The Evolution of Income-Related Inequalities in Health Care Utilization in Switzerland Over Time

This study investigates equity in access to health care in Switzerland over time, using nationwide representative survey data from 1982, 1992, 1997 and 2002. Both simple quintile distributions and concentration indices are used to assess horizontal equity, i.e. the extent to which adults in equal need for medical care appear to have equal rates of medical care utilization. Looking at each of the four survey years separately the results indicate that by and large, there is little or no inequity in use except with respect to specialist visits which are clearly pro rich distributed as in most other OECD countries. We neither find much significant variation over time despite the fact that the share of health care has grown from close to 8% to more than 11% over this period and that a major reform of the health care system has taken place in 1996.


Introduction
Economists traditionally focus on efficiency. However, political decisions are notoriously cast by distributional considerations. This is most notably the case for social policy in general and health policy in particular. In the latter area the egalitarian view predominates that access to health care is every citizen's right (Williams 1993). This is reflected by the fact that a key policy objective in all OECD countries is to achieve adequate access to health care by all citizens on the basis of need. In line with the other OECD countries this egalitarian view predominates in Switzerland as well, at least as far as treatment aspects and not just amenities such as a (semi-)private hospital room are concerned. The overwhelming concern in the political debate seems to be that a two-class society with respect to health care should be avoided at all cost. This attitude is mirrored in the major reform of the Federal Health Insurance Law (KVG) enacted in 1996 which increased insurance coverage to 100%, extended comprehensiveness of the compulsory basic insurance package and increased solidarity between the healthy and the sick, the rich and the poor and men and women. Equal access was an explicit goal of this reform.
This study investigates whether and to what extent these ideologies are reflected in actual utilization patterns. While the focus of our earlier research was on cross-country comparisons (see for example Van Doorslaer et al. 2000) we analyze in this paper the evolution of horizontal equity in health care utilization over time, using nationally representative survey data from 1982, 1992, 1997 and 2002. In the literature horizontal equity is usually interpreted to require that people in equal need of care are treated equally, irrespective of individual characteristics such as income, place of residence, race, etc. It is this principle of horizontal equity that the present study uses as the yardstick for the comparison over time as well. The method we employ to describe and measure the degree of horizontal inequity in health care delivery is conceptually identical to the one used in Wagstaff and van Doorslaer (2000b), van Doorslaer et al. (2000), van Doorslaer et al. (2004) and van Doorslaer and Koolman (2004a) for cross-country comparisons. The method proceeds by comparing the actual observed distribution of medical care by income with the distribution of need. The results can be interpreted readily as performance measure of the Swiss health care system with respect to equity.
The paper is organized as follows: Section 2 provides a rather comprehensive description of the major features of the Swiss health care system as it emerged after the 3 1996 reform, including the relevant changes relative to the legislation preceding the reform. Section 3 outlines the estimation methods used. Data and variable definitions are discussed in section 4 while the results are displayed in section 5. Section 6 concludes.

