Analogue to digital in the NHS: is the shift within reach? – with Holly Krelle and Erik Mayer

Jennifer Dixon:

We're now very familiar with three shifts the secretary of state is aiming for in health care. From sickness to prevention, from hospital to community and from analogue to digital. We also know that a chunk of capital investment was directed in the NHS's way in the autumn Budget, and the 10-Year Plan working group is developing plans on digital and technology.

But if we want to move from analogue to digital, where are we on the journey? Are we near, far and what do we need to do to achieve this shift? Well, with me to discuss all this, I'm pleased to welcome two guests: Holly Krelle, who is Assistant Director at the Center for Healthcare Innovation and Delivery Science at New York University's Langone Medical Centre, and Erik Mayer who is a Consultant Surgeon at Imperial and he's also Director of Imperial's clinical analytics unit. Welcome both. 

If you look back over time, even just back to 2019, you see a lot of work in the NHS, a lot of strategising to indeed make that shift. Just a roll call of some of the initiatives here. You've got the [NHS] Long Term Plan of 2019, which set out route map, if you like, for analogue to digital. You then had the Wade-Gary review talking about what it would take to affect that change. And then you had Goldacre part one and then Data Saves Lives. And then in 2022, a plan for digital and social care, but that was by the Department of Health. And then you've got all the developments on top of that by NHS England, which includes the app, the shared care records, front-line digitalisation programme, virtual wards, GP access, digital first, etc, etc. 

So this is the latest in this journey. Given all of that, where do you think we are and is progress happening

Erik Mayer:

For me goes further back to the Global Digital Exemplar Programme and in one or two NHS trusts in which I still currently work, we went live with our electronic patient record back in 2014. There's this terminology around that, a revolution. But I see this now as an opportunity to accelerate the evolution of this analogue to digital. And I know the previous government set targets around onboarding acute NHS trusts to using electronic patient records for the direct delivery of care. And the 90% target was hit in December '23. The ambition is to raise that to 95% by March 2025 and hit a 100% by March 2026. 

Some of the initiatives you mentioned there are about centralising data for then repurposing for research and development in a de-identified form. But of course, fundamentally, that data comes from the appropriate use of electronic patient records. But those EPRs and the target set, of course, are to just start the use of EPRs in those acute trusts to a minimum standard. And I think there is a lot more to think through about the digital maturity of those electronic patient records, how they are appropriately used to then generate the right data, which we can then use to really start to transform and challenge ourselves around service and care delivery.

Jennifer Dixon:

If you look at all the big bits of strategy that need to be in place to realise this shift, digital maturity, electronic records and their use, but you also see there's a number of other things in there as well, isn't there? There's skills that is necessary, there's capital investment and so on and so forth. And again, the policy seem to say point to fragmentation of strategising across all the elements that are needed to really make this shift. Is that your take on it or do you think that's an uncharitable assessment?

Erik Mayer:

Broadly across health care delivery, there is always that danger of fragmentation and I think, yes, that has happened in part, but just drawing on a couple of things that have launched recently. So, NHS England data research for R&D programme for search development programme around the sub-national secure data environments is aligning with what we are doing at a local NHS level around providing that de-identified routine data for research and development. That is becoming increasingly aligned with research infrastructure funding, so through the National Institute for Health and Social Care Research and their digital health informatics programme. And they all come together to realise that you can't have silos because it's not affordable, let alone for the NHS community, the treasury, but also the funding bodies and the charities as well.

We have to build on what we've got and align what we've got because it's not cheap to build this digital infrastructure. It's not cheap to keep on developing it because, of course, it's a fast moving space and it's not cheap to make it financially sustainable. So we have to align these initiatives.

Jennifer Dixon:

And I suppose it's the alignment that needs to happen when a lot of national organisations that are responsible for some of this strategy are coming and going, aren't they with NHSX and NHS Digital and all the changes. 

So, Holly, and I know you work at NYU now, but you did for a while, work in the UK government and must have seen some of this churn at national level.

