Out-of-Pocket Maximum Rules under a Compulsory Health Care Insurance Scheme: A Choice between Equality and Equity

Transcription

1 Document de travai Working paper Out-of-Pocket Maximum Rues under a Compusory Heath Care Insurance Scheme: A Choice between Equaity and Equity Thierry Debrand (Irdes) Christine Sorasith (Irdes) DT n 34 Novembre 2010 No copying aowed in other websites, but direct ink to the document is authorized: Institut de recherche et documentation en économie de a santé IRDES - Association Loi de rue Vauvenargues Paris - Té. : Fax :

2 Institut de recherche et documentation en économie de a santé 10, rue Vauvenargues Paris Té: Fax: E-mai: pubications@irdes.fr Director of pubication: Yann Bourguei Scientific adviser: Thierry Debrand Transator: Véronique Dandeker Copy editing: Franck-Séverin Cérembaut - Anne Evans Layout compositor: Khadidja Ben Larbi Distribution: Suzanne Chriqui, Sandrine Bequignon These current findings wi be submitted for pubication in a peer-reviewed journa. This document is mainy addressed at the scientific and academic community. Authors are soey responsibe for content disseminated.

3 Out-of-pocket Maximum Rues under a Compusory Heath Care Insurance Scheme: A Choice between Equaity and Equity 1 Thierry Debrand a, b and Christine Sorasith a Abstract Using the microsimuation mode ARAMMIS, this study attempts to measure the impacts of introducing an out-of-pocket (OOP) maximum threshod, or a safety net threshod, on consumer copayments for heath care financed by the aboition of the Long-term Iness Regime (ALD) in France. The anaysis is based on a comparison of different safety net threshod rues and their redistributive effects on patients OOP payments. We attach particuar importance to indicators that bring to ight changes in OOP payment eves and measure their impact on the equity of OOP distribution. The first section outines the French Nationa Heath System to provide a better understanding of the stakes invoved in reforming the heath care reimbursement rues under the Compusory Heath Care Insurance scheme. In the second section, we describe the hypotheses retained, the database and the microsimuation mode. The fina section presents key findings, measuring the impact of the reform at both individua and system eves. Keywords: Microsimuation, Heath expenditure, Out-of-pocket payment. Codes JEL: I18, H51, D63. a IRDES (Institut de recherche et de documentation en économie de a santé), Paris, France. b Corresponding author: debrand@irdes.fr; te: +33 (0) The authors woud ike to thank a those that participated in the PSE and IRDES seminars, Groink Insurance at IRDES, the JESF 2009 conference and the 2010 Irdes Workshop. We woud particuary ike to thank T. Van Ourti, B. Dormont and P-Y. Geoffard for their commentaries. This study stems from research carried out by A. Couierot, S. Chambaretaud, T. Debrand and L. Rochaix in 2007 and coaboration with R. Lega. Usua discaimers appy Document de travai n 34 - IRDES - Novembre

4 Résumé Boucier sanitaire : choisir entre égaité et équité 2 Une anayse à partir du modèe ARAMMIS Cet artice cherche à mesurer, à aide du modèe d Anayse des réformes de Assurance maadie par micro-simuation statique (ARAMMIS), es effets de a mise en pace d un boucier sanitaire financé par a suppression du régime des affections de ongue durée (ALD). Notre étude repose sur a comparaison des conséquences redistributives de différentes règes de bouciers sur es restes à charge des patients dans e secteur ambuatoire en France. Nous attachons une importance particuière aux indicateurs permettant de mettre en évidence es modifications des restes à charge et de mesurer évoution du système en termes d équité. Nous présentons, dans une première partie, e cadre généra du système de santé en France pour mieux comprendre e contexte et es enjeux d une refonte du mode de remboursement ié à Assurance maadie obigatoire. Dans une deuxième partie, nous décrivons es hypothèses retenues, a base de données et e modèe de micro-simuation. Enfin, nous consacrons a dernière partie à a présentation des principaux résutats mesurant impact de a réforme tant au niveau des individus qu au niveau du système. Mots-cefs: Micro-simuation, Dépenses de santé, Restes à charge. Codes JEL: I18, H51, D63. 2 Les auteurs tiennent à remercier es participants aux séminaires de a Paris Schoo of Economics et des mardis de Irdes, du Groink Assurance à Irdes et des Journées des économistes de a santé français. Nous tenons pus particuièrement à remercier A. Evans, B. Dormont et P-Y. Geoffard pour eurs commentaires. Ce travai trouve son origine dans es travaux réaisés par A. Couierot, S. Chambaretaud, T. Debrand et L. Rochaix en 2007 et d une coaboration avec R. Lega. Document de travai n 34 - IRDES - Novembre

