Limp NHS productivity and what to do about it – with Anita Charlesworth and Neil Sebire

Jennifer Dixon:

The NHS is treating fewer patients compared to before the pandemic. And this, despite increased funding and staffing levels. Why is productivity lower than expected in the NHS and will investment in tech help? Well, with me to explore this today is Anita Charlesworth, Chief Economist and director of the REAL Centre here at the Health Foundation and former Treasury Official. And Dr. Neil Sebire, who is professor of Paediatrics at University College Hospital and Great Ormond Street and he serves in GOSH DRIVE, which is a unit dedicated to innovation using data and digital tech. Welcome both. 

If you read the press, what you hear is NHS got load more money, load more staff. How come activity in the NHS hasn't recovered since the pandemic? What's going on I think is the big question and why, Anita, are national policymakers so worried about this issue? And indeed what is productivity, what are they talking about?

Anita Charlesworth:

When we talk about productivity, what we mean is if you look at the inputs that are provided to health care, so that's primarily the staff who work in it, but all the drugs and goods that we buy and also running all of our buildings and facilities. And you compare those inputs with how much care is provided, over time, we've provided more care for any given level of inputs, but in the pandemic there was a big hit to productivity, inputs grew. We actually provided less care as we had to manage the pandemic. And so operations didn't go ahead, people only saw their GP for very urgent things. And getting back up to running a health service operating normally is proving quite a big challenge. In one sense it's almost shocking that we're shocked by that. When you think back to the scale of impact of the pandemic on all aspects of our life, but particularly in the NHS, the idea that the day after the vaccine had been rolled out, everything could return exactly to how it was before is probably pretty fanciful.

Actually we are now recovering quite fast. But perhaps the biggest issue is not so much how we recover from the pandemic, but if we want to have a thriving NHS out to its 100th anniversary, it's really important for its affordability, for patient experience, for the quality of staff experience that we use the inputs into the health service really well to get the maximum possible value in terms of amount of care. Productivity is much in the news at the moment, so the post-pandemic effect, but actually that's something of a distraction really. What we should be really profoundly thinking about is how do we, on a continued basis, use all those inputs to health care as well as we can to make the most difference to people's lives.

Jennifer Dixon:

I think everybody knows that productivity across the wider economy has been limp, particularly compared to [the ] international picture. We haven't really recovered have we since the pandemic? Is there anything you can say about how the NHS compares to the rest of the economy in terms of productivity?

Anita Charlesworth:

Yeah so if the pandemic is the big shock for the NHS, the big shock for the wider economy was the 2008 crash. Since the 2008 crash, productivity in the economy as a whole has been very, very slow and below that of our peers. And that might sound like a techie, nerdy issue, but in the documents for the Budget this year, there was a figure that really brought what that means home to roost. So productivity really determines our economic growth and wage growth. And if productivity had continued growing at the rate in which it was before the financial crash in the period since 2008, average earnings would be £14,000 a head higher than they are today. And obviously if you think of £14,000 per employee, that's an enormous amount of quality of life resilience to the cost of living crisis, but also that feeds through into how much money we have for key public services like the NHS.

Productivity in the wider economy is incredibly important. Actually these things are also then quite interconnected because one of the things that is proving to be a constraint on our productivity growth is lots of people out of the labour market due to ill health. So the health service depends on the productivity of the economy and the economy is also dependent on how well the health service is operating. Actually the NHS through that period did better in productivity than the whole economy. The NHS isn't some basket case outlier of poor productivity either globally or all compared to wider economy.

Jennifer Dixon:

Neil, you're at the front line in many respects, aren't you, working at UCL and also Great Ormond Street Hospital? What sort of conversations are there locally about this issue and this characterisation you see in the press, which is NHS got a lot more money, not more staff, but not much more activity. Do you recognise that?

Neil Sebire:

Well certainly I recognise the discussion because this is talked about more or less all the time, particularly kind of at board level. There's the issue of at the organisational level, how do you increase productivity, which is primarily how do you get to see more patients with the same resources. Then there's also the discussion about kind of individual productivity. How do you get, for example, a clinician to be able to either see more patients in a clinic or if you take someone like myself, my clinical background, I'm a diagnostic pathologist. Ideally when I'm working, you want me doing the thing I'm trained to do most of the time and you don't want me spending lots of time doing things that I'm not very well-trained to do or is not adding much value. And there's an interesting phenomenon I suppose around the wider more general introduction about electronic patient record systems, particularly from the data from the US. Admittedly they use them slightly differently there where these things were designed to make individual tasks for clinicians easier are actually generating quite a lot of additional work which is not directly related to patients. 

