For our most recent HTN Now webinar on the role of data and digital in supporting population health management (PHM), we were joined by a panel including Victoria Townshend, portfolio director (associate) with the GIRFT Elective Team; Mayur Vibhuti, CCIO and GP clinical lead for digital at Kent and Medway ICB; and Harry Thirkettle, director of health and innovation from Aire Logic.
Our panellists explored and discussed approaches to PHM, successes, challenges, what works and what doesn’t, through to measuring the impact of PHM interventions.
Victoria shared her journey to date in this space, including five years leading on the implementation of PHM infrastructure and change management at Lincolnshire ICS. “Lincolnshire had very little data and digital infrastructure in place at that time, and I think from my perspective, starting with a blank sheet and a short brief of seeing whether this was something to pursue, it gave us a huge amount of freedom to act and a simple place to start from,” she reflected. Working with an infrastructure partner, her team built a person-level linked dataset that went from primary care through to community, mental health, and acute trust.
“That gave us a whole population view to help us understand the ebb and flow of our population through our services and to identify opportunities for change,” Victoria noted. “That’s where my experience came in as a change manager, and I think the programme went well – it handed over to BAU in July, and at that time I was invited to join the GIRFT team nationally to bring PHM and health inequalities methodology onto the national stage and to pursue the objectives of the 10-Year Plan.”
Mayur shared that his main role is now CCIO, and he spends one day per week working clinically as a GP in Kent. “I’ve been working with our PHM and data analytics teams to understand the direction of travel for us in Kent and Medway. We already had really good data sharing agreements in place across general practice, our trusts were sharing data including social care, into our shared care record, and we had our data warehouse.” The decision was made to implement the Johns Hopkins PHM risk stratification tool into the shared care record, he went on, which helped develop a single clinical language for risk across the system.
That represented a culture shift that the ICB worked on with its implementation team and with Optum, with the aim of bringing that risk stratification number, which is the patient needs group (PNG), into the EMIS clinical record. Having that in place within the clinical system that clinicians use every day, has meant that “people are now starting to talk about PNGs”, according to Mayur.
Harry told us how he had started out as a doctor, completing core surgical training, before getting interested in health tech and moving across around ten years ago. “I’ve worked in a variety of health tech roles, helping to lead large-scale digital transformations, and then working with Aire Logic for the last four years, a health tech consultancy that primarily works with the NHS.” Aire Logic’s role is as facilitators of PHM, he highlighted, working through integration, interoperability, and data optimisation, having helped to build the NHS Spine.
“We also help with data standards, including writing some of those data standards, such as FHIR and HL7,” Harry shared. “More recently, we’ve worked on the vaccinations programme, which is an important part of population health, and we’ve been running that for the last three or four years with NHSE.” Considering his reasons for being passionate about PHM, he said: “If we continue as we are with current trends, we’re forecasted to spend over $47 trillion globally on chronic disease in the next five to ten years, more than the GDPs of the six biggest countries in the world added together. So it’s not a “nice to have” – it’s absolutely imperative that we get better at PHM and prevention. I’m excited for this conversation, because I think it’s as important as it gets for the future of our healthcare system.”
Reflections on PHM in the 10-Year Plan
Mayur reported feeling “pretty positive” about the contents of the 10-Year Plan. “From my perspective, I’m keen we accelerate its ideas forward, because they’re all things we should be doing. I think there’s always a tension in how local areas might approach implementation, but if we all have the vision of moving care closer to home and being more digital-first, building relationships with all the layers of the system, we can stop thinking so organisationally and become more person-centred.” Whilst some level of autonomy is good to have, the biggest barriers are around how quickly we can shift to working collaboratively, he summarised.
“I couldn’t agree more,” said Victoria. “The 10-Year Plan sets out a clear ambition, and a lot of the content is what people like us have been working on and advocating for, for a long time. I think we need to focus on how the national and local teams and people with subject matter expertise around digital, intelligence, and analytics can support dialogue with those who have the task of delivery.” Part of moving forward is building understanding and accessibility for PHM, she noted, “as a lot of colleagues will feel terrified if being asked to come and look at a dashboard”. Making the data and intelligence accessible and able to be manipulated at the front end without risks of breaking anything is key, which is where the Pathfinder tool was used in Lincolnshire, she added.
