In 1988 the dutch government began to implement radical marketoriented

Transcription

1 Health-Adjusted Premium Subsidies In The As it implements managed competition, the Dutch government has taken steps to avoid risk selection in its sickness funds. by Wynand P.M.M. van de Ven, René C.J.A. van Vliet, and Leida M. Lamers ABSTRACT: The Dutch government has decided to proceed with managed competition in health care. In this paper we report on progress made with health-based risk adjustment, a key issue in managed competition. In 24 both Diagnostic Cost Groups (DCGs) computed from hospital diagnoses only and Pharmacy-based Cost Groups (PCGs) computed from outpatient prescription drugs are used to set the premium subsidies for competing risk-bearing sickness funds. These health-based risk adjusters appear to be effective and complementary. Risk selection is not a major problem in the. Despite the progress made, we are still faced with a full research agenda for risk adjustment in the coming years. In 1988 the dutch government began to implement radical marketoriented reforms in health care. Central government planning was to be replaced by a system of managed (or regulated) competition. 1 Competing health insurers were to act as prudent buyers of care on behalf of their members. However, the transformation of a centrally planned health care system into managed competition appeared to be politically, technically, and institutionally complex. 2 Workable competition cannot be introduced overnight. It requires prolonged investments in developing an adequate system of risk adjustment, product classification and quality management, an appropriate consumer information system, and, last but not least, an effective competition policy. None of these preconditions was in place in 1988, which explains the substantial time gap between the adoption of the market-oriented reform plans and their actual implementation. In the past fifteen years, however, successive governments consistently worked to realize these preconditions. Much progress has been made for example, with respect to competition policy and risk adjustment. In this paper, as a follow-up to our earlier paper, we report on progress toward health-based risk adjustment in the in the past decade. First, we discuss some key elements of the Dutch financing system and the rele- Wynand van de Ven (vandeven@bmg.eur.nl) is a professor of health insurance at Erasmus University Rotterdam, The. René van Vliet is an associate professor of health insurance, and Leida Lamers, an assistant professor of medical technology assessment there. HEALTH AFFAIRS ~ Volume 2, Number 45 DOI 1.177/hlthaff Project HOPE The People-to-People Health Foundation, Inc.

2 Costs & Competition vance of good risk adjustment. We next report on five different risk-adjustment models successively applied in the. Finally, we discuss the health policy conclusions and some new risk-adjustment issues for the next years. Relevance Of Health-Based Risk Adjustment In 1992 a radical reform took place in the Dutch mandatory sickness fund insurance, which covers some 1.5 million people with incomes below a certain threshold (66 percent of the Dutch population). During all sickness funds received a retrospective, full reimbursement for all of their actual health care expenses out of the Central Fund, to which all working Dutch citizens contribute a mandatory income-related amount. Starting in 1992, sickness funds receive a prospective risk-adjusted premium subsidy per enrollee out of the Central Fund and bear the financial risk of deficits and surpluses. The subsidy equals the (national) predicted per capita expenses in the risk group to which the enrollee belongs, minus a fixed amount. The per capita subsidy is independent of the sickness fund the consumer chooses. On average these subsidies equal 78 percent (2) of the total expenses. In addition, consumers pay a premium contribution directly to their sickness fund. Because the risk-adjustment model was imperfect, the government required premium contributions to be the same for all enrollees in the same sickness fund. This prevents sickness funds from charging their chronically ill enrollees a high premium contribution. In addition, there is an annual open enrollment period that is, sickness funds are not allowed to refuse new applicants. Each sickness fund is free to set its own premiums, which results in premium competition among the sickness funds. Adverse effects of selection. Although