5 Introduction The present Socia Security system in France officiay came into being with the Ordinance of 1945, which estabished a pact of soidarity between a French citizens. A Socia Security system has been instituted to guarantee workers and their famiies protection against socia risks of whatever nature susceptibe of reducing or impeding their abiity to earn a iving, to cover maternity costs or famiy expenses. Its aim is to protect individuas from socia risks reated to heath, empoyment, the famiy and retirement. In terms of heath, socia protection is provided by a Compusory Heath Care Insurance scheme, to which contributions are mandatory, based on the fairy simpe principe that each pays according to their abiities and receives according to their needs. To this basic principe are added a certain number of schemes that permit heath care costs partiay covered by Socia Security (Sickness Insurance) to be whoy reimbursed for certain diseases or under specific circumstances (e.g., ong-term iness, or the ALD scheme, pregnancy, and occupationa accidents). The system exhibits specific reimbursement characteristics: independent of income and remaining out-of-pocket payments (OOP) can be excessive (OOP payments are the difference between actua heath expenditures and the amount reimbursed by Compusory Heath Care Insurance). Excessive or catastrophic OOP payments can be defined in two ways. In the first definition, catastrophic OOP payments are associated with a period of iness that temporariy or duraby deteriorates a patient s heath status and requires extremey expensive care that is not aways reimbursed by the Compusory Heath Care Insurance scheme. In the second definition, OOP payments are considered as an absoute vaue measured against income or an individua s abiity to pay (Abu Naga and Lamiraud, 2008). In this case, OOP payments are quaified as catastrophic if they ower a househod s standard of iving or if they represent a significant percentage of the househod income over and above a certain critica threshod (Xu et a., 2003; Wagstaff and van Doorsaer, 2003; van Doorsaer et a., 2007; Pradhan and Prescott, 2002 and Fores et a., 2008). Economic iterature provides a number of studies that estimate this threshod at between 5% (Berki, 1986) and over 40 % (Xu et a., 2003) according to country and socia protection systems. Catastrophic OOP payments are an indicator of a heath insurance system s performance and can revea insufficient risk coverage (Schei-Adung et a., 2006). The notion of equity in a reimbursement system effectivey impies that househods shoud be protected against such excessive expenditures. Moreover, a ink between poverty eves and excessive heath expenditures has aready been estabished (Abu Naga and Lamiraud, 2008). According to Whitehead, Dahgreeb and Evans (2001), an increase in OOP payments pushes some famiies down to the poverty eve and aggravates hardship among famiies aready at the poverty eve. These additiona heath expenditures repace short-term consumption and ong-term savings, thereby reducing famiies we-being. In France, even if the majority of excessive OOP expenses are captured by the specific heath insurance schemes, high OOP expenses persist (Tabuteau, 2006). The way OOP payments are distributed between patients and the inequaities they generate have ed pubic poicy-makers and economists to question the sustainabiity of the current system and to imagine possibe evoutions. One of these aternatives invoves substituting the ong-term iness scheme (ALD: system that provides 100% coverage for patients with a recognized chronic disease), with one that provides an Document de travai n 34 - IRDES - Novembre

6 OOP maximum threshod or a copayment safety net threshod 3 (Bras, Grass and Obrecht, 2007). Briet and Fragonard (2007) present the interests, impacts and imitations of instituting a safety net system introducing a threshod on OOP payments. These authors consider that an OOP maximum threshod woud protect househods from catastrophic heath expenditures, reduce the number of househods foregoing care by protecting ow-income househods and, finay, simpify the current compexities of the reimbursement system whist faciitating the contro of pubic heath spending. The nature of the OOP maximum threshod thus depends on the chosen definition of catastrophic OOP payments. Based on the first definition of catastrophic OOP, the safety net woud introduce a uniform, fixed OOP maximum appicabe to a and independent of income. If one retains the second definition, in which OOP payments are reated to income, the safety net threshod woud necessariy be determined according to income eve. The notion of equaity in the distribution of OOP payments is defined as the equa treatment of a individuas, and equity is defined as the notion of proportiona equaity, suggesting that a individuas are treated equay but according to income eve. Our objectives in this study are to simuate hypothetica reform modes and to measure their redistributive effects. Particuar importance is thus attached to indicators that enabe us to measure the evoution of equity within the system under anaysis. In this context, the study is structured as foows: The first section outines the French Nationa Heath System to provide a better understanding of the stakes invoved in reforming the heath care reimbursement rues under the Compusory Heath Care Insurance scheme. The second section describes the retained hypotheses, the database and the microsimuation mode, ARAMMIS (Anayse des Réformes de Assurance Maadie par Microsimuation Statique; Anaysis of Heath Insurance Reforms by Static Microsimuation). ARAMMIS is a static, exogenous microsimuation mode buit by IRDES with the aim of simuating different reforms by modifying the variabes used in cacuating the financia burden of the insured, such as reimbursement rates, financia contributions and the possibiity of aboishing one or severa forms of exoneration. The fina section is devoted to presenting the key findings, using different toos to measure the impact of the reform at both individua and system eves. 1. The stakes invoved in reforming the reimbursement system In France, heath insurance is mainy provided by the Compusory Heath Care Insurance scheme. Essentiay Bismarkian, the system was originay financed in the majority by empoyer-empoyee contributions. Around the mid-1970 s, with the arriva of mass unempoyment, the reduced share of wages in the nationa revenue, the aging of the popuation and a certain desire for greater socia justice, the system evoved to adopt a more Beveridgian ogic, as it effectivey became more and more dependent on financing from taxes (Paier, 2005). The heath insurance scheme is essentiay financed by socia contributions deducted from earnings, income tax and other taxes, such as the Genera Socia Tax (CSG). This suppementary income tax, created in 1990, accounts for approximatey 30% of Compusory Heath Care Insurance funding. Contributions are thus income-reated rather than risk-reated, which can be interpreted as a redistribution mechanism that shoud guarantee a degree of equaity between the insured in terms of heath care costs. These statutory contributions cover a percentage of the heath expenditures of a the insured. Other sources of heath insurance funding come from 3 For us, in this artice an OOP maximum threshod and a copayment safety net threshod are synonymous. Document de travai n 34 - IRDES - Novembre

