A recent CSW round table examined the potential of Big Data to improve outcomes and cut costs in our healthcare system. Colin Marrs reports. Photos by Paul Heartfield
The NHS budget may have been ring-fenced to protect it from central government cuts, but fast-rising delivery costs and patient demand are still ratcheting up the pressure on our healthcare system. Casting around for ways to cut delivery costs and improve outcomes, many healthcare professionals have identified opportunities around the ‘Big Data’ agenda: can analysing large datasets help commissioners and providers to weather the financial and demographic storms buffeting the NHS? To answer that question, Civil Service World got together with analytics provider SAS and hosted a round table at which commissioning, data, policy and clinical professionals discussed how healthcare organisations can realise the potential of new and emerging datasets.
James Shield, project manager of the cancer population evidence programme at Macmillan Cancer Support, pointed out that Big Data is not a new idea in healthcare. Macmillan, he pointed out, was founded in 1911 by a former civil servant who combed through paper records to track the prevalence of cancer. “There is an incredibly rich and long history of [data-based research] in the cancer field,” he said. “Big Data is in some ways a very old technology; it is just that we have different ways of working with it now.”
“The traditional randomised control trial is pretty hopeless to check whether service redesign works,” he said. “We have an advantage with Big Data in that we can set up large natural experiments and collect enormous amounts of data very quickly.” Dr Peter Brindle
Indeed, health workers have long led the rest of the public sector in publishing and reusing data. Tim Kelsey, the Department of Health’s national director for patients and information, made his name in the field by founding Dr Foster – a publisher of comparable data on health providers – back in 1999, and by 2006 the NHS was already so convinced of the power of data that it bought half the company.
There are many potential benefits of collecting and analysing health data. Clinical information can be used to spot examples of best practice or poor care, or to assess the effects of new treatments or differing delivery systems. Financial data can be scoured to pick out particularly efficient or costly teams or bodies, and combining organisations’ management information can identify opportunities to avoid duplication and share assets.
As an example Dr Peter Brindle, a GP in Bristol and R&D programme director at the Avon Primary Care Research Collaborative, said that comparing care outcomes data from different geographical areas can produce evidence on the efficacy of treatments much more quickly than traditional research methods. “The traditional randomised control trial is pretty hopeless to check whether service redesign works,” he said. “We have an advantage with Big Data in that we can set up large natural experiments and collect enormous amounts of data very quickly.”
Performance data, said David Paynton – the clinical lead at the Royal College of General Practitioners’ Centre for Commissioning – can help pick out the most successful health providers, enabling others to learn from the best. The college is creating a dashboard showing the relative performance of 35 general practices, he said: “This allows them to see where they are in comparison with their peers. There are sometimes debates that involve some practices pointing out the differences between themselves and better performers, but overall standards are improving as a result of it.”
“Everyone is terrified about being beaten in the national press, so everyone errs on the side of not giving anything away. If I give my data to you and it ends up in the hands of a drug company, then that is my fault.” Amanda Cheesley
Asked to provide more examples of Big Data driving down costs or improving outcomes in the NHS, however, the panel struggled to name UK case studies. Tessa Richards, the British Medical Journal’s patient partnership editor, said that a Swedish initiative relating to rheumatoid arthritis is collating both clinical data and “patient reported experience” information, producing stats that provide a rounded picture of each care provider and enabling patients to take a more active role in shaping their own treatment. And David Downing, director of health at SAS UK, mentioned an American scheme run by healthcare firm Kaiser that combines datasets to highlight patients who should be checked for a bone condition. “They defined a business question and got buy-in from managers and clinicians to solve it,” he said. “Then they did a lot of X-rays and scanning of people they predicted were at risk of osteoporosis, improved hip fracture rates by 38%, and saved tens of millions of dollars by getting to them early.”’
