{"id":14834,"date":"2023-01-30T15:07:00","date_gmt":"2023-01-30T08:07:00","guid":{"rendered":"https:\/\/news.adeptasiaconsulting.com\/?p=14834"},"modified":"2022-06-19T07:13:12","modified_gmt":"2022-06-19T00:13:12","slug":"google-completes-controversial-takeover-of-deepmind-health","status":"publish","type":"post","link":"https:\/\/news.adeptasiaconsulting.com\/google-completes-controversial-takeover-of-deepmind-health\/","title":{"rendered":"Google completes controversial takeover of DeepMind Health"},"content":{"rendered":"
\n

Google <\/a> has completed a controversial take-over of the health division of its UK AI acquisition<\/a>, DeepMind.<\/p>\n

The personnel move had been delayed as National Health Service <\/a> (NHS) trusts considered whether to shift their existing DeepMind contracts \u2014 some for a clinical task management app, others involving predictive health AI research \u2014 to Google.<\/p>\n

In a blog post <\/a>yesterday Dr Dominic King, formerly of DeepMind (and the NHS), now UK site lead at Google Health, confirmed the transfer, writing: \u201cIt\u2019s clear that a transition like this<\/a> takes time. Health data is sensitive, and we gave proper time and care to make sure that we had the full consent and cooperation of our partners. This included giving them the time to ask questions and fully understand our plans and to choose whether to continue our partnerships. As has always been the case, our partners are in full control of all patient data and we will only use patient data to help improve care, under their oversight and instructions.\u201d<\/p>\n

The Royal Free NHS Trust<\/a>, Taunton & Somerset NHS Foundation Trust<\/a>, Imperial College Healthcare NHS Trust<\/a>, Moorfields Eye Hospital NHS Foundation Trust<\/a> and University College London Hospitals NHS Foundation Trust<\/a> all put out statements yesterday confirming they have moved their contractual arrangements to Google.<\/p>\n

In the case of the Royal Free, patients\u2019 Streams data is moving to the Google Cloud Platform infrastructure to support expanding use of the app which surfaces alerts for a kidney condition to another of its hospitals (Barnet Hospital).<\/p>\n

One NHS trust, Yeovil District Hospital NHS Foundation Trust, has not signed a new contract \u2014 and says it had never deployed Streams, suggesting it had not found a satisfactory way to integrate the app with its existing ways of working \u2014 instead taking the decision to terminate the arrangement. Though it\u2019s leaving the door open to future health service provision from Google.<\/p>\n

A spokeswoman for Yeovil hospital sent us this statement:<\/p>\n

We began our relationship with DeepMind in 2017 and since then have been determining what part the Streams application could play in clinical decision making here at Yeovil Hospital.<\/p>\n

The app was never operationalised, and no patient data was processed.<\/p>\n

What\u2019s key for us as a hospital, when it comes to considering the implementation of any new piece of technology, is whether it improves the effectiveness and safety of patient care and how it tessellates with existing ways of working. Working with the DeepMind <\/a> team, we found that Streams is not necessary for our organisation at the current time.<\/p>\n

Whilst our contractual relationship has ended, we will remain an anchor partner of Google Health so will continue to be part of conversations about emerging technology which may be of benefit to our patients and our clinician in the future.<\/p><\/blockquote>\n

The hand-off of DeepMind Health to Google, which was announced just over a year ago<\/a>, means the tech giant is now directly providing software services to a number of NHS trusts that had signed contracts with DeepMind for Streams; as well as taking over several AI research partnerships that involve the use of NHS patients\u2019 data to try to develop predictive diagnostic models using AI technology.<\/p>\n

DeepMind \u2014 which kicked off its health efforts by signing an agreement with the Royal Free NHS Trust in 2015, going on to publicly announce the health division in spring 2016 \u2014 said last year<\/a> its future focus would be as a \u201cresearch organisation\u201d.<\/p>\n

As recently as this July<\/a> DeepMind was also touting a predictive healthcare research \u201cbreakthrough\u201d \u2014 announcing it had trained a deep learning model for continuously predicting the future likelihood of a patient developing a life-threatening condition called acute kidney injury. (Though the AI is trained on heavily gender-skewed data from the US department of Veteran Affairs.)<\/p>\n