7 copayments, the percentage of a patient s heath expenditures that are not reimbursed by the insurance system (cf. Ebaum, 2008, for a more detaied presentation of patient contributions to heath expenditures in France). Since the beginning of the 2000 s, the distribution of heath expenditures reimbursed by the Compusory Heath Care Insurance system has remained reativey stabe; 65% of expenditures concern ambuatory care (outside hospita care expenditures), even if the atest statistics indicate a sight drop. This reative stabiity in fact hides radica changes that occurred in the wake of two contradictory trends. The first was an increase in patients individua contributions, which effectivey decreased coective contributions. The second came from the introduction of the Long-Term Iness scheme (ALD), through which an increasing number of insured were abe to benefit from 100% reimbursement of heath care expenditures. The two together atered the nature of the reimbursement system. Stabiity is obtained through the seection of risks: the system provides ess protection from sma risks by increasing patients financia contributions and, on the contrary, provides better risk and expense coverage for patients that are chronicay i. Progressivey, this risk seection eads to the non-avowed questioning of the 1945 soidarity pact that in turn radicay modifies the system s redistributive effect Heath insurance: increase in expenditures and patients financia contributions The Nationa Heath Insurance system designers instituted patient copayments from the beginning with the introduction of the ticket modérateur, a fixed copayment system. Uniformy appied to a of the insured, it can nevertheess have a negative impact on access to heath care. The ticket modérateur was egitimized, on the one hand, by the desire to reduce socia injustices and, on the other, by its potentia to reduce ex-post mora hazard by instiing a greater sense of consumer responsibiity. Driven by technoogica progress, heath care consumption behaviors and use rates, and to a esser extent the impact of an aging popuation, heath spending nevertheess continued to outpace nationa weath (Grignon, Huber and Dormont, 2006). Pubic heath expenditures increased more rapidy than pubic revenues (Ebaum, 2008). Since 1950, successive governments have initiated a series of recovery pans. The Vei, Barrot, Seguin, Evin, Bianco (and the foowings) reform pans, initiated between 1967 and 1995, concerned increases in contribution eves, cut-backs in the reimbursement of prescribed medicines, raising copayments for hospita care and the creation of a fixed copayment for hospita services. These reforms proved to be financiay ineffective (the Heath Insurance budget deficit sti exists) and sociay unjust (affecting a individuas whatever their income). Subsequent governments introduced more structura reforms with the aim of reguating the heath system and modifying agents behaviors (Juppé Ordinances of 1995). The 2004 Douste-Bazy Reform Pan instituted the introduction of the fixed 1 copayment for consutations, increased copayments for hospita services from 12 to 16 and introduced the Persona Medica Fie aimed at coordinating a patient s care path. In 2006, a fixed, universa copayment of 18 on medica acts costing over 91 was introduced and, in 2008, deductibes on prescribed drugs, paramedica care and transport. Document de travai n 34 - IRDES - Novembre

8 It shoud not be forgotten, however, that a reduction in Socia Security reimbursements designed to more or ess offset budget deficits can have an adverse effect on owerincome popuations that may be incined to forego care in the face of increased costs. It can aso significanty increase OOP payments for the chronicay i Heath insurance and heath and socia risk coverage There are three main insurance systems, either pubic or private, to cover these two types of risk. The first, a pubic insurance system, is the Universa Heath Insurance Coverage (CMU- Couverture Maadie Universee). It aows the most disadvantaged popuation group (individuas whose monthy income per consumption unit is beow 598 ) to benefit from 100% heath expenditure 4 coverage that, in theory, exempts them from OOP expenses. The effects of this universa coverage are not anayzed in this study. The second, a private insurance system, concerns suppementary heath insurance. It competes the reimbursement eves offered by the statutory heath insurance scheme 5 (Grignon, Perronnin and Lavis, 2008). Even if 93% of the French have suppementary insurance coverage (of which 7% thanks to the CMU), househods without suppementary insurance are often the most fragie with the owest income eves. Transferring the weight of reimbursements from Nationa Heath Insurance to suppementary insurance, however, may resut in increasing heath inequaities and reinforcing the CMU threshod effect. The introduction of the ACS (Aide à a Compémentaire Santé) scheme, providing financia assistance for the acquisition of suppementary insurance, aims at imiting this threshod effect (Grignon and Kambia- Chopin, 2009). Moreover, it shoud not be forgotten that suppementary insurance is paid directy or indirecty by individuas and that a the additiona heath care services covered by these private insurance schemes are rapidy transated into higher premiums paid by the insured. This is a the more probematic because the premiums are not proportiona to income. Finay, the third system, a pubic insurance scheme that wi be deat with more specificay in this study, concerns the Long-Term Iness Scheme (ALD). Individuas suffering from a chronic iness are covered at 100% for a expenses reated to that specific iness. For any other unconnected heath event, reimbursements come under the genera insurance statutory insurance scheme. The ALD is not a medica concept but a medico-administrative one. Its aims are economic, on the one hand, by attempting to neutraize catastrophic heath costs and medica on the other hand, by ensuring the better foow-up care of patients recognized as suffering from a ong-term iness. The ist of recognized diseases does not incude a chronic, costy diseases but rather focuses on diseases with high ong-term therapeutic costs. The number of patients admitted to the ALD scheme increases by 3.5% every year (Païta and Wei, 2009). This scheme currenty generates 62.3% of Nationa Heath Insurance reimbursements, whereas it ony concerns 14.6% of the popuation insured, that is, 8.3 miion individuas (Païta and Wei, 2008), and wi represent over 70% of expenditures in 2015 (Obrecht, 2009). Despite its usefuness and importance, the ALD scheme is not without its shortcomings. Firsty, it creates ratchet effects: few individuas eave the scheme vountariy. Secondy, 4 Heath professionas are obiged to appy binding tariffs to individuas benefitting from the CMU (price fixed by the government without the possibiity of charging above the statutory fee). 5 Suppementary insurance offers incompete coverage. It does not, for exampe, cover deductibes. Document de travai n 34 - IRDES - Novembre

9 the diseases covered by the ALD scheme are extremey heterogeneous, and there is no correation between the cost and gravity of an iness and its incusion in the ALD scheme. Thirdy, it can be a source of inequaities. It is the Genera Practitioner (GP), in a one-to-one reationship with the patient in a specific environment, who decides whether or not the patient is eigibe for ALD. Finay, it does not competey resove the probem of catastrophic OOP payments for a the patients benefiting from the ALD scheme (Geoffard, 2006). This two-fod observation: the trend dynamics that point to an increase in the number of persons benefiting from ALD and the continued existence of arge OOP payments ead to refections on possibe ways of improving the current system (Bras, Grass and Obrecht, 2007). One possibiity woud be to repace the ALD scheme with a system of OOP maximums. Briet and Fragonard (2007) proposed the safety net threshod for OOP payments, a system aready in pace in numerous European countries. 2. The reform: hypotheses, simuation toos and fieds of investigation The idea is based on a system that woud impose a safety net threshod for OOP payments. In other words, once the threshod is reached, additiona heath expenditures woud be universay reimbursed at 100%, regardess of the disease or care motive. The utimate goa is to find an equitabe reimbursement system that is not simpy aimed at curbing pubic heath costs to the point of penaizing the sick. In Europe, countries such as Switzerand, Germany, Sweden, the Netherands and Begium have aready instituted a system of OOP maximums either gobay or by sector (Chambaretaud and Hartman, 2007). In Begium and Germany, OOP maximums are fixed according to annua househod income. In these countries, where copayments are intended to increase user responsibiity with regards to heath care consumption, recourse to suppementary insurance is imited, prohibited in Switzerand and non-existent in the Netherands and Sweden. Compared to these countries, France appears to be behind in deaing with this issue but has severa targeted protection mechanisms reated to heath status (insured through the ALD scheme) or status (e.g., pregnant women, the disabed and occupationa accidents). These two systems (genera insurance scheme with OOP maximums or specificay targeted poicies) are seemingy incompatibe 6. In European countries, another criterion aside from revenue can intervene in the definition of OOP maximums: heath status. In Germany and Begium, specific schemes for patients suffering from chronic diseases exist without a predetermined ist of eigibe diseases having been estabished beforehand. 6 In a of these countries, however, there exist targeted protection mechanisms aimed at pregnant women and the edery. Document de travai n 34 - IRDES - Novembre