Holly Krelle:

I think for sure it's had an impact and you listed a lot of different strategies and it sounds like a lot of strategies and not a lot of progress. But what I would flag from the NYU experience is this just is a really hard thing to do. Like Erik said, the first step is to get the EPRs up and going. NYU is one institution doesn't have to deal with government strategies, has all of these separate strands in itself, has a huge budget, but it's still taken NYU 15 years to get to where we want to be with an EPR. So it is just hard. It takes a while.

Jennifer Dixon:

But presumably NYU leadership has been pretty solid and not so churny as we've seen over here.

Holly Krelle:

Exactly, and I think that's the key advantage NYU has had. There has been an ambition set when the last dean took over, which is in 2008, that the development of the EPR was going to be the main thing that NYU did. It was going to be a national leader, it was going to have a seamless patient experience, it was going to use it to drive OR and quality improvement. And it's been 15 years of a consistent focus and investment on that. But again, NYU is one hospital, one hospital system. It doesn't have to deal with all of the separate strategies and policies and departments that the NHS does.

Jennifer Dixon:

Erik, as you say, come may, regardless of all the other different bits of strategy you need to have working together to really make this shift. There's some basics here and that is the electronic patient record for which you've seen some progress over time as you were describing. So does that give you hope that even with all sorts of stuff happening nationally, you can still get on and do some basics and there's full insight as to the need to do that and consensus across national and local?

Erik Mayer:

There is, and as you just referenced there, this is also about local leadership as well, both within acute trust but across now what we now know as integrated care systems to have bravery around this because the challenge has been around the funding streams. And of course, I mentioned this has been over 10-year journey for us, it's been drawing on small amounts of money at the beginning to get things going and build the safe infrastructure that we need. And then of course, the larger centrally funded pieces of money have started to flow.

It does need that local vision and enthusiasm and I suppose culture initially to get this off the ground. But you can also do an awful lot in a short time. And just reflecting on your question at the outset, COVID undoubtedly has been an accelerator around this. Whilst we were making progress, certainly COVID allowed us to make a huge amount of progress very quickly and certainly from my own perspective, the secure data environment that I'm the custodian of, it's allowed us really to then draw not just on our Imperial trust data, but very much then draw and import databases across our 2.8 million population in North West London, which had more population health focus but integrates all the primary care electronic patient record data. So it certainly wasn't an accelerator for us.

Jennifer Dixon:

How much of this agenda, this shift, will have to be built from the bottom up? And is there a consensus about the proper role of the Centre given that we have run into trouble with very centralised strategies that didn't work quite frankly and cost a lot, is there a sort of working consensus now between what should be nationally done and what the locals can develop and grow from the bottom up?

Erik Mayer:

I think there is and there is that alignment, but fundamentally because we have to make sure that the patients and our public for who we are serving and we're providing health and social care, have trust in us providing access to data to support the ongoing research development innovation within the NHS and see the benefit to them as individual patients or as a population that it can bring.

And so we started that conversation. So in North West London, we are lucky that we can utilise that primary care record EPR data, but the secondary care EPR data in a de-identified form for research development across a North West London population as part of an information sharing agreement.

And of course, until we went round each of the individual health social care, mental health providers to add them sign up to an agreement to put that data in one place. We had to make sure that we had public and patients on our side and continue to do so. It's not a one-off process. It has to be continuing, particularly now around the world of artificial intelligence, where there is the black box concerned what's coming out, is it right, what is my data being used for? And of course, then the interaction with industry.

Jennifer Dixon:

And I want to come back to you on that, but maybe I'll ask Holly that question. Presumably, some of this is solved because NYU Medical Centre is a centre in of itself. Do you have such investment in patient consent here or is it just part of the deal with private health insurance and Medicare and Medicaid?