the intended effect of the restrictions on premium contributions is to achieve solidarity (or fairness), they also create predictable losses for sickness funds on high-risk patients and thereby create incentives for risk selection. Three severe adverse effects of such incentives can be distinguished. First, sickness funds have a disincentive to respond to the preferences of high-risk consumers. Consequently, the chronically ill might receive poor quality of care or poor service. Although the findings in the literature do not warrant final conclusions, sickness funds could encounter strong financial incentives to be unresponsive to the preferences of the chronically ill, and this would be cause for concern. 4 Second, to the extent that some sickness funds attract low-risk consumers, these selection activities result in market segmentation, wherein high-risk patients are left in plans with high premiums and low-risk patients self-select into plans with low premiums. That is, selection may threaten solidarity. Third, at least in the short run, selection will be more profitable than improving efficiency. In sum, the restrictions on premium contributions that are intended to increase solidarity instead provide incentives for selection that may threaten solidarity, quality of care, and efficiency. Prevention of selection. The most effective strategy to avoid selection is to adjust the premium subsidy to an enrollee s health status: health-based risk adjust- 46 May/June 24

3 ment. A second strategy is risk sharing between the Central Fund and the sickness funds. 5 Risk sharing implies that the Central Fund retrospectively reimburses the sickness funds for some of their members actual costs. However, risk sharing also reduces the funds incentive for efficiency and consequently causes a trade-off between risk selection and efficiency. Below we discuss how the Dutch government used both strategies to prevent selection. Toward Health-Based Risk Adjustment Since 1992 the Dutch government has used several sets of risk adjusters to calculate the premium subsidies. Here we report the results of our empirical analysis, in which we simulated five adjustment models (Exhibit 1). The empirical analyses are based on a database of 2.8 million enrollees of five sickness funds. 6 The health expenses include the costs of inpatient room and board (except for fixed costs), inpatient and outpatient specialist care, dental care, physiotherapy, ancillary services, and prescribed drugs. Exhibit 1 presents for each set of risk adjusters the average predicted losses in year t for the 1 percent of patients with the highest expenses in year t 1. The first column of Exhibit 1 indicates that if there were no risk adjustment, the funds would only receive 875 (that is, the overall expected per capita expenses in year EXHIBIT 1 Predicted Accuracy Of Different Risk-Adjustment Models, The Model No risk adjusters Demographic I Demographic II PCG DCG PCG + DCG R2 x 1% for each model (percent explained variance) Outpatient expenses Inpatient expenses a Total expenses a Predicted losses in year t of 1 percent of people with highest expenses in year t 1 Average predicted expenses (Euro) Average predicted losses (Euro) Average predicted losses as percent of average actual expenses.%.. 6.9% % % % % ,558 75% 1,5 2,8 61% 1,44 1,99 58% 1,967 1,466 4% 2,46 1,87 4% 2,418 1,15 % SOURCE: L.M. Lamers et al., Risk-Adjusted Capitation Payment System for Health Insurance Plans in a Competitive Market, Expert Review Pharmacoeconomics Outcomes Research, no. 5 (2): NOTES: Average expenses in year t are,4 Euro. The five risk-adjustment models are as follows. Demographic I: Age and sex (used in ). Demographic II: Age, sex, urbanization, and eligibility status (for example, employed or not employed, disabled) and age x eligibility status (used in ). PCG: Demographic II plus Pharmacy-based Cost Groups (PCGs) (used in DCG: Demographic II plus Diagnostic Cost Groups (DCGs). PCG + DCG: Demographic II plus PCGs and DCGs (preliminary version of the model to be used starting in 24). a Excludes fixed hospital costs. HEALTH AFFAIRS ~ Volume 2, Number 47