10 2.1. Hypothetica OOP safety net threshod modes The reforms tested are based on severa hypotheses: the aboition of the Long-Term Iness scheme (ALD), the creation of an OOP expenses safety net threshod and the neutraity of the insurance system (the heath insurance system wi neither gain nor ose from the reform). Mathematicay, the condition statement optimization program consists of finding the OOP maximum that confirms the foowing equiibrium: where i represents the insured in our database, OOPcurrent are OOP payments before reform and OOPafter are OOP payments after reform. This statement condition optimization aows us to focus soey on the redistributive nature of OOP payments cacuated according to a maximum threshod rue. The OOP maximum is defined as the threshod that protects the insured against catastrophic expenses as, over and above that threshod, expenses woud be covered at 100%. In this anaysis, we chose to finance the safety net threshod scheme soey through the aboition of the ALD scheme. The CMU threshod rues are not atered. Where income is taken into account in the anaysis, income eve is taken as being above the CMU threshod. This being the case, the different safety net threshod eves ony change in form and nature. We thus propose anayzing three OOP-safety net threshod modes. - The uniform threshod (UT) appies a uniform OOP threshod to a of the insured regardess of their incomes. This can be compared to the guaranteed stop oss offered by certain private insurance companies in the United States (Cuter and Zeckhauser, 2000). In the framework of the French debate, the idea of a sickness deductibe was put forward by Simon 7 (2007): - The income-reated threshod (IRT) fixes an individua OOP threshod cacuated according to gross annua income by consumption unit. The rue of proportionaity is written as foows: - The income-reated threshod with an increasing margina effect (IRTM) is based on the preceding rue but marginay increases the proportionaity coefficient by cacuating OOP threshods by income bracket: weathier individuas wi have a marginay higher OOP threshod before being eigibe for 100% coverage: If income is beow 1,200 : If income is between 1,200 and 1,800 : If income is between 1,800 and 2,400 : If income is over 2,400 : 7 The proposition made by Simon (2007) was more ambitious in that it was based on the assumption that the current reimbursement system (heath insurance reimbursement + copayment) woud be repaced by a fixed, uniform deductibe. In a of our simuations, we maintained the current system to observe the effect of aboishing the ALD scheme excusivey. The Simon simuation might be envisaged in the future. Document de travai n 34 - IRDES - Novembre

11 The income-reated maximum threshod is very simiar to the system of Maximum Biing (MAB) that was introduced in Begium in 2002 (Schokkaert et a., 2008) Threshod coefficients α UT, α IRT and α IRTM are cacuated to guarantee the neutraity of the different reforms with regards to the Compusory Heath Care Insurance scheme. The three OOP maximum modes are each based on a different ogic. The UT assumes that a individuas are treated identicay regardess of income. Thus, it respects the principe of uniformity underying the 1945 soidarity pact, ensuring that a individuas have an identica maximum risk protecting them from excessive OOP payments. The UT hypothesis, however, fais to take into account individuas abiity to pay: the financia participation reated to income demanded from individuas with more modest incomes is greater than for those with higher incomes. The OPP threshod proposed by the UT mode wi not protect ower income individuas from catastrophic OOP payments in reation to their income. The IRT and IRTM modes differ in the fact that threshods are cacuated taking income eves into account. For these two modes, the copayment threshod increases according to income eve, but for the IRTM, this increase is noninear 8 (it wi be ower for the ower income brackets and higher for the higher incomes than in the case of the IRT mode). To take heath status into account, as is the case in Germany (Busse and Riesberg, 2004), we equay simuate threshod rues that distinguish the ALD from the non-ald popuation (here, we use the indicator being covered by ALD as a heath status proxy). Each rue wi therefore incude a variant taking into account the ALD status of the insured. The threshod or coefficient of proportionaity wi be two times ower for the popuation on ALD. Athough it may seem surprising to use the ALD criterion in estabishing the OOP maximum rue when the ALD scheme is assumed to have been aboished, the idea is not to reinstate the ALD scheme but rather to use it as a medica criterion permitting the heath status of the most chronicay i to be taken into account Interest of microsimuation modes in heath poicy issues Microsimuation modes have proved efficient in anayses measuring the effects of administrative or fisca reforms on individua agents: a case that appies to the projected OOP threshod reform. Microsimuation is a method of investigation based on a representative sampe of microeconomic units. The microsimuation method in its appication to economic and socia poicies was originay defined from the end of the 1950 s by Orcutt (1957). His aim was to study the evoution of a system: to study the impact of a new reform by using the characteristics of these microeconomic units. The microsimuation mode thus stems from a database of individuas that aggregates the resuts obtained for each of the units in order to study the system as a whoe. In genera, one can distinguish between two main types of microsimuation modes: static and dynamic 9. The static mode, the type used in this anaysis, uses a cross-section database and a date t. Static modes are essentiay used to measure the immediate or short-term 8 Mathematicay, the second derivative of the income-reated IRT threshod is equa to zero, whereas, for the IRTM threshod, it has a positive vaue. 9 The dynamic modes use cross-section or ongitudina data, taking into account the demographic evoutions of the micro-units. The characteristics of each individua are updated on each timeine on the basis of assumed evoution (e.g., matrices, state transition equations and institutiona rues). The number of rues thus evoves through time, taking into account marriages, births and deaths. Dynamic modes are essentiay deveoped in the framework of ong-term pubic poicy (e.g., pension reform and detaied demographic modes). Document de travai n 34 - IRDES - Novembre