Harry picked up on Mayur’s point about getting different organisations to work together, pointing to the challenges which often accompany disparate systems and data siloes between and across organisations. The private sector definitely has a role in helping to rectify that, he considered, “and the way we do that at Aire Logic is by being very open and very agnostic”. There is a need, potentially, for NHSE and the DHSC to “get a bit tougher with vendors”, he suggested, “and I think open standards and data should be mandated in order to work with the NHS”. Where it has worked well, in his experience, is where there’s agreement about how data will be coded, getting the architecture right, and having private sector vendors involved at the table. “That’s when you can start to do useful analytics and tie that back into changes in clinical pathways.”
What works for PHM
“I think a lot of suppliers actually want that clarity in terms of what is required and what we are trying to achieve,” agreed Mayur. Having a conversation about the need to integrate with current systems and prerequisites for delivery is really helpful, he acknowledged, “and having one way of doing PHM across the ICB is similarly helpful – I wouldn’t say our risk stratification tool is the best out there, but it’s a tool we can gather around and say that it’s better than doing nothing”. In general practice, for example, organising long-term condition call and recall based on month of birth was previously widely accepted, but suggesting it is done instead using data on clinical need rather than essentially an organisational algorithm has been well received, he said.
Working on building understanding of what PHM actually means and how it looks to deliver care on the frontline according to risk stratification is important, Mayur told us. “I run a digital champions network in primary care and a CCIO network across our hospital trusts, and just consistently talking about things like risk stratification or approaching things from a waiting list perspective really helps.” Getting everybody on board, and then wrapping the technology, architecture, and design around that to make it accessible to everyone, is key, he went on, “and everyone in our system can go onto the shared care record, bring up a patient, and see what category that patient is in, so they can deliver care differently. That is different from a dashboard – it’s about making that behavioural change at the point of care.”
Managing expectations can be a challenge, according to Victoria, as leadership often expects huge deliverables very quickly. “By very quickly, I mean within a year, which might be a long time if you’re planning a holiday, but isn’t much time at all when you’re changing how a whole system works, and requiring people to be familiar with what you are introducing.” Working in areas such as health inequalities and disparities in access alongside more traditional targets allows everyone to get involved in the conversation, she noted. “You put the infrastructure in place, you get your board in on it and build the understanding.”
Victoria shared an example of a project in Lincolnshire looking at a PCN to identify frequent visitors to A&E. “We didn’t want to target them at the front door of A&E as the national initiative on high intensity users had done; we can re-ID within our linked dataset, add a flag, and do something different in terms of intervention. The PCN designed a completely non-medical intervention because it wasn’t just health, it was social care, there were mental health challenges, and so on. They were supported by non-clinical link workers, and their A&E attendances dropped significantly.” As the linked dataset is costed, a business case could be built demonstrating how much had been saved, and this attracted social investment from Macmillan as a partner, she continued, “and that shows how you can make a really scalable change due to infrastructure you put in for PHM”.
Harry highlighted the challenges for early-stage innovators when trying to align with the funding model of delivering in-year cash savings. “If you think about pre-diabetes, for example, if an intervention costs thousands of pounds, but can stop that person developing Type 2 diabetes, over the course of their lifetime that saving will be huge. But often that falls flat, because a funding board or CFO will be under pressure and looking at how to save money in this years’ budget.” A “radical rethink” is required, he suggested, “and I don’t know what the answer is, but it feels to be mission critical”.
It’s interesting to consider where the line can be drawn between PHM and personalised healthcare, Harry told us, talking about work with Leeds Teaching Hospitals using AI on referral to surgery to look at the GP record and estimate ASA surgical risk grade. “We showed we could accurately give an ASA grade based on the data we already had access to, that allowed risk stratification and pre-op assessment to indicate who was high risk. We’re hoping now to scale that up and take some pressure off pre-op assessment clinics.” An agile approach and a product mindset are key, he continued, clearly identifying the problem to be solved and the simplest solution to solve it. “Start small, and make sure you can actually solve that problem, and then when it’s working, you can scale it,” he suggested.
“The deeper you go, the greater the chance is you’ll uncover more stuff you haven’t thought about,” Mayur said. “Having the detail in the data is an ongoing mission, and you could keep pushing for more and more detail, but you need the people on the ground who understand what it’s telling them and who have the skills to talk to people and find out what is happening. If you try to just control people and “do to”, it’s not going to work, so co-production is important. We have lots of people all serving the same population, and they all have a different role, so it’s talking to each other and learning from each other, knitting that together to create a population health shift.”
