Big Data,precision medicine and private insurance: A delicate balancing act

Abstract

In this paper, we discuss how access to health-related data by private insurers, other than affecting the interests of prospective policy-holders, can also influence their propensity to make personal data available for research purposes. We take the case of national precision medicine initiatives as an illustrative example of this possible tendency. Precision medicine pools together unprecedented amounts of genetic as well as phenotypic data. The possibility that private insurers could claim access to such rapidly accumulating biomedical Big Data or to health-related information derived from it would discourage people from enrolling in precision medicine studies. Should that be the case, the economic value of personal data for the insurance industry would end up affecting the public value of data as a scientific resource. In what follows we articulate three principles – trustworthiness, openness and evidence – to address this problem and tame its potentially harmful effects on the development of precision medicine and, more generally, on the advancement of medical science.

This article is a part of special issue. Below disclaimer should be added after keywords as per style in both PDF and online XML: This article is a part of special theme on Health Data Ecosystem. To see a full list of all articles in this special theme, please click here: https://journals.sagepub.com/page/bds/collections/health_data_ecosystem.

Introduction

In 2018, in light of progress in genome sequencing and given the growth of direct-to-consumer (DTC) genetic testing, the Swiss Parliament has discussed amendments to the Federal Act on Human Genetic Testing, in force since 2004 (Schweizer Nationalrat, 2018). To the surprise of many, a parliamentary committee suggested to lift the ban on the use of genetic information in private life insurance. While the proposal was ultimately not approved by the Swiss Parliament, it revamped the debate on the use of personal health data in private insurance.

In this paper, we discuss how access to health-related data by private insurers, other than affecting the interests of prospective policy-holders, can also influence their propensity to make personal data available for research purposes. We take the case of national precision medicine initiatives as an illustrative example of this possible tendency. Precision medicine initiatives, such as the All of Us research cohort in the US, pool together unprecedented amounts of genetic data, phenotypic data, environmental and also lifestyle data collected by participants themselves through smartphone apps and wearable devices. Through ever more sophisticated data analysis techniques, including artificial intelligence methods such as machine learning (Vayena et al., 2018), researchers can uncover patterns and correlations in such data, thus generating unprecedented amounts of predictive information about the health of data donors (Alyass et al., 2015). Private insurers could be prompted to claim access to such rapidly accumulating biomedical Big Data about their prospective customers, in order to generate more accurate estimates of their health risks. Should that be the case, out of fear of being denied insurance coverage, people may become more reluctant to donate data for precision medicine research. In this way, the economic value of personal data for the insurance industry would undermine the public value of data as a scientific resource. In what follows we articulate three principles – trustworthiness, openness and evidence – to address this problem and tame its potentially harmful effects on the development of precision medicine and, more generally, on the advancement of medical science.

A growing data ecosystem for precision medicine

Building on progress in pharmacogenetics, pharmacogenomics and targeted therapy, precision medicine (PM) aims at integrating multiple data sources to ‘tailor medical treatment to the individual characteristics of each patient’ (National Research Council, 2011). The development of precision medicine is therefore premised on the collection of large repositories of data including electronic health records (EHRs), standard clinical measurements, genome sequences, environmental data and lifestyle data collected over time even through mobile devices and apps. Most national precision medicine initiatives are pulling disparate data types from hundreds of thousands of citizens (Ginsburg and Philips, 2018). For instance, the U.S. ‘All of Us’ research cohort will comprise data from at least one million individuals – representative of the ethnic diversity of the country – with the aim of making such data available for biomedical research to uncover correlations about biology, lifestyle, environmental exposures and health outcomes. In other words, precision medicine initiatives aggregate personal data as a publicly valuable resource that should be made accessible for the generation of scientific knowledge through constantly evolving data processing tools. Such an understanding of the public value of Big Data for human health contrasts with a long standing barrier to effective data sharing, namely, the fact that different data types are conventionally locked in separated, non-communicating and often privately owned silos (Blasimme et al., 2018). The ubiquity of sensor-equipped devices enabling extensive and relatively inexpensive data collection, coupled with developments in data analytics in recent years, and in particular with rapid progress in machine learning (Vayena et al., 2018) conferred momentum to the growth of precision medicine as a form of Big Data science.