There were far fewer UK examples, illustrating the fact that – despite the arrival of Dr Foster some 15 years ago – Big Data is still in its infancy in the health field. Where there has been progress, it’s often been the result of a specific policy initiative that’s pushed public bodies to work together. As an example Claire Oatway, chief operating officer of Beacon Medical Group, cited the cross-departmental Every Child Matters initiative: “It allowed the mining of data to identify the most vulnerable children and develop health promotion interventions,” she said – with the result that professionals in various fields reached out across sectoral and organisational boundaries.
If proof of success was thin on the ground, though, plenty of organisations are pursuing the agenda – so the evidence on efficacy should mount up fast. The Airedale NHS Foundation Trust, for example, has overseen a system revamp that brings together data from hospitals, GPs and the local authority, explained Simon Platt, the service improvement manager at the North West Commissioning Support Unit. “It has led to them being able to share documents all around the system,” he said, suggesting that the trust has been brave to make this leap to the cutting edge: “The costs of changing 50 to 70 systems just so you can all be on the same page could have been seen as a huge risk.”
It is, of course, at least as risky to do nothing – and several panellists were enthused by the potential benefits of bringing together data from linked fields such as health, social care and even transport modes. However, people also pointed out that there are significant hurdles to sharing data. One big obstacle is organisations’ concerns about breaching the Data Protection Act. Another is their fear that data they pass to other bodies will be lost or accidentally released, said Amanda Cheesley, professional lead for long-term conditions and healthcare at the Royal College of Nursing: “Everyone is terrified about being beaten in the national press, so everyone errs on the side of not giving anything away. If I give my data to you and it ends up in the hands of a drug company, then that is my fault.”
“I think most would say they are happy for their information to be used for research and other NHS purposes, but they don’t want it to be sold to insurance companies and pharmaceutical firms" Isabel Hodkinson
Emerging technologies can help overcome such fears, suggested Isabel Hodkinson, clinical champion for care and support planning at the Royal College of General Practitioners. The college is wary of giving away raw data, she said, “so we created a space where you can go and play with the data. It is all anonymised, but you can’t take the raw data out. There are some ways you can begin to get the value of the big data, but [users] can’t draw down [data on] the individual people.”
These reputational challenges facing the Big Data agenda were thrown into sharp relief recently by the groundswell of opinion against the care.data programme, which aimed to give academic and pharmaceutical firm researchers access to anonymised patient records to test the efficacy of different treatments and processes. “The furore it produced has pushed it back, so it is less likely to happen,” said Richard Carthew, programme manager at the Health & Social Care Information Centre – which was responsible for care.data. “We have to address why there was a furore.”
Patients wondered who would get hold of the information, panellists suggested. “I think most would say they are happy for their information to be used for research and other NHS purposes, but they don’t want it to be sold to insurance companies and pharmaceutical firms,” said Hodkinson. But there was agreement that many patients would actively welcome wider sharing of data between NHS bodies; most members of the general public, panellists suggested, assume that GPs, consultants and hospitals already have access to a shared medical record.
The public, having witnessed more than a few breaches of data security in the NHS, probably don’t realise just how far the technology has come on. As Platt commented: “In other industries, security is mostly a solved problem – the issue is more about how people perceive their data is being used.” He added that the NHS is moving towards a “data package” approach, which sidesteps the problems around organisations sharing data: “The idea is that the patient owns everything, and various bodies will then approach them for access for that information rather than the IP being contested constantly,” he said.
Paynton urged health organisations to invest as much effort and money in training and supporting staff to work with data as they do in buying the necessary IT capabilities. “We buy very expensive systems and plonk them on people’s desks and say: ‘Use it!’, and then wonder why they don’t use it”, he said, pointing to provider Kaiser Permanente and the United States as an example of good practice. “They put as much support and effort into the training as the kit.”
Another major obstacle lies in the perverse incentives created by the NHS funding system, under which providers are paid according to the number of treatments delivered – giving organisations a disincentive to pursue preventive remedies. As Steph Edusei, chief executive of Healthwatch Newcastle, said: “If improvements mean patients end up being treated in community settings, what is the incentive for the hospitals to actually do anything? Their income then drops and tips an already borderline, unstable organisation over the edge.”
