the rich tended to spend relatively more on health than the poor, which means that out-of-pocket payments were slightly progressive, but in 2001 and 2002 the situation was reversed.

Figure 2.

Out-of-pocket payments across income quintiles

These graphs show that over the period, not only did the amount of out-of-pocket payment increase but the burden became heavier for the poorest part of the population and lighter for the richest part. Since the share of out-of-pocket spending in total health spending also increased during this period, it is likely that the distribution of the overall burden of funding the health system (not merely the out-of-pocket payments) became less equitable.⁶

The structure of out-of-pocket payments in 2001 and in 2002 is similar. The largest proportion of out-of-pocket expenditures was spent on medicines (both prescription and over-the-counter items), which represented 52% of total out-of-pocket expenditures in 2001 and 56% in 2002. The second largest item was outpatient services (mostly dental care) which counted for 28% of out-of-pocket expenditures in both years. Very little was spent on inpatient care, which is largely covered by public insurance (see description of co-payments above).

The structure of this expenditure varies across income quintiles. The poorest quintile spent almost exclusively on medicines (Figure 3). The rich spent relatively more on outpatient services. A combination of factors could explain these findings, namely that (a) the higher income groups used more private services and outpatient (mostly dental) care, perhaps including some element of ‘discretionary’ services; (b) the low income groups forewent needed outpatient (e.g. dental) health services or could not obtain the whole course of treatment.

Figure 3.

Components of out-of-pocket payments in 2002 across income quintiles

The household financial burden of out-of-pocket expenditure

The financial burden of a household is measured by the proportion that out-of-pocket expenditure represents of the household's capacity to pay. On average, this proportion has increased continuously from 1995 to 2002. Figure 4a presents the proportion of households who spend more than given proportions of their capacity to pay in 1995, 2001 and 2002. For instance, it shows that in 1995, 11% of households, once their food needs were covered, spent more than 10% of the remaining capacity on health care, 3.3% spent more than 20% and 0.3% more than 40%.⁷

Figure 4.

Level of out-of-pocket payments over the years, and proportion of households with high health payments or impoverished across income quintiles

Focusing on the 20% threshold, the proportion of households bearing a high burden increased from 3.3% in 1995 to 6.5% in 2001 and 7.4% in 2002. Figure 4b shows that the lower the income, the more frequently households spend more than 20% of their capacity to pay. For instance, in 2002, 13% of households do so in the first quintile versus 2% in the higher income group.

Over the same period more households were pushed under the poverty line after paying for health services. In 1995, 1% of the population fell below the poverty line as a result of health care payments, and the proportion increased to 1.3% and 1.4% in 2001 and 2002, respectively. In other words, the results show that in 2002, approximately 42 000 households faced high health expenditure and approximately 8200 were pushed into poverty because of out-of-pocket health payments, mainly those who were poorer (Figure 4c).

Factors associated with a high financial burden

Results from the multiple logistic regression are presented in Table 3. A positive value of coefficient or an odds ratio (OR) greater than 1 means the factor increases the risk of incurring high health payments, while a negative value of a coefficient or an OR smaller than 1 indicates a decrease in risk (see Figure 5). The results suggest that household income, family structure and the characteristics of the household head are related to the probability of a household facing high expenditure. Variables on urban/rural locations and education level of the household head were also tested in the model but none of them was statistically significant.

Table 3.

Multiple logistic regression of high health payments

Variables	Coef.	Std. err.	P > z
Old65	1.14	0.16	0.00
Quintile	−0.25	0.05	0.00
Male	−0.23	0.13	0.08
Employed	−0.72	0.19	0.00
Child	−0.91	0.21	0.00
_cons	−1.81	0.18	0.00

Number of obs = 5499.

χ²(5) = 298.56.

Log-likelihood = −1217.2.

P < 0.001.

Figure 5.