The Swiss health care system
The Swiss health care system, instituted 1911 and revised 1996, is characterized by the following features: a) All permanent residents have to buy a compulsory basic health insurance policy from one of the officially acknowledged health insurance companies. Compulsory insurance policies can be freely changed by the end of each year. In addition, there is the possibility to buy supplementary health insurance (see below). b) There is a wide variety of compulsory policies. In the traditional fee-for-service sector which still accounts for over 90% of all insurance contracts policies may be differentiated by deductible levels. By contrast, there is no deductible for HMO plans which closely manage access to providers. A further possibility is to choose a "bonus" plan, a five-year policy that rewards enrollees who do not use insurance with substantial reductions in premiums. c) Compulsory insurance policies cover outpatient care, including a wide variety of providers, hospital care (100% coverage on the general ward of a public or publicly subsidized hospital in the enrollees' canton, except for a co-payment of CHF 10 per day), prescription drugs listed, preventive vaccinations, prescribed treatments in health resorts and alternative medicine. In addition, there are contributions to certain preventive tests, home care, glasses and medical devices as well as transportation and salvage expenses. In emergency cases up to twice the rate of the cantonal tariffs are paid for treatments abroad. Dental care, by contrast, is only covered in case of accident or severe dental problems. Compulsory insurance in the fee-for-service sector offers direct access and free choice of physician for outpatient care (general practitioners and specialists). Except for emergency cases hospitalisation requires referral by a physician. There is no choice of physician in the hospital. However, patients may freely choose among all hospitals which are included on the cantonal eligibility list. d) Supplementary insurance covers additional treatments and check-ups, all drugs, extended home care, provides generally higher benefits and up to 100% universal coverage world wide. Most important it provides access to the private ward of all public and private hospitals in a one or two-bedroom and free choice of physician in the hospital (medical or assistant medical director), depending on insurance package.
e) The overall number of insurance companies has decreased massively over the last decades, from 1'100 in 1960 to 93 in 2003. Since not all of these companies were active in all cantons consumers could choose from between 51 to 72 companies in each canton. Some of these companies have existed for more than 100 years, others are relatively new. Some have less than five thousand enrollees, others more than one million. There is free competition among the health insurance companies within the boundaries set by the Federal Health Insurance Law. f) Insurance companies offering compulsory insurance policies are heavily regulated.
They may not be profit-oriented, have to offer the same basic insurance package, and premiums as well as benefits are tightly controlled by the federal government. While insurers can reject supplementary insurance applicants, they must accept all those who apply for compulsory insurance. Premiums for compulsory insurance policies may only be differentiated by canton, type of region (urban, rural or suburban) and age (reduced premiums for children and young adults). To reduce the amount of risk selection induced by this type of premium regulation all insurance companies have to participate in a risk-adjustment pool. Depending on the risk structure of their enrollees insurers pay or receive contributions from this pool. Supplementary insurance policies may, by contrast, be adjusted for the enrollees' risk, and insurers may be for-profit.
g) The Swiss health care system is financed out of four sources, compulsory health and casualty insurance (31.9% in 2001), supplementary health insurance (10.2 %), out-ofpocket payments and donations (32.7%) and taxes (25.2%). Roughly 58% of the tax revenues were provided by the cantons, 31% by the communities and 11% by the federal government. As mentioned above, health insurance premiums do not depend on income. To ease the financial hardship associated with per capita premiums, government provides means -tested subsidies to low income residents. In 2001 roughly one third of the insured were subsidized to some extent, and about 15% of all 5 enrollee premium payments were paid for by the government. Subsidies are paid if premiums exceed a certain percentage of household income -usually 8%-10%. The maximum amount paid is typically the mean of the premiums of all insurance companies in the canton. Essentially, this means that the poor get free care. In the feefor-service sector patients have to pay a deductible of their choice (within a range set by the government) and a coinsurance rate of 10%. This explains the relatively high share of out-of-pocket payments. Insuring compulsory out-of-pocket payments is not permitted. Premium reductions associated with higher deductibles are regulated by government as well.
h) Licensed physicians are free to choose their location for outpatient care. Government is restricted to control the formal educational requirements for licensing. Physicians' fees for compulsory policies in the fee-for-service sector are negotiated between their union and the Swiss insurance association on the cantonal level and are overviewed by the government. If a consensus cannot be reached fees are set by the cantonal governments. By contrast, fees are not regulated with respect to supplementary insurance. Hence anti-trust law applies. Physicians can bill only for services that are covered by the insurer and may not supplement their bills for compulsory enrollees.
In the fee-for-service sector every licensed physician can bill every compulsory insurance policy (enforced contracting), i.e. the insurer cannot choose preferred providers except in the context of managed care, while providers cannot refuse treatment to any patient with a compulsory insurance policy. i) In 1999 there were 249 public or publicly subsidized and 143 private hospitals in Switzerland. The former accounted for roughly 80% of total acute care bed capacity.
Public and publicly-subsidized hospitals are operated by the cantons or the communities in which they are located or by some other non-profit organization.
They are guaranteed deficit coverage and/or subsidies from public funds. Capital investments, education, research expenses and at least 50% of the operating costs are financed by cantonal tax revenues. Since 1996, cantons can impose fixed budgets on public or publicly-subsidized hospitals. Private hospitals which may be organized as for-profit or non-profit are not eligible for public funds and have to substitute tax subsidies by payments from supplementary health insurance and patients. Prices (tariffs) for public hospitals are negotiated between the hospital association, the insurers association and the government on the cantonal level. Private hospitals 6 negotiate their own fees with the insurers and may not agree on a universal tariff because of the anti-trust law. Insurance companies usually pay hospitals per diem fees. In a number of cantons APDRGs (all patient diagnosis-related groups) have been introduced to pay public hospitals a flat rate per patient. The higher per diem rates for "hotel" costs and services in the private ward are covered by supplementary health insurance or out-of-pocket payments. j) Although the Swiss health care system is basically consumer-driven, government plays an important role. The system mirrors the federal structure of the country in that the cantons are responsible for securing the provision of health care according to the constitution. The health insurance law on the federal level provides the basic guidelines within which the cantons, the communities and the various other players can operate. In addition, the federal government participates in the financing of the means-tested premium subsidies operated by the cantons. The bulk of the tax revenues flowing into the health sector stems from the cantons, however. They are the dominant player in the hospital sector. Together with the communities they own and operate a majority of hospitals, finance more than 50% of hospital costs, participate in the negotiation of fees and tariffs, and set the tariffs if negotiations are not successful.
Last but not least they set up the eligibility list which is crucial for the remuneration of hospitals. Listed hospitals are entitled to tax subsidies and can bill every compulsory insurance policy for all services included in the basic package. Since the cantons control together with the communities on average about 80% of total bed capacity their position in the provision of inpatient care is very powerful, determining the allocation of hospital resources to a large extent.
k) In comparison with other highly developed countries like Germany, Great Britain, Canada or even the US the population in Switzerland generally has more health care resources available (OECD 2002). While this may be seen as an achievement -the share of GNP for health care amounts to roughly 11% -economists have argued that the level of provision is in some areas inefficiently high, in particular with respect to hospital care. The positive side of this ample capacity is that queuing is not a major problem and that shortages of care in rural areas do not occur to any significant degree despite large differences in regional health care supply. Quality of care is generally regarded as high and health outcomes measured by selected indicators compare favourably with those of other countries (WHO 2000, OECD 2002. 24 % of 7 the population rate their health status as very good, 62 % as good, 11 % as average and 3 % as poor or very poor (Swiss Health Survey 2002).
Most of the institutional features described above have been brought about by the 1996 reform of the Health Insurance Law. Three main goals were pursued with this major reform: a) Extension of the insurance coverage for the entire population (universal coverage, expansion of the compulsory basic insurance package); b) increase in solidarity between the healthy and the sick, the rich an the poor and men and women (flat insurance premiums, premium subsidies and universal access to compulsory health insurance); c) reduction in cost inflation. There is widespread agreement that the first two goals have been achieved to a considerable degree, while cost inflation remains a major problem.
Extended insurance coverage and increased solidarity can be expected to have increased equity in utilization. In addition, supplementary insurance coverage which is known to contribute to inequities in use has decreased since 1996 due to the extension of the compulsory insurance package. By contrast, the instruments of the revised law aiming at reducing costs, in particular increased cost-sharing and managed care options tend to reduce equity. Since neither explicit rationing nor sizable queues have been observed so far we do not expect, on the other hand, that the possibility of imposing a fixed budget on public hospitals has influenced the distribution of health care. Overall, we are left without a clear a priori hypothesis of whether the 1996 reform has increased or decreased equity in utilization. Tackling this question therefore boils down to a purely empirical exercise, using nationwide representative survey data before and after 1996 and keeping in mind that other factors may have exerted an influence over this extended period of time as well.

