Eur J Health Econ DOI 10.1007/s10198-014-0572-x
ORIGINAL PAPER
On the calculation of the Israeli risk adjustment rates Amir Shmueli
Received: 17 July 2013 / Accepted: 5 February 2014 Ó Springer-Verlag Berlin Heidelberg 2014
Abstract Objective The Israeli risk adjustment formula, introduced in 1995 and which serves for the allocation of the health budget to the sickness funds, is unique compared to countries with a similar national health insurance system in that it is not calculated on the basis of actual cost data of the sickness funds but on the basis of quantities retrieved from surveys. The current article aims to analyze the implications of the Israeli methodology. Methods The article examines the validity of the Israeli methodology used to set the 2004 risk adjustment rates and compare these rates with the ‘‘correct’’ ones, which are derived from the 2004 internal relative cost scales of the sickness funds. Results The Israeli methodology ignores services provided by the sickness funds and assumes constant unit cost across the sickness funds, an assumption which is implausible. Comparing the actual and the ‘‘correct’’ rates, it turns out that the actual rates over-compensate all the sickness funds for members in age 0–14, and under-compensate them for insurees aged 55?. In age 0–4, the over-compensation per capita is about NIS 1,500 while the undercompensation in age group 75? reaches NIS 1,600. Conclusions The current risk adjustment formula distorts the intended competition on good quality care among the sickness funds, and turns it into a competition on profitable members. After 18 years of using incorrect rates, the Israeli risk adjustment rates should be calculated, as is common in other systems, based on individual cost data from the sickness funds.
A. Shmueli (&) Department of Health Management and Economics, The Hebrew University-Hadassah School of Public Health, POB 12272, 91120 Jerusalem, Israel e-mail: [email protected]
Keywords Risk adjustment National health insurance Risk selection Israel JEL Classification
I11 I13 I18
Introduction In all health systems where the collection of the citizens’ contributions (premiums, social security fees or taxes) is separate from the delivery of health services, a special mechanism is introduced by which the ‘‘national health budget’’ is allocated among the decentralized units responsible for the provision of care. In competitive schemes, these units are ‘‘health plans’’ or ‘‘sickness funds’’, while in non-competitive schemes these are the ‘‘provinces’’, ‘‘local health areas’’ or ‘‘district health authorities’’. All mechanisms, however, use risk-adjusted allocations, to assure an equitable medical care distribution and an ‘‘appropriate’’ competition in competitive systems. The calculation of the risk adjustment rates differs from system to system, shaped to a large extent by data availability. The presently common statistically driven (‘‘correct’’) derivation of the risk adjustment scheme typically uses a (huge) random sample of individual administrative claims data. Examples include the American Medicare risk adjustment scheme which is based on the diagnoses cost groups, which defines groups and relative risk using medical diagnoses and cost information, or the Dutch scheme which is based on pharmacy cost groups, a similarly heavy data mining product based on the use of pharmaceuticals in individual claims data covering the Dutch population. The Israeli way to set the risk adjustment rates has been totally different. It is based on national group-specific
123
A. Shmueli
relative use of the various services (visits to physicians, hospitalization days, etc.). These service-specific relative uses are weighted according to the shares of the expenditure on the service in total expenditure. The purpose of this paper is to critically evaluate the Israeli practice with regard to the ‘‘correct’’ rates derived from the sickness funds’ internal relative cost scales. It is important to underline that, by ‘‘correct’’ and ‘‘incorrect’’, we refer to the practice in most countries, namely, using statistically-derived cost weights. Statistically-derived relative cost scales represent actual (or recent) relative medical expenditures and not ‘‘need’’ [1] nor ‘‘optimal risk adjustment’’ rates which are derived from a maximization of economic welfare [2]. While need-derived and optimal risk adjustment rates provide a conceptual-theoretical basis for the rates, they provide no guidance on how to measure and calculate them. The statistically-derived rates aim to predict accurately the cost in various groups in the population and to alleviate the problem of risk selection.
The Israeli national health insurance system The national health insurance scheme which was introduced in January 1995 consists of a managed competition model [3], where four private non-profit sickness funds compete on the quality of medical care covered by a uniform package of benefits defined by the law. The package of benefits is comprehensive and includes primary, secondary, and inpatient care, as well as diagnostic and pharmaceutical care. The budget of the package of benefits is determined annually by the government, partially being indexed to changes in input prices, demography, and technological advances. It is financed by an earmarked health tax and transfers from the general revenues. The scheme is compulsory (all are insured) and universal (no rejection). Citizens are free to switch sickness funds yearly, however, the switching rate is low (1–2 % annually). There is no direct premium paid by the members to the sickness funds. The main source of income for the sickness funds are the risk-equalized payments from the budget of the package of services. Risk equalization consists of two separate components: a prospective age adjustment specifying fixed rates for each of 9 (11 since 2005) age groups (governing the allocation of 94 % of the budget), and a retrospective risk sharing arrangement—governing 6 % of the budget— by which the sickness funds receive an annual fixed payment per person who is sick with one of five ‘‘severe conditions’’—Renal failure on Dialysis, Thalasemia major, Gaucher, AIDS, and Hemophilia. Until 2010, the prospective risk adjustment was based on age only. In 2010, two additional risk-adjusters were
123
introduced—gender and peripheral status. In this paper, we focus on the year 2004, when the risk adjustment mechanism was based on age only, defined by nine age groups. The issue of the (poor) set of risk adjusters is discussed elsewhere [4] and is beyond the scope of this paper.
