Sarah Wilkinson, Chief Executive of NHS Digital, writes exclusively for Hospital Times outlining how the data revolution could, and should, transform healthcare across the NHS. 

The UK has an extraordinary, and internationally unique, resource at its disposal: the data within the NHS. 

As a single system that serves 65 million people from their birth and through all the health-impacting events of their lives, the NHS has rich longitudinal health records. These records provide a comprehensive summary of an individual patient’s clinical history, as opposed to encounter-based or provider-based records. 

“The more people who use NHS data, the better.” Sarah Wilkinson, Chief Executive of NHS Digital

There is also a national culture of goodwill and support for the NHS and for medical research. The UK Biobank alone has 500,000 volunteer participants. Last year, the National Institute for Health Research (NIHR) reported that 875,250 people took part in clinical research studies across England, the equivalent of 2,383 per day. 

Increasing access 

The more people who use NHS data, the better. When patients have access to their own data, they have an increased understanding of their conditions and hence are better able to manage them. 

When clinicians can see connected data from multiple sources they are able to customise and optimise care for a patient. When research communities analyse this data, they identify patterns in diseases and treatments for particular populations. This leads to the optimisation of clinical pathways and the development of new drugs and treatments. 

It is also true that the more people who use this data, the richer the data becomes. When patients have access to their data, they might see inaccuracies and report them. The increasing digitisation of the health system allows data validation at the point of entry and throughout the lifecycle of a patient record. But there are significant challenges in enabling the effective use of this data. 

Joining the dots 

In recent years, digitisation of the NHS, particularly in secondary care, has occurred through independent local selection and deployment of systems and technologies. This has had minimal coordination of approach or consistency of data or technical standards. The result is a hugely heterogenous technology environment, characterised by minimal and poorly formed points of integration between systems, services and locations. When this data is brought together, in NHS Digital’s central national repositories or elsewhere, much work is necessary to ensure the data is consistently described and of equally high quality. 

As far as possible, simple algorithms are used to identify quality issues by comparison with baselines or historical trends, and to predict and propose corrections. In some cases, human curators with clinical expertise will have a deep understanding of data representations of diseases. They can review records manually and work with clinicians, coders and administrators across the NHS to correct and complete those records manually. Increasingly, artificial intelligence (AI) is used to carry out complex quality assessment work. But there are considerable challenges with this approach. 

Speaking a common language 

There are often multiple interactions between the many diverse events that take place over the course of a patient’s life. However, each patient has a unique profile. When specific disease characteristics are considered along with genetic profiles, exposure and lifestyle risk factors, demography and treatment and medication histories, it becomes difficult to identify erroneous outcomes from analytical processes, which are an important aspect of honing the algorithms. 

Even quality systems have inherent limitations, as human curation is expensive and depends on rare talent. While sophisticated machine processing is relatively new, and there are many complexities to be overcome in order to make it mainstream, the UK must invest and foster these approaches. 

Increased standardisation at the point of data collection would, of course, be transformative. Speaking a common language within the NHS system, in terms of the way we describe symptoms, diagnosis, prognosis, medication and many other aspects of care is critical to safety and understanding. 

Assessing, correcting and coding data downstream from the point of input will never be entirely satisfactory. There are some hugely exciting fields of technology that promise to transform the ease with which this is possible, in particular voice technologies underpinned by natural language processing algorithms.  

We can’t forget that parts of the NHS remain undigitised, even in 2019. These parts are not only hugely expensive and challenging to operate, but also create holes in longitudinal records that impact care and detract from our ability to clean and interpret a clinical record. 

The extent to which we look backwards or forwards in our data curation work is a conundrum. Fixing forward is the most efficient route to a highly curated data set that starts from this point in history, but it is the historic data that will yield rich insights and that will represent this incredible national industrial opportunity. Both are therefore necessary. 

Finally, but most critically, all uses of clinical data must fall within legislative guidelines, be ethically acceptable when viewed through multiple lenses and have the broad support of the public. This requires a detailed, sophisticated and ongoing dialogue with the public that must be given the significant focus, time and funding necessary. 

NHS data is extraordinary. Making better use of it will ultimately improve the lives of patients and clinicians in many ways. It is also a unique national asset with the potential to attract significant international investment. It remains important to invest continually in this as a resource and ensure full commitment to the diligent work necessary to leverage the potential of this unique asset.