12 impacts of fisca or socia reform. New reforms can thus be simuated to measure the effects of poicy on income distribution at the individua eve and estimate the impact at the microeconomic eve. These modes are frequenty used in Ango-Saxon countries as an aid to poicy-makers (for a more genera presentation, see Gupta and Harding, 2007). One of the areas in which microsimuation modes have been appied is pubic heath poicy (Breui-Grenier, 1999). In France, unfortunatey, even if a number of modes demonstrated an interest in using this method to unrave the compexities of the heath system (Lachaud, Largeton and Rochaix, 1998), these were temporary, study-specific rather than permanent modes. The ARAMMIS mode was created by IRDES to fi this gap. Specificay, one of our aims is to create a microsimuation mode that is permanent, easy to hande and that can take into account severa decision variabes so as to simuate a diversity of possibe reforms. We chose to buid an exogenous, static microsimuation mode. The mode is static, as it aows us to evauate the reform by characterizing the distribution of the financia burden before and after its impementation in a given year without modifying the popuation structure. It is exogenous in that it assumes that individua behavior remains unchanged in the face of the new reform. Mora hazard is not controed. Nevertheess, simiar to Keeer Newhouse and Pheps (1977), we simpy assume that demand (...) are ikey to be insensitive to the size of the deductibe above certain range and that individua behavior remains unchanged Database Our work base was constructed using 2006 data from the Heath and Socia Protection Survey (ESPS) database (socio-economic data) and the Permanent Sampe of Sociay Insured (EPAS: data on medica consumption). The EPAS data were used at the nonaggregated eve (in other words, at the eve of each unit of consumption for each individua), so as to be abe to recacuate the reimbursement and OOP payment variabes on the hypothetica suppression of ALD-reated exonerations. For each insured, the new simuated data essentiay concern the amount reimbursed, copayments, charges exceeding statutory fees and OOP payments. After having combined EPAS and ESPS data, we isoated agents eigibe for CMU and those not having consumed heath care. Our fina database was made up of a sampe of 6,960 individuas. Our study is imited to ambuatory services and standard charges not exceeding statutory fees. The modes of reimbursement and patients financia contributions differ consideraby according to whether they concern office-based care or hospita care. Reimbursements for ess costy medica procedures administered via office-based care are effectivey imited. On the contrary, hospita care and, notaby, care reated to chronic iness and diseases requiring expensive treatments are reativey we-reimbursed in France, as in the other European countries. In addition, the hospita sector captures the highest expenditures but ony concerns 10 to 15% of the insured, whereas 85% of the sampe used ambuatory care services. Creating an OOP maximum rue on the basis of cumuated tota expenditures woud have the effect of raising the OOP threshod and woud carry the risk of penaizing insured agents that do not consume hospita services. On that basis, ambuatory care woud practicay no onger be reimbursed. This type of risk seection might chaenge the genera acceptabiity of the system as a whoe. In view of this, we thus concentrated on discipines within the ambuatory sector, such Document de travai n 34 - IRDES - Novembre

13 as medica procedures practiced by GPs, medica auxiiaries, bioogica procedures, prescribed drugs or heath-reated transport. In addition, the anaysis of patient OOP payments excudes charges exceeding statutory fees. In this way, we remain within the Socia Security framework that does not reimburse charges exceeding statutory fees. Taking these additiona fees into account in the threshod rue can equay have the effect of generating mora hazard: patients may be ess incined to contro their heath expenditures, and GPs may be tempted to increase their fees. 3. Evauating reforms: from who are the winners and osers? to the redistributive characteristics of the compusory heath care insurance scheme To anayze the possibe effects of this reform, our anaysis is deveoped in three phases. In the first phase, we present the overa resuts obtained for the different scenarios tested. In a second phase, we identify the winners and the osers, and finay, we observe the redistributive effects on the heath insurance system in terms of equity Initia observations We initiay concentrated on the distribution characteristics of current OOP payments. The current average OOP payment amounts to 223, the maximum to 3,607, and with equa standard deviation, 254 (Tabe 1), which suggests an excessivey broad dispersion of OOP payments. In addition, the average OOP payment is constant according to revenue decies (Figure 1). The financia participation reated to income, or the OOP payment-toincome ratio, thus decreases according to revenue (Figure 2). It is approximatey three times higher for individuas in the first decie in reation to individuas in the third decie. Current OOP payments are ow for the majority of the popuation and extremey high for a sma percentage (Figure 3). OOP payments are therefore concentrated among a sma number of individuas: 40% of OOP payments weigh on 10% of the popuation. The first descriptive resuts enabe us to obtain criteria permitting comparisons between the current situation and the potentia situation for a given OOP threshod (Tabes 1 and 2). The coefficients enabing us to cacuate OOP threshod vaues are the foowing: α UT =544, α IRT = 0.092, α IRTM = When the ALD criterion is taken into account, the coefficients for non-ald individuas are the foowing: α IRT = 804, α UT = 0.138, α IRTM = They are two times ower than for individuas with ALD. As expected, the OOP threshod coefficients wi be higher for non-ald individuas when the ALD criterion is taken into account. It produces a compensation effect from non-ald to ALD. These coefficients aow us to cacuate absoute and reative vaues for the different OOP maximum threshods (Tabe 1-bis). Consequenty, the absoute vaue for the uniform threshod (UT) does not depend on income, whereas its reative vaue decreases according to income eve. Weathier househods woud therefore have reativey ower OOP payments than individuas with more modest incomes. Contrary to the uniform threshod, absoute vaues for the IRT and IRTM modes increase with income eve. For threshods cacuated according to income eve, the absoute and reative vaues for the IRTM mode wi be ower for ow-income househods than the IRT mode. Inversey, for higher income househods, the reative and absoute vaues for the IRTM wi be higher than for the IRT mode. Document de travai n 34 - IRDES - Novembre