Setting the Israeli risk adjustment rates The calculation of the risk adjustment rates has been based on using quantities of care rather than expenditure data. The main reason (in 1995) was the limited availability and quality of individual (cost) data at the sickness funds. Later on, several arguments for the continued use of the original methodology were raised (and discussed below). The medical expenditure within the budget of services was decomposed in 1995 into three ‘‘expenditure heads’’: visits to physicians, visits to outpatient clinics, and general inpatient days. For each head, the relative quantities (uses) are calculated for each risk adjustment cell (age group), and a scale presenting the mean use in a specific cell relative to the grand mean was constructed. Data on agespecific mean number of visits to doctors was obtained from the 1993 central bureau of statistics (CBS)’s use of health services survey. Data on outpatient visits was obtained from the government hospitals for the year 1993, and age-specific mean number of inpatient days was obtained from the CBS’s 1987 hospitalizations data file. The three scales were then combined with weights reflecting the share of the expenditure head in the national health expenditure. The 1994 weights were: visits to doctors—45 %, outpatient visits—13 %, and inpatient days— 42 %. In 2005, a fourth expenditure head was added— visits to emergency departments, and the data on the other expenditure heads were updated using the 1999 CBS’s use of health services survey, and the 2002 Ministry of Health (MOH)’s hospitalizations file. The weights (out of the total sickness funds’ expenditures) were: visits to doctors— 50 %, outpatient visits—8 %, inpatient days—40 %, and visits to ER—2 %. In 2010, a third update was performed, still keeping the 1995 methodology. A fifth expenditure head was introduced—the number of prescription drugs. The weights were: visits to doctors—40 %, prescription drugs—10 %, outpatient visits—9 %, ER visits—3 %, and inpatient days—38 %. Use of hospital services was obtained from the 2007 MOH’s hospitalizations file. The number of prescription drugs by age was obtained from the sickness funds. To see the logic behind the decomposition used in Israel, denote the mean medical cost in age group g by c(g), and the overall mean cost by c. Suppose there are two health services (expenditure heads) in the package of benefits, denoted by 1 and 2. Denote by q1(g) and q2(g) the mean
On the calculation of the Israeli risk adjustment rates
number of units (quantities) used of the two services by members of age group g, and by q1 and q2 and overall mean use. Denote the services’ unit cost by p1 and p2 respectively. Using these notations, the age-specific mean cost is c(g) = q1(g) p1 ? q2(g) p2 and the overall mean cost is c = q1p1 ? q 2p2. The risk adjustment rates are defined as c(g)/c. Inserting the above definitions, we have:
The third problem is theoretical and is a source of a methodological bias when the 1995 methodology is used. There are several reasons why the unit costs are expected to differ across sickness funds, at least for the three main expenditure heads: 1.
cðgÞ=c ¼ ½q1 ðgÞ p1 + q2 ðgÞ p2 =½q1 p1 + q2 p2 ¼ ½q1 ðgÞ=q1 w + ½q2 ðgÞ=q2 ½1 w Namely, the risk adjustment rates are in fact a weighted average of the quantity-scales [(q1(g)/q1) and (q2(g)/q2)] with weights being the relative cost of each service in total cost, w = q1p1/(q1p1 ? q2p2), 1 - w = q2p2/(q1p1 ? q2p2). Evaluating the Israeli way to set the risk adjustment rates No doubt, the 1995 methodology of setting the Israeli rates enabled the regulator to enact of the national health insurance law without available data on the sickness funds’ expenditures which are needed to calculate the risk adjustment rates. This methodology suffers, however, from several weaknesses; each by itself rendering the calculated rates incorrect: 1.
2.
3.
2.
The total cost used (q1p1 ? q2p2) is the total cost of supplying the services, namely the total expenditures of the sickness funds and not the national health expenditure, as was used in 1995 (this was corrected in 2005). There are services (expenditure heads) which are omitted from the calculation, such as medicines (until 2010), imaging, preventive services, and physical therapy. This omission causes differential biases in different age groups, depending on the relative use of these services. The unit costs (p1 and p2) are assumed constant across the sickness funds. If the unit costs vary across the sickness funds, a similar decomposition is still possible, but it is much more complicated, involving a weighted mean of the quantity-scales in the different sickness funds, and requiring (aggregate) cost data from the sickness funds (we will elaborate further on that point in the last section).
The second problem originates from the limited information on the omitted services available in population surveys and hospitalization data. The complete information is available only in the sickness funds’ data.
3.
Visits to physicals: There is a large mix in the ways doctors are paid by the sickness funds (for details see [5]). In Clalit, the largest sickness fund in Israel, doctors are mainly salaried. The physicians contracted by the other three sickness funds serve mainly as independent doctors. The sickness funds differ not only in the level of reimbursement for a given payment method (e.g., capitation) but also in the payment methods. Furthermore, the doctors visited represent a mix of specialties differing in their level of payment. If the mix is not identical across the sickness funds, the ‘‘average’’ payment will not be similar. Finally, the productivity of doctors varies across sickness funds and payment methods. As a result, it is hardly likely that the unit cost of a visit to a doctor will be similar across sickness funds. Inpatient days: While the MOH’s prices of an inpatient day and of the ‘‘differential activities’’ (a prospective price system for selected procedures) serve as benchmark in the system, the sickness funds do not bear a similar unit cost of inpatient days. First, the Clalit owns general hospitals while the other three sickness funds purchase inpatient care from the government and public hospitals. Several studies showed [6, 7] that the cost structure and the length of stay of the Clalit’s hospitals differ from those of the other general hospitals, resulting in different unit cost per inpatient day. Second, many of the inpatient days and procedures are paid within global bilateral contracts between hospitals and sickness funds. These contracts are likely to differ among the sickness funds, leading to different unit cost per inpatient day. Third, since 1994, the sickness funds’ payments to the hospitals are subjected to the ‘‘capping arrangement’’. This arrangement specifies the average price per inpatient day based on the volume of days transacted between each sickness fund and each hospital. A sickness fund which directs the hospitalizations of its members to one hospital will achieve a lower average unit cost per day than a sickness fund which contracts several hospitals. Furthermore, the Clalit owns general hospitals and the effect of the capping arrangement on its unit cost is different than that in the remaining sickness funds. Prescription drugs (introduced in 2010): The unit costs of prescription drugs are typically determined in large global agreements between each of the sickness funds and the pharmaceutical firms. Volume is clearly very
123
A. Shmueli Table 1 The internal relative risk scales of the sickness funds and the national risk adjustment scale (2004) Sickness fund
A
B
C
D
The national risk adjustment scale
0–5
0.72
0.72
0.87
0.66
1.18
5–15
0.31
0.43
0.43
0.25
0.45
15–25 25–35
0.43 0.60
0.54 0.70
0.45 0.49
0.36 0.53
0.39 0.54
35–45
0.72
0.75
0.69
0.69
0.69
45–55
1.11
1.07
1.14
1.13
1.14
55–65
1.85
1.68
1.96
2.12
1.75
65–75
3.05
2.56
3.16
3.05
2.72
75?
4.03
3.35
4.31
3.95
3.42
Overall mean cost = 1
important in these agreements, and probably the Clalit gets better terms than Maccabi (the second largest sick fund) which gets better terms than the two smaller sickness funds. Consequently, it is hardly likely that all sickness funds bear equal unit costs of prescription drugs. The conclusions from the above discussion are that the methodology used for the derivation of the Israeli risk adjustment rates is probably flawed and results in incorrect rates. In the next section, the internal relative risk rates [c(g)/c] of the sickness funds circa 2004 are used to calculate the ‘‘correct’’ risk adjustment rates which are then compared to the actual 2004 rates.