Holly Krelle:

Yeah, I would say it's part of the deal. NYU is driven by desire to get patients. It's a much more competitive environment and New York, so NYU is trying to get patients off Mount Sinai and Northwell and all these other systems. The way NYU decided they are going to do it is by having the best digital product and having the most efficient electronic health record and using that to drive really high quality. I'd say it's less driven by the kind of governmental strategy. Medicare is one of our payers, but more driven by this desire to desire attract patients to our hospital rather than the hospital down the road. I would just also add on the NHS point. So I've just come off five, six conversations with different NHS trusts about the maturity of their EPRs. And one of the things that struck me is there are a lot of trusts like Erik's at Imperial who are super advanced, who are super tailored to the level community who are developing things innovatively, but that's not everyone. And it tends to be those trusts that are linked to universities or who have sources of funding beyond the typical NHS sources of funding.

And as ever is the challenge nationally of thinking about how do you let those hospitals go ahead and do their work themselves and drive themselves forward, but then also the kind of hospitals that struggle a bit more that don't have those resources, how are you getting them working in this way? How are you encouraging them?

Jennifer Dixon:

And when you talk about engagement of the public, both of you, there's one aspect which is engaging then on consent and trust issues, but there's also a sort of market research engagement as well. What sort of digital products would really help you in this patient pathway or to manage your disease? What's the balance then in NYU, Holly?

Holly Krelle:

Probably more towards the patient experience market side of things, so it's very much... And you'll see a lot of my friends who've ever used NYU will say, oh my God, I went and then suddenly everything was just on this app. NYU has its own app, but it was just all immediately there and all my information was there and the whole process was totally seamless and the digital technology was just way better than any other hospital.

And I think that's what they've decided is the way to get patients. The question of consent, I don't know, it doesn't come up as much. I think perhaps patients in the US don't expect to trust their health care providers in the same way that we expect to trust the NHS. They view them as commercial entities because they kind of are. And so I think they're acting more like a consumer than they would in the NHS context anyway.

Jennifer Dixon:

And how valuable are they in deciding and co-producing what might be needed in terms of innovation?

Holly Krelle:

I think there's a huge user experience team within NYU working with a classic tech company, essentially doing user research about how the app feels, how the patient-based interface feels. Then there's also the more academic side of things, the quality improvement, all of those kind of projects where you would anyway be doing, talk to your patients about how their experience of cancer care is and how we might improve it. But [inaudible] we're saying both of those ways.

Jennifer Dixon:

And Erik, how does that contrast with how patients are being engaged at your end?

Erik Mayer:

So for me, they are front and centre from providing input to project or data access project approvals for audit service evaluation through to your point around co-designing potential solutions. At the end of the day, they're the ones that the change is being put in place for and it's amazing and Eaton's a practicing clinician, I think I know it all, but I don't sit on the other side of the fence for that.

For instance, diabetic patient who will be the end-benefit or otherwise or the intervention we're trying to put in place, because we have to remember that healthcare is such a complex environment and there are often unintended consequences that no one, as a single person, can think through all of them. And so having that public and patient engagement is really, really critical. And they get you thinking in a very different way and challenging, which is really helpful.

Jennifer Dixon:

And it's Imperial unusual or usual in that respect because the NHS isn't necessarily historically been known for good market research and acting on it.

Erik Mayer:

I wouldn't say that we are unique and I can think of many other good examples across the NHS, but to Holly's point, often they are NHS organisations, not always but often that have a strong academic partner because certainly again through funding bodies and charities, that is the expectation.

It's not a nice to have having a public and patient engagement component or co-design aspects to it. It needs to be front and centre to the point that I have patients who are formal co-investigators, the investigators named on grants that I submit and often one's not enough.

It should be two, it should be three. It should indeed be a community potentially. And certainly I know that as part of our Biomedical Research Centre application that we put in 3 years ago, we did a lot of public and patient engagement across the whole of North West London to shape the whole digital health theme that we're now delivering on.

Holly Krelle:

With traditional QI improvement teams. They do tend to involve patient feedback. If I'm going to do a quality improvement project on, I like to think about recently on our check-in process. Of course, my first step is to go and sit in waiting rooms and watch patients check in and ask them about all the problems with the process. So even the kind of QI and operational research things will involve patients. That's just how you find the answers.

