A couple of years ago I attended a lecture on anxiety disorders. It was given by an eminent academic psychiatrist. It was fascinating, but I was struck by a repetition as he outlined the anxiety disorders. 'We sometimes use treatment X or treatment Y, for this condition, but we don’t have much good evidence as to which is most effective.'

At medical school we were taught the importance of evidence-based practice, but in the real world, the Randomised Control Trial (RCT) just hasn’t always been done. When it has, it often doesn’t apply to my patient, with their particular comorbidities, lifestyle, genetics, etc. In the meantime, we still have to treat patients as best we can, so we make educated guesses.

I’d stopped listening to the lecture and was daydreaming about all of those educated guesses that must be happening around the world every day. My mind slipped back to a time before I studied medicine, when I was a software engineer. If all of those patient characteristics and all of those single participant trials were captured in a huge database, then surely we could perform retrospective observational studies into the real world effectiveness of any treatment for any group of patients, at the click of a button? And was an Electronic Health Record (EHR) not just such a database?

Of course, this is not an original idea and its methodology and implementation would be nowhere near as simple as I have suggested. But the idea of a health care system that really learns from every patient who it treats, a so called Learning Healthcare System, is a real possibility and early examples are already up and running.

Optum Labs, in Boston, are conducting observational studies on a database containing details of 150m patients. They can do hundreds of observational studies for the price of one RCT.

The US Food and Drug Administration have established a network of almost 180m patients. It can pull data about treatment side effects and interactions from patient records, allowing them to quickly identify side effects from new treatments.

In the UK, the QSurveillance network tracked the last flu pandemic using GP electronic record entries, providing real-time updates on spread.

Collecting outcome data on patients can help us manage their care, but in aggregate, it can help identify ‘positive deviants’ - really good providers. Everyone else can then learn from them. The International Consortium for Health Outcomes Measurement has begun the process of defining outcomes measurement sets for every major condition. They hope to cover half of the disease burden in developed countries by 2017.

Computer algorithms can predict the likelihood that particular events will occur, based on many factors that have been recorded in EHRs. For example, Geisinger Health System employ a team of modellers and have had success in predicting and mitigating costly and risky events, such as patients not attending appointments.

OpenClinical.net, a UK based collaborative, have developed technology to create machine-readable guidelines. Organisations such as NICE could eventually distribute their guidelines in this form, so that hospital decision support systems could monitor quality of care and suggest pathway optimisations that clinicians may not have been aware of.

Many of these technologies are already in place; it is the social and cultural challenges of embedding them in a way that benefits patients and staff that will be hardest to overcome. Here too there has been much progress in the fields of behaviour change and implementation science.

These and many other issues are discussed in our new report on The Potential of Learning Healthcare Systems. We outline systems that are already available for providers to implement. We also address the major policy implications that need to be considered now, including a new ethics framework that might oblige patients to share their data and the implications for workforce planning, cost, quality and value in health care.

In the month that CCGs begin to prepare their Digital Roadmaps, NHS England discusses its Digital Maturity Index, Jeremy Hunt launches a landmark review of the NHS Digital Future and DH submit a £5.6bn funding request for technology, the future may be closer than you think. Whether we repeat past mistakes or use technology to improve our care for patients will depend on the decisions that we are about to make at every level in the health care system. Join the debate.


Tom is a doctor, academic and ex-software engineer. He has written reports on health care policy issues for a variety of organisations including the Royal College of Psychiatrists, Centre for Mental Health and NHS England, and for the last year, he has been investigating Learning Healthcare Systems, through a project funded by the Health Foundation.

You might also like...


Webinar: NHS reform – what do the proposals mean for general practice?


How the government's proposed NHS reforms will affect primary care and why general practice should...


Is it time for another Wanless Review?


Episode 9. Given the pandemic, do we need another Wanless Review to assess the NHS's future funding...


Mind the gaps: adult social care workforce policy during the pandemic


Has national support for adult social care staff in England during the pandemic been adequate?...

Kjell-bubble-diagramArtboard 101 copy

Get social

NHS Test and Trace celebrated its first birthday last month. We've tracked its up and downs and reported on its su… https://t.co/Iro6fcRdgH

Follow us on Twitter
Kjell-bubble-diagramArtboard 101

Work with us

We look for talented and passionate individuals as everyone at the Health Foundation has an important role to play.

View current vacancies
Artboard 101 copy 2

The Q community

Q is an initiative connecting people with improvement expertise across the UK.

Find out more