Estimation methods
This section draws heavily on the OECD Health Working Paper No. 14 by van  Jones, 2000, for a review). While these models have certain advantages over OLS specifications, their intrinsic nonlinearity makes the (linear) decomposition method described in section 3.2 impossible. In order to restore the mechanics of the decomposition, one has to revert to either decomposing inequality in the (latent variable) propensity to use (rather than actual use) or to a re-linearization of the models using approximations (see Van Doorslaer, Koolman and Jones, 2003, for an example). However, Van Doorslaer et al. (2000) have shown that the measurement of horizontal inequity hardly differs between OLS-based TPMs and non-linear TPM specifications such as the logistic model combined with a truncated negative binomial model.
Following the OECD paper, we have therefore chosen a pragmatic approach. We use simple OLS estimation for the decomposition based measures and we check the sensitivity of the HI indices and quintile distributions by comparing these with the indices and distributions obtained using non-linear specifications. We obtained "needed" health care use based on a generalized negative binomial model for total consumption, a logistic specification for the probability of use, and a truncated negative binomial model for the conditional positive use. In comparing the HI indices obtained using linear versus nonlinear models, we found that the estimates are extremely similar and that in only very few cases, the linearly and non-linearly estimated indices differ significantly (not shown).
This provides some reassurance that our results are not conditional on the choice of the linear standardization model.
For the four survey years, cross-sectional sample weights were used in all computations in order to make the results more representative of Switzerland's population. Robust standard errors were obtained using the Huber/White/Sandwich estimator.