The sickness funds’ internal relative risk rates The sickness funds’ internal rates are private-commercial information and are not made public. However, the Ministries of Health and Finance’s team which was appointed to review the risk adjustment mechanism in 2005 invited the sickness funds to submit material which could assist the team’s discussions. Clalit and Maccabi submitted their own internal relative (age-based) risk rates to the team. The two smaller sickness funds did not. Following a long negotiation, the two smaller sickness funds agreed to supply their internal rates for this study, provided the sickness funds are not identified. The data for this study consists, therefore, of four internal relative risk scales from the four sickness funds. The year of calculation (2002–2004) differs across the sickness funds, and each scale was adjusted to the corresponding sick fund’s population size in 2004 (so that the population size equals the number of age-adjusted persons). Table 1 presents the unidentified four scales along with the national risk adjustment rates for the year 2004. Recall that until 2010, age was the sole risk adjuster and
123
until 2005, age was recorded in nine age groups. All five scales are normalized so that 1 = population mean cost. The scales clearly differ across the four sickness funds. At age 0–4, for example, sickness fund A and B have a relative cost value of 0.72, in sickness fund C, it is 0.87, and in sickness fund D, it is 0.66. Among elderly persons (age 75?), the range is 3.35 (B) to 4.31 (C). These differences represent ‘‘real’’ differences in the production of health, care management, input prices etc., but also differences in the way overhead costas are allocated and possibly additional ‘‘technical’’ differences. Comparing the national risk adjustment scale to the sickness funds’ scales, two observations emerge: the rate for age group 0–4 on the national scale (1.18) is higher than all the sickness funds’ rates (0.66–0.87), and the rates for elderly persons aged 65–74 and 75? (2.72 and 3.42, respectively) are lower than all the sickness funds’ rates (2.79–3.82 and 3.35–4.31). In the next section, we will show that the main distortion introduced by the actual risk adjustment rates affects these age groups.
The derivation of the ‘‘correct’’ risk adjustment scale from the sickness funds’ internal relative risks scales As was mentioned earlier, the common way, used in most countries, to derive the risk adjustment scale is to have a big representative sample of the population from all the sickness funds, and to calculate the mean cost—as retrieved from the sickness funds’ individual administrative records—in each group defined by the set of risk adjusters. The rates are obtained by the ratios of the groupspecific mean cost and the overall mean cost. When age is the only risk adjuster recoded into nine age groups, we need to calculate the mean cost in each age group [c(g)] and the overall mean [c]. The risk adjustment rates are obtained as c(g)/c. Suppose now that there are two sickness funds, denoted by 1 and 2. Denote the mean cost in age group g in sickness fund i by ci(g) and the group size—by ni(g). The mean cost in sickness fund i is ci, and the overall mean cost in the population is c. Denote the size of age group g in the population by n(g) [=n1(g) ? n2(g)]. Since c(g) = c1(g)n1(g)/n(g) ?c2(g)n2(g)/n(g), c(g)/c = [c1(g)/ c1][c1/c][n1(g)/n(g)] ? [c2(g)/c2][c2/c][n2(g)/n(g)] namely, c(g)/c is a weighted average of the internal relative risk rates (ci(g)/ci is the internal relative risk in sickness fund (1). The average costs in the sickness funds and in the entire population (c1, c2 and c) are routinely reported by the MOH, and the age group sizes in the sickness funds and in the population as a whole [n1(g), n2(g) and n(g)] are routinely reported by the National insurance institute. Figure 1 presents the ‘‘correct’’ risk adjustment rates which are based on the sickness funds’ cost data and the
On the calculation of the Israeli risk adjustment rates
Fig. 1 The Israeli actual and ‘‘correct’’ (calculated) risk adjustment scales (1 = overall mean)
Fig. 2 Age-specific profit (?)/loss (-) born by the sickness funds per Israeli citizen when the national risk adjustment is used (2004 NIS)
Table 2 The improvement in the 0–4 and 75? national risk adjustment rates over time
scale is getting closer to the ‘‘correct’’ scale derived from the sickness funds’ cost data.
Age group
1995
2005
2010
0–4
1.18
1.08
1.02
75?
3.42
3.67
3.73
actual risk adjustment rates calculated using the methodology described above. Between age 15 and 54, the scales are almost similar. They differ markedly for ages 0–14 and 55?. For ages 0–14, the ‘‘correct’’ scale is lower than the actual one, while for ages above 55, the actual scale is lower than the ‘‘correct’’ one. As was mentioned above, three aspects of the methodology used to derive the actual risk adjustment scale account for these differences—incorrect weights of the expenditure heads, variable unit costs across sickness funds, and omission of services. It is impossible to identify the contributions of the various sources of error in these differences. However, the incorrectly low rate for the elderly is clearly caused by the omission of services (expenditure heads) which are not available in the methodology used to calculate the age-specific mean costs and are used relatively intensively by the elderly. These include prescription drugs (added in 2010), diagnostic and laboratory tests, and physical therapy. The omission of these expenditure heads lowers the mean cost of care among the elderly more than among other age groups and in total. Consequently, the rate for the elderly is underestimated and the rates for the other age groups are overestimated, in particular for the infants who use these omitted services rarely. A partial confirmation of the biases in the actual scale argued above is obtained from Table 2. The table shows how the rate for the 0–4 age group has dropped—and the rate for the 75? age group has increased—with the revisions of the scale in 2005 and 2010. These revisions obviously aimed at increasing the accuracy of scale derived by the methodology launched in 1995, and the resulting
Implications and conclusions The budget per capita of the package of benefits provided by the sickness funds, which constitutes their mean revenue, was NIS 3,102 in 2004. This is the revenue per one age-adjusted citizen, and it is multiplied by the risk adjustment scale to set the per capita age-specific revenue of the sickness fund (e.g., the 2004 revenue per capita in age group 35–44 was 3,102 9 0.69 = NIS 2,140). If we calculate the age-specific difference between the actual risk adjustment and the ‘‘correct’’ relative cost scales and multiply the difference by NIS 3,102 we obtain the per capita age-specific profit or loss born by the sickness funds in Israel. These profits/losses are presented in Fig. 2. According to the figure, the Israeli sickness funds make significant profits on children aged 0–14, but have significant losses on persons aged over 55. On infants aged 0–4, the sickness funds make a profit of about NIS 1,500 on average. The mean profit on children aged 5–14 is more modest, NIS 300. At the other end of the age scale, on persons aged 55–64, there is a mean loss of NIS 350, and this loss increases to NIS 700 in the 65–74 age group, and to NIS 1,600 in the upper age group (75?). Under such circumstances, the sickness funds have a clear incentive to aim for large groups of children and small groups of elderly persons. In fact, according to the rates given in Table 1 and the sickness funds’ mean costs for 2004 [8], all the sickness funds make a profit on children and lose money on the elderly. Since the sickness funds cannot reject applicants by law, they are incentivized to exercise implicit (risk) selection. One of the most obvious tools for such selection is marketing campaigns. And, indeed, in recent years all the sickness funds have launched aggressive marketing campaigns aimed to attract infants and children (Fig. 3). The advertisements are in
123
A. Shmueli Fig. 3 The four sickness funds’ selective advertisement