Jennifer Dixon:

Just to finish off some of the basics, digital maturity, Holly, you mentioned that and if you look across the country and just look at the Darzi report, I think its data maturity assessment is out of five and the average score of NHS providers is two. Is that really a function of capital development and lack of investment in skills or there's limited money? And then why so slow? What could be done to accelerate this?

Erik Mayer:

So I think money does come into it, to be honest. We talk about the digital transformation within the NHS. You think about how the annual budget is apportioned across the different divisions and directorates within the trust. Digital gets a relatively small... I think the average is about three or 4% of the annual trust budget.

So we do have to think about how we need to be brave because we need to redirect funding to support what you've just described is actually the true digital transformation of the acute trust. And so naturally, with the financial constraints that can only come with a, because you are pulling on different funds or capital funds that are made available. So I think that is an issue. I think also though we also forget the workforce here because when I think about innovation, we think about the theories of Rogers from the early 1960s and you've got your innovators and your early adopters. And of course, innovation in health care often has been traditionally quite safe in a way. Think about robotics and surgery. You get a few innovators, they learn and a bit like bees coming to the hive slowly, the early doctors come on, the early majority come on, meet a critical mass and you bring on the others. But everyone's learning from the innovators. When you go live with an EPR, it's a big bang Saturday morning, Sunday night, Monday morning. From one day to the next, the entire workforce is working in a different way and that includes everyone. From the innovators and the digital champions to those late majority stroke laggards is a term that's used isn't it, in that curve. And everyone's got to work in a different way. So often what happens if people find a way around to almost work in a digital paper record as opposed to a true electronic patient record, they'd not leverage the true benefit of it. So I think that is a challenge as well.

Jennifer Dixon:

Yeah. Holly, you recently did a really interesting briefing on data maturity, didn't you, in the NHS for us?

Holly Krelle:

Yeah and I would say I pretty strongly agree with what Erik had said. I think there's two reasons for the lack of maturity for those trusts that were struggling more. So number one was they made that massive upfront investment in getting the EPR up and running and it was a huge amount of money. And then I think they expected, well it's done now. We've spent all this money, surely it's working. And they didn't want to keep spending the money. And actually, the problem is the set of the EPR is like the first stage. You then have to maintain it and then you also have to build all the infrastructure on top of it in order to be able to extract the data from it to use the tools that it has; to do all of the actually useful stuff doesn't happen until it's been embedded and being run properly.

And I think that can lead to a slightly vicious circle, to Erik's point, about the kind of workforce. You get this big bang, it's switched on, your information is digitised. But if you're not, then immediately using that data in interesting ways, you're not pulling the data on your patients to understand them better, you're not making use of this great online booking system because you haven't invested in that stage of tool, then you don't get the benefits and you are not motivated to keep developing and keep improving it.

Earlier a point about COVID being a real benefit in some ways because we sort of jumped forward and the shock really helped people to innovate fast. I think that's true. And I know at NYU it wasn't COVID, it was Hurricane Sandy. NYU went live with the CHR in 2008, but it didn't really have big bang progress till 2012 when Sandy took down the entirety of NYU: it flooded the hospital, it closed the hospital, it got rid of all of its electronics, all of its clinics.

And they used that to when the clinics came back online, they made them go back online on an Epic on digital health record. And from there on, they progressed really fast. And obviously, we've had COVID now we hopefully don't have another crisis, but how do we get that same kind of motivation and progress without a crisis like that?

Erik Mayer:

Just thinking about how we use electronic patient record, there is also the challenges around the productivity piece because, undoubtedly, it does take longer to work in an EPR. So take the outpatient setting, is that with all the elements and the steps that I need to do to generate all the data points that Holly just mentioned around problems and diagnosis, information for coding. Of course, the letter ordering, the tests, the follow-ups, etc. It does take longer per patient and yet the system is still trying to make you see as many patients as you would have done with almost a paper record, almost doing a dictation, then signing off your letter a week or two later. So we have to pull that in and work smarter. So I think the system needs to adapt to that as well.

