Good data: with Ben Goldacre

Best known for his ‘Bad Science’ column in ‘The Guardian’, Ben Goldacre started a new role last March at Oxford University’s Centre for Evidence Based Medicine.

He wants to use a new ‘evidence based medicine data lab’ to build a range of NHS data tools, with help from a £500,000 funding grant.

“I have been reasonably successful with academic funding, but it seems to me there’s a structural problem with the way we fund activity in healthcare,” he says.

“There are funds in place for very straightforward, obvious academic work where the end product is an academic paper in a journal read by a small number of people. But when you try to use your academic skills to put data into action in the real world there aren’t really any funding streams available.

“That’s why we urgently need a fund, just the same as for academic research; a competitive fund where coders collaborating with academics can apply for funding to build functioning live data tools for the NHS – as there’s no point sharing data if it sits there unused.”

OpenPrescribing.net

His most recent project, OpenPrescribing.net, is an example of the impact that a relatively a small investment – £50,000 in this case – can have in terms of making publicly available data truly accessible to people.

It is a new tool that enables users to quickly anaylse the huge GP datasets published by the Health and Social Care Information Centre. Goldacre says it is a great example of what can be achieved when academics and coders work together on innovate, real-life data tools targeted at the NHS.

He argues that tools like OpenPrescribing save people a lot of wasted time trying to use spreadsheets and other cumbersome methods to produce reports.

“I believe humans should do what humans are good at and we should automate anything that can be automated. That’s what OpenPrescribing sets out to do,” he tells Digital Health.

The push for open data

The regular publication of government-held data, for free, and in machine readable form, is something open data campaigners and media supporters such as The Guardian have been pushing for a number of years.

Recent governments have embraced the idea, launching the open data website data.gov.uk in 2010 and encouraging the release of data sets as a way of boosting transparency and growing the economy.

On the one hand, the idea is that open data can be used by campaigners, public bodies and charities to approach problems in new ways while, on the other, it can be used by entrepreneurs and small companies to deliver new tools and to charge for support and services.

In the NHS, the transparency agenda has resulted in the publication of outcomes data for physicians and ratings data for all GPs in England, among other examples.

But the open data movement has failed to gain significant traction (despite, or perhaps because of, the existence of well-established companies delivering proprietary business intelligence, risk management, data and analysis services to commissioning bodies).

The Health and Social Care Information Centre has also been releasing GP practice-level prescribing data every month since September 2011, in line with the government’s open data agenda.

While this is all very important and a good first step, Goldacre argues that if you do not have an ecosystem of tools around using that data, it is ultimately “meaningless.”

The GP prescribing dataset, for example, amounts to 100GB and there are around four million rows for each file, making analysing it time consuming and challenging. Hence the need for a system to produce useful tools for using this and similar data releases.

Open data open source

Goldacre explains how the open source agile development approach taken by the developers of OpenPrescribing has already paid dividends by getting other people to invest their time, free of charge.

When first launched, there were complaints from people that it was missing data from Wales. He replied to this criticism on Twitter, saying it would take about 20 days of coding to import the data, which was available but in a slightly different format to the data in England, and this work was not funded.

In response, a coder used his own time to rework the code and the methods necessary to import the Welsh data. “That’s a really interesting illustration of the non-commercial approach that can be taken using open data and software,” Goldacre explains.

The creation of a prototype also got a reaction in Scotland, where tweeters questioned why similar data was not available for Scottish GPs. Goldacre says he had approached NHS Scotland about releasing their prescribing data and was told it would not be done.

However, when a Scottish journalist got a hold of the story and ran it in the Sunday Express, NHS Scotland “saw the writing on the wall” and announced that the data would be released by this April.

“That’s the value of it being an agile produced working tool: getting it out there, getting it used and getting feedback about how to drive forward quality,” Goldacre argues.

“One of the really valuable things about developing an open access tool that anyone can use is that you can empower all of these people who aren’t in a formal senior role in a CCG or CSU [commissioning support unit] but are really interested in engaging with data.

“Those people are our best allies for getting better use of data in healthcare.”