Pharma needs better data, but there are obstacles

Originally published at Pharmafield

Data for the diary

There is incredible potential in advanced healthcare analytics. Imagine a scenario where we can intervene to save a patient, just before a heart attack occurs, because of predictive modelling with ‘big data’ and ‘machine learning’.

This kind of advance requires data. Not aggregated data in data tables, but rows upon rows, fields upon fields of clean, recent, suitably anonymised or pseudonymised (where data in identifiable fields are replaced with pseudonyms) data. Data with a longitudinal structure that can be linked to other datasets to add value – which may mean it must first be linked together before it is pseudonymised. Stay with me!

That’s why Google’s ‘AI’ venture ‘DeepMind’ worked so hard to secure a partnership at the Royal Free London NHS Foundation Trust late last year, to power the app for patients at risk of acute kidney injury. Access to vast amounts of granular, detailed healthcare data is exceptionally valuable.

The question that has repeatedly arisen from this kind of endeavour is, ‘do we trust these kinds of organisations with patient data?’ There’s clearly a perception problem. The care.data furore ended with the initiative closed down last year. Both DeepMind and the Royal Free came under fire for striking the deal, after concerns were raised about the sheer scale of data being analysed.

Worse still, NHS organisations across the country speak in hushed tones of ‘The Daily Mail problem’ – the dilemma that while releasing data may have demonstrable benefit to patients, there’s a fear of being taken to task by the national media for playing fast and loose with patient data, even for valuable projects, no matter how considered the approach.

A palpable lack of trust around industry accesses real world healthcare data for research presents a massive obstacle, both to innovation and proving the value of innovation within an outcomes-based framework. It means innovators and healthcare organisations have to dance around the issue, compromising on what data they work with, thereby stalling the future and the kind of advances we might all benefit from.

The point is that the way DeepMind – and others like it – manage NHS data will have an impact on us all. We’re fortunate in the UK to have access to datasets like the Clinical Practice Research Datalink, but the Systemic Anti-Cancer Therapy Dataset has already been interminably delayed and pharma has even more to lose if data access is restricted, especially now that NICE has introduced a £20 million limit on the cost of new drugs.

There have been many discussions about the need for good governance around data, with technology that ensures security and transparency such as ‘blockchain’ (the technology that powers Bitcoin). The next step, however, must be about building trust so patients understand the value of allowing their healthcare data to be used for research, even industry-led research.

Data analytics in life sciences offers huge promise and the consequences for innovation without it don’t bear thinking about.