Hebrew but the message is clear, in particular among the relatively big Orthodox Jewish and Arab families. Naturally, no parallel marketing campaign has ever been launched trying to attract elderly persons, since, on average, they constitute predictable loss. The (new) risk adjustment scale in Germany underpays the elderly and also overpays the young [9]. However, the German rates are calculated from the sickness funds’ expenditure data, and the reasons for these gaps are different. The underpayment for the elderly is the result of not annualizing the costs of the deceased persons, resulting in an underestimation of the annual cost. The overpayment for the young who are in good health (chronic conditions are rare)—and probably the underpayment for the elderly—is due to some misspecification of the relationship between cost and morbidity (the scale underpays persons with multi-morbidity). The Israeli gaps originate from an incorrect calculation and omission of services’ expenditures. The biggest danger of an inappropriate risk adjustment scheme is that it distorts the (managed) competition among the insurers—which is supposed to be a competition on quality of care and services experience—and provides incentives for distorting the quality and quantity of health services [10]. This is clearly a regulation failure, leading to social inefficiency and inequity. It must be admitted that the Israeli risk adjustment scheme suffers from an even more serious shortcoming, namely, it lacks any healthbased risk adjusters, while many other countries (e.g.,
123
Belgium, The Netherlands, Germany, and Switzerland) now have (or are going to have) such an adjuster. An illustration of the implications to competition of this shortcoming (and of the inefficiency and inequity which might result from inappropriate risk adjustment scale) in presented in [11]. It is argued that, since the Israeli risk adjustment scheme does not include health state, sick insurees present a predictable loss while healthy ones present a predictable profit. Consequently, in ‘‘healthy (and usually rich) towns’’, the competition between the sickness funds is intensive and the physicians’ office hours per inhabitant are high. In ‘‘sick towns’’, competition is low and the physicians’ office hours per inhabitant are low. The Israeli methodology to calculate the risk adjustment rates in 1995 was the only one possible, considering the unavailability of individual cost data in the sickness funds. Medical information technology has advanced tremendously ever since, and the sickness funds currently have detailed and (almost) complete individualized data on costs. The ‘‘correct’’ risk adjustment rates are derived using the actual cost data of the sickness funds. This is the methodology which is used in all the countries having a national health insurance scheme such as the Israeli one. The 2010 team for the review of the risk adjustment rates concluded against basing the scale on the sickness funds’ cost data. Two main arguments were raised: first, that the cost data are not ‘‘uniform’’ across the sickness funds due to accounting and calculation issues; and second, that the sickness funds might manipulate the data. The first
On the calculation of the Israeli risk adjustment rates
argument is technical and can be accommodated by creating a detailed accounting protocol on how to prepare and submit the cost data. The MOH already publishes a yearly report on the (comparative) financial status of the sickness funds (The Supervision of the Sickness Funds and Complementary Health Services Department, The Public Report on the Operation of the Sickness Funds, various years, in Hebrew), without any calibration. However, this argument touches upon an important issue in risk adjustment—what are the acceptable costs for which the sickness funds are held responsible and are reimbursed for and for which factors the payments are not adjusted [10]. While this issue has been discussed in the literature, in practice it is very difficult to quantify the variation in the unacceptable factors, and hence to decompose the costs into acceptable and unacceptable costs. Differences in the production of health, in care management, and in geographical dispersion are examples of factors for which the payments are not adjusted. At first glance, it might seem that using common unit prices across the Israeli sickness funds (p1 and p2) does that; namely, that these are the acceptable unit costs. However, this is not the case. First, the weights of the services’ relative quantities in the decomposition of the ‘‘correct’’ rates, w = q1p1/(q1p1 ? q2p2), 1 - w = q2p2/ (q1p1 ? q2p2), have been calculated using total—rather than acceptable—costs; namely, including the non-adjusting factors. Second, the unacceptable costs cannot be ignored. They enter the risk adjustment formula as the national means rather than the particular levels of the sickness fund’s risk adjusters. The second argument is also questionable—the entire regulation and tax system is built on the firms’ costs, revenues, and profits data reports. There is no apparent reason why the sickness funds would manipulate cost data more than they do in reporting financial results for tax purposes. Naturally, verification and monitoring system should function to minimize these threats. After 18 years of using the 1995 methodology for calculating the Israeli risk adjustment rates, it is about time for a change. Since the allocation of the health budget is a
zero-sum game for the sickness funds, some sickness funds would support such a change, while others, expecting a drop in their share of the budget, will resist it. The MOH should exercise its regulatory power to set ‘‘correct’’ risk adjustment rates for the benefit of the sickness funds’ market, i.e. the welfare of the population. Acknowledgments I am grateful to Sari Dotan, Francis Wood, Shuli Brammli-Greenberg and the participants in the 13th risk Adjustment Network (RAN) meeting in Tel Aviv, June 2013, for their valuable comments.
References 1. Rice, N., Smith, P.: Approaches to Capitation and Risk Adjustment in Health Care: An International Survey. University of York, The Centre for Health Economics (1999) 2. Glazer, J., McGuire, T.: Optimal risk adjustment in markets with adverse selection: an application to managed care. Am. Econ. Rev. 90, 1055–1071 (2000) 3. Enthoven, A.C.: Consumer choice health plan: a national health insurance proposal based on regulated competition in the private sector. NEJM 298, 709–720 (1978) 4. Shmueli, A., Chernichovsky, D., Zmora, I.: Risk adjustment and risk sharing: the Israeli experience. Health Policy 65, 37–49 (2003) 5. Rosen B. et al.: Israel health system review, the European observatory on health systems, Health Sys. Trans. (2009) 6. Chernichovsky, D., Zmora, I.: A hedonic prices approach to hospitalization costs: the case of Israel. J. Health Econ. 5, 179–191 (1986) 7. Shmueli A. and Dor A.: The structure of general hospitals’ costs in Israel: implications from panel data analysis, Discussion Paper, The Gertner Insitute, (1992) 8. MOH: The national health insurance law: Statistical Data 1995–2011, prepared by Arieli D., Horev T., Kaidar N., Jerusalem, (2012) 9. Buchner, F., Goepffarth, D., Wasem, J.: The new risk adjustment formula in Germany: implementation and first experiences. Health Policy 109, 253–262 (2013) 10. Van de Ven, W., Ellis, R.: Risk adjustment in competitive health plans markets. In: Culyer, A.J., Newhouse, J.P. (eds.) Handbook of Health Economics. Elsevier, Amsterdam (2000) 11. Shmueli, A., Nissan-Engelcin, E.: Local availability of physicians’ services as a tool for implicit risk selection. Soc. Sci. Med. 84, 53–60 (2013)
123
ORIGINAL PAPER
On the calculation of the Israeli risk adjustment rates Amir Shmueli
Received: 17 July 2013 / Accepted: 5 February 2014 Ó Springer-Verlag Berlin Heidelberg 2014
Abstract Objective The Israeli risk adjustment formula, introduced in 1995 and which serves for the allocation of the health budget to the sickness funds, is unique compared to countries with a similar national health insurance system in that it is not calculated on the basis of actual cost data of the sickness funds but on the basis of quantities retrieved from surveys. The current article aims to analyze the implications of the Israeli methodology. Methods The article examines the validity of the Israeli methodology used to set the 2004 risk adjustment rates and compare these rates with the ‘‘correct’’ ones, which are derived from the 2004 internal relative cost scales of the sickness funds. Results The Israeli methodology ignores services provided by the sickness funds and assumes constant unit cost across the sickness funds, an assumption which is implausible. Comparing the actual and the ‘‘correct’’ rates, it turns out that the actual rates over-compensate all the sickness funds for members in age 0–14, and under-compensate them for insurees aged 55?. In age 0–4, the over-compensation per capita is about NIS 1,500 while the undercompensation in age group 75? reaches NIS 1,600. Conclusions The current risk adjustment formula distorts the intended competition on good quality care among the sickness funds, and turns it into a competition on profitable members. After 18 years of using incorrect rates, the Israeli risk adjustment rates should be calculated, as is common in other systems, based on individual cost data from the sickness funds.