14 The average OOP payment for each OOP maximum mode being very cose to 223, the condition of equiibrium, or neutra effect on the insurance system, is vaidated. In effect, average osses and gains are equa to zero for each of the different safety net modes; the Compusory Heath Care Insurance neither gains nor oses. Concerning the other descriptive statistics, the OOP maximum is equa to the UT threshod, whereas the maxima can be higher for income-reated modes (IRT = 3,638 and IRTM = 4,504 ). The standard deviations aso change consideraby according to the safety net mode. Heterogeneity is reduced with the UT and IRT modes, whereas it tends to increase with the IRTM. Whatever the OOP maximum mode, the proportion of winners and osers is reativey ow, at between 20 and 25% 10. A greater number of insured wi be in a neutra position on the impementation of the Uniform Threshod; the OOP maximum is reativey high and the majority of individuas wi never reach the threshod. The percentage of individuas impacted increases if income is taken into account in the safety net threshod definition, with the percentage of winners becoming higher than the percentage of osers (UT: 9.6% winners and 10.9% osers; IRTM: 14.1% winners and 10.1% osers). If one concentrates on the popuation with ALD, the percentage of neutras is negigibe regardess of the reforms envisaged. Individuas in ALD are essentiay osers, but 15 to 27% are winners. OOP payments increase on average by 62 for the UT and by 232 for the IRTM, taking ALD into account. To better anayze the redistribution factor, we define I1 as the redistribution average and I2 as the standard deviation of redistributions as foows: I1 reveas that the redistribution average is ceary higher for the income-reated threshod modes. The same appies for I2: redistribution is higher with ALD. According to the I1 and I2 indicators, the redistribution of OOP payments is higher among ALD than non-ald individuas when safety net threshods are reated to income. OOP payment distribution according to income decie changes for the IRT and IRTM modes, whereas the curve remains cose to that of current OOP payments for the UT; in other words, it remains constant for each income decie (Figures 1-4). For the incomereated safety net modes, OOP payments increase according to income decie: OOP payments for poorer individuas wi be ower than the current eve and respectivey higher for weathier individuas (Figure 1). Therefore, the financia participation reated to income curve for income-reated safety net threshods tends to fatten and form a be curve, contrary to the financia participation reated to income curve for the UT, which remains unchanged in reation to the current financia participation reated to income. The differences in financia participation reated to income between income decies are ess for income-reated heath OOP maximum rues. 10 According to the hypotheses retained, the osers can ony be individuas on ALD. This does not, however, mean that a individuas on ALD are osers. Document de travai n 34 - IRDES - Novembre

15 3.2. Who oses? Who gains? How much? These initia resuts incite us to ook in more detai at the determinants of gaining or osing with the reforms impemented and the amounts transferred. In a first phase, we estimated the probabiity of being a winner or a oser using a ogistic regression mode (Tabe 2). For the three safety net threshod modes, estimates show that the effect of age, being femae and having suppementary heath coverage increases the probabiity of gaining, whereas being on ALD with a poor heath status has a negative effect on the probabiity of gaining. There is a non-inear effect of age on OOP payment eves. In addition, resuts ceary demonstrate the differences that distinguish the UT, IRT and IRTM modes concerning the income effect. This effect is neutra for the UT and significanty non-inear for the IRT and IRTM. Furthermore, the singe fact of being on ALD has a significanty positive effect on the probabiity of being on the osing side. As for the income effect, it becomes significanty positive for the IRT and IRTM: individuas wi have a higher probabiity of being osers if their income is high and the OOP maximum threshod is reated to income. In a second phase, we used a inear regression mode to estimate the amounts gained and ost (Tabe 3). For the UT, the gains are higher if the individua is young, femae and has suppementary heath insurance. For the IRT and IRTM, the income effect is significant and higher for the IRTM mode. The ALD status is significanty negative for both these modes: an individua on ALD wi gain ess than a non-ald. Heath status is significant ony for the IRTM; the effect of heath status is negative and increasing. Consequenty, an individua gains ess, the poorer his/her heath status. In terms of amounts ost, the estimated coefficient for the ALD status becomes positive and extremey high for a three safety net modes and the osses higher for individuas on ALD. For the IRT and IRTM, a positive income effect is added and shows that osses are a the greater, the higher an individua s income Anaysis of equity and redistributivity The anaysis in terms of equity is compementary to the initia descriptive anayses. One of the motivations behind these OOP safety net threshod anayses is to find a better redistributive equity and better risk coverage for the heath insurance system. To achieve this, we use three different methods: the Kakwani index, ALJ decomposition of the redistributive effect and second-order stochastic dominance. The Kakwani index The odest way of measuring equity uses the difference between the Gini indices before and after the introduction of a reform or tax to measure the redistribution effect (Musgrave and Thin, 1948). This effect is defined as a owering of the Gini coefficient. Kakwani (1977) demonstrated that this method expains the redistributive effect without measuring its progressivity. It fais to distinguish between the effect of a change in the average tax rate and its eve of progressivity concerning income distribution. The Kakwani index is the difference between the concentration curve for OOP payments (C OOP ) and the concentration curve for income (C Inc.). To measure the impact of a reform, we cacuate the difference between the Kakwani index before and after the reform as foows: Document de travai n 34 - IRDES - Novembre

