Ban the use of genomic data by UK insurers

Last week, AstraZeneca unveiled MILTON, an artificial intelligence (AI) system capable – among other things – of augmenting techniques for understanding the relationship between a person's DNA and their likelihood of developing particular diseases (see BioNews 1256).

As well as being impressive in its own right, MILTON is emblematic of two broader trends. First, it is the latest in a series of AI-driven advances in our ability to use genomic data to make predictions about people's future health. Second, as research drawing on data from the UK Biobank and undertaken by an Anglo-Swedish firm, it is testament to the UK's interest in these techniques and their application to healthcare.

On the same day that MILTON was revealed to the world, the Ada Lovelace Institute and the Nuffield Council of Bioethics published a report assessing the implications of these wider trends. Predicting: The Future of Health? draws on two years of research into 'AI-powered genomic health prediction', exploring what this emerging capacity might mean for the NHS and broader society, and what needs to be done to ensure the impacts are benign.

Our research shows that the potential benefits of advances in genomic health prediction could be substantial. But as a capability promising to provide insight into individuals' likelihood of developing particular diseases over their lifetime, genomic health prediction also poses several risks, as well as numerous challenges for policymakers.

Many of these challenges – such as those around the possibility of genomic predictions undermining the social solidarity undergirding the NHS, and about the creation of novel forms of discrimination not caught by current equalities laws – are complex, and unlikely ever to be resolved perfectly.

Strikingly, though, one of the biggest risks associated with genomic health prediction is both far simpler and far more tractable – the use of genomic health predictions by health insurers.

A common theme among the experts, practitioners and members of the public with whom we engaged was concern that health insurers might potentially use genomic health predictions to inform the price of premiums, with higher prices for those at higher risk of disease, and lower prices for those at reduced risk.

This practice could be problematic, for several reasons. First, the idea that those most in need of health coverage should end up having to pay more for that coverage sits uneasily with the ideal of society pooling the risk of ill health.

Second, while it is true that the existence of the NHS means that nobody would end up priced out of healthcare as a result of the practice, the increased ability of the private sector to entice those at lowest genomic risk of disease with lower premiums could create broader, more structural problems. As our report explains, a situation in which those with higher disease risk are more likely to use the NHS (and those with lower disease risk more likely to go private) would likely mean that the average NHS patient would require more care than otherwise. Unless NHS funding were adjusted upwards to address this potential change, the uneven distribution of disease risk between the NHS and private healthcare could place substantial strain on the NHS.

Finally, and perhaps most immediately, people's fears about the possibility of genomic data they share being used by insurers are a real deterrent to participation in genomic medicine and research, and are likely to most deter the groups who are already underrepresented in genomic datasets. For a country like the UK, whose ambitions for genomic science and medicine will require a dramatic expansion in public participation in genomic testing, these concerns need to be taken especially seriously.

In recognition of these problems, many other countries – including the USA, France, Germany and Canada – have enacted laws that ban insurers form taking into account the results of genomic tests.

By contrast, the UK has to date refrained from introducing any such legislation, opting instead to rely on a voluntary code of practice. Under this arrangement, the UK insurance industry commits to refrain from using genomic testing in all but a very limited set of circumstances (see BioNews 1202).

While the Code is often cited as de facto equivalent to the laws that exist in other countries, it fails to provide a comparable level of protection. There is no legal sanction for a UK insurance company that fails to comply with the Code, and – as stated in government commentary – its restrictions may be relaxed in the event of the insurance industry having significant reason to make use of genomic predictions. The non-binding, open-ended nature of the Code means it cannot provide the public with genuine reassurance about how their information might be used.

The good news is that addressing this problem does not need to be difficult. As we argue in our report, fixing this issue could be a simple as placing the existing Code onto a statutory footing, passing legislation making compliance with the Code a legal requirement for all UK insurers. In so doing, government could close one of the biggest gaps in our legal protections against genomic discrimination, provide genuine reassurance to those who are considering sharing their genomic data, and address what could become one of the biggest barriers to the advancement of genomic science in the UK.

Advances in genomic health prediction like MILTON are set to pose a multitude of complex, difficult questions for UK policymakers. Fortunately, the issue of genomics and health insurance isn't one of them.