A defining trait of precision medicine is that research participants who contribute their data to large-scale repositories acquire a more engaged role with the data they provide. While until now research participants would simply make their data and biological samples available for research, precision medicine initiatives tend to emphasize the possibility of enabling participants to access their data, and to receive relevant findings about their health (Blasimme and Vayena, 2016, 2017). This form of empowerment can potentially grant large numbers of people access to relevant health information about themselves, including estimates concerning their risk of developing diseases in the future. From the perspective of private insurers, this can give rise to an information asymmetry between them and their policy-holders who – by enrolling in a precision medicine study – may acquire a more accurate picture of their health-related risks.

Information asymmetry in economic markets: The case of genetics and insurance

In economic transactions, information asymmetry occurs when one party has more or better information than the other. Neoclassical micro-economics considers equal access to information by all trading parties (i.e., buyers and sellers) as a precondition to ‘make markets work’ (Arrow, 1963). This is particularly relevant in insurance markets, in which parties trade risk coverage contracts.

When individuals know they bear high risks of some sort, they might be more inclined to obtain private insurance coverage against those risks. This phenomenon is known as adverse selection due to informational asymmetry (Akerlof, 1970; Arrow, 1963) and it might lead to an imbalance in the portfolios of the insurers if they have an excessive amount of high-risk policy-holders. As a result insurance companies pay out disproportional claims with respect to collected premiums. To cope with adverse selection and informational asymmetry, insurance companies differentiate between risk groups through the process of risk selection, in order to ensure that the premium offered to customers statistically reflects the overall risk in their portfolio of clients.

Regulatory restrictions pose limits to insurers’ risk selection practices, as in the case of genetic information. These types of regulations were first introduced in the 1990s, when the advent of genetic testing prompted fears of ‘genetic discrimination’ by insurance companies. Genetic discrimination is defined as ‘discrimination directed against an individual or family based solely on an apparent or perceived genetic variation from the “normal” human genotype’ (Billings et al., 1992: 476). Private insurance companies can commit genetic discrimination if they refuse to insure individuals at risk of a genetic disease, or if they increase the premiums for their policies.

Fears of genetic discrimination have been reported in numerous studies examining participation in genomic research and clinical use of genetic testing (Geelen et al., 2012). Such fears can lead to refusal to participate in scientific research, a decline in genetic testing, and disincentives to disclose test results to health care providers and relatives (Wauters and Van Hoyweghen, 2016). For instance, a recent study in Australia found that colorectal cancer patients who are strongly opposed to genetic testing mostly cite insurance concerns as the reason for their attitude (Keogh et al., 2017).

Since the 1990s, many countries worldwide have enacted regulation aimed at restricting or prohibiting the use of genetic information in private life and/or health insurance (Quinn et al., 2014). Canada has recently enacted legislation prohibiting insurance companies from accessing any genetic test results (GNA, 2017). Other countries have followed a moratorium approach, such as in the UK, where an agreement between the government and the insurance industry prohibits access to genetic test results, except for Huntington’s Disease and for policies worth over £500,000 (HM Government and ABI, 2018). Interestingly, according to this agreement, British insurers do not expect findings produced in the context of scientific research to be disclosed, but this exemption applies only to genetic data. Many European countries such as Belgium, Austria, Denmark, France, Germany, Lithuania, Norway, Portugal, and Sweden have implemented outright bans. In addition to national regulation, international bodies have produced soft law (e.g., the UNESCO Universal Declaration on the Human Genome and Human Rights of 1997; the 2004 UNESCO International Declaration on Human Genetic Data; and the 1997 European Convention on Human Rights and Biomedicine of the European Council) rejecting any form of discrimination on the grounds of one’s genetic heritage. Such regulations, however, have not alleviated concerns about genetic discrimination (Wauters and Van Hoyweghen, 2016, 2018). Moreover, private insurers have in turn shifted their focus to so-called ‘lifestyle underwriting’ as a new ‘disruptive’ business space (Meyers and Van Hoyweghen, 2017). Lifestyle underwriting amounts to insurance contracts that are tailored to the lifestyle of the customer in terms of physical activity, smoking habits, diet and other lifestyle factors known to have an impact of health-related risks. Health and life insurers thus offer reduced fees to those policyholders that allow their lifestyle to be monitored, for instance through a fitness band or a smartphone app tracking what a person does to improve her fitness. This model applies also outside the health sector. An increasing amount of car insurance companies, for example, offer incentives and reduced fees calculated on the data collected by a small device installed in the car, that provides the insurance company with a variety of behaviour-based measures (e.g. driving habits, speed, car use and the like).