Holly Krelle:

I think there's a lot of simple training that goes into that. NYU has training over three months for clinicians on things like the shortcuts that make the EPR easier to use. There are shortcuts and clever links that can be built into the EPRs. But if you're not actually training your staff on them, they're going to be doing these long-winded ways of doing things and they're going to be getting frustrated and it's going to be taking them longer.

Erik Mayer:

And then of course, they don't do the coding, they don't have the information as you said about how many patients they've seen with this particular condition, for audit service evaluation or even just their annual appraisals.

Holly Krelle:

One trust that I spoke to, they gave the example of coding something basically like smokers. So you might think now you've got an electronic health care record, okay, great, I can pull all the data on all the smokers because I have in my trust. Why not do that and see something interesting about them?

There's a box to tick in the EPR that says this patient is a smoker except when they actually pull that data, only 10% of doctors coded in the box. Most of them just wrote it in the notes. So in order to therefore pull your data on smokers, you can't just pull that box, you have to pull the notes, then you have to use a large language model or natural language processing to pull out the word ‘smoker’ from the note. You have to pull it all together. It's very complicated before you can do anything useful with that data. Whereas if you've trained your provider, to Erik's point, to use the box, then well, you've suddenly got really good structured data that you can do something with.

Jennifer Dixon:

So we've got a picture, in my mind anyway at least, we've got varying digital maturity and then within different providers, you've got different levels of functionality of some of these electronic patient records because of the availability of money or availability of interest and skill in developing certain aspects of functionality. And then on top of that, what you've got coming onto AI, now you've got a number of vendors approaching trusts in random order, interesting trusts in their AI thing that they want to sell and use. So this strikes me as being a bit random, if I can put it that way in so far that it's not a curated journey, it seems, particularly towards trialling and testing of AI because so much of it depends on what's coming forwards, but also what is the infrastructure on which it can work? Is that a fair assessment, Erik?

Erik Mayer:

Absolutely. Every day the Imperial Clinical Analytics Research evaluation secure data environment/ digital collaboration space have people knocking on the door saying, I've got this product, this product, this product. Now where do you think you can apply it? Where might it be useful?

As opposed to say, what are the three priorities we want to address this year? And let's put it out there and let's find out what's in the market and what's maybe best suited to our needs. But equally, to your point, and you've touched on what's the infrastructure to actually then start to implement and evaluate, let alone start to understand how those tools, particularly around AI have been developed and on what data sets they've been developed. Because there was a big difference between a tool having been developed on a very cleansed, curated de-identified dataset and then it working properly on real world data that is messy and not de-identified. So you then start to have issues around AI bias, etc.

So what we do have, what we provide in in our secure data environment is both an AI test bed where piloted AI tooling can then be tested and validated on de-identifiers in a secure way, bringing tools into the environment, validated on real world data. But also thereafter, we have the ability, because we have linked back into the direct electronic patient record to actually start to implement it in the dark, so it's not directly affecting care. The clinicians and our health care professional aren't seeing it.

We can start to understand how it then performs in a dark runway. But most importantly thereafter, if you want to start running some digital trials, for instance, changing a risk assessment, we can also then track how the behaviours of the front-line health care professionals are changing in response to that and measure that as part of a learning health system or better care loop.

So having that ability to development, validating, testing and obviously ongoing evaluation is critical. That's going to be an ongoing need because fundamentally, I know AI is a broad term, but no, for some of the more deep learning neural network AI, they're continuously learning. So as an organisation, the challenge to me from our chief executive was: how do I know that when you implement it tomorrow, one, it's not going to cause harm within the first 24 hours, but more importantly in a month, in 3 months, in 6 months, in 12 months, if it's constantly learning and changing. I think that for me is a big challenge.

Jennifer Dixon:

So are you saying then that even if the trust develops very good set of priorities to encourage the use of AI, that may or may not be possible depending on the state of maturity of the underlying electronic patient records?

