Google can reportedly see if UK patients are HIV positive under a data-sharing agreement with the NHS

The extent of a partnership between Google DeepMind and the NHS was understated, according to a data-sharing agreement obtained by New Scientist.

The data-sharing agreement between DeepMind and the Royal Free NHS Trust reportedly gives DeepMind access to a "wide range" of healthcare data on the 1.6 million patients who use Barnet, Chase Farm, and the Royal Free hospitals in London annually.

DeepMind will reportedly be able to see, for example, whether people are HIV-positive, or whether they have overdosed on drugs, or had an abortion. Hospitals will also share the results of certain pathology and radiology tests, according to New Scientist. Google DeepMind was not immediately available to comment.

DeepMind announced its partnership with the NHS back in February, saying it had created an app called Streams that could help hospital staff to monitor patients with kidney disease. However, the data-sharing agreement reportedly implies that DeepMind could also be looking at other illnesses.

The agreement is said to state that DeepMind, which was acquired by Google for a reported £400 million in 2014, will have access to patient data going back five years.

Sam Smith, who runs health data group MedConfidential, told New Scientist that this is the first we've heard about DeepMind getting access to historical medical records. "This is not just about kidney function," he said. "They're getting the full data."

The agreement is said to note that Google cannot use the data in any other parts of its business. The data itself will reportedly be stored in the UK by a third-party contractor, not at DeepMind's office in King's Cross, London. DeepMind will have to delete the data when the agreement expires at the end of September 2017, according to New Scientist.

Google reportedly said that there is no separate dataset for people with kidney conditions, meaning DeepMind needs to access all of the data in order for its Streams app to work.

DeepMind receives the NHS data under strict information governance rules, just like the other 1,500 recorded third-party organisations who have certification to process patient data at present.

Nathan Benaich, an investor at venture capital firm Playfair Capital, told Business Insider: "I don't think the risk profile of DeepMind Health should be any different from the likes IBM or GE Healthcare that have been in the business for longer."

DeepMind insists that patient data will be kept completely separate from Google accounts, products, and services. But New Scientist states that some people are still likely to be concerned, highlighting that Google is being given logs of day-to-day hospital activity, including records of the location and status of patients, in addition to who visits them and when.

New DeepMind platform

The data-sharing document is also said to show that DeepMind is developing a platform called Patient Rescue, which will reportedly provide data analytics services to NHS hospital trusts. The document reportedly says that Patient Rescue will use data from hospitals to build new tools that can carry out real-time analysis of clinical data and support diagnostic decisions.

The data that DeepMind is collecting will allow it to make predictions about any disease it wants, according to Smith. "What DeepMind is trying to do is build a generic algorithm that can do this for anything - anything you can do a test for," he said.

Ross Anderson, an academic at the University of Cambridge, questioned whether Google should have so much control over health analytics. " If Google gets a monopoly on providing some kind of service to the NHS it will burn the NHS," he told New Scientist.

Google doesn't currently have any plans to commercialise DeepMind's work with Royal Free, according to New Scientist.

A panel comprised of government tech leaders and healthcare experts is set to scrutinise the work of Google DeepMind's healthcare team at an upcoming meeting, which could be held within the next month.