مدیریت آزمون های ژنتیکی، نظارت، و طب پیشگیری تحت یک سیستم بیمه سلامت عمومی

There is a prospect in the medium to long term future of substantial advancements in the understanding of the relationship between disease and genetics. We consider the implications of increased information from genetic tests about predisposition to diseases from the perspective of managing health care provision under a public health insurance scheme. In particular, we consider how such information may potentially improve the targeting of medical surveillance (or prevention) activities to improve the chances of early detection of disease onset. We show that the moral hazard implications inherent in surveillance and prevention decisions that are chosen to be privately rather than socially optimal may be exacerbated by increased information about person-specific predisposition to disease.

It is fair to say that genomic science is now well into its second phase since current research involves not only the identification of so-called “disease genes” or, more appropriately, “disease alleles”, but also the understanding of how specific sequences of genes interact with each other and with environmental factors to affect the onset and influence the treatment of diseases. Claims in the scientific literature and the media suggest that advancements made in genetic information will lead to significant improvements in the effectiveness of prevention and treatment of disease. A rough road map of the human genome has been available since 2003 and currently, according to the NIH-sponsored web site genetests.org, there are over 1600 genetic tests used clinically. With the prospect of the so-called $1000 genome close to reality (see Davies, 2010), whole genome sequencing may soon become the norm for developed countries. The information that can be gleaned from an individual's whole genome has the potential to revolutionize the practice of medicine with population wide genome sequencing forming the basis of so-called P4 medicine (i.e., medicine that is Predictive, Preventive, Personalized and Participatory). Although the future of P4 medicine has many proponents, not least of whom is Leroy Hood through his P4 Medicine Institute (p4mi.org), there is some controversy over the pace of its progress.1 Once the relationships between specific genes, environment, and diseases are better understood, harnessing this information to create improved health outcomes in a cost effective manner requires a good understanding of how individuals will behave in the context of such individualized informational change. We provide insight into this debate by focusing on how individuals’ incentives for use of surveillance (monitoring) technologies, such as colonoscopies or mammograms, change in the presence of risk-type specific information about the likelihood of onset of disease. It has been debated in the literature whether population wide screening for diseases such as colon cancer or breast cancer is cost effective and whether monitoring should be restricted to those at higher risk as identified, for example, by family history. As genetic tests become more wide ranging and less costly, there is the potential of substantial improvements to the targeting of surveillance techniques such as colonoscopies with the potential of improved overall health outcomes in a more cost effective manner. However, we show that the usual moral hazard problems associated with insurance coverage may interact with improved knowledge of individual risks in a way that could blunt the potential for such improvements. Through the use of simple models, we develop a series of results which characterize the possible outcomes that could develop as more genetic information becomes available. Many genetic tests continue to be expensive and so choosing which tests to make available through health insurance plans, be they private or public, represents a challenge. Insurance or health care providers are concerned about the possibility of escalating costs due to the adoption of more genetic tests (e.g., see report by Miller et al. (2002) funded by the Ontario Ministry of Health and Long Term Care) while others believe improved targeting of surveillance and preventive measures will ultimately reduce health care costs.2 It is this aspect or phase of growth in genetic testing and related knowledge that we address here. In particular, we study the implications of improved genetic information about risk of disease in terms of the socially optimal management of surveillance and related health care strategies for public health insurance systems. The results of this exercise can be helpful in developing guidelines to use in determining which genetic tests to offer within the coverage of the public health system. Some aspects of what we find could also be applied to a population covered (or partly covered) by private health insurance, although there are some important differences to consider. Many of the papers that model the effects of improved information about risk classification involve the private insurance market and exogenously specified (fixed) probabilities of disease and/or financial loss (e.g., Rothschild and Stiglitz, 1976, Wilson, 1977, Hoy, 1982, Hoy, 1984, Crocker and Snow, 1985, Crocker and Snow, 1986, Tabarrok, 1994, Hoel and Iversen, 2002 and Rees and Apps, 2006).3 Our model also involves exogenously determined differences in the probabilities of onset of disease. However, we allow for the possibility of early or late detection of disease through individuals’ choices of level of surveillance. For many diseases, early detection leads to improved treatment and outcomes. Information from genetic tests creates (or increases) differential assessment of risk of disease onset across individuals. Thus, although probability of onset may be fixed by genotype, choice of level of surveillance creates endogenous determination of detection being late or early (i.e., at least probabilistically). The possible benefit of a genetic test in this context arises from potential improvements in targeting of surveillance strategies for early detection of onset of disease. The important management issue is in determining the extent to which higher (lower) risks should increase (decrease) surveillance and then trying to encourage the appropriate responses from individuals. We show that a model of differential use of preventative medicine based on genotype is very similar and so determination of the value of genetic tests follows a similar pattern relating to improved targeting of such strategies. 4 It is well known that in the presence of health insurance, be it public or private, individuals face incentives that lead to actions that are not necessarily socially optimal. In our context, we presume that individuals do not pay for the financial costs of surveillance or treatment, should onset of disease occur. The result is that individuals may either over-use or under-use medical surveillance or prevention. The moral hazard problems due to insurance are complicated by the introduction of information about differential risk of disease onset. We characterize how genetic testing can lead to changes in the pattern of over- and under-use of surveillance. We find, under a broad range of scenarios, that at least one group (i.e., the average, high or low risk types) will tend to want to over-use surveillance relative to the socially optimal decision. The relative extent to which over-use (or under-use when it occurs) of surveillance reduces social welfare can vary across the groups in counter-intuitive ways. Overall efficiency may fall as improved knowledge about risk type interacts with the standard moral hazard implications of insurance leading to a reduction in social welfare. In the following section, we introduce a simple model of surveillance, which is also referred to as screening or monitoring. The basic model describes the decision for intensity of monitoring taken by the individual and compares that to the socially optimal decision. In Section 3, we present our results regarding the implications of introducing genetic tests and in Section 4, we briefly consider the case of private insurance and implications of explicitly accounting for costly genetic tests. We provide a discussion, conclusion, and suggestions for further research in the final Section 5.