“Maybe the answer is that small scale local innovation is the answer to the NHS problems instead of trying to nationalise innovation.” Will Smart
In Northumberland, she said, a “gain-share” arrangement has been introduced which tries to change those incentives – but to introduce the system, some healthcare providers have had to take a financial hit. This isn’t necessarily the case, suggested Downing, who pointed out that a shared pot – the “Leeds Pound” – has been introduced in West Yorkshire to try to ensure that parts of the system are not punished when they work together to reduce demand for medical treatments. Will Smart, director of information management and technology at the Royal Free London NHS Foundation Trust, suggested that the early adopters are going to have to take some pain in order to generate the evidence that could lead to system change in the future. “Payment mechanisms are always going to follow innovation,” he said. “Not the other way round.”
Gleams of light in this area were picked out by Paynton, who suggested that the current financial climate is beginning to encourage a more collaborative approach. “We used to have commissioners trying to get one over on providers, and providers trying to get one over on commissioners,” he recalled. “You could cope with that when you had 7% [income] growth each year, but now we are in flat cash. That is going to require us all to share information and work in a very different way.”
Looking to the future, panellists stressed the importance of thinking about how data might be used before deciding how to collect and store it. Eric Robinson, clinical trials manager at Oswestry Outcome Centre, noted that hospitals generate vast amounts of data, but much of it is not easy to search or feed into other IT systems. “We gather data on all our procedures – it is a huge amount of data, and if someone asks you for information on a how a particular hip replacement is performing, it is a huge amount of work to pull out the patients,” he said.
Ben Ryan, technical client manager at SAS, added that practitioners should begin by identifying fairly narrow datasets which can be used to produce clear results without drawing in too many variables or collaborators. “If we were looking for the ideal scenario where we were pulling in data from primary care, social care, it would take forever and wouldn’t be possible. Having everything in one place is not going to happen in the first instance,” he said, suggesting the use of projects with “smaller cohorts and short timescales in the first instance, to grow into bigger projects over time.”
Indeed, the Department of Health has long since stepped back from building the kind of nationwide system that could facilitate broad and early data-sharing across the NHS. The failure of the National Programme for IT, which attempted to create a national care records system, has put the onus for innovation back onto local bodies, suggested Smart: “Maybe the answer is that small scale local innovation is the answer to the NHS problems instead of trying to nationalise innovation.”
“There isn’t a systematic way of finding out whether your stuff works and for everyone else to find out about it.” Dr Peter Brindle
Others agreed, but there was a feeling that such fragmentation could also lead to waste. “What is frustrating though is that I won’t know about your innovation in kidney injury that may well work, so our local hospital will be trying to innovate in a slightly different way,” said Brindle. “There isn’t a systematic way of finding out whether your stuff works and for everyone else to find out about it.”
Platt pointed his peers to an existing website called Learning Environment, run by the NHS Commissioning Assembly for Clinical Commissioning Groups. “It is full of case studies and searchable under every subject under the sun,” he said. “But many people don’t know it exists, even in CCGs.” But he urged practitioners to make progress before lengthy research has been undertaken to prove that better use of data generates better patient outcomes: “I appreciate the need for genuine evidence but if any kind of model has been standing for more than five years and hasn’t gone belly up and generally the local population supports it, then that is pretty good evidence for me that it works.”
Overall, the panel concluded that the barriers to data sharing are surmountable, and the approach has the potential both to improve outcomes and to save money. But hard-pressed NHS staff will struggle to make rapid progress unless some are given the time and resources to concentrate on the Big Data agenda – hopefully thereby lessening those pressures. “In health and social care, we are so busy pulling people out the river that nobody has time to go up river and shoot the person throwing them in,” said Edusai. “We don’t have the time to actually take a step back and do the thinking and research because we are too busy saving lives.”
Download this free whitepaper on delivering integrated care