Avoiding bias in PHM
Harry posed a question to Mayur and Victoria, asking for their views on social equality and bias in PHM. COVID highlighted disparities in some populations where people weren’t engaging with services, he shared, meaning data held on them will be poor. “When we do the risk stratification and prioritise interventions to certain groups, does that mean we might be missing some of those in the greatest need of additional support?”
Mayur talked about the need to do something differently. “Health inequality is a structural problem, and everything we’ve done so far has got us to this point,” he explained. “But we need to have one way of dealing with 80 percent of the population and a different way of dealing with the 20 percent who are disengaged or underserved, who are also driving behaviour that is most expensive to the system. Ultimately, we need to work both angles and have the personalised care teams working with the population health and health inequality teams to understand what the impact is of those interventions.” As better data is fed in, that will help to direct understanding of the population and their needs, as well as the work that is required in response, he added .
“The only thing I would add is the GIRFT tagline about reducing unwarranted variation,” Victoria shared. “The unspoken part of that is understanding warranted variation, and where that variation should be delivered proactively – we always used reverse logic modelling, starting from outcomes and working it backwards.” Listening to patient advocates and those who work on the personalisation programme are also recommendations, she continued, “because that is how we bring the voice of the people into the decisions we make from strategic commissioning downwards – that’s how we address inequities that are baked into our systems and change our decision-making process so we don’t continue to build them in”.
Looking at PHM from a tech perspective means it’s often possible to overlook non-functional requirements on access, Harry conceded, but Aire Logic has been investing in areas such as real-time translation, with a digital forms engine that can translate forms as they are sent out and then translate them back into English as they come back in. Another recent idea has been working with an ambulance service to test the feasibility of real-time translation whilst someone is on the phone with a call handler. “It’s about what we can do to reduce those barriers and communicate effectively with people,” he said.
Next steps and future considerations
When considering what she would change to make an immediate impact, Victoria mentioned giving leaders more headspace to learn new skills and to consider PHM. “Often, we expect people to move very fast, with no headspace to consider what the unintended consequences might be. I would like to see a bit of protected space for people to start to think about how to do things differently.”
For Mayur, a “nice to have” would be the infrastructure to facilitate two-way communication across Kent and Medway’s two million people. “It’s the old adage of delivering the care you would deliver to your mother – how do I know, if I can’t speak to my mother so she can tell me honestly and truthfully? I think if we’re going to do large-scale PHM interventions, I’d like large-scale feedback, essentially, so we could drive that iterative change.”
Seeing the NHS define what the minimum tech stack is for an ICB and having some standardisation around that would be beneficial, Harry noted. “We’ve seen some success with the shared care record programme, and as a vendor, it’s knowing that if you’re going to work with an ICB, you need to plug into that. I’d also go back to the point I made earlier about getting a bit tougher with contracting, so every vendor has to put data in the same place – I think that would unlock some amazing analytics, and it would be really powerful.” Funding is also an important piece, he went on, “as being held to show how you’re going to deliver year-in cash savings is stifling innovation, and we need to think more long-term in how we’re procuring and reimbursing some of these great interventions out there that are dying before they get to patients because they can’t find the funding.”
When it comes to AI, Mayur highlighted the importance of making sure the basics are in place and the data is correct. Having huge datasets where everything is joined up across the system has the potential to be “revolutionary” in terms of decision-making, he continued, but we’re not there yet as far as the data is concerned. “The Secure Data Environment is a good test bed to run projects we might want to do in real life, and things like the Federated Data Platform – using that structural dataset that’s coming nationally, I can see it all linking up together.”
AI readiness is a major challenge, according to Harry, with lots of organisations still relying on on-premises servers, who haven’t moved to the cloud, and who have siloed systems with poor-quality data and a lack of standardisation around how it’s stored. “AI has got to be part of the solution to that $47 trillion problem I outlined at the beginning of this webinar, but we need those building blocks in first. Somewhere AI is having an impact in other industries is also in the automation of processes – if you’re going to handle the data from 60 million people, and you know that if they hit certain criteria then they need an email to be sent out to them, that’s where that could come in.” Putting that kind of framework in would help roll out interventions at scale, he added.
We’d like to thank our panellists for taking the time to share these insights with us.