A. Shmueli (&) Department of Health Management and Economics, The Hebrew University-Hadassah School of Public Health, POB 12272, 91120 Jerusalem, Israel e-mail: [email protected]
Keywords Risk adjustment National health insurance Risk selection Israel JEL Classification
I11 I13 I18
Introduction In all health systems where the collection of the citizens’ contributions (premiums, social security fees or taxes) is separate from the delivery of health services, a special mechanism is introduced by which the ‘‘national health budget’’ is allocated among the decentralized units responsible for the provision of care. In competitive schemes, these units are ‘‘health plans’’ or ‘‘sickness funds’’, while in non-competitive schemes these are the ‘‘provinces’’, ‘‘local health areas’’ or ‘‘district health authorities’’. All mechanisms, however, use risk-adjusted allocations, to assure an equitable medical care distribution and an ‘‘appropriate’’ competition in competitive systems. The calculation of the risk adjustment rates differs from system to system, shaped to a large extent by data availability. The presently common statistically driven (‘‘correct’’) derivation of the risk adjustment scheme typically uses a (huge) random sample of individual administrative claims data. Examples include the American Medicare risk adjustment scheme which is based on the diagnoses cost groups, which defines groups and relative risk using medical diagnoses and cost information, or the Dutch scheme which is based on pharmacy cost groups, a similarly heavy data mining product based on the use of pharmaceuticals in individual claims data covering the Dutch population. The Israeli way to set the risk adjustment rates has been totally different. It is based on national group-specific
123
A. Shmueli
relative use of the various services (visits to physicians, hospitalization days, etc.). These service-specific relative uses are weighted according to the shares of the expenditure on the service in total expenditure. The purpose of this paper is to critically evaluate the Israeli practice with regard to the ‘‘correct’’ rates derived from the sickness funds’ internal relative cost scales. It is important to underline that, by ‘‘correct’’ and ‘‘incorrect’’, we refer to the practice in most countries, namely, using statistically-derived cost weights. Statistically-derived relative cost scales represent actual (or recent) relative medical expenditures and not ‘‘need’’ [1] nor ‘‘optimal risk adjustment’’ rates which are derived from a maximization of economic welfare [2]. While need-derived and optimal risk adjustment rates provide a conceptual-theoretical basis for the rates, they provide no guidance on how to measure and calculate them. The statistically-derived rates aim to predict accurately the cost in various groups in the population and to alleviate the problem of risk selection.
The Israeli national health insurance system The national health insurance scheme which was introduced in January 1995 consists of a managed competition model [3], where four private non-profit sickness funds compete on the quality of medical care covered by a uniform package of benefits defined by the law. The package of benefits is comprehensive and includes primary, secondary, and inpatient care, as well as diagnostic and pharmaceutical care. The budget of the package of benefits is determined annually by the government, partially being indexed to changes in input prices, demography, and technological advances. It is financed by an earmarked health tax and transfers from the general revenues. The scheme is compulsory (all are insured) and universal (no rejection). Citizens are free to switch sickness funds yearly, however, the switching rate is low (1–2 % annually). There is no direct premium paid by the members to the sickness funds. The main source of income for the sickness funds are the risk-equalized payments from the budget of the package of services. Risk equalization consists of two separate components: a prospective age adjustment specifying fixed rates for each of 9 (11 since 2005) age groups (governing the allocation of 94 % of the budget), and a retrospective risk sharing arrangement—governing 6 % of the budget— by which the sickness funds receive an annual fixed payment per person who is sick with one of five ‘‘severe conditions’’—Renal failure on Dialysis, Thalasemia major, Gaucher, AIDS, and Hemophilia. Until 2010, the prospective risk adjustment was based on age only. In 2010, two additional risk-adjusters were
123
introduced—gender and peripheral status. In this paper, we focus on the year 2004, when the risk adjustment mechanism was based on age only, defined by nine age groups. The issue of the (poor) set of risk adjusters is discussed elsewhere [4] and is beyond the scope of this paper.
