Privacy must come before profit if we are to benefit from NHS data
The writer, a former head of the Number 10 Policy Unit, is a Harvard senior fellow
A new gold rush has begun and the prospectors are seeking our health data. The combination of genomics, machine learning and digital records offers the hope of discoveries which could benefit us all. But only if the data is comprehensive enough and citizens are willing. Britain’s NHS is a potential gold mine but it is also jealously guarded by voters, some of whose worries are justified.
The UK government has had to pause its plan to pool GP medical records from 55m patients after threats of legal action. It was one thing to ask citizens to share medical records in the name of science. It was quite another to add, a little too breathlessly, that the UK will make billions out of selling that information to companies. The use and abuse of NHS data is a fraught enough subject without government conflating the issues of privacy and profit.
Data saves lives. We’ve known that ever since Dr John Snow discovered that cholera was waterborne. Machine learning has accelerated what can be achieved. Last year, Oxford researchers used NHS GP data to identify dexamethasone as highly effective in reducing deaths from Covid-19, a discovery which has saved 22,000 lives in the UK and about 1m more worldwide. In Israel, doctors have spotted patterns in electronic health records which should aid the earlier diagnosis of colorectal cancer.
The potential to do good is enormous. But insiders tend to laugh hollowly at the routine description of NHS data as a “treasure trove”. Outside, Dr Google might be salivating at the thought of 55m patient records, each with a unique NHS number, estimated by EY to be worth nearly £10bn a year. But inside, all is chaos and fragmentation. Around one in four NHS hospitals still run entirely on paper records; we can’t judge things such as whether a GP is good at diagnosing; and it sometimes feels like ours is the only healthcare system where winter comes as an annual shock. We may worry that Tesco knows too much about our shopping habits but at least it can match demand to supply.
This makes it all the more imperative that we digitise the information and use it wisely. In 2014, a project called care.data tried to link all NHS data about all patients together. But it foundered on worries about confidentiality and workability. The public accounts committee claimed the contractor had “taken advantage” of the NHS. Now, the same worries are resurfacing.
Some concerns seem overblown. The fear that health insurers might change your premium if they knew your full medical history ignores the fact that they already insist on exactly that. Concerns about Big Pharma can also miss that the process of licensing new drugs often rightly involves studies on large numbers of patient records.
But the public is right to distrust glib statements about privacy. We are promised that our data will be “pseudonymised”, with names and addresses erased. But data analysts say that this doesn’t guarantee protection. A stranger could still find your unique health record if they knew your birth year and if you’d tweeted about being treated in a particular place for cancer, for example. They might then find something you didn’t want to tweet: a depressive episode or a miscarriage.
The answer, surely, is for the NHS to retain the data centrally and give access to researchers only under strict conditions. Rather than letting others download patient records, they could be required to carry out their analyses within the central platform. Making analysts come to the data is commonplace elsewhere: that’s what has happened for two decades with the population census, which can only be accessed through the Office for National Statistics. Why should health be any different?
Amid the hysteria, a sophisticated analytics platform called OpenSAFELY is quietly showing what could be done. Pioneered by Dr Ben Goldacre and others, OpenSAFELY hosts hundreds of open-source research endeavours where researchers write their analyses using dummy versions of NHS data, before their code is executed against the real thing. No outside researcher actually touches the real patient record. Once each analysis is completed, the codes are published. This lets others build on the analysis or — crucially — point out mistakes which could have skewed the outcome.
Two weeks ago, OpenSAFELY analysis showed that fewer black people than white people have been vaccinated against Covid-19 in every age group. That’s a powerful finding for policymakers.
The government now has an opportunity to build professional and public trust. It should declare, unambiguously, that it will create a central mechanism with patient protections along these lines. We all want to see the NHS benefit from investment, but privacy concerns must come before profit. If a sceptical public chooses to opt out, the enormous potential to do good will be lost, perhaps forever.
The problem is partly that we are trying to run and walk at the same time. The NHS urgently needs to do the prosaic work of cleaning up reams of data. It needs to educate its managers in how to use information wisely, find ways to protect it properly and then figure out how best to share profits.
No one doubts the size of the prize. If we can link genomic data to health records at scale, we might spot patterns which could let us vanquish everything from diabetes to Parkinson’s. But no one wants their health information splattered around. If that happens, the new gold rush will be halted before it has even begun.