We have provided a model that characterizes moral hazard implications resulting from an individual's choice of surveillance/monitoring for disease (e.g., use of colonoscopies, FOBT, and other techniques for detecting colon cancer). We have shown that, although a higher predisposition to a disease will always lead to an increase in the privately optimal level of surveillance chosen by individuals, this is not necessarily the case for the socially optimal level. If the continuing lifetime costs of treatment for those incurring disease are higher under early detection than under late detection, which is a common phenomenon, then individuals will over-utilize surveillance relative to what would be socially efficient. However, in some cases early detection of disease leads to a cure or preemption of disease (e.g., detection and removal of polyps through colonoscopy) and this is not as costly compared to the lifetime costs of treating the disease conditional on late detection. An example would be detecting colon cancer in situ or at stage 1 rather than at a later stage. In such a case, under-utilization of surveillance may occur. Given all of these possibilities, it is clear that, in the context of genetic testing, careful attention must be paid to the problem of which individuals should be encouraged to either decrease or increase their level of surveillance and by how much. This relationship will depend on the specifics of the disease, including; (i) perceived (personal) health benefits of early versus late treatment of disease; (ii) relative financial costs of treatment under early versus late detection; (iii) personal (including psychological) and financial costs of the monitoring technologies available; (iv) relative effectiveness of these technologies at detecting disease early and how this effectiveness varies with intensity of surveillance across risk types.22 The main contribution of our paper is to develop a method to analyze the welfare implications of a genetic test (or other diagnostic test) that creates improved information about person-specific risk-type (i.e., predisposition to disease onset). We show how such information interacts with the moral hazard phenomenon regarding choice of surveillance intensity in a way that may lead to a reduction in social welfare even for a costless genetic test. Whether welfare will be enhanced or reduced depends crucially on the curvature of the equilibrium cost function. This function describes how the per capita overall cost of health care provision depends on the relationship between a person's perceived probability of onset of disease and her privately optimal choice of surveillance (in conjunction with other health care parameters). If this function is concave in the probability of disease onset (ρ), then the introduction of a genetic test will be welfare improving. However, if this function is strictly convex then welfare may fall with the introduction of a genetic test even when the test is costless. The intuition underlying this result is that a genetic test is essentially a mean preserving spread in the population wide perceived probability of onset of disease and so, if the cost function is convex in this probability, then the introduction of improved information leads to an increase in the overall cost of providing health care. Although the improved information leads to individuals choosing surveillance levels that reflect improved targeting from the perspective of their personal health benefits, individuals ignore the financial cost implications of both their decision to obtain a genetic test and their surveillance choices. Therefore, if health care delivery costs increase sufficiently, these advantages to improved information may be insufficient to result in a welfare improvement. Our model for assessing the benefits of genetic testing in the context of “improved” surveillance decisions can be adapted to a model for prevention.23 Instead of individuals choosing a level of intensity of surveillance, which affects the probability that a disease which has already been incurred will be detected early, consider a scenario in which individuals choose a level of preventive care. The higher the level of preventive care, the greater the likelihood that there will be no onset of disease. There are a multitude of diseases for which genetic tests can lead to potential improvements in choice of level of prevention (i.e., so-called multifactorial genetic diseases). For example, a woman who discovers she has the BRCA1 gene may choose to have prophylactic surgery that can significantly reduce the probability of later onset of breast or ovarian cancer. There is a similar moral hazard phenomenon associated with preventive activity and so over-utilization and under-utilization are possibilities. Similar analyses follow in that the way decisions of individuals regarding privately optimal levels of preventive care change as a result of information from genetic tests and this may lead to either an increase or decrease in social welfare. Our model could also be adapted to scenarios in which relevant health care strategies involve both surveillance and prevention activities. Suppose an individual is diagnosed with (relatively) early stage breast cancer and this occurs at an early age. A genetic test may reveal whether this person has one of the so-called breast cancer genes (BRCA1/2). If the test is positive, then the individual may choose an aggressive treatment for the disease (e.g., double mastectomy) while if the test is negative the individual may choose a less aggressive treatment (lumpectomy). The result of the genetic test may also affect the intensity of surveillance going forward in the individual's life in order to detect any possible reoccurrence of the disease at an early stage. The continuing lifetime costs depend on both the person's genetic type and the treatment option where (current) treatment behaves also as a preventive measure against reoccurrence. A major challenge in organizing health care is to decide which programs of surveillance and prevention are worthwhile. Cohen et