In November 2015, the Royal Free NHS Foundation Trust transferred over 1.6 million identifiable patient records to DeepMind, an artificial intelligence subsidiary of Google. Earlier that year, the Trust had privately signed an agreement commissioning DeepMind to develop an early warning system to detect Acute Kidney Injury (AKI). The resulting smartphone application, called Streams, is now in clinical use across the Royal Free and will soon be rolled out to other NHS Foundation Trusts.
When news of the deal broke, privacy advocates expressed astonishment at the private agreement. They were outraged by the Trust’s willingness to share confidential patient data without consent and other necessary protections. The Information Commissioner’s Office subsequently launched a year-long investigation into the deal, concluding in July 2017 that the Royal Free had violated the 1998 Data Protection Act. Despite the ruling, DeepMind has since partnered with multiple NHS Trusts. The company’s CEO, Mustafa Suleyman, has declared his ambitions to expand into the NHS and develop a digital platform that could support artificial intelligence technologies in the future.
Criticism of the DeepMind-Royal Free partnership has centred around privacy rights and data protection issues. While privacy concerns are important, our research suggests that a narrow focus on privacy has thwarted attention to related human rights concerns, particularly around the right to the highest attainable standard of health.
For over 20 years, public-private partnerships (PPFS) have been touted as a mutually beneficial financing mechanism. However, corporations often have a legal, fiduciary duty to maximise shareholder returns; DeepMind face mounting pressures to deliver profitsin the wake of looming shareholder ‘earnout’ payments next year. Profit incentives can come at the expense of public interests, particularly when CEOs are incentivised to link their own pay and bonuses to these returns.
The assumption that PPPs provide value for money also lacks substantiation. Indeed, research carried out at Manchester and Essex Universities has shown that Private-Financing Initiatives (PFIs), a type of public-public partnership, have been financially burdensome for UK taxpayers. In 2016, PFI debt totalled £300 billion for projects worth £55 billion. Today, DeepMind provide Streams to the Royal Free for free, though they will eventually sell the technology back to NHS Trusts at a currently unspecified price. This ex post facto pricing provides no guarantee that taxpayer money will be well-spent and risks disadvantaging less prosperous NHS trusts and their patients.
The UK has clear right to health obligations under the International Covenant on Economic, Social and Cultural Rights (ICESCR). This includes legal obligations to provide an equitable healthcare system and to use the maximum of their available resources to fulfil the right to health. Partnerships like DeepMind-NHS threaten advancement of the right to health by implementing a business model that could exacerbate existing inequalities in the NHS system and prioritise corporate profits at the expense of public interest.
The DeepMind-NHS case illustrates that large tech companies are playing a greater role in accessing NHS patient data – a commercial win that may impinge on the delivery of equitable care. The benefits and risks of companies such as Alphabet moving into healthcare systems drained and fragmented by austerity measures are still unclear. Our research is centred on clarifying the perceived perils and benefits of data-sharing partnerships by drawing novel comparisons to earlier private-public partnerships, including the PFI model adopted under Labour and Conservative governments since the 1990s.
Under data-sharing partnerships, public service providers and profit-driven corporations deal in patient data. Just like economically burdensome PFIs, these new partnerships could jeopardise the government’s ability to fulfil its right to health obligations under international law. Behind the ‘value for money’ rhetoric championed by advocates of a business-centric approach to health care, our research asks: does DeepMind actually benefit more from patient data than patients do themselves? Does ‘value for money’ conceal the problem of ‘value for monopolists,’ and if so, what are the human rights implications of entrenching corporate wealth concentration in the name of ‘sharing’ agreements?
Disclaimer: The views expressed herein are the author(s) alone