Different socio-economic factors influencing the risk of facing high health expenditures (20%) measured as odds ratios

One of the main determinants of high expenditure is having senior (above 65 years old) family members. Since the elderly are statutorily covered by public insurance, we can attribute this result to a limited depth of coverage⁸ rather than a lack of coverage per se. In other words, those at risk do not appear to be so much the uninsured, but insured individuals for whom the coverage does not offer sufficient protection against out-of-pocket expenditures. The probability of facing high expenditures for a household having members above 65 years is 3.14 times (coefficient 1.14) that of a household without senior members.

Controlling for other variables, the higher the income the smaller is the likelihood of facing catastrophic expenditure. This confirms the results from the descriptive analysis. A household headed by a male who is currently employed is less likely to face catastrophic expenditure than a household headed by someone non-employed or a female. The ‘protective effect’ of having children under 16 years suggests that they were not a frequent source of high health expenditures compared with adults and older people.

Discussion and policy implications

Bearing in mind some limitations of this study, we can nevertheless draw some conclusions. First, compared with many middle-income countries, and in particular the other two Baltic countries, Estonia relies relatively less on out-of-pocket payments, and as a result the health financing system provides adequate financial protection to the vast majority of the population. Drawing on the results of Xu et al. (2003a), in terms of ‘catastrophic’ expenditure (40% threshold), 0.3% of households faced catastrophic expenditure in Estonia in 1995 (a lower level than that in 2002, at 1.6%). This compares with 2.8% in Latvia (1997–98) and 1.3% in Lithuania (1999). By comparison, among the 59 countries studied, this proportion reaches 10% for Vietnam and Brazil, but remains below 0.1% in a number of Western European countries (France, Germany, Belgium, Denmark or the UK) (Xu et al. 2003a).

But our results also show clearly that the financial protection of the population has eroded over the period under consideration, particularly for poorer Estonians. We believe that there is sufficient cause for concern, particularly with regard to the trend, to bring this to the attention of policy-makers. Some obvious limitations to this study result from constraints imposed by the data, but they do not fundamentally challenge the results. For instance, no information is available on the volume of health service utilization or on health status. The messages that coverage is eroding over time and that low income groups are bearing too high a burden would only be reinforced if we could take into account these factors. Because people of low income tend to be in worse health in Estonia, as in other countries (Kunst et al. 2002), the fact that they spend relatively more is unlikely to result from choice. In addition, evidence is available that for a given health status, low income groups tend to use less care (Habicht and Kunst 2005) than high income groups. In other words, the observed financial burden probably understates the ‘true’ financial barriers because poorer people are more likely to forego needed care because of its cost (and this non-use is not captured by the expenditure data collected from the survey). This lower rate of service utilization relative to need is, by definition, contrary to the objective of solidarity, and more specifically equity in utilization.

Some additional insight into the underestimation of the financial barriers can be drawn from an analysis of other data sources on the various reasons that people give for not consulting a doctor, since the data we have on expenditures do not allow for this. The 2004 EU-SILC survey in Estonia (Statistical Office of Estonia 2005) shows that 2% and 3% of people experienced the need for care but did not consult their GP or outpatient services, respectively, over the preceding year due to economic reasons. The highest level of not consulting a doctor due to economic reasons was for dental care, reaching almost 13%. And as previously seen, the pattern of non-use of services (in the case of all three service categories) that persons reported needing is not distributed evenly; decisions to seek care were almost three times lower among poorer compared with richer quintiles (Statistical Office of Estonia 2005). Still, when this pattern of access is compared with that in other countries of the Former Soviet Union (Balabanova et al. 2004), Estonia seems to be in a more favourable situation. To fully understand the challenge that out-of-pocket payments represent to both financial access and financial risk, the ideal data source would be a household survey that includes both utilization and expenditure questions. When only expenditure questions are available (as in the surveys we analyse for this paper), we can be confident that the observed response understates the true barriers that exist, because those who are deterred from any service use for financial reasons do not have any expenditures that are captured by the survey. By connecting our survey analysis to the other available survey data in Estonia on utilization patterns, we believe that the case for concern among policy-makers about the implications of deteriorating financial protection is strengthened.