al. (2008) point out that there is often excessive optimism about the ability of preventive health measures and technologies (including screening programs) to reduce health care costs. This raises questions as to which measures and technologies are reasonable investments. Our paper offers a methodology to aid in determining how specific preventive medicine programs may be improved through the use of genetic tests in conjunction with targeted screening directives. As they note (Cohen et al., 2008, p. 661), for example, “the efficiency of cancer screening can depend heavily on both the frequency of the screening and the level of cancer risk in the screened population.” Suppose, in the context of our model, that the existing level of screening that occurs for an entire population with probability of disease ρ0 (i.e., without genetic testing) represents over-utilization. A genetic test will identify those at lower risk (ρL), who are predicted by the model to reduce surveillance, which in this scenario is likely to improve welfare. Those who are discovered to be high risk (ρH) will increase their intensity of surveillance and since targeting surveillance to more risk prone individuals is often more efficient, this may lead to an improvement in welfare for this group as well. However, as our model points out, whether or not these changes will represent an overall improvement in welfare is not straightforward even though the “direction of change” seems to be appropriate for both risk groups. Our model provides a starting point for conceptualizing a cost-benefit approach that includes behavioural responses to assess the value of the introduction of any particular genetic test. We also offer directions for future research. Our model presumes individuals are ultra-rational in that they are expected utility maximizers who understand well the various probabilities in the model (i.e., probability of onset of disease and probability of early detection at various levels of intensity of surveillance). Applied studies need to consider how individuals interpret risks such as genetic predispositions to disease (e.g., see O’Doherty and Suthers, 2007) and how, conditional on genetic risk type, they understand how different intensities of prevention and surveillance affect the probability of onset of disease or early detection. Non-EU or behavioural models of decision making also deserve attention in the context we have described in this paper. It should be recognized that counselling services from health care professionals may help individuals understand the outcomes of tests and strategies but that in itself does not correct for the wedge between privately and socially optimal decisions. For example, suppose many of those who perfectly understand the extent of the increased risk of a disease resulting from a positive genetic test wish to increase surveillance beyond what is socially optimal. This can be a perfectly rational individual decision. Our model demonstrates that, in such a case, the health care system should, in principle, respond to such pressures by denying the full demand for increased surveillance. Doctors are often modeled as gatekeepers of medical services but such practices can sometimes be difficult to carry out when patients’ desires are in conflict with the socially optimal provision of services. In other instances, individuals under-use surveillance – possibly only after a genetic test is received – and the socially optimal level of surveillance can only be achieved by encouraging an individual to submit to a higher level of surveillance than the individual wishes. This can also create conflict between doctors and patients and lead to the erosion of doctor–patient trust. Methods for dealing with these problems are beyond the scope of this paper but certainly worthy of consideration. We have ignored many potential sources of heterogeneity of individuals in our model. Family history can often be used to create different subpopulations facing different likelihoods of receiving a positive genetic test result. This specific feature is not difficult to include in our analysis as it simply implies the relevant parameters vary across such subgroups as would the value of a genetic test (e.g., see Hoy and Witt, 2007 for such an approach in a different context). A more problematic type of heterogeneity is the range of different personal preferences one would expect over the physiological benefits and costs of surveillance or prevention. For example, some people may simply have a higher disutility from certain surveillance procedures. This makes it difficult to determine, for example, which individuals should receive higher surveillance levels and which lower surveillance levels in order to obtain a social optimum. Such information about preferences is intrinsically private and so this represents a serious challenge to so-called one-size-fits-all health insurance plans. But this feature, admittedly, is not unique to our problem. Many countries with substantial coverage through private health insurance plans have prohibitions on risk-rating of premiums as well as mandatory coverage of certain items. Thus, our models and results can directly provide some guidance for such private schemes. Private insurance regimes, however, are often more open to user fees and co-payments. These features offer additional instruments for influencing private choices of surveillance or prevention and so may offer some interesting avenues for future research. However, since the burden of co-payments and user fees would fall differentially across (genetic) risk types, use of such instruments would create a phenomenon akin to premium risk. It is also worth integrating models that highlight the risk premium problem or adverse selection costs associated with genetic testing in related markets such as life insurance. These models demonstrate that such information may be welfare reducing (see, for example, Hoy and Polborn, 2000 and Hoy and Ruse, 2005 and Hoy (2006)) and such effects can be compared to the potential welfare improvements associated with improved health care decisions.