Measuring inequity
This study measures distributions of actual and needed use of care by income quintiles. These are groups of equal size, each representing 20% of the total (adult) population, but ranked by their household income from poorest to richest. The "needed" health care use is computed by running a regression on all individuals in the sample, But these quintile distributions are difficult to compare across a large number of several data points and types of care use. It is therefore useful to summarize the degree of inequality observed using a concentration index. It is defined as (twice) the area between a concentration curve and a line of perfect equality. A medical care concentration curve plots the cumulative proportion of medical care against the cumulative proportion R of the sample, ranked by income (Wagstaff and Van Doorslaer, 2000a and b).
A concentration index of a variable y can be computed using a simple "convenient covariance" formula, which looks as follows for weighted data: where m y is the weighted sample mean of y, cov w denotes the weighted covariance and R i is the (representatively positioned) relative fractional rank of the ith individual, defined as : where w i denotes the sampling weight of the ith individual and the sum of w i equals the sample size (n).
Testing for differences between concentration indices requires confidence intervals.
Robust estimates for C and its standard error can be obtained by running the following convenient (weighted least squares) regression of (transformed) y on relative rank: It is worth emphasizing that coinciding concentration curves for need and actual use provide a sufficient but not a necessary condition for no inequity. Even with crossing curves, one could have zero inequity if, for example, inequity favoring the poor in one part of the distribution exactly offsets inequity favoring the rich in another

Decomposing and explaining horizontal inequity
It is possible to estimate the separate "contributions" of the various determinants and their relative importance. Using the regression coefficients k γ , (partial) elasticities of medical care use with respect to each determinant k can then be defined as: It has been shown (Wagstaff, Van Doorslaer and Watanabe, 2003) that the total concentration index can then be written as: where the first term denotes the partial contribution of income inequality, the second the (partial) contribution of the need variables, and the third the (partial) contribution of the other variables. The last term is the generalized concentration index of the error term ε.
In other words, estimated inequality in predicted medical care use is a weighted sum of the inequality in each of its determinants, with the weights equal to the medical care use elasticities of the determinants. The decomposition also makes clear how each determinant k's separate contribution to total income-related inequality in health care demand can be decomposed into two meaningful parts: i) its impact on use, as measured by the use elasticity (η k ), and ii) its degree of unequal distribution across income, as measured by the (income) concentration index (C k ). This decomposition method therefore not only allows us to separate the contributions of the various determinants, but also to identify the importance of each of these two components within each factor's total contribution. This property makes it a powerful tool for unpacking the mechanisms contributing to a country's degree of inequality and inequity in use of health care.