4 Costs & Competition t) for each person belonging to the top 1 percent in year t 1 and would have an expected loss of 2,558 for each of them. Demographic models. In 1992 the risk-adjusted premium subsidies were based only on age and sex. These risk adjusters resulted in sizable incentives for risk selection. For example, for the 1 percent of patients with the highest expenses in the previous year, the average predictable losses in the current year are 61 percent of the average actual expenses (Exhibit 1). In other words, for these patients the sickness fund s predictable losses equal 15 percent of its revenues. Pharmacy-based Cost Groups (PCGs). A weakness of the demographic model is that predicted expenses are not adjusted for the large differences in individual health status within each age/sex group. A way to improve the predictions is to extend the set of risk adjusters with health status proxies based on prior use for example, using diagnostic information of prior hospitalization or information derived from prescription drug usage. The Dutch government decided to implement PCGs as new risk adjusters in 22. The basic idea is that prescribed medication can be an indication of existing chronic illness. In the, pharmacy claims contain a code from the Anatomical Therapeutic Chemical Classification Index (ATC code). 7 Based on these codes, we identified people with pharmacy claims that are indicative of chronic conditions. As a starting point, we used the list of chronic conditions of the revised Chronic Disease Score (CDS) as developed by Daniel Clark and colleagues. 8 Although pharmacy data provide good predictions of future spending, they could cause problems with respect to incentives and fairness. First, the additional subsidy for a PCG-classified enrollee may be two or three times the costs of the prescribed drugs that form the basis for PCG assignment. 9 These perverse incentives should not be allowed to frustrate sickness funds attempts to motivate physicians to prescribe efficiently. Second, because of the large amount of variation in prescribing practices, it would not be fair for sickness funds that selectively contract with doctors with cost-effective prescribing habits to be penalized by lower premium subsidies. Third, fairness and incentive problems would exist if a higher premium subsidy would be given for medications that are prescribed for both a major chronic disease and a minor temporary health problem. To overcome these problems, the Dutch government has been selective in rewarding prior medication use with higher premium subsidies (Exhibit 2). First, more than half of those with an outpatient prescription (77 percent of the population) are excluded because their medication could not to a sufficient extent be uniquely related to a specific medical condition according to the Chronic Disease Score. 1 Second, a person with only one or two prescriptions per year may be an incidental user rather than having a chronic condition. Therefore, to increase the probability of identifying only chronically ill people, the PCG classification is based on at least four prescriptions per year or at least 181 prescribed daily doses, both of which yielded similar results. This, again, reduced the number of PCG- 48 May/June 24

5 EXHIBIT 2 Percentage Of People With An Outpatient Prescription Per Year Versus Percentage Of PCG-Classified People, The Percent of population At least one outpatient prescription per year Has condition according to the Chronic Disease Score, based on at least one prescription per year Has condition according to the Chronic Disease Score, based on at least four prescriptions per year Classified in one of the 22 PCGs according to the Chronic Disease Score, based on at least 181 prescribed daily doses per year Classified in one of the 1 selected PCGs as used in the in SOURCE: L.M. Lamers and R.C.J.A. van Vliet, Health-Based Risk Adjustment: Improving the Pharmacy-Based Cost Group Model to Reduce Planning Possibilities, European Journal of Health Economics 4, no. 2 (2): NOTES: PCG is Pharmacy-based Cost Group. Chronic Disease Score based on the work of Daniel Clark and colleagues; see Note 8 in text. classified people by more than 5 percent. Finally, some PCGs with a low additional premium subsidy were excluded because of a lack of clear consensus concerning the use of the medications involved and because of large practice variations, which may reduce their validity as indicators of chronic illness (for example, hypertension, depression, and hyperlipidemia). 