16 The Kakwani index measures the proportionaity gap between a tax system and taxpayers abiity to pay. Wagstaff et a. (1999) used this index to quantify the progressivity or regressivity of a heath system. They demonstrate that the redistribution effect of the heath insurance system in the Netherands transfers income from the poor to the rich, whereas in Great Britain or the United States, it is transferred from the rich to the poor. The Kakwani index cacuated for the current situation is negative (Tabe 4). The reimbursement system is thus regressive or pro-rich in that the distribution of OOP burden in reation to income favors the weathiest (Wagstaff et a., 1999). In other words, the poorest individuas have a greater OOP burden proportiona to their income than the weathiest individuas. The Kakwani index cacuated after the reform changes according to the safety net threshod mode tested and varies from for the UT to for the IRTM 11. The regressivity of the system is accentuated with a uniform threshod (UT), but the effect diminishes when income inequaities are taken into account. The IRT and IRTM modes redistribute the OOP burden in a way that is more favorabe to ow-income earners, contrary to the uniform threshod, which has a regressive redistribution effect in favor of the rich. If the ALD criterion is taken into account, the system characteristics (progressive or regressive) are unatered. The graphic anayses of cumuative OOP payments and income concentration curves equay confirm our resuts (Figure 5). The curve for current OOP payments concentration superimposes on the bisecting ine. The situation is egaitarian without being equitabe, in that individuas have the same OOP burden regardess of their income eve. The OOP curve with a uniform threshod superimposes the current OOP payments curve. The OOP payments are, in effect, not dependent on income. On the contrary, with the IRT and IRTM modes, the concentration curves remain, for the most part, above the income curve but remain cose. The system remains regressive but becomes marginay more equitabe. Moreover, for 20% of the popuation with the owest incomes, the cumuated sum of OOP payments is proportionay ower than the cumuated sum of their incomes (Figure 5). The income-reated threshod rues favor the owest income popuation for whom the system becomes progressive. AJL Anaysis One of Kakwani s hypotheses, however, is that individuas with equa incomes are faced with the same tax. The reaity is, however, far more compex. Aronson, Johnson and Lambert (1994) demonstrate that the differences in concentration indices before and after a reform can be written in the foowing manner (the AJL decomposition): where V stands for vertica equity, H for horizonta equity, and R for the reranking effect. The vertica effect can aso be decomposed with g, the average tax rate, and K, the Kakwani progressivity index that measures the extent of income distortion before financing the tax. 11 The coser the coefficient is to 0, the more the system wi be redistributive. Document de travai n 34 - IRDES - Novembre

17 Concerning the anaysis of heath insurance systems, V measures the vertica effect of redistribution. It depends on progressivity but aso on the average rate g that corresponds to the percentage of income devoted to average heath expenditures. The higher the average rate, the greater the redistribution effect. The vertica effect shows how househods with different incomes are affected by the reimbursement method. The horizonta effect measures the inequaity generated among househods with the same income, whereas reranking quantifies the change in OOP payments distribution occasioned by the reform (Zhong, 2009). The hypothesis underying the income-reated OOP threshod modifies the distribution of the OOP burden such that it takes into consideration inequaities stemming from income distribution. Consequenty, bringing the distribution of the OOP burden coser to a vertica equity situation, this distribution must be the most unequa. It is thus necessary to obtain the greatest possibe vertica effect 12 (because V is negative). From the point of view of vertica equity in heath, the non-equa must be treated differenty. Individuas with different incomes shoud not benefit from the same reimbursement eves and thus shoud have a different OOP burden. The horizonta effect measures equity between groups of individuas with equa incomes 13. H is measured as the weighted sum of Gini indices of income diminished by OOP on sub-popuations with equa incomes. It is expressed as foows: where G j OOP reform represents the Gini index of income diminished by OOP payments according to the different threshod modes for a group of individuas β j with the equa incomes, and is the product of the percentage of the popuation in group k and OOP payments proportiona to income after the introduction of the associated threshod mode. By construction, the H component is defined as being non-negative. Horizonta equity can thus ony reduce the redistribution effect but not increase it. Individuas with simiar financia resources shoud have the same heath insurance benefits. The horizonta effect is an indicator that refects the way in which cose-equa individuas are treated. In the attempt to improve equity, H must have the highest possibe vaue. The R component captures the reranking effect occasioned by changes incurred when OOP payments distribution is taken before and after impementation of the reform. It is measured by: where G OOP reform represents the income concentration index diminished by OOP payments, and C OOP reform the concentration index for OOP payments after the introduction of an OOP threshod, cacuated by cassifying individuas by equa income sub-popuations and OOP payments eves within each sub-popuation group. The R component cannot be negative. It is reated to the Atkinson-Potnick reranking measure (Atkinson, 1980 and Potnick, 1981). If the desired objective is to redistribute the OOP 12 This reasoning is the opposite to that used in anayzing income tax effects on income, for which the significant vertica effect means that the income curve after tax is more egaitarian and therefore more equitabe in this particuar case. 13 One important question is: are horizonta inequities truy inequitabe? Indeed, if we think the differences in OOP payments due to differences in heath status vs differences in heath preferences or if the differences in OOP payments due to no heath care needs vs. postponement of heath care consumption or unmet needs, distributiona consequences and poicy impications wi not be the same. One soution wi be to introduce in our mode the eve of heath care use. Document de travai n 34 - IRDES - Novembre

18 burden to make it more equitabe, it is necessary to maximize this indicator. To cacuate R, it is thus necessary to define income intervas. The bandwidth determines the magnitude of the horizonta and reranking effects (van de Ven, Creedy and Lambert, 2001 and Biger, 2008). The wider the bandwidths, the esser the horizonta effect (due to the size of popuation sub-groups), and the greater the reranking effect (due to the higher number of popuation sub-groups). It is for this reason that we have cacuated this decomposition with four different bandwidths 14. The current redistribution effect (RE) is negative (Tabe 4). This means that the redistribution effect reated to the system of cacuating OOP payments according to income favors the higher income group (Van Doorsaer et a., 1999). If one ooks at the current distribution between V, H and R cacuated from income intervas represented by centies, the vertica effect represents 66% of the redistribution, the horizonta effect 12% and the reranking effect 27% (for bandwidths equa to 100). Concerning the distribution impact after introducing the OOP threshods, the redistribution effect (RE) for the UT mode is cose to the current situation. On the contrary, for the income-reated heath OOP threshod, IRT and IRTM, the redistribution vaues increase but remain negative. Moreover, the vertica equity (V), horizonta equity (H), and the reranking effect (R) change according to the OOP threshod mode being tested. In comparison with the current situation, V and R decrease with the UT. With the IRT and IRTM, V and R increase. This confirms that the situation becomes more regressive with the UT and ess regressive, and thus more equitabe, with the incomereated threshods. Regardess of the threshod mode, vaues reating to horizonta equity are reativey stabe, which refects ow iniquity within popuation casses with equa income. Individuas with simiar incomes pay the same amount of OOP payments. The anaysis of distribution between vertica equity (%V = V / RE), horizonta equity (%H = - H / RE) and the reranking effect (%R = - R / RE) in percentages of redistribution confirm these resuts 15. In effect, compared with the current situation, the percentage of V increases and the percentage of R decreases in the UT mode, whereas for the IRT and the IRTM, the percentage of V decreases and the percentage of R increases. There is an inversion of progressivity between the percentage of the vertica effect and the percentage of the reranking effect. An increase in the percentage of V is synonymous with a more regressive system, which means that OOP payments distribution wi be ess redistributive. As the percentage of H remains reativey stabe, if the percentage of V increases, then the percentage of R wi diminish and consequenty ower the reranking effect. The anaysis of the redistribution effects of the different threshod modes on the Kakwani index and ALJ decomposition reveas that the consequences in terms of redistribution wi not be identica, according to the type of threshod mode chosen. The UT wi tend towards greater inequaity and, as a resut, further dissociate OOP payments from income. Inversey, the IRT and IRTM modes give a more unequa redistribution but in favor of poorer individuas, which, in this context, woud make the heath insurance benefits system more equitabe. 14 Ten intervas; 696 individuas on average per interva, 50 intervas; 139 individuas on average per interva, 100 intervas; 70 individuas on average per interva and, 250 intervas; 28 individua on average per interva. 15 Evoutions in terms of eve and percentage of V wi be different because we are in a regressive distribution system. Document de travai n 34 - IRDES - Novembre