Yet it\u2019s now become clear that it\u2019s handed off several of its key NHS research partnerships to Google Health as part of the Streams transfer.<\/p>\n

In its statement<\/a> about the move yesterday, UCLH writes that \u201cit was proposed\u201d that its DeepMind research partnership \u2014 which is related to radiotherapy treatment for patients with head and neck cancer \u2014 be transferred to Google Health, saying this will enable it to \u201cmake use of Google\u2019s scale and experience to deliver potential breakthroughs to patients more rapidly\u201d.<\/p>\n

\u201cWe will retain control over the anonymised data and remain responsible for deciding how it is used,\u201d it adds. \u201cThe anonymised data is encrypted and only accessible to a limited number of researchers who are working on this project with UCLH\u2019s permission. Access to the data will only be granted for officially approved research purposes and will be automatically audited and logged.\u201d<\/p>\n

It\u2019s worth pointing out that the notion of \u201canonymised\u201d high dimension health data should be treated with a healthy degree of scepticism<\/a> \u2014 given the risk of re-identification.<\/p>\n

Moorfields <\/a>also identifies Google\u2019s \u201cresources\u201d as the incentive for agreeing for its eye-scan related research partnership to be handed off, writing: \u201cThis updated partnership will allow us to draw on Google\u2019s resources and expertise to extend the benefits of innovations that AI offers to more of our clinicians and patients.\u201d<\/p>\n

Quite where this leaves DeepMind\u2019s ambitions to \u201clead the way in fundamental research applying AI to important science and medical research questions, in collaboration with academic partners, to accelerate scientific progress for the benefit of everyone\u201d, as it put it last year<\/a> \u2014 when it characterized the hand-off to Google Health as all about \u2018scaling Streams\u2019 \u2014 remains to be seen.<\/p>\n

We\u2019ve reached out to DeepMind for comment on that. Update: <\/strong>The company told us it is now purely focused on fundamental science research areas, which includes medical questions, rather than applied healthcare \u2014 citing its research into protein folding<\/a> as an example. Whereas it said Moorfields and UCLH are interested in the translation of research into applicable technologies beyond fundamental research, making Google Health a better fit.<\/p>\n

Co-founder Mustafa Suleyman, who\u2019s been taking a leave of absence from the company, tweeted yesterday to congratulate the Google Health team.<\/p>\n

\n
\n

When we started DeepMind Health 3 years ago it was because we believed good research and smart software could make a difference to patients & nurses & doctors. Proud to be part of this journey. Huge progress delivered already, and so much more to come for this incredible team. https:\/\/t.co\/zynBrrlgUc<\/a><\/p>\n

\u2014 Mustafa Suleyman (@mustafasuleymn) September 18, 2019<\/a><\/p><\/blockquote>\n

DeepMind\u2019s NHS research contracts also transferring to Google Health suggests the tech giants wants zero separation between core AI health research and the means of application, using its own cloud infrastructure, of any promising models it\u2019s able to train off of patient data and commercialize by selling to the same healthcare services providers as apps and services.<\/p>\n

You could say Google is seeking to bundle access to the high resolution patient data that\u2019s essential for developing health AIs with the provision of commercial digital healthcare services it hopes to sell hospitals down the line, all funnelled through the same Google cloud infrastructure.<\/p>\n

As we reported at the time, the hand-off of DeepMind Health to Google is controversial<\/a>.<\/p>\n

Firstly because the trust that partnered with DeepMind in 2015 to develop Streams was later found by the UK\u2019s data protection watchdog to have breached UK law. The ICO said there was no legal basis<\/a> for the Royal Free to have shared the medical records of ~1.6M patients with DeepMind during the app\u2019s development.<\/p>\n

Despite concerns being raised over the legal basis for sharing patients\u2019 data throughout 2016 and 2017 DeepMind continued inking NHS contracts for Streams \u2014 claiming at the time that patient data would never be handed to Google. Yet fast forward a couple of years and it\u2019s now literally sitting on the tech giant\u2019s servers.<\/p>\n