Data and variable definitions
The data used in this paper were taken from four nationwide representative crosssectional data sets, the survey on Socio-Medical Indicators for the Population of Switzerland (SOMIPOPS) which was collected in 1982, and the Swiss Health Surveys (SHS) 1992, 1997 and 2002. While the latter three use almost identical questions, there are some notable differences between SOMIPOPS and the SHS data sets. First, while the information on household income in SOMIPOPS is derived from official tax records (see Leu et al. (1986)) income in the SHS data is self-reported. We cannot exclude that the substantial increase in income from 1982 to 1992 is at least partly due to this difference in data source. In the present context, this causes a problem only if the two income variables lead to a different ranking of individuals by income. Second, the self-assessed health question in SOMIPOPS has only four categories (very good, good, fair, bad) while the SAH questionnaire contains five items ranging from "very good" to "very bad". The other two health variables used are coded as dummy-variables. The first is derived from questions on the existence of twelve different physician-diagnosed but self-reported chronic illnesses which were included in all surveys. The second addresses problems with activities of daily life (ADL) and is derived from comparable questions on the existence on a physical or mental health problem which hampers individuals in daily life. One problem with the 1992 SHS data set is that it only contains information on all physician visits which cannot be disentangled into GP and specialist visits. Table 1 presents descriptive statistics for the four data sets.

The evolution over time
In We further disentangle total use into the probability of accessing a specific care category at all and the conditional number of visits with a physician given an initial contact has taken place. The evolution of the HI indices for all physician visits, GP visits, specialist visits and hospital nights is also illustrated by figures 1a to 1d which display the The decomposition graphs show that the most important determinant of higher health care utilization among the poorer individuals is the unequal distribution of health across the population. It should be mentioned, however, that Switzerland exhibits one of the most equal distributions of health, as can be seen by comparing results from a study on inequalities in health in Switzerland over time by Leu and Schellhorn (2004)  Whether this negative ME indicates incentive effects or reflects unobserved aspects of health is a much debated issue in the literature (see e.g. Schellhorn 2001, 2002a,b and Werblow and Felder 2003 and subject of ongoing research. There is also substantial inequality in utilization across regions. The varying ME of the various regions on utilization can be partly explained by the different specialist densities across these regions.

The Swiss results in the international context
To put the Swiss results into a meaningful context we compare them with the results of recently published studies by van Doorslaer et al. (2004) and van Doorslaer and Koolman (2004b) which compares income-related inequality in health care utilization across 21 OECD countries and 12 EU countries respectively. With respect to all physician visits there are no significant inequities (controlling for need) in a majority of these countries including Switzerland. The distribution is pro-poor in Ireland and Belgium, and pro-rich in the US, Finland, Sweden and Austria. In about half of these countries the distribution is pro-poor for GP visits. In Finland, it is pro-rich while no significant differences can be found in the remaining countries including Switzerland.
The situation is very different for specialist visits. In every country for which the necessary data exist the rich are more likely to see a specialist, after controlling for need differences, and in most countries the rich visit a specialist also more frequently. The HI-Index for Switzerland takes a midfield position in the group of countries showing prorich inequity. Finally, no significant inequities can be found with respect to hospital nights in the 21 countries excepting Canada (pro-poor) and Mexico (pro-rich). The HI index for Switzerland is among the most pro-poor.

Conclusions
We have used both simple quintile distributions and concentration indices estimated using regression models to assess the extent to which adults in equal need for medical care appear to have equal rates of medical care utilization in Switzerland. In particular, we have focused on the evolution of horizontal equity in health care utilization over the last two decades, using nationally representative survey data for 1982, 1992, 1997 and 2002. We believe that this approach is justified because the horizontal equity principle "equal treatment for equal need" is a widely accepted policy goal in Switzerland as far as compulsory health insurance is concerned.
Except for specialist visits (and hospital nights in 1982) we have found no significant differences in the distribution of medical care by income. For all those who share the egalitarian interpretation of horizontal equity this is good news. However, a crucial question which cannot be tackled with survey data is whether and to what extent the remaining differences translate into inequities in health outcomes. A related question is whether treatment quality is the same even when there are no inequities in utilization, measured by physician visits and hospital days. For example, it might be that the better off (the better educated) get their treatment in better hospitals or are treated by better qualified physicians once they experience a hospital spell (in particular, when they have supplementary insurance). The described inequity in accessing specialists might have such an effect as well.
In line with the international literature we find that the most important contributor to income-related inequalities in health care utilization is an unequal distribution of need. The