Setting the Israeli risk adjustment rates The calculation of the risk adjustment rates has been based on using quantities of care rather than expenditure data. The main reason (in 1995) was the limited availability and quality of individual (cost) data at the sickness funds. Later on, several arguments for the continued use of the original methodology were raised (and discussed below). The medical expenditure within the budget of services was decomposed in 1995 into three ‘‘expenditure heads’’: visits to physicians, visits to outpatient clinics, and general inpatient days. For each head, the relative quantities (uses) are calculated for each risk adjustment cell (age group), and a scale presenting the mean use in a specific cell relative to the grand mean was constructed. Data on agespecific mean number of visits to doctors was obtained from the 1993 central bureau of statistics (CBS)’s use of health services survey. Data on outpatient visits was obtained from the government hospitals for the year 1993, and age-specific mean number of inpatient days was obtained from the CBS’s 1987 hospitalizations data file. The three scales were then combined with weights reflecting the share of the expenditure head in the national health expenditure. The 1994 weights were: visits to doctors—45 %, outpatient visits—13 %, and inpatient days— 42 %. In 2005, a fourth expenditure head was added— visits to emergency departments, and the data on the other expenditure heads were updated using the 1999 CBS’s use of health services survey, and the 2002 Ministry of Health (MOH)’s hospitalizations file. The weights (out of the total sickness funds’ expenditures) were: visits to doctors— 50 %, outpatient visits—8 %, inpatient days—40 %, and visits to ER—2 %. In 2010, a third update was performed, still keeping the 1995 methodology. A fifth expenditure head was introduced—the number of prescription drugs. The weights were: visits to doctors—40 %, prescription drugs—10 %, outpatient visits—9 %, ER visits—3 %, and inpatient days—38 %. Use of hospital services was obtained from the 2007 MOH’s hospitalizations file. The number of prescription drugs by age was obtained from the sickness funds. To see the logic behind the decomposition used in Israel, denote the mean medical cost in age group g by c(g), and the overall mean cost by c. Suppose there are two health services (expenditure heads) in the package of benefits, denoted by 1 and 2. Denote by q1(g) and q2(g) the mean
On the calculation of the Israeli risk adjustment rates
number of units (quantities) used of the two services by members of age group g, and by q1 and q2 and overall mean use. Denote the services’ unit cost by p1 and p2 respectively. Using these notations, the age-specific mean cost is c(g) = q1(g) p1 ? q2(g) p2 and the overall mean cost is c = q1p1 ? q 2p2. The risk adjustment rates are defined as c(g)/c. Inserting the above definitions, we have:
The third problem is theoretical and is a source of a methodological bias when the 1995 methodology is used. There are several reasons why the unit costs are expected to differ across sickness funds, at least for the three main expenditure heads: 1.
cðgÞ=c ¼ ½q1 ðgÞ p1 + q2 ðgÞ p2 =½q1 p1 + q2 p2 ¼ ½q1 ðgÞ=q1 w + ½q2 ðgÞ=q2 ½1 w Namely, the risk adjustment rates are in fact a weighted average of the quantity-scales [(q1(g)/q1) and (q2(g)/q2)] with weights being the relative cost of each service in total cost, w = q1p1/(q1p1 ? q2p2), 1 - w = q2p2/(q1p1 ? q2p2). Evaluating the Israeli way to set the risk adjustment rates No doubt, the 1995 methodology of setting the Israeli rates enabled the regulator to enact of the national health insurance law without available data on the sickness funds’ expenditures which are needed to calculate the risk adjustment rates. This methodology suffers, however, from several weaknesses; each by itself rendering the calculated rates incorrect: 1.
2.
3.
2.
The total cost used (q1p1 ? q2p2) is the total cost of supplying the services, namely the total expenditures of the sickness funds and not the national health expenditure, as was used in 1995 (this was corrected in 2005). There are services (expenditure heads) which are omitted from the calculation, such as medicines (until 2010), imaging, preventive services, and physical therapy. This omission causes differential biases in different age groups, depending on the relative use of these services. The unit costs (p1 and p2) are assumed constant across the sickness funds. If the unit costs vary across the sickness funds, a similar decomposition is still possible, but it is much more complicated, involving a weighted mean of the quantity-scales in the different sickness funds, and requiring (aggregate) cost data from the sickness funds (we will elaborate further on that point in the last section).
The second problem originates from the limited information on the omitted services available in population surveys and hospitalization data. The complete information is available only in the sickness funds’ data.
3.
Visits to physicals: There is a large mix in the ways doctors are paid by the sickness funds (for details see [5]). In Clalit, the largest sickness fund in Israel, doctors are mainly salaried. The physicians contracted by the other three sickness funds serve mainly as independent doctors. The sickness funds differ not only in the level of reimbursement for a given payment method (e.g., capitation) but also in the payment methods. Furthermore, the doctors visited represent a mix of specialties differing in their level of payment. If the mix is not identical across the sickness funds, the ‘‘average’’ payment will not be similar. Finally, the productivity of doctors varies across sickness funds and payment methods. As a result, it is hardly likely that the unit cost of a visit to a doctor will be similar across sickness funds. Inpatient days: While the MOH’s prices of an inpatient day and of the ‘‘differential activities’’ (a prospective price system for selected procedures) serve as benchmark in the system, the sickness funds do not bear a similar unit cost of inpatient days. First, the Clalit owns general hospitals while the other three sickness funds purchase inpatient care from the government and public hospitals. Several studies showed [6, 7] that the cost structure and the length of stay of the Clalit’s hospitals differ from those of the other general hospitals, resulting in different unit cost per inpatient day. Second, many of the inpatient days and procedures are paid within global bilateral contracts between hospitals and sickness funds. These contracts are likely to differ among the sickness funds, leading to different unit cost per inpatient day. Third, since 1994, the sickness funds’ payments to the hospitals are subjected to the ‘‘capping arrangement’’. This arrangement specifies the average price per inpatient day based on the volume of days transacted between each sickness fund and each hospital. A sickness fund which directs the hospitalizations of its members to one hospital will achieve a lower average unit cost per day than a sickness fund which contracts several hospitals. Furthermore, the Clalit owns general hospitals and the effect of the capping arrangement on its unit cost is different than that in the remaining sickness funds. Prescription drugs (introduced in 2010): The unit costs of prescription drugs are typically determined in large global agreements between each of the sickness funds and the pharmaceutical firms. Volume is clearly very
123
A. Shmueli Table 1 The internal relative risk scales of the sickness funds and the national risk adjustment scale (2004) Sickness fund
A
B
C
D
The national risk adjustment scale
0–5
0.72
0.72
0.87
0.66
1.18
5–15
0.31
0.43
0.43
0.25
0.45
15–25 25–35
0.43 0.60
0.54 0.70
0.45 0.49
0.36 0.53
0.39 0.54
35–45
0.72
0.75
0.69
0.69
0.69
45–55
1.11
1.07
1.14
1.13
1.14
55–65
1.85
1.68
1.96
2.12
1.75
65–75
3.05
2.56
3.16
3.05
2.72
75?
4.03
3.35
4.31
3.95
3.42
Overall mean cost = 1
important in these agreements, and probably the Clalit gets better terms than Maccabi (the second largest sick fund) which gets better terms than the two smaller sickness funds. Consequently, it is hardly likely that all sickness funds bear equal unit costs of prescription drugs. The conclusions from the above discussion are that the methodology used for the derivation of the Israeli risk adjustment rates is probably flawed and results in incorrect rates. In the next section, the internal relative risk rates [c(g)/c] of the sickness funds circa 2004 are used to calculate the ‘‘correct’’ risk adjustment rates which are then compared to the actual 2004 rates.