Erik Mayer:

You do need an infrastructure that sits within the NHS obviously for security reasons, but the ability to do this development validating testing evaluation because you can't just plug and play. You want to plug and play into one of the EPR providers, the first question is who owns the system? You, we've had a question around proprietary software, those sort of things. You need the ability to have almost a parallel running environment to start to do this pipeline implementation, and that's critical.

Jennifer Dixon:

But to even begin to test intelligently, you need a certain level of maturity or quality of data to do that, yes?

Erik Mayer:

Absolutely. Yeah, you do need, and a lot of the AI. It is about the volume of data but also the quality of the data. So again, back to what I said, you can't have 50% of a particular data point missing for instance, because how it works and how it learns is going to be then very biased. And we need to have the data that's diverse as well.

Jennifer Dixon:

I mean if you think about the NHS and the fierce headwinds ahead morbidity wave, demographic wave and the pressures upon it, we can't really waste time here, can we? We've got to do things faster than we've done before. I'm just wondering, is there a place for national big ticket demonstrations on things that are high priority, whether they're clinical or non-clinical, tested in the best possible site with the most data mature sites where there's the most skills, where there's enough patient engagement or whatever the assets are, to vanguard sites with more pristine data that you can at least forge your way towards the future more quickly, than relying on a network of local trusts to do things that according to local priorities. That's necessary. But is there an additional role that the centre could play in developing some best-in-class test sites to demonstrate what is likely to be needed across the system quickly?

Erik Mayer:

Yeah, you're absolutely right. And that is the sort of direction of travel that has started and through both, again, I mentioned the NIHR, BRCs, but also through the work around health data research, UK's HGR UK, and we have done some pilot projects where we have across a group of five to eight organisations done what you've just described. And that doesn't have to be about sharing data to be clear that. You can federate across those sites. But you're right, you can demonstrate the feasibility of it. That's the first thing, to be honest. The pressures are now and here, let alone down the road and got, we have to move quickly. We can't spend 3 years developing a solution.

But that does also, I think, mean that we need to think about coming back to the clinical trials perspective in the academic world, the evidence level you want is often around that double-blinded placebo controlled trial type approach. That can't be the case in around its rapid implementation. So we do need to think about how we are pragmatic but safe and do get some evidence generation about how these things will land and obviously benefit our patients from the public.

Jennifer Dixon:

And then turning to Holly. Obviously, these pressures apply to NYU just as much. And in addition, there's the competitive environment that you described. So how is it developing? Have you developed new methods for speeding up what might be, as Erik says, more pragmatic, rapid, robust, believable, credible evaluations of some of these innovations in AI in particular?

Holly Krelle:

So yeah, I think we have, and it's a core part of what my team does. So I think the first thing to say is your earlier question around AI and people shopping different AIs to different NHS trusts, I think that sometimes worries me because it sounds like AI is search of a problem. Whereas what you actually want to do is as an organisation, find out what your problem is and then work out which of your many tools, maybe it's AI, maybe it's simple dashboard actually solves that problem. And that's the kind of way we approach it at NYU.

So a front-line team comes to us with a particular problem and then we work out the best way to solve it and the best way to evaluate whether or not our solution was actually effective. So recent example that we're working on at the moment would be our kidney transplant team came to us.

So as you probably know, kidneys typically you get listed on a wait list on a kidney according to various criteria, and then you just work down the list as a kidney becomes available. But in the US there are also these, so-called open kidneys, which are sort of lower quality kidneys that you wouldn't necessarily want to give someone who's waiting on that list, but they could be given. And these offers typically come in the middle of the night, they will come to say, NYU is the first provider and our clinicians will be given an hour to work out who, if anyone on our list, could take this kidney, and that could be 300 people. So they came to us, this is the problem, can we come up with a solution and then test the solution?