11 All of these steps resulted in the thirteen PCGs that were added as risk adjusters in 22 (Exhibit ). To remove the incentive for prescribing additional drugs for people who are already PCG-classified, people can be assigned to only one PCG (their most costly condition). When looking at these thirteen conditions, we have the impression that the concerns about incentives and fairness have largely been addressed. As a result of all of the above steps, only 11 percent of people with a prescription are classified in a PCG. This is an important aspect of the Dutch PCG classification. For example, in the pharmacy-based risk adjustment model for public programs as developed by Todd Gilmer and colleagues, 75 percent of people in the U.S. Temporary Assistance for Needy Families (TANF) program received a higher subsidy because of their prior prescription drug usage. 12 Adding PCGs to the demographic model greatly increased the predictive accuracy. The R2 nearly doubled, in particular as a result of a better prediction of outpatient expenses. The predictable losses for the 1 percent of people with the highest expenses in the previous year were reduced by more than 25 percent (Exhibit 1). In particular, the PCG model reduces the disincentive for sickness funds to invest in high-quality care for people with chronic conditions. In our study we did not explicitly pay attention to manipulation by upcoding, such as if a well-controlled Type II diabetic were moved from oral blood glucose lowering drugs to insulin. This emphasizes the need for a constant, intensive monitoring and improving of the PCG model and for regular updates. Updates are also necessary because of new drugs entering the market and because of HEALTH AFFAIRS ~ Volume 2, Number 49

6 Costs & Competition EXHIBIT Additional Annual Premium Subsidy For Individuals Classified In PCGs And DCGs, The Risk group (PCG) Risk group (DCG) Reference group Asthma/COPD Epilepsy Rheumatism Heart disease Crohn s disease/c. ulcerosa Stomach diseases Diabetes (insulin dependent) Parkinson s Organ transplants Cancer Cystic fibrosis HIV/AIDS Severe kidney problems (including ESRD) Reference group Ongoing pregnancy with complications Serious perinatal problem affecting newborn Ectopic pregnancy Central nervous system disorders Vertebral fracture without spinal cord injury Cardio-respiratory failure and shock Brain injury Myocardial infarction Colon cancer Gastrointestinal obstruction/perforation Liver disorders Rectal cancer Congestive heart failure Hypertension, complicated Neurologic disorders Brain/nervous system cancers Chronic obstructive pulmonary disease Mouth/pharynx/larynx/other respiratory cancer Chemotherapy Diabetes with chronic complications Pulmonary fibrosis and bronchiectasis HIV/AIDS Renal failure/nephritis Additional annual premium subsidy (Euro) 876 1,51 1,176 1,495 1,58 1,92 2,87 2,65 4,6 4,796 5,82 11,455 18, ,214 1,8 1,46 1,75 1,71 2,261 2,886,487,66,578 4,491 5,9 6,165 5,548 6,984 7,591 7,288 8,6 9,78 24,2 SOURCE: R.C.J.A. van Vliet, Diagnostic Cost Groups and the Risk-Adjusted Payments for Sickness Funds (in Dutch), WOVM-report 468, ibmg, Erasmus University Rotterdam, January 22. NOTES: PCG is Pharmacy-based Cost Group. DCG is Diagnostic Cost Group. Based on the PCG + DCG model (see text and Exhibit 1). Average annual per capita expenditures equal 874 Euro. For DCGs only one diagnosis per DCG is mentioned here. COPD is chronic obstructive pulmonary disease. ESRD is end-stage renal disease. 5 May/June 24

7 ever-changing treatment protocols. Although this process of continuously updating and improving the risk adjusters might be considered a disadvantage because of its additional costs (which are not that high), it has a clear advantage that may outweigh its costs. The uncertainty about the precise definition of next year s risk adjusters, in combination with the certainty that government aims at the best achievable model, reduces the sickness funds long-term profits from specific risk-selection strategies (such as avoiding certain groups of chronically ill patients). This year s unprofitable patients may be next year s profitable patients, especially in efficient sickness funds. Diagnostic Cost Groups (DCGs). PCGs, which are based on outpatient prescriptions, can be considered an outpatient health status proxy. Therefore, DCGs, as an inpatient health status proxy, might be an ideal complement to PCGs. We adapted the methodology developed by Greg Pope and colleagues. 