The sickness funds’ internal relative risk rates The sickness funds’ internal rates are private-commercial information and are not made public. However, the Ministries of Health and Finance’s team which was appointed to review the risk adjustment mechanism in 2005 invited the sickness funds to submit material which could assist the team’s discussions. Clalit and Maccabi submitted their own internal relative (age-based) risk rates to the team. The two smaller sickness funds did not. Following a long negotiation, the two smaller sickness funds agreed to supply their internal rates for this study, provided the sickness funds are not identified. The data for this study consists, therefore, of four internal relative risk scales from the four sickness funds. The year of calculation (2002–2004) differs across the sickness funds, and each scale was adjusted to the corresponding sick fund’s population size in 2004 (so that the population size equals the number of age-adjusted persons). Table 1 presents the unidentified four scales along with the national risk adjustment rates for the year 2004. Recall that until 2010, age was the sole risk adjuster and
123
until 2005, age was recorded in nine age groups. All five scales are normalized so that 1 = population mean cost. The scales clearly differ across the four sickness funds. At age 0–4, for example, sickness fund A and B have a relative cost value of 0.72, in sickness fund C, it is 0.87, and in sickness fund D, it is 0.66. Among elderly persons (age 75?), the range is 3.35 (B) to 4.31 (C). These differences represent ‘‘real’’ differences in the production of health, care management, input prices etc., but also differences in the way overhead costas are allocated and possibly additional ‘‘technical’’ differences. Comparing the national risk adjustment scale to the sickness funds’ scales, two observations emerge: the rate for age group 0–4 on the national scale (1.18) is higher than all the sickness funds’ rates (0.66–0.87), and the rates for elderly persons aged 65–74 and 75? (2.72 and 3.42, respectively) are lower than all the sickness funds’ rates (2.79–3.82 and 3.35–4.31). In the next section, we will show that the main distortion introduced by the actual risk adjustment rates affects these age groups.
The derivation of the ‘‘correct’’ risk adjustment scale from the sickness funds’ internal relative risks scales As was mentioned earlier, the common way, used in most countries, to derive the risk adjustment scale is to have a big representative sample of the population from all the sickness funds, and to calculate the mean cost—as retrieved from the sickness funds’ individual administrative records—in each group defined by the set of risk adjusters. The rates are obtained by the ratios of the groupspecific mean cost and the overall mean cost. When age is the only risk adjuster recoded into nine age groups, we need to calculate the mean cost in each age group [c(g)] and the overall mean [c]. The risk adjustment rates are obtained as c(g)/c. Suppose now that there are two sickness funds, denoted by 1 and 2. Denote the mean cost in age group g in sickness fund i by ci(g) and the group size—by ni(g). The mean cost in sickness fund i is ci, and the overall mean cost in the population is c. Denote the size of age group g in the population by n(g) [=n1(g) ? n2(g)]. Since c(g) = c1(g)n1(g)/n(g) ?c2(g)n2(g)/n(g), c(g)/c = [c1(g)/ c1][c1/c][n1(g)/n(g)] ? [c2(g)/c2][c2/c][n2(g)/n(g)] namely, c(g)/c is a weighted average of the internal relative risk rates (ci(g)/ci is the internal relative risk in sickness fund (1). The average costs in the sickness funds and in the entire population (c1, c2 and c) are routinely reported by the MOH, and the age group sizes in the sickness funds and in the population as a whole [n1(g), n2(g) and n(g)] are routinely reported by the National insurance institute. Figure 1 presents the ‘‘correct’’ risk adjustment rates which are based on the sickness funds’ cost data and the
On the calculation of the Israeli risk adjustment rates
Fig. 1 The Israeli actual and ‘‘correct’’ (calculated) risk adjustment scales (1 = overall mean)
Fig. 2 Age-specific profit (?)/loss (-) born by the sickness funds per Israeli citizen when the national risk adjustment is used (2004 NIS)
Table 2 The improvement in the 0–4 and 75? national risk adjustment rates over time
scale is getting closer to the ‘‘correct’’ scale derived from the sickness funds’ cost data.
Age group
1995
2005
2010
0–4
1.18
1.08
1.02
75?
3.42
3.67
3.73
actual risk adjustment rates calculated using the methodology described above. Between age 15 and 54, the scales are almost similar. They differ markedly for ages 0–14 and 55?. For ages 0–14, the ‘‘correct’’ scale is lower than the actual one, while for ages above 55, the actual scale is lower than the ‘‘correct’’ one. As was mentioned above, three aspects of the methodology used to derive the actual risk adjustment scale account for these differences—incorrect weights of the expenditure heads, variable unit costs across sickness funds, and omission of services. It is impossible to identify the contributions of the various sources of error in these differences. However, the incorrectly low rate for the elderly is clearly caused by the omission of services (expenditure heads) which are not available in the methodology used to calculate the age-specific mean costs and are used relatively intensively by the elderly. These include prescription drugs (added in 2010), diagnostic and laboratory tests, and physical therapy. The omission of these expenditure heads lowers the mean cost of care among the elderly more than among other age groups and in total. Consequently, the rate for the elderly is underestimated and the rates for the other age groups are overestimated, in particular for the infants who use these omitted services rarely. A partial confirmation of the biases in the actual scale argued above is obtained from Table 2. The table shows how the rate for the 0–4 age group has dropped—and the rate for the 75? age group has increased—with the revisions of the scale in 2005 and 2010. These revisions obviously aimed at increasing the accuracy of scale derived by the methodology launched in 1995, and the resulting
Implications and conclusions The budget per capita of the package of benefits provided by the sickness funds, which constitutes their mean revenue, was NIS 3,102 in 2004. This is the revenue per one age-adjusted citizen, and it is multiplied by the risk adjustment scale to set the per capita age-specific revenue of the sickness fund (e.g., the 2004 revenue per capita in age group 35–44 was 3,102 9 0.69 = NIS 2,140). If we calculate the age-specific difference between the actual risk adjustment and the ‘‘correct’’ relative cost scales and multiply the difference by NIS 3,102 we obtain the per capita age-specific profit or loss born by the sickness funds in Israel. These profits/losses are presented in Fig. 2. According to the figure, the Israeli sickness funds make significant profits on children aged 0–14, but have significant losses on persons aged over 55. On infants aged 0–4, the sickness funds make a profit of about NIS 1,500 on average. The mean profit on children aged 5–14 is more modest, NIS 300. At the other end of the age scale, on persons aged 55–64, there is a mean loss of NIS 350, and this loss increases to NIS 700 in the 65–74 age group, and to NIS 1,600 in the upper age group (75?). Under such circumstances, the sickness funds have a clear incentive to aim for large groups of children and small groups of elderly persons. In fact, according to the rates given in Table 1 and the sickness funds’ mean costs for 2004 [8], all the sickness funds make a profit on children and lose money on the elderly. Since the sickness funds cannot reject applicants by law, they are incentivized to exercise implicit (risk) selection. One of the most obvious tools for such selection is marketing campaigns. And, indeed, in recent years all the sickness funds have launched aggressive marketing campaigns aimed to attract infants and children (Fig. 3). The advertisements are in