So we, first of all, developed them a simple dashboard that just presents all of the information all their more easily so that they can filter it and get quickly down to the kind of 15 patients that will then be eligible. On top of that, we will then layer something like GPT-4 or ChatGPT in order to try and produce a quick summary of all of that patient's notes to give you some of the more qualitative and up-to-date information you might need to make that judgment . And then my team does is we'll run some kind of rapid randomised tests to see if this was effective. So we will randomise some portion of our patients to get this ChatGPT or to not, and we'll check for a couple of weeks. What impact does this have if we show doctors this new summary of the patient, are they making better decisions? Are they increasing inequalities in any way? Are they doing anything wrong?

So we're actually rapidly testing whether what we're doing worked, and that doesn't need to be a formal clinical trial. It doesn't need to go on for several years, but it uses the kind of randomisation in order to learn something about whether or not our AI is actually useful.

Jennifer Dixon:

And one way of developing and testing AI is indeed when some clinical group comes to you with a problem like that. But does NYU take a strategic view of saying what do we anticipate in the next 5 years? How much of it is the corporate sort of approach as opposed to a clinical team based, the troubleshooting as it were?

Holly Krelle:

Yeah, I think it's a good question. I think the bulk of the corporate approach is to actually build the capacity internally. So almost all of our AI models are built by us. So we have an AI team, a machine learning team. There's very little that we are buying in from outside. I think precisely the kind of approach that Erik outlines because we want to be able to develop something on our own data, test it on our own historic data, run it silently in our population to see if it's working or not, then turn it on, do rapid tests about does it work or not manipulate it. So for example, think of another one. We're alerting doctors to whether or not a patient is in the last 6 months of life. We can have a really good model, an AI model that predicts these patients, but how do we get doctors to act upon that? In order to do that, we need to be trying different alerts, we need to think about how we're talking to them. All of that is stuff we want to run as part of the overall way the hospital runs, not as some tool we've bought in from outside. So I would say the main thrust of NYU's development here is to basically build the capacity internally rather than buy it in.

Jennifer Dixon:

And this is where the NHS has critical mass, isn't it, Erik? You said 5 to 8 organisations can link up federate and try this.

Erik Mayer:

This is exactly where people can link up. But we've learned a lot from the work Holly and team have done to shape some of our work around AI, particularly around generating clinical summaries, a part of the discharge summary process. And we're going through that sort of rapid cycle evaluation at the moment. Doing it across other NHS organisations thereafter is really important. But also just to highlight that although you might not be buying a tool off the shelf, the work with industry is still really critical because there are partially developed tools out there. Holly mentioned open AI services. That is a tool in itself, but it still needs to be applied to a specific problem.

So working in partnership with industry, and I choose that term very carefully as part of life sciences agenda and life sciences cluster [inaduble] life sciences is really, really important because industry then starts in working partnership with a number of partners.

They start to understand the problems, not about the clinical problems, but the problems with deployment and some of the factors that we've already been talking about here. This is not plug and play technology. And so it each shapes their future roadmap as well in terms of how they set up their teams and understand where the need is. Because often in industry, they develop something and they don't think about the deployment piece of it. And so it does shape their thinking. So we have to move together a part of this journey to keep up because we've mentioned the financial envelope for the NHS is restricted. Industry and private income is a source of income to help accelerate the sort of issues that we need to address. So we can't shy away from leveraging industry in this space.

Jennifer Dixon:

No, very good. And monetising some of the value of the data assets as well as the adaption of the AI is also an agenda, isn't it, for the future. Just two quick questions I think to finish on.

One is the shift of care out to primary and community, and the extent to which both of your work in the digital space has accelerated that. Mostly what we've talked about is hospitals, they are the big cost centre, so that's why it's important to talk about hospitals. But what about the shift to primary and community?

Erik Mayer:

This is the critical bit, isn't it? And more than ever, the information is going to have to follow the patient wherever they move. In some ways, that shift community is increasing the age-old dilemma of transitions of care. So the information will need to follow the patient. We've obviously had announced recently the promise of the digital patient passport by the NHS app. That's somewhere to centralise all the patient's health information, test results, letters, etc. It's not a new concept. It was around 2007, 2008, but the NHS digital infrastructure wasn't there to support it at the time. I think we're in a much better place now. But again, if you're putting all the information from all these different systems in the same place, we do need to think carefully about reconciling where there are differences. 

