کژ گزینی و تبعیض طبقه ای در بازارهای بیمه سلامت: اثرات، آزمون های ژنتیکی

In this paper, the effects of new methods for risk classification, e.g., genetic tests, on health insurance markets are studied using an insurance model with state contingent utility functions. The analysis focuses on the case of treatment costs higher than the patient's willingness to pay where standard models of asymmetric information are not applicable. In this case, the benefit from signing a fair insurance contract will be positive only if illness probability is low. In contrast to the common perception, additional risk classification under symmetric information can be efficiency enhancing. Under asymmetric information about illness risks, however, there can be complete market failure.

Insurance market equilibria under asymmetric information1 and welfare effects of imperfect categorical discrimination2 have been extensively studied in rather generally designed models. In these models, utility functions are assumed to be state-independent and financial loss does not exceed wealth. Hence, a risk averse person would always choose to sign an insurance contract at a fair premium. If the population consists of two different risk types and there is asymmetric distribution of information about the insureds risk types (insureds are informed about their illness risks, while insurance companies are not informed), the market equilibrium is found to be a Nash-equilibrium providing high risk types with full insurance and low risk types with partial insurance (known as Rothschild–Stiglitz–equilibrium).3 In case of symmetric information about risk types, insureds would prefer not to receive additional information about their illness risks because of the additional income risk caused by more risk adequate insurance premiums.4 For health insurance markets, however, these standard models are not generally applicable, mainly for the following reason: loss in health insurance implies non-financial loss of well-being by getting ill. If there exists a method of treatment, this loss of utility can be transformed into a financial loss by visiting a physician and undertaking the costs of the treatment. Without health insurance, a rational patient would compare expected benefits from being treated to the price he would have to pay for the treatment. He would choose to bear his sufferings and not to be treated if treatment costs were higher than his willingness to pay.5 The structure of possible losses of well-being (and consequently of wealth) in the health sector is quite heterogeneous. Roughly we can distinguish between acute (curable) diseases and chronic (incurable) diseases. In the latter case, treatment prevents worsening and alleviates complaints, but effectuates no cure and often there is no significant change in life expectancy. Still, costs of long-term treatment can be quite high and, hence, exceed an uninsured patient's willingness to pay (or even his wealth). Examples for this phenomenon are numerous: bypass surgery in cardiovascular diseases, disc surgery or visits to the health resorts as a treatment of a slipped disc as well as treatment of many internal diseases like rheumatism or subsequent diseases in diabetes can be named here. Since many infectious and other acute diseases are becoming less of a threat (because of new drugs, operation methods and intensive care medicine), life expectancy keeps on rising and people tend to accumulate chronic diseases while aging. Consequently, in the last decades, there has been a trend towards spending an increasing proportion of the total health care expenditures for treatment of chronic diseases.6 In order to construct a model suitable for the description of the health insurance market, this specific fact has to be taken into account. In Section 2of this paper, a simple health insurance market model will be presented. State contingent utility functions are used,7 and patients in case of illness are assumed to have the choice between undergoing a treatment or suffering from their diseases. The analysis thereby focuses on the case in which treatment costs are higher than an uninsured person's willingness to pay and lie within a certain range so that low risk types benefit from a fair insurance contract while high risk types do not. The effects of genetic testing have become of increasing political importance.8 In recent years there has been a rapid development of new biochemical and microbiological screening methods (e.g., genetic tests) rendering possible not only the diagnosis of diseases that have already occurred, but also enabling the prediction of a higher risk of the onset of specific diseases years later.9 From the insurance point of view, these diagnostic methods have to be considered as new possibilities for more exact risk classification of insured persons, and also as a possible source of new asymmetry in the distribution of information between insurer and insureds. Within the next decade, we especially expect gene technology to strongly push this development by offering (at least to some people) a quite detailed forecast about what kind of illness one is likely to suffer from in the future. Among these potentially predictable diseases are such common chronic diseases as Alzheimer disease,10 diabetes (types 1 and 2),11 malignant neoplasias,12 alcoholism,13 and many more internal,14 orthopedic (osteoarthritis)15 and psychiatric diseases.16 In light of these developments, a controversial debate on whether private insurance companies should be allowed to gather genetic information from applicants and to calculate insurance premiums according to the insured person's genetic risk has started.17 In the framework of our model, we contribute to this debate by considering the market outcome under two scenarios: in the first scenario, all information is publicly available. This is discussed in Section 2. From a regulatory point of view, this corresponds to a `laissez-faire' regulation: insurer are allowed to ask for and demand genetic tests from their customers. In Section 3the case of private information is analysed, where insurer are not allowed to use information resulting from genetic tests. This scenario seems to become more and more relevant. As discussed by Chuffart (1996), there is a tendency by policy makers in many countries to forbid insurer to use genetic information. We derive the following results. If information is symmetrically distributed, then, in contrast to standard models, the screening of insurance applicants for illness risks can enhance efficiency. The welfare effect depends on the precision of separating high and low risk types, reaching an optimum at a certain degree of imperfection of the test. More precisely, the highest efficiency gain is achieved if the probability of a false positive test result (indicating high illness probability though the individual has a low risk) is minimized and the probability of a false negative test result is significant. Some genetic tests seem to have quite similar properties.18 On the other hand, if the information is private, it will be demonstrated that in case of treatment costs higher than willingness-to-pay, asymmetry in information about the insureds risk types not only could cause signaling costs (i.e., partial insurance for low risk types) as in a Rothschild–Stiglitz- or a Wilson–Miyazaki–Spence-type market equilibrium, but could cause the whole market to break down. In Section 4of the paper conclusions for a legal framework to regulate the health insurance market are drawn. We have to be aware that the validity of the results is restricted to only a part of the health insurance sector, i.e., to treatment methods for which some people would not be willing to pay for without insurance. In this area of health insurance, the implementation of a private, voluntary insurance market, where risk classification using any kind of costless information is allowed, would be efficient. For all remaining treatment methods, a basic (perhaps national) insurance without risk classification is more appropriate from an efficiency point of view.