123
A. Shmueli Fig. 3 The four sickness funds’ selective advertisement
Hebrew but the message is clear, in particular among the relatively big Orthodox Jewish and Arab families. Naturally, no parallel marketing campaign has ever been launched trying to attract elderly persons, since, on average, they constitute predictable loss. The (new) risk adjustment scale in Germany underpays the elderly and also overpays the young [9]. However, the German rates are calculated from the sickness funds’ expenditure data, and the reasons for these gaps are different. The underpayment for the elderly is the result of not annualizing the costs of the deceased persons, resulting in an underestimation of the annual cost. The overpayment for the young who are in good health (chronic conditions are rare)—and probably the underpayment for the elderly—is due to some misspecification of the relationship between cost and morbidity (the scale underpays persons with multi-morbidity). The Israeli gaps originate from an incorrect calculation and omission of services’ expenditures. The biggest danger of an inappropriate risk adjustment scheme is that it distorts the (managed) competition among the insurers—which is supposed to be a competition on quality of care and services experience—and provides incentives for distorting the quality and quantity of health services [10]. This is clearly a regulation failure, leading to social inefficiency and inequity. It must be admitted that the Israeli risk adjustment scheme suffers from an even more serious shortcoming, namely, it lacks any healthbased risk adjusters, while many other countries (e.g.,
123
Belgium, The Netherlands, Germany, and Switzerland) now have (or are going to have) such an adjuster. An illustration of the implications to competition of this shortcoming (and of the inefficiency and inequity which might result from inappropriate risk adjustment scale) in presented in [11]. It is argued that, since the Israeli risk adjustment scheme does not include health state, sick insurees present a predictable loss while healthy ones present a predictable profit. Consequently, in ‘‘healthy (and usually rich) towns’’, the competition between the sickness funds is intensive and the physicians’ office hours per inhabitant are high. In ‘‘sick towns’’, competition is low and the physicians’ office hours per inhabitant are low. The Israeli methodology to calculate the risk adjustment rates in 1995 was the only one possible, considering the unavailability of individual cost data in the sickness funds. Medical information technology has advanced tremendously ever since, and the sickness funds currently have detailed and (almost) complete individualized data on costs. The ‘‘correct’’ risk adjustment rates are derived using the actual cost data of the sickness funds. This is the methodology which is used in all the countries having a national health insurance scheme such as the Israeli one. The 2010 team for the review of the risk adjustment rates concluded against basing the scale on the sickness funds’ cost data. Two main arguments were raised: first, that the cost data are not ‘‘uniform’’ across the sickness funds due to accounting and calculation issues; and second, that the sickness funds might manipulate the data. The first
On the calculation of the Israeli risk adjustment rates
argument is technical and can be accommodated by creating a detailed accounting protocol on how to prepare and submit the cost data. The MOH already publishes a yearly report on the (comparative) financial status of the sickness funds (The Supervision of the Sickness Funds and Complementary Health Services Department, The Public Report on the Operation of the Sickness Funds, various years, in Hebrew), without any calibration. However, this argument touches upon an important issue in risk adjustment—what are the acceptable costs for which the sickness funds are held responsible and are reimbursed for and for which factors the payments are not adjusted [10]. While this issue has been discussed in the literature, in practice it is very difficult to quantify the variation in the unacceptable factors, and hence to decompose the costs into acceptable and unacceptable costs. Differences in the production of health, in care management, and in geographical dispersion are examples of factors for which the payments are not adjusted. At first glance, it might seem that using common unit prices across the Israeli sickness funds (p1 and p2) does that; namely, that these are the acceptable unit costs. However, this is not the case. First, the weights of the services’ relative quantities in the decomposition of the ‘‘correct’’ rates, w = q1p1/(q1p1 ? q2p2), 1 - w = q2p2/ (q1p1 ? q2p2), have been calculated using total—rather than acceptable—costs; namely, including the non-adjusting factors. Second, the unacceptable costs cannot be ignored. They enter the risk adjustment formula as the national means rather than the particular levels of the sickness fund’s risk adjusters. The second argument is also questionable—the entire regulation and tax system is built on the firms’ costs, revenues, and profits data reports. There is no apparent reason why the sickness funds would manipulate cost data more than they do in reporting financial results for tax purposes. Naturally, verification and monitoring system should function to minimize these threats. After 18 years of using the 1995 methodology for calculating the Israeli risk adjustment rates, it is about time for a change. Since the allocation of the health budget is a
zero-sum game for the sickness funds, some sickness funds would support such a change, while others, expecting a drop in their share of the budget, will resist it. The MOH should exercise its regulatory power to set ‘‘correct’’ risk adjustment rates for the benefit of the sickness funds’ market, i.e. the welfare of the population. Acknowledgments I am grateful to Sari Dotan, Francis Wood, Shuli Brammli-Greenberg and the participants in the 13th risk Adjustment Network (RAN) meeting in Tel Aviv, June 2013, for their valuable comments.
References 1. Rice, N., Smith, P.: Approaches to Capitation and Risk Adjustment in Health Care: An International Survey. University of York, The Centre for Health Economics (1999) 2. Glazer, J., McGuire, T.: Optimal risk adjustment in markets with adverse selection: an application to managed care. Am. Econ. Rev. 90, 1055–1071 (2000) 3. Enthoven, A.C.: Consumer choice health plan: a national health insurance proposal based on regulated competition in the private sector. NEJM 298, 709–720 (1978) 4. Shmueli, A., Chernichovsky, D., Zmora, I.: Risk adjustment and risk sharing: the Israeli experience. Health Policy 65, 37–49 (2003) 5. Rosen B. et al.: Israel health system review, the European observatory on health systems, Health Sys. Trans. (2009) 6. Chernichovsky, D., Zmora, I.: A hedonic prices approach to hospitalization costs: the case of Israel. J. Health Econ. 5, 179–191 (1986) 7. Shmueli A. and Dor A.: The structure of general hospitals’ costs in Israel: implications from panel data analysis, Discussion Paper, The Gertner Insitute, (1992) 8. MOH: The national health insurance law: Statistical Data 1995–2011, prepared by Arieli D., Horev T., Kaidar N., Jerusalem, (2012) 9. Buchner, F., Goepffarth, D., Wasem, J.: The new risk adjustment formula in Germany: implementation and first experiences. Health Policy 109, 253–262 (2013) 10. Van de Ven, W., Ellis, R.: Risk adjustment in competitive health plans markets. In: Culyer, A.J., Newhouse, J.P. (eds.) Handbook of Health Economics. Elsevier, Amsterdam (2000) 11. Shmueli, A., Nissan-Engelcin, E.: Local availability of physicians’ services as a tool for implicit risk selection. Soc. Sci. Med. 84, 53–60 (2013)
123