From genetics to precision medicine: Re-opening the conundrum

In recent years, the insurance industry has clearly regarded the growing availability of predictive health information as increasing the risk of adverse selection – that is the tendency of those who bear higher risks to buy insurance policies. For instance, the insurance industry has expressed concerns that direct-to-consumer (DTC) genetic testing and the clinical use of next generation sequencing (NGS) may eventually damage its commercial interests due to information asymmetry (Swiss Re, 2017). Recent reports by insurance industry think tanks (Geneva Association, 2017), actuarial professional organizations (Actuaries Institute, 2017) and re-insurance companies (Swiss Re, 2017) have discussed the implications of progress in genomic testing for their business, re-opening the debate on regulatory restrictions to insurers’ use of genetic information. A recent actuarial report in Australia found that insurance claims are expected to rise as a result of genetic testing because customers, armed with more knowledge about their health risks, will likely decide to buy more insurance coverage (Actuaries Institute, 2017).

Insurers support the increased use of genetic testing for clinical purposes as it contributes to early identification of diseases and encourages the adoption of healthier lifestyles – which can be taken into account in insurance premiums. To this aim Swiss Re (2017) proposes not to require people to undergo genetic testing as part of a life insurance application, but to allow insurers to access genetic information that is available and known to the applicant (Swiss Re, 2017: 4). The growth of precision medicine, with its specific emphasis on empowering research participants by granting them access to health-related information is likely to add to this phenomenon. Notably, precision medicine projects this problem beyond the realm of clinical genetics. While legal provisions exist to limit private insurers’ access to genetic tests, clear safeguards have not been in place to restrict access to predictive information and risk estimates that are likely to emerge from precision medicine research (Van Hoyweghen, 2018). Insurers may thus in principle claim access to health information obtained in the course of precision medicine research, which – albeit being generated also through processing genetic data – is not necessarily covered by existing regulations on access to genetic test results. In this case, individuals may fear that participating in precision medicine research may put them at disadvantage, should insurers be allowed to access predictive information about their health. People’s willingness to join precision medicine research cohorts may be affected by this fear, with possible effects on the scientific prospects of precision medicine itself.

Towards a new framework

Proposing a definitive solution to the long standing problems of information asymmetry and adverse selection in insurance markets falls beyond the scope of this commentary. Yet, progress needs to be made in creating the conditions to mitigate the effects of new phenomena taking shape at the intersection of biomedical Big Data research and private insurance markets. While empowerment in precision medicine is a desirable development, information asymmetry can be detrimental to insurance markets due to the risk of adverse selection. Similarly, the fear of being denied access to services – such as private life and health insurance – and, more generally speaking, the fear of being discriminated against for reasons having to do with one’s health status can decrease people’s willingness to participate in research and thus severely curtail precision medicine’s chances of success. Such a state of affairs calls for initiative to strike a socially acceptable balance between the need to promote precision medicine research, participant empowerment and the interest insurers have in preventing the effects of information asymmetry and adverse selection.

What we propose here is a set of three principles to guide the creation of governance mechanisms and to mitigate the potentially deleterious effects of information asymmetry generated in the context of precision medicine initiatives.

Trustworthiness: to encourage research participation and to ensure sustained participation rates in precision medicine cohorts, research biobanks and data-repositories should clearly state that they will not share data with private insurers.

Openness: to limit the possible risk of adverse selection, private insurers can be allowed to propose premium-based incentives to encourage policy-holders to voluntarily disclose health-related information generated in the context of research.

Evidence: to ensure equitable use of disclosed information, insurance companies should issue and follow scientifically validated standards for risk prediction models based on integrated information from genetic, genomic, lifestyle and medico-actuarial data.