Secondly, the study shows that medicines comprise the largest share of out-of-pocket payments for those who are most at risk of experiencing financial hardship. Furthermore, the data analysis shows that poor households with elderly members are the most vulnerable. By considering this evidence together, we conclude that the ‘problem to be addressed’ is to extend greater protection against outpatient medicines costs for elderly persons (especially those with chronic conditions requiring medicines) from poor families.⁹ Lack of health insurance coverage cannot be an explanation for the heightened risk of impoverishment for the elderly because all persons of pensionable age are automatically covered. So the problem has to do not with the presence of insurance coverage but rather the depth of that coverage (i.e. the extent to which patients have to co-pay for various services, especially medicines) and the absence of targeted exemptions from co-payments for the poor elderly. We are confident in this conclusion even though the data did not include variables to capture specific concerns such as health insurance status or health service needs, or even just utilization.

Thirdly, in 2003 the co-payment policy was changed in many respects (see Table 1). On the one hand, additional financial protection was provided to those who face high pharmaceutical expenditure: the EHIF now reimburses 50% of a yearly cost of prescription between €383 and €639, and 75% beyond, up to a limit of €1278. Our data do not allow us to establish whether the high expenditure on medicines comes from prescription or over-the-counter drugs. Yet, general sales statistics show that prescription medicines account for about 75% of the entire market, so this measure could have improved the situation. In any case, whether this measure led to an improvement of the population's depth of coverage still needs to be evaluated in view of all aspects of the reforms and their overall impact on out-of-pocket payments. The 2003 reform package included the introduction of reference prices for pharmaceuticals to make generic drugs more accessible and applied a ceiling for each prescription which could increase out-of-pocket expenditures in real terms. This would depend considerably on the behaviour of doctors, pharmacies and patients, and is not yet evaluated.

Additional cost-sharing on services is included in the benefits package (primary health care home visit fee, inpatient bed-day fee, and regulated maximum visit fee applied by private entities), as well as the exclusion of most adult dental care from the benefit package. The situation in 2004 has been studied in relation to primary care home visit fees and it was found that visit fees on average are a barrier to fewer patients than in previous years. However, it has been reported that the home visit fee at its current level (€3.2) is limiting access to care for 70% of the lowest income group compared with 28% among the highest income group (Faktum 2004). The only regulation setting rules for private patients (when services are not covered by the EHIF) states that the fees applied have to be ‘reasonable’, with this being open to interpretation by providers.

While we cannot determine a causal link, National Health Account data (Figure 1) show that out-of-pocket expenditure increased by more that 30% in 2003. From 2002 to 2004, more people had health-related expenditures, with prescription and over-the-counter medicines being the most common expenditures (Faktum 2004). In this context, it is possible but doubtful that the burden of out-of-pocket payments shifted dramatically in a pro-poor direction, and that the additional coverage introduced on high pharmaceutical expenditure provides financial protection to patients with chronic conditions. Recent survey evidence shows that older patients, especially those over 65, consider the price of pharmaceuticals as more important than their clinical efficiency when a doctor is prescribing medicines (Faktum 2004). We believe that our analysis should be replicated to assess the distribution of these increased private payments and to monitor the impact of the changes in co-payment policy.

Fourthly, and closely related to the increase in patient obligations for co-payments, trends in Estonia show that as government spending on health declined in both real terms (slightly) and as a share of total government spending (significantly), the role of out-of-pocket spending increased. The burden of this spending has fallen most heavily on the poor. While the solution to the problem of improving financial protection for the poor will not come just from increasing government health spending, it is likely that the success of reforms to improve financial protection, e.g. targeted exemptions from co-payment, will have to be supported by increased public spending in order to ‘purchase’ this extra protection for the poor. One suggested option is to extend the revenue base of the health insurance to ensure long-term sustainability and the availability of funds (Couffinhal and Habicht 2005). For targeted exemptions, different options are available, varying from indirect measures such as regulating the list of prescription medicines for those whose economic situation or health condition is particularly poor, to direct ones such as having an annual medicines spending ceiling on the costs incurred by patients. However, as pharmaceuticals are increasing the costs for both insurer and patient, the impact of such arrangements will have to be evaluated, which is outside the scope of the current study.