1 People with multiple hospital admissions in a year were assigned to only the most costly DCG of their various diagnoses. So we use the principal inpatient (PIP) DCGs. Hospital admissions of one or two days were excluded to prevent undesired substitution of outpatient or day surgery care for inpatient care. In our database, 5.5 percent of the population had at least one hospitalization of at least three days. About 4 percent of them were classified in one of twenty-three DCGs (Exhibit ). The other 6 percent of admissions were not included in the DCGs because either (1) the diagnosis was highly discretionary or vague (such as stomachache), (2) it was a temporary health problem (such as a broken leg), or () the diagnosis was not the main reason for hospitalization (such as fever or headache). Adding DCGs to the demographic model greatly increased the predictive accuracy. The R2 doubled, in particular as a result of a better prediction of inpatient expenses. The predictable losses for the 1 percent of people with the highest expenses in the previous year were reduced by percent (Exhibit 1). PCG + DCG model. The simultaneous predictive accuracy of PCGs and DCGs is illustrated in Exhibit 1. In comparison to the demographic model, the predictable losses for the 1 percent of people with the highest expenses in the previous year were reduced by 5 percent, and the R2 nearly tripled. In addition, PCGs and DCGs are quite complementary in their ability to predict future expenses. The complementary character of PCGs and DCGs is also illustrated in Exhibit 4. Exhibit presents the additional annual premium subsidies for people classified in PCGs and DCGs as calculated in our PCG + DCG model. Risk Sharing To prevent selection, the Dutch government also applies risk sharing in particular, a combination of outlier and proportional risk sharing. 14 In 199 the risk sharing was structured so that the sickness funds were responsible for only percent of their financial losses/profits; the Central Fund was responsible for the remainder. Over time there was a gradual increase of the sickness funds financial HEALTH AFFAIRS ~ Volume 2, Number 51

8 Costs & Competition EXHIBIT 4 Percentage Of Sickness Fund Populations Classified In A DCG Or A PCG, Or Not Classified In Either, The Classified in a DCG (%) No Yes All Classified in a PCG No Yes All SOURCE: L.M. Lamers and R.C.J.A. van Vliet, Financing of Sickness Funds (in Dutch), Het Verzekerings-Archief 1, no. 1 (2): NOTES: DCG is Diagnostic Cost Group. PCG is Pharmacy-based Cost Group. risk; that is, the proportion of efficiency gains or inefficiency losses that after ex post risk sharing, on average, is reflected in sickness funds financial statements. In 24 sickness funds financial risk is 5 percent (Exhibit 5). An advantage of this incremental increase in sickness funds financial risk, commensurate with the improvement of the risk adjusters, is that risk selection has not been an issue in the in the past fifteen years. This is in sharp contrast with Germany and Switzerland, for example, where demographic riskadjustment models are combined with 1 percent financial risk for the sickness EXHIBIT 5 Financial Risk Of Dutch Sickness Funds, Year Outpatient expenses (%) Inpatient expenses (%) Total expenses (%) SOURCE: Authors analysis. NOTES: The financial risk is the proportion of efficiency gains or inefficiency losses that after ex post risk sharing, on average, is reflected in the financial result of a sickness fund. Outpatient expenses are between 9 percent (24) and 41 percent (2) of total health care expenses. Total and inpatient expenses include fixed hospital costs, which are about half of inpatient costs. 52 May/June 24

9 funds. As a consequence, risk selection (and its adverse effects) is one of the most pressing health policy issues in Germany and Switzerland. 15 New Risk-Adjustment Issues Further improving the risk adjusters. A challenging question is, How can the health-based risk-adjustment model be improved even more? One option would be to take diagnoses in the DCG model from all medical encounters, not just from hospitalizations. Another strategy could be to use multiyear diagnostic information from prior hospitalizations. 16 Or we could use a better indicator of invalidity for example, an indicator based on the prior use of certain medical aids or appliances. Acceptable costs. By acceptable costs we mean the cost of the set of services and intensity of treatment that government has decided is acceptable for subsidy. 