It\u2019s that U-turn that led the DeepMind to Google Health hand-off to be branded a trust demolition<\/a> by legal experts when the news was announced last year.<\/p>\n

This summer the UK\u2019s patient data watchdog, the National Data Guardian, released correspondence<\/a> between her office and the ICO which informed the latter\u2019s 2017 finding that Streams had breached data protection law \u2014 in which she articulates a clear regulatory position that the \u201creasonable expectations\u201d of patients must govern non-direct care uses for people\u2019s health data, rather than healthcare providers relying on doctors to decide whether they think the intended purpose for people\u2019s medical information is justified.<\/p>\n

The Google Health blog post<\/a> talks a lot about \u201cpatient care\u201d and \u201cpatient data\u201d but has nothing to say about patients\u2019 expectations of how their personal information should be used, with King writing that \u201cour partners are in full control of all patient data and we will only use patient data to help improve care, under their oversight and instructions\u201d.<\/p>\n

It was exactly such an ethical blindspot around the patient\u2019s perspective that led Royal Free doctors to override considerations about people\u2019s medical privacy in the rush to throw their lot in with Google-DeepMind and scramble for AI-fuelled predictive healthcare<\/a>.<\/p>\n

Patient consent was not sought for passing medical records then; nor have patients\u2019 views been consulted in the transfer of Streams contracts (and people\u2019s data) to Google now.<\/p>\n

And while \u2014 after it was faced with public outcry over the NHS data it was processing \u2014 DeepMind did go on to publish its contracts with NHS trusts (with some redactions), Google Health is not offering any such transparency on the replacement contracts that have been inked now. So it\u2019s not clear whether there have been any other changes to the terms. Patients have to take all that on trust.<\/p>\n

We reached out to the Royal Free Trust with questions about the new contract with Google but a spokeswoman just pointed us to the statement on its website \u2014 where it writes: \u201cAll migration and implementation will be completed to the highest standards of security and will be compliant with relevant data protection legislation and NHS information governance requirements.\u201d<\/p>\n

\u201cAs with all of our arrangements with third parties, the Royal Free London remains the data controller in relation to all personal data. This means we retain control over that personal data at all times and are responsible for deciding how that data is used for the benefit of patient care,\u201d it adds.<\/p>\n

In another reduction in transparency accompanying this hand-off from DeepMind to Google Health, an independent panel of reviewers that DeepMind appointed to oversee its work with the NHS in another bid to boost trust has been disbanded.<\/p>\n

\u201cAs we announced in November, that review structure \u2014 which worked for a UK entity primarily focused on finding and developing healthcare solutions with and for the NHS \u2014 is not the right structure for a global effort set to work across continents as well as different health services,\u201d King confirmed<\/a> yesterday.<\/p>\n

In its annual report last year<\/a> the panel had warned of the risk of DeepMind exerting \u201cexcessive monopoly power\u201d as a result of the data access and streaming infrastructure<\/a> bundled with provision of the Streams app. For DeepMind then read Google now.<\/p>\n

Independent experts raising concerns about monopoly power unsurprisingly doesn\u2019t align with Google\u2019s global ambitions in future healthcare provision.<\/p>\n

The last word from the independent reviewers is a Medium post<\/a> penned by former chair, professor Donal O\u2019Donoghue \u2014 who writes that he\u2019s \u201cdisappointed that the IR experiment did not have the time to run its course and I am sad to say goodbye to a project I\u2019ve found fascinating\u201d.<\/p>\n

\u201cThis was a fascinating exploration into how a new governance model could be applied to such an important area such as health,\u201d he adds. \u201cIt\u2019s hard to know how this would have developed over the years but\u2026 what is clear to me is that trust and transparency are of paramount importance in healthcare and I\u2019m keen to see how Google Health, and other providers, deliver this in the future.\u201d<\/p>\n

But with trust demolished and transparency reduced Google Health appears to have learnt exactly nothing from DeepMind\u2019s missteps.<\/p>\n

\n

Google gobbling DeepMind\u2019s health app might be the trust shock we need<\/a><\/p><\/blockquote>\n