While these principles are not new in health policy, what we propose is a new articulation of the demands deriving from these principles. Each of the three principles caters to the needs of a different stakeholder group. Trustworthiness addresses the needs of the scientific community and society more in general, and is grounded in the public value of scientific research for present and future generations. Its aim is to ensure that science has access to vital resources such as data to carry out its activities – which can only be achieved if there is trust in the scientific enterprise and its role (Vayena and Blasimme, 2017). Restricting access to sensitive personal data and information increases trust by limiting the likelihood that data will be used for private purposes that research participants may see at odds with their interests and rights. Reducing the risk of potentially negative consequences of data misuse for research participants is likely to increase public trust in particular for scientific activities involving the collection and analysis of personal data.

The second principle – openness – addresses the need of the insurance market to cope with the possible risk of adverse selection. It is based on the value of fair market transactions and aims at contributing to the sustainability of economic activities in the insurance sector. Premium-based incentives to voluntary disclosure allow insurers to reduce the uncertainty caused by the risk of adverse selection, while at the same time rewarding customers who decide to disclose. Moreover, openness reduces the need for insurers to actively look for information or, in the longer run, to lobby for changing existing regulatory protections – which as we showed in our opening vignette is a concrete possibility.

The third principle – evidence – is intended to protect individual data subjects whose data are used in the context of risk selection practices. Its aim is to prevent arbitrary uses of predictive information on the part of insurers. Even within existing regulatory constraints, risk selection practices can be unfair if they are based on biased datasets or inaccurate risk models. Growing reliance on Big Data is likely to exacerbate this problem due to the fast-paced development of this still relatively young field. High standards of evidence can for instance prevent that health-related predictions generated in the course of precision medicine research – and disclosed to participants – are prematurely incorporated into risk selection practices. Quality standards in Big Data actuarial research, along with validated predictive risk models are therefore needed to promote fairness and to ensure equitable processing of personal data and information.

The implementation of our three principles will require continuous monitoring. Interdisciplinary research including actuaries, legal scholars and social scientists should study how such measures affect people’s and insurers’ attitudes towards health-related information in an era in which data are becoming more abundant and accessible than ever before. Precision medicine research – with its emphasis on participants’ empowerment – offers new opportunities to access one’s health-related information. Yet, social groups having more to fear in terms of discrimination may be discouraged from participating. This would create a socio-economic gradient on access to medically relevant information, therefore potentially increasing health inequalities (Hausermann et al., 2018). Analogously, voluntary disclosure may induce insurers to increase premiums for non-disclosers, thus putting less affluent people at a further disadvantage. Furthermore, premium-based incentives may be more attractive to certain social groups, namely to those that, possibly due to their socio-economic conditions, are anyway better off in terms of both health prospects and financial capacity to afford coverage. Close monitoring of such phenomena is therefore part and parcel of the framework we have presented so far.

Another set of observations is in order. Our framework seems to put on a par interests that rest on indeed very different moral grounds. In particular, it may be argued that the public value of data as enablers of scientific progress and the right of individuals to be protected from discrimination are intrinsically more important than the commercial interests of private insurers in avoiding adverse selection. We certainly concede that. However, the integrity of the insurance market is a pre requisite for individuals, families and societies at large to be able to cope with a variety of risks. In other words, while private insurers are moved by profit-based motives, their activity fulfils a social function, that of supporting those harmed when certain risks materialize. For this mechanism to work, market incentives – appropriately balanced with key societal needs – have to be in place.

Precisely because in this long-debated context trade-offs are inevitable, any viable balancing act should recognize that all parties have at least some legitimate interests and that they are linked by intricate forms of interdependency.

Trustworthiness, openness and evidence, as we said, may not provide a definitive solution to the problems resulting from the increased availability of health-related predictive information, namely: the risk of reduced participation in research; the risk of adverse selection; and the risk of discrimination. Yet, they can represent a roadmap for the development of appropriate governance mechanisms to cope with such risks.

Conclusions

As inhabitants of the information society, we will have to become more accustomed to the risk of information asymmetry in virtually every sphere of social interaction. This problem is likely to transform the way in which each of us exchanges information with a variety of other social actors. As personal data is used to produce predictive health-related information, it acquires tangible economic value – for instance in the context of private insurance. Such economic value, as we have seen, may enter in tension with the value of data as a fundamental asset for scientific research, as well as with the rights of protecting people from unauthorized or discriminatory uses of their personal information. As a consequence, these processes cannot be left ungoverned. Trustworthiness, openness and evidence, we have shown, can work together to promote a delicate balancing act between the different interests at stake, particularly in the emerging domain of precision medicine research.