17 In the past the basic benefit package and acceptable costs were clearly defined. In recent years, however, individual sickness funds got more and more flexibility to determine the entitlements of their enrollees. In the coming years, when sickness funds will become more active in managed care, this flexibility is expected to increase further. The question then becomes how to define the acceptable costs that is, the national cost level that forms the basis for calculating the premium subsidies. A related question is, Should all risk factors that are observed by the Central Fund be used to calculate the risk-adjusted subsidies? In December 2 the Dutch Social Economic Council explicitly advised the government to adjust the premium subsidies only for age and health status, and not for other risk factors, such as providers (in)efficiency, overcapacity, and the price levels of contracted providers. A fundamental policy question then is this: Are the sickness fund allowed to differentiate their premium contributions based on the observed risk factors for which the premium subsidies are not adjusted? If not, sickness funds have incentives for selection based on these risk factors. To answer this question, policymakers must make a trade-off between solidarity and (the adverse effects of) selection. Similar questions arise in the case of a voluntary deductible, as recently (again) proposed by the Dutch government. If it appears that the deductibles greatly reduce moral hazard, the issue then is whether the acceptable cost level should be based on those with a deductible or those without. In other words, are the additional expenses attributable to moral hazard acceptable to be subsidized or not? 18 Pay-for-performance. If (in the future) sickness funds pay providers for their performance for example, measured by the change in health status of their patients over time there might be an incentive problem. The better the providers perform in terms of improving health status, the more a sickness fund pays to providers but the lower the next year s premium subsidies that the sickness fund receives. 19 In sum, despite the progress the Dutch have made so far, we are still faced with a full research agenda for health adjustment in the coming years. HEALTH AFFAIRS ~ Volume 2, Number 5

10 Costs & Competition Conclusions And Policy Implications In the, as in several other countries, proposals for managed competition in health care have been launched. To combine managed competition with solidarity requires an appropriate health-based risk adjustment model. If sickness funds encounter strong financial incentives to be unresponsive to the preferences of chronically ill people, the disadvantages of a competitive sickness-fund market may outweigh its advantages. Good health adjustment is the only effective strategy to reduce these incentives for risk selection without reducing solidarity and efficiency and without disturbing competition among risk-bearing sickness funds. Adequate health adjustment. In the past decade much progress has been made toward implementing health adjustment. Good health adjustment not only reduces the predictable losses from high-risk consumers, but it also increases the sickness funds costs of risk selection as well as the uncertainty about their net benefits of selection. Although not perfect, the PCG + DCG model in combination with some risk sharing between the Central Fund and the sickness funds appears adequate to the task of managing competition among risk-bearing sickness funds. Along with improvements in the risk adjustment model, the financial risk of the competing sickness funds has risen from percent in 199 to 5 in 24, without any evidence of adverse effects resulting from risk selection. Toward managed competition, as soon as justifiable. In the past five years the Dutch government increasingly encountered problems with the classic way of controlling health care expenses: government s central capacity planning and price regulation. The public no longer accepts the long waiting lists and public poverty, with private wealth. Because of several court rulings, the macro budget for health care is no longer effective as tool to enforce cost containment by creating a centrally planned scarcity. According to the declaration of policy of the Dutch government that took office in May 2, The central planning by government has failed and will be replaced by managed competition as soon as justifiable. With these last four words, government on the one hand stresses the urgent need for reform and on the other hand indicates that not all preconditions for managed competition are yet fulfilled. Only time will tell whether, after fifteen years of health care reform toward managed competition, the halfway point has already been passed. The authors thank three anonymous referees for their valuable comments on a previous draft. Only the authors are responsible for the content of this paper. NOTES 1. A.C. Enthoven, The Theory and Practice of Managed Competition (Amsterdam: North-Holland, Elsevier, 1988). 2. F.T. Schut and W.P.M.M. van de Ven, Rationing and Competition in the Dutch Health Care System (Paper prepared for the IMPACT Project, London School of Economics, December 22); and J.K. Helderman et al., Market-Oriented Health Care Reforms and Policy Learning in the (Unpublished paper, Erasmus University Rotterdam, 21 March 2). 54 May/June 24