The health insurance model presented in this paper is different from standard insurance models that have been widely used to study problems of informational asymmetry and efficiency effects of categorical discrimination. While in liability insurance most of the time there is a definite financial loss, this is not the case in health insurance. In case of illness at first there is a loss of quality of life. This loss can only be transformed into a financial loss by undertaking a treatment. However, most of the time there are many alternatives for treatment, which might differ a lot in respect of their costs and effects. In fact, there is always another diagnostic measure or step of treatment that still has a marginal positive effect but causes such high costs that uninsured individuals would not be willing to pay for it, especially in treatment of chronic diseases. Besides that, treatment of manifested chronic diseases sometimes is also preventive treatment,45 and successful treatment is often part of the diagnosis.46 Therefore, in contrast to liability insurance companies, health insurance companies only refund treatment costs and do not pay out money to sick patients who reject undergoing treatment.47 Building a model for the health insurance market, these characteristics have to be integrated and simplified without being too restrictive. The use of state contingent utility functions seems to be quite an appropriate way of substantiating the transformation of the loss of well-being into a financial loss. The only restriction caused by this assumption might be the exclusion of acute life-threatening diseases from the analysis. However, if treatment only reduces mortality marginally, the model can also be applied to acute diseases. The quality of the results obtained would not be different, if the assumptions were dropped that treatment provides perfect restoration of health. In fact, it is the assumption that treatment costs are higher than willingness-to-pay which drives the main results. Still, rational individuals with low risks will demand insurance that refunds costs of those kinds of therapies where T>R(W0), although they would not be willing to pay for the treatment in case they were ill. This might be quite an important aspect in health insurance that so far has not received much attention in the analysis of insurance markets under asymmetric information. As has been shown in this paper, results from standard insurance market models may not simply be transferred to health insurance markets. On the basis of the present model, policy conclusions for the regulation of health insurance markets have to be thought over. 4.2. Efficiency effects of new methods for risk classification In Section 2.3, it has been argued that under certain conditions costless categorical discrimination can enhance efficiency on private health insurance markets. In contrast to standard models, making the results of genetic tests available to the insurer might be welfare improving. The efficiency gain is maximized if carrier have a very large illness risk, while non carriers still face some non-negligible risk. This is a characteristic that some genetic tests seem to have in deed, since many common diseases (e.g., osteoarthritis48) have a hereditary, genetically detectable variant as well as a genetically not detectable variant. Efficiency gains from risk classification concerning these diseases have to be regarded as relatively high. National health insurance fails to reach efficiency in cases where costless categorical discrimination is able to enhance welfare. Under national insurance, the insurance contract for high risk types covers treatment methods that informed high risk types would not want to be included in their optimal health risk insurance bundle, so that high risk types have to be regarded as overinsured. In contrast, low risk types to this extent remain underinsured. As follows from the analysis in Section 3, results from predictive tests can have disastrous efficiency effects on private insurance markets if there is asymmetric distribution of information. Depending on the legal framework, insurance market equilibria can be of the Rothschild–Stiglitz-, Wilson E2- or Wilson–Miyazaki–Spence-type, in all of which the ex ante expected utility is lower than under the condition of no information about risk types.49 In case of treatment costs higher than willingness-to-pay, the insurance market can even collapse. Now, if we try to draw conclusions from the health insurance model, we have to be aware that in the real situation there are many different kinds of diseases and treatment possibilities so that there is a multidimensional problem. Indeed, we can regard real health insurance as a bundle of numerous unique health insurance contracts each covering costs of one specific way of treatment applicable in one specific diagnosis. If we assume that individuals are able to freely and independently combine essential contracts in order to form their optimal health insurance bundle, then the feasible bundle will become smaller if the market collapses for one of those essential contracts. In the case of asymmetric information this leads to an efficiency loss because of underinsurance of low risk individuals, since they cannot include all the desired essential contracts into their insurance bundle. As has been shown in Section 3.1, this underinsurance of low risk individuals is even more distinctive if the legislation framework admits only pooling (and no separating) contracts. In those cases, forcing insurance companies to sell pooling contracts in health insurance has to be considered as an inadequate way to reach efficiency.