Medications is one example that springs to mind. The medication that's held in acute trust may be very different to what's held in primary care, which version is right, which version is wrong. And those sort of issues are a patient safety risk. But wearables clearly put us in a much better place as well for this sort of thing. And integrating that wearables data into any of the electronic patient records is critical and key.

Jennifer Dixon:

And, Holly, how much effort... I don't know the setup at NYU whether it's an integrated system or not, how are you reaching out into primary care?

Holly Krelle:

Yeah, it's an integrated system. So when I go to my primary care doctor at NYU, they can see every appointment I ever had in NYU, every test result, every X-ray, all of it is just right there. It's one integrated EHR. I would say the caveat to that is two things. One, primary care isn't really a very integral part of system, well, in the US generally, and especially in New York, most people don't have a primary care doctor. There's no primary care gatekeeping. And number two is the integration is only the case. If you're within the NYU system. There's some sharing, but many people will have their primary care doctor in another hospital system and then come and use the hospital NYU, and that won't necessarily share. So the patient experience is still pretty fragmented unless you happen to just decide to use all your care in one.

Jennifer Dixon:

And then the last question, which is probably quite big. So we've got the 10-year plan. One of the enabling groups is obviously on tech and digital, and they will be cooking up some ways to accelerate this shift. What would you like to see coming out of that 10-year plan in this area?

Erik Mayer:

Yeah, one of the agenda's for that group that you mentioned is about how we can make life easier and more productive. I don't quite like the word productive or more effective or efficient for the workforce in the NHS and therefore, to be able to use the data more efficiently. I think for me, that is the critical bit. And we mentioned rights much earlier around data fragmentation. I mean, you log into patient's record and it does take time to find a bit of information you really need. But I think there's a real agenda here about utilising tooling. And that can be AI potentially, but drawing on not just the structured data, but all the unstructured data that we have in the electrical patient records, which makes up about 75% of the information in the NPR to action through the data quality. Because you can triangulate across those date sets to make sure then the information we're providing is to the right person at the right time at the right place.

So people are spending less time around operational activity within the electronic patient record. That of course, then will support operational efficiencies. And for instance, particularly in the nursing world, but in the clinical world, the amount of time now spent documenting and filling out forms. And it's very easy to generate a form in a EPR, but then it has to be filled out. So you are then distracting people from time to care. And so for me, I think that needs to be a real focus of that working group. And therefore, by also improving the data quality, we can be assured and confident that then using it to simulate changes in how we provide services. And so those simulation models are real, and we can then think about how we can provide care more efficiently. 

Jennifer Dixon:

Get the basics right, is what I'm hearing throughout all of that, which sounds right because it's on which everything else is constructed, doesn't that? And, Holly, do you have any thoughts on that given where your perch at NYU?

Holly Krelle:

It's developing those teams that allow you to extract the data from your EPR and then take action based upon it, which is what really seem to be missing from my conversations with trusts. So places like Imperial, like great Ormond Street, like UCLH already have these teams because they're relying on and leveraging money from academia or from partnerships or from their charitable arms, but other trusts don't.

And therefore, they just haven't unlocked this whole part of their EPR that they could be doing, even just basic UI projects that they could be doing that they're not. And I would say focus on building those teams first before thinking about bringing AI, etc, etc. Bring those other trusts up to the level of places like Imperial and Guys, and then think about the more complex second order challenges of the EPR.

Jennifer Dixon:

We must leave it there. Thank you very much to Holly and Erik for all those insights into quite a rich area. So join us next month for the traditional highlights of the year. We'll be doing our normal Christmas roundup of all the choice excerpts from the year’s podcasts, which is quite fitting because it's our 50th podcast, can you believe? So we'll be very, very happy to do that over a mince pie in a glass of brandy. 

And in the meantime, a big thank you to our team here at the foundation that's Leo and Sean, and to Paddy and team at Malt Productions. And it's goodbye from me, Jennifer Dixon.