It's the data, stupid

Unlocking innovation in the NHS

At a conference of NHS research and development managers in Birmingham this week I talked about what I had learned as a patient and a journalist about innovation in the health service. Then from the audience came a stunning question: “So far the millions we’ve invested in research have produced little in the way of innovative new treatment for patients,” said a manager. “How would you change that?”

Oh fine, I thought, no biggie - just deliver the recipe for an innovative and forward looking NHS that countless Health Secretaries and their advisors have failed to come up with in decades. You won’t be surprised to hear that I didn’t have a simple solution either. But after attending a workshop on the challenge of digital transformation before my session, I had an idea that part of the answer must lie in data.

A dozen or so people running research departments from across the UK had gathered to hear the results of a survey which showed just how wide the gulf is between the those trusts forging ahead with introducing AI into their systems and those which were only just getting round to moving to electronic patient records.

It was a reminder that, far from being a monolithic bureaucracy, the NHS is made up of over 200 trusts, in effect city states, each with their own priorities. That means exploiting the health service’s unique asset, its vast pool of data, is beset with difficulties, with trusts either unwilling to share or frightened of falling foul of data protection laws.

The managers were well aware that data had to be at the heart of their efforts to innovate. But setting up new AI projects was taking up to a year when, said the man leading the discussion, it should be done in one to two months. A major part of the blame for that related to an acronym new to me, IG or Information Governance. In other words jumping through all sorts of regulatory hoops to convince nervous hospital executives that the data would be safe from prying eyes. (One R&D manager admitted they were loathe to use the term AI in funding bids because it was scary to risk averse executives, instead talking of “machine learning” or even “applied statistics”.)

There was rueful laughter when discussion turned to the salaries trusts were able to offer the data scientists who were now essential to most AI projects but who could earn six figure sums in the private sector even straight out of university: “We’ve got this huge data just stood there,” said one woman. “Can’t do anything because I can’t recruit.” And she went on: “Our hospital isn't very sexy, very exciting. If you're in this world, why on earth would you come to work with us? “

But it wasn’t all doom and gloom. One successful clinician and entrepreneur with a number of AI projects under his belt said partnerships with the private sector were the way forward:

“It doesn't say anywhere that the NHS has to do all this development work. What the NHS has is a lot of data on which development work can take place. And I think in many cases, our role is to make that data available in an understandable way, which can then be worked on by an external partner who does have the wherewithal to attract those necessary skills.”

I think such partnerships are going to be essential in bringing an innovation culture to the health service, even though there will be understandable caution about giving tech companies - and in particular the American giants - access to our valuable data.

But if it’s to understand what it is buying from the private sector, the NHS does need to have some in-house data science skills. As I listened, I thought back to the early successes of GDS, the Government Digital Service, a kind of guerrilla band of software developers trying to change a culture where mandarins handed billion pound contracts to tech giants for IT systems they did not understand and which turned out to be duds. Perhaps the NHS needs something similar to shake things up.

Certainly, if the people I met in Birmingham are anything to go by, there is no lack of appetite for change, and in a few places, rapid innovation is happening. The trick is going to be making sure the pioneers are able to spread what they learn across the health service but that means that every one of those city states has got to be open to new ideas. “Not invemted here” or “but we’ve always done it like this” won’t cut it in the AI era.