The National Health Service is sitting on a goldmine of patient information, but identifying how best to tap its rich seams is very much a work in progress. SA Mathieson reports. Photography by Paul Heartfield
Data analytics has great potential in healthcare but most NHS organisations have yet to explore it in a meaningful fashion, according to participants at a recent round table hosted by Civil Service World in conjunction with Leidos, a provider of science and technology solutions and services to the defence, intelligence, national security, civil, and health markets.
Like many other health service professionals, those working with data are often overwhelmed by routine work. “We have to sort-of feed the beast,” said Angel Shrestha, deputy head of informatics at Colchester Hospital University NHS Foundation Trust.
NHS organisations have to collect and submit data to national bodies such as NHS England and NHS Improvement, he explained, and such submissions may increase as a result of rising collaboration with other parts of the NHS. While greater automation of routine data collection processes may help, Shrestha added that increasing amounts of data processing are also demanded by local management teams.
He noted that the data being collected is not always relevant or useful, and can change regularly. “I have been through many of these so-called executive dashboards and scorecards,” he said, adding that it is common for managers to order the calculation of key performance indicators that are rarely used, but keep being collected. He added: “We’ve had to produce KPIs out of KPIs.” It makes sense to review such indicators, and stop the use of those which have limited relevance, he said.
Another reason to focus on the relevance of data was given by Dr Lesley Jones, a nurse fellow for digital at NHS Improvement. She said that clinicians can “shut down” if given large amounts of raw information. Providing a small subset of clinically relevant data through screens and mobile devices works much better, she added, citing one organisation's presentation of real-time data on patients in wards. “They have stripped back a lot of the rubbish they were generating,” she said, and as well as clinicians the data also helps in workforce planning, which can be based on the mix of patients on a ward at that time.
Dr Jones added that clinicians may not trust raw data. “People question it, they wonder about the timeliness of it,” she said. “They will say, ‘that doesn’t feel like how the clinic felt this week’.” As a result, data is often more powerful when linked to patient stories and other anecdotal material that helps to explain it.
Jones’ point was supported by Michael Fleming, analytical programme manager at the Department of Health and Social Care, who said that stories often hold the key to advances in understanding. “This is how science actually develops,” he said. “Your information is in your outliers.” He agreed that this vital information it can be hidden when lots of data is presented in a standardised way.
Ravi Hubbly, vice-president of data analytics products and services for Leidos Health, agreed that trusts had to involve those working at the point of care when exploring data-enabled tools. “The people who need to be empowered are dependent on what is deployed,” he said. If data is not standardised or there is no way to visualise it, it may be of little use. And those frontline workers can help to improve the data, he said: “One of the key best practices I have seen is to bring the policymakers to work with the data themselves.”
Despite the challenges, it’s clear that sophisticated use of data can help trusts to improve, especially if they are able to use it to work with others. NHS England’s RightCare project is aiming to help clinical commissioning groups to get more sophisticated in their analysis of data, by matching them with similar CCGs across England based on a range of data, rather than just their geographic neighbours. Terunnum Shakeel, senior analytical manager for the programme, explained that this means CCGs can compare themselves with others in similar circumstances, with action plans based on these comparisons.
“It’s looking in a completely different way, where you’re actually focused on what needs to change,” she said of RightCare. “I’ve been in the NHS for over 20 years looking at public health data, and it’s always been the case that somebody’s decided this is the focus, we produce a report, and it disappears somewhere.”
Another barrier to better analysis of patient data is a lack of standardisation. Dr Martin Chapman, a research associate at the School of Population Health and Environmental Sciences in King’s College London, is working on how data from different electronic health record (EHR) software can be converted into standard formats, as well as using older EHR data. The latter means handling changes in the clinical codes used for some conditions: “Depression was recorded in a different way in the past,” he said. “A cursory examination of those EHRs would probably tell you that there was no-one with depression before 2000, which is obviously not true.”
Peter Oliver, head of strategy for health in the UK at Leidos commented that “standardisation, as well as the ability to rapidly process unstructured data, would be key to supporting the effective implementation and governance of new models of care across England in the most cost effective way; as we seek to benefit from the utilisation of increasing sources of data across organisational boundaries”.
Chapman added that distributed ledger technologies, also known as blockchain, may well have a role in improving the sharing of patient data. Distributed ledger records are permanent and can be digitally signed, so the patient or organisation can assert ownership and control their use. However, he added that scalability is a major problem with such systems.
No discussion of data analytics would be complete without touching on machine learning and AI. Christopher Geary, innovation lead for digital health at Innovate UK, said that about a third of its applications for funding mention these technologies, although some appear to make only superficial use of them. He added that AI software is being trained to triage medical scans in pathology, choosing the images that need to be viewed by human clinicians: “We can get more accurate, faster and cheaper – that’s the dream.” Such systems are more likely to reduce human workloads than end them. “It is clinician augmented, rather than replacing the human,” he said. But Geary added that the NHS is hindered in using AI by having relatively small amounts of data in limited locations, which he described as “puddles” rather than the large pools of data that are generally required by AI systems.
There are more basic problems with data analysis at many NHS organisations, stemming from the fact that a lot of information is still held on paper. “We are trying to force GP referrals to come in electronically,” said Colchester’s Angel Shrestha. “That’s posing so many challenges.”
NHS Improvement’s Dr Jones has visited most of the English health service’s 16 “global digital exemplars” – leading trusts that have been given extra funding to push forward with digitisation, to act as trailblazers for the rest of the NHS. She said that removing paper makes a big difference to hospital wards, with clinicians entering data directly on mobile devices rather than scribbling on scraps of paper then queuing to use a ward computer.
But she added that these exemplars are taking different routes to digitisation, with the best listening closely to their staff on how to do so. This must be the model for the rest of the NHS: “Don’t do it ‘to’, do it ‘with’,” she urged, adding that trusts need to publicise what does not work as well as what does. “If something fails, we’ve got to have that out on the table,” she said. “We need the bad news stories as well as the good news stories, to prevent the NHS from repeating mistakes.”