19 Second order stochastic dominance The fina criterion to better characterize the system does not measure equity but rather the notion of individuas risk aversion in reation to insurance. Second-order stochastic dominance is used to measure the preferences of the risk-phobic insured when faced with a change in the reimbursement system (Geoffard and De Lagasnerie, 2009). Using the vei of ignorance hypothesis, whereby individuas ignore their heath status, secondorder stochastic dominance asserts that individuas with risk aversion wi prefer one form of OOP payments distribution to another if OOP payments distributions have the same average and if the Lorenz curves associated with the two types of distribution ony bisect once (Rothschid and Stigitz, 1970). Therefore, if a reform reduces risks (in the second-order stochastic dominance sense), a of the agents with risk aversion shoud prefer it. The Lorenz curves obtained from the UT stochasticay dominate the current distribution pattern at second order (Figure 6 and Tabe 4). We confirm this graphic anaysis by using the Komogorov-Smirnov test. This is not the case for income-reated threshods. Individuas with risk aversion wi then prefer the uniform threshod soution because the maximum OOP burden remains reativey ow and thus protects them from catastrophic risks. This confirms the descriptive anayses showing that the standard deviation for the owest OOP payments was obtained with the UT mode (Tabe 1). Concusion The aim of this work is to measure the redistributive effect of an integra reform of the reimbursement rues for heath care costs. The idea is to test the repacement of the current 100% reimbursement system for patients with ong-term inesses with an OOP maximum threshod to imit catastrophic out-of-pocket payments. Different safety net modes, either based on a uniform threshod or an income-reated threshod, are proposed. The resuts are obtained using an exogenous, static microsimuation mode. It is thus assumed that individua behavior does not change. This hypothesis may seem farfetched, but it enabes a first round observation of OOP payments transfers without making other hypotheses concerning behavior. It woud be possibe to measure the sensitivity of our resuts by taking into account modifications in behavior. With the ARAMMIS static microsimuation mode, IRDES armed itsef with a powerfu, permanent anaysis too that enabes a better understanding of the effects of extremey specific reforms. The simuations presented are not aimed at finding the right reform but rather at describing reform impacts and, in so doing, providing poicy-makers with an objective viewpoint. The mode is in the construction phase in that we aim to compete it with additiona modues concerning hospitas, suppementary insurance and contributions. Our resuts ony concern heath insurance reimbursements and do not take individuas statutory income-reated contributions into account. The notion of equity discussed here is aso ony partia. In a future deveopment of this study, we wi aso mode socia security contributions so as to study the equity of the system as a whoe. The aim in this paper being to study the evoution of reimbursement rues, the ony variabes required were those on the expenses side, that is, the modes of covering expenditures, as the resources side is invariant. In addition, our resuts do not take into account OOP payment coverage for CMU beneficiaries, which, by its very nature, favors the poorest Document de travai n 34 - IRDES - Novembre

20 members of the popuation. It thus strongy modifies the redistributive characteristics of the heath expenditure reimbursement system. Initia resuts indicate that a the envisaged scenarios ead to an increase in OOP from 62 to 232 per year for patients on ALD. A of the estimations were carried out using identica budgetary constraints. Naturay, there is a change in OOP distribution, but not a patients on ALD woud be osers. According to the different safety net modes retained, from 15% to 27% of ALD beneficiaries woud be winners, namey, patients currenty paying high OOP payments. In effect, the heterogeneity of OOP is greater among ALD beneficiaries than non-ald beneficiaries. These first resuts indicate that the characteristics of winners and osers are highy dependent on the OOP threshod modue. An income-reated threshod, for exampe, woud favor ower income groups, whereas, with a uniform OOP threshod, income has no importance in the determination of winners and osers. The effects of these two main reform concepts, uniform or income-reated OOP-safety net threshods, currenty being broached in the pubic debate are not identica and ead to contrary concusions. In reducing the risk of being faced with a very high OOP burden, uniform threshods eve out the heterogeneity of situations and appear to suit individuas with high risk aversion. Inversey, income-reated threshods increase the heterogeneity of OOP burdens but have a ess regressive redistributive effect - moving from an egaitarian system to a more equitabe system. Impementing reforms such as these woud inevitaby raise a number of questions: an OOP threshod for individuas or househods? An OOP payment threshod concerning ambuatory care ony or the totaity of expenditures (ambuatory + hospita care)? Woud denta and optica expenditures be incuded? What roe shoud be payed by suppementary heath insurance? A of these questions need to be studied in detai because, as we have demonstrated, the devi is in the detais! The different heath systems throughout the word partiay refect the way individuas perceive the notion of socia justice. The poitica choice concerning the way OOP payments are cacuated wi, in the same way, refect what French society considers to be sociay acceptabe and fair in terms of heath insurance (Raws, 1971). In that context, French society wi have to resove the diemma between equaity and equity: at what point does the search for absoute equaity become inequitabe? Document de travai n 34 - IRDES - Novembre