Fifthly, some methodological comments. The tools we used to measure the extent of the problem are contextualized for Estonia: we look at ‘high’ out-of-pocket expenditure (20% of capacity to pay) rather than ‘catastrophic’ (40%) expenditure. On the other hand, our poverty line is higher than the national one, and therefore we may be underestimating the impact of out-of-pocket payments on poverty. In other words, the monitoring tool used here was adapted to reflect differing value judgements on what constitutes a problematic situation. Such flexibility with regard to different country contexts and value assessment is an important asset for a monitoring tool.

Finally, this study suggests that even with general population surveys not specifically oriented to the health system, and with a fairly simple tool, but a good understanding of how a country's health system is organized and knowledge of its reforms, policy-makers can monitor something that is extremely important to the health care system: the degree of financial protection provided. Further, the analysis can be disaggregated for ‘subgroups’ or for type of care. Hence, we could reach quite specific conclusions about the nature of the problem that has to be addressed, and hence target policy responses to improve financial protection against the risk of catastrophic health care costs. In Estonia, the results of this analysis have been conveyed to policy-makers in a manner that they could easily interpret (i.e. Figure 5 showing odds ratios for high health expenditures) and are now playing a role in the policy debate. Thus the results have raised awareness among specialists and politicians about the financial protection provided within the health system and who are the persons at risk in Estonia, as well as what the possible causes are. We believe that the relevance of this approach is not limited to Estonia but could be used anywhere that regular household surveys are implemented with some questions on health expenditures.

¹In their analysis of Indonesian health care utilization and expenditure data, Pradhan and Prescott (2002) note that conventional survey-based measures of exposure to catastrophic financial risk understate the actual risk faced by poorer households that do not seek care because of its cost (and thus reported zero health expenditures in the survey).

²Data are only available from 1996 on, and are not official for the first 2 years, as the National Health Account (NHA) system was launched in Estonia in 1998.

³The first (decline in public spending on health) perhaps induced the second (increased co-payments).

⁴Ignoring the fact that the authors of the different papers do not use the same measures of the burden as ours.

⁵More detailed explanations about how to compute these measures can be found on the World Bank website [http://www.worldbank.org] under Topics; Poverty and Health; Publications; Quantitative Techniques for Health Equity Analysis: Technical Notes.

⁶This conclusion has to be true unless the funding of the prepaid part of the health system became much more equitable during the same period. Although we cannot measure this from the available data, there is no reason to believe that there was any significant change in the distribution of payroll and other tax contributions for health.

⁷Some numbers may appear not to add up due to rounding.

⁸Depth of coverage is the extent to which services are covered from prepaid, pooled sources (i.e. available without out-of-pocket co-payment). The other dimension of coverage is breadth, which is the proportion of the population that is ‘covered’, i.e. has some degree of protection against the costs of care from prepaid sources. In other words, ‘depth’ refers to service coverage, whereas ‘breadth’ refers to population coverage (see Kutzin 1999, 2000).

⁹At the same time, it is important to keep in mind that improvements of the populations’ financial protection can be achieved by other means than decreases in co-payment rates, for instance, by influencing physicians’ prescription practices through guidelines or by encouraging the development and use of generics. Disease management programmes could also be used to address both quality and cost issues for the chronically ill.

Acknowledgements

This research was co-funded by the National Institute for Aging (grant no. Y1-A9–9421/OPHS-9–062), which had no role in the study design, execution or reporting. The authors would like to thank Anneli Taal, Anna Mylena Aguilar, Jan Klavus and Kei Kawabata for their valuable suggestions and their encouragements, a well as Kadri Ruusma for the preliminary analyses of the data. The authors would additionally like to thank the Statistical Office of Estonia for providing datasets for analysis. Raul Kiivet, Maris Jesse, Triin Habicht and Kees de Joncheere provided very useful comments on earlier versions of this document.

The views expressed in the article are those of the authors alone and do not necessarily reflect the decision or stated policy of the World Health Organization.