11 . W.P.M.M. van de Ven et al., Risk-Adjusted Capitation: Recent Experiences in the, Health Affairs 1, no. 4 (1994): L. Nelson et al., Access to Care in Medicare HMOs, 1996, Health Affairs 16, no. 2 (1997): ; G.F. Riley, M.J. Ingber, and C.G. Tudor, Disenrollment of Medicare Beneficiaries from HMOs, Health Affairs 16, no. 5 (1997): ; J.E. Ware et al., Differences in Four-Year Health Outcomes for Elderly and Poor, Chronically Ill Patients Treated in HMO and Fee-for-Service Systems, Journal of the American Medical Association 276, no. 1 (1996): ; K. Davis and C. Schoen, Assuring Quality, Information, and Choice in Managed Care, Inquiry 5, no. 2 (1998): ; and R.H. Miller, Health Care Organizational Change: Implications for Access to Care and Its Measurement, Health Services Research, no., Part 2 (1998): E.M. van Barneveld, R.C.J.A. van Vliet, and W.P.M.M. van de Ven, Risk Sharing between Competing Health Plans and Sponsors, Health Affairs 2, no. (21): ; and R.C.J.A. van Vliet, A Statistical Analysis of Mandatory Pooling across Health Insurance, Journal of Risk Insurance 67, no. 2 (2): L.M. Lamers, R.C.J.A. van Vliet, and W.P.M.M. van de Ven, Risk Adjusted Capitation Payment System for Health Insurance Plans in a Competitive Market, Expert Review Pharmacoeconomics Outcomes Research, no. 5 (2): WHO Collaborating Centre for Drug Statistics Methodology, Anatomical Therapeutic Chemical (ATC) Classification Index (Oslo: WHO Collaborating Centre for Drug Statistics Methodology, 2). 8. D.O. Clark et al., A Chronic Disease Score with Empirically Derived Weights, Medical Care, no. 8 (1995): L.M. Lamers, R.C.J.A. van Vliet, and W.P.M.M. van de Ven, Pharmacy-Based Cost Groups: A Risk Adjuster for Capitation Payments Based on the Utilization of Prescribed Drugs (in Dutch) (Rotterdam: Department of Health Policy and Management, Erasmus University, 1999). 1. L.M. Lamers and R.C.J.A. van Vliet, The Pharmacy-Based Cost Group Model: Validating and Adjusting the Classification of Medication for Chronic Conditions for the Dutch Situation, Health Policy (forthcoming). 11. L.M. Lamers and R.C.J.A. van Vliet,, Health-Based Risk Adjustment: Improving the Pharmacy-Based Cost Group Model to Reduce Planning Possibilities, European Journal of Health Economics 4, no. 2 (2): T. Gilmer et al., The Medicaid Rx Model: Pharmacy-Based Risk Adjustment for Public Programs, Medical Care 9, no. 11 (21): G.C. Pope et al., Principal Inpatient Diagnostic Cost Group Model for Medicare Risk Adjustment, Health Care Financing Review 21, no. (2): L.M. Lamers, R.C.J.A. van Vliet, and W.P.M.M. van de Ven, Risk Adjusted Premium Subsidies and Risk Sharing: Key Elements of the Competitive Sickness Fund Market in the, Health Policy 65, no. 1 (2): W.P.M.M. van de Ven et al., Risk Adjustments and Risk Selection on the Sickness Fund Insurance Market in Five European Countries, Health Policy 65, no. 1 (2): L.M. Lamers and R.C.J.A. van Vliet, Multiyear Diagnostic Information from Prior Hospitalization as a Risk-Adjuster for Capitation Payments, Medical Care 4, no. 6 (1996): W.P.M.M. van de Ven and R.P. Ellis, Risk Adjustment in Competitive Health Plan Markets, in Handbook of Health Economics, vol. 1, ed. A.J. Culyer and J.P. Newhouse (Amsterdam: Elsevier Science BV, 2), For the complications in case of a voluntary deductible in the context of a demographic risk adjustment model, see, for example, K. Beck, S. Spycher, and A. Holly, Risk Adjustment in Switzerland, Health Policy 65, no. 1 (2): For a discussion on this issue, see, for example, W. McClure, On the Research Status of Risk-Adjusted Capitation Rates, Inquiry 21, no. (1984): 25 21; H.S. Luft, Modifying Managed Competition to Address Costs and Quality, HealthAffairs 15, no. 1 (1996): 2 8; and van de Ven and Ellis, Risk Adjustment in Competitive Health Plan Markets, 782. HEALTH AFFAIRS ~ Volume 2, Number 55