There is a lot of interest in how we can better use data, for example, to schedule how you make optimal use of operating theatres and clinics, etc. And then there is a more general thing which I think hospitals are quite bad at doing compared to other organisations about logistics. If you take the example of a supermarket, they will know exactly where everything and probably every person is and therefore it's relatively easy to then use tools like machine learning to optimise. If the analogy was narrowed to a hospital to make sure that you've got the right medications or the right bits of equipment where they need to be. We generally don't have that in the NHS.

There are little pockets where things are tracked, but if you went to a general hospital more or less anywhere in the country and you said, do you know where every single piece of your equipment is as of this moment? I suspect the answer almost certainly will be no. How much more you can do that is tractable still remains slightly unresolved. Just as an example, we did some internal work looking at scheduling in theatres and we basically took all of the cases that had been done over a couple of months period and said, well, if you rearranged these in the optimal way, could you get more operations done at the same time? And the answer is, you could, but it was actually better than we thought. It was about 90% efficient as it was. And then when we looked at the reasons why there were inefficiencies, so to speak, the vast majority of those were things that are outside people's control. It was where someone was ill, it was where a patient couldn't turn up because of a problem with transport or some of these unpredictable issues.

What is often not discussed as much is, well, are these things tractable? Can we do anything about them? Because there will always be element of unpredictability.

Jennifer Dixon:

Can I just bring you back to the organisational level, Neil? At Great Ormond Street or places like that in UCH, would there be a board level strategy on productivity? And if there is, how much of that really filters through down to the clinical leaders who are operating the front line?

Neil Sebire:

Productivity in terms of activity, waiting lists, activities through clinic, etc. These are things that are discussed or reported up to board level, I'm sure in every NHS organisation at every meeting. Whether there is a coherent strategy for how to improve this productivity is a slightly different matter. It's one thing being able to report, but there is less emphasis then on saying, okay, we've reported this, there is an issue here with let's say waiting lists. Rather than just say everyone work a bit harder, get the waiting list to come down. What are strategies we could try to have an evidence base for saying this approach does or does not work for this productivity issue.

Jennifer Dixon:

Just going back to Anita now, what are the factors that influence productivity in the health system and what does the evidence tell us so far?

Anita Charlesworth:

Often people think about productivity as how hard people were working. Productivity is not how hard people are working and we don't improve productivity by making people work harder and beating them up to do it. That's not what we mean by productivity. What productivity is about is it's a system that helps people to work smart. And to work smart, what you need is all the different things that enable people to do their job well, to be aligned and in place. Typically, there are three big things that are really important to productivity. One is having the right number of workers with the right skills organised in the right way, well motivated and understanding how to do a good job and work as a team. And we know in the NHS, we've got a lot of work to do on that. Got a workforce plan which has started to address the numbers issues, but actually one of the things the pandemic did is to really shock our system in terms of some of the structures of teams. We've got a lot more new and inexperienced staff, the number of more experienced staff leaving went up and a lot of people report in the NHS now that they're not in such stable teams. And then you've got all the challenges of burnout and stress and industrial action as well impacting on that. Letting the workforce right is really, really important. The second issue then is capital investment, having the kit in the right place of the right quality to do the job. I'll give you one example of how that might work in the NHS. Talking to an NHS leader this week who was saying through the investment in the Community Diagnostic Programme, they'd bought some new CT scanners that scan in 10 minutes, whereas the old scanner scanned in 30 minutes. Obviously with that modern piece of kit, the staff can scan many more people in a working day.

And that's been, over time, the big driver of productivity. We've systematically under-invested in capital in the NHS. We've got big maintenance backlogs, we've got old and out of date equipment, insufficient equipment. But also then the other issue that is really important is not having enough capacity. We've got bed occupancy rates way above the level that you need to be able to manage the normal fluctuations in demand. Perversely for productivity, you need a little bit of spare capacity, you need a bit of buffer. People might think being productive means running everything 100% hot at the edge of capacity. It is not. In a complex system with interrelationships, you need to make sure that you are not running everything too hot, and we are. And then the final thing, which is the lifeblood of productivity, is innovation. We've seen enormous reductions in length of stay over the last two decades, which has been a big driver of our productivity with innovation.

The big question in the wider economy and now for health care is the innovation of the 21st century is the digital age, AI and digital technologies. At the moment, the paradox in productivity is that across the globe in our economy and in health care, we're living through this absolute explosion in innovation, but in a period in which our productivity performance has declined and no one quite understands why, but everyone understands, I think, that the prize for all of us now is to get the digital revolution to work well for us in health care.

Jennifer Dixon:

And of course the Chancellor in the spring statement has allocated £3.4bn to the NHS over the next 5 years starting next year for this type of technology and innovation with productivity in mind. Neil, what sort of technologies are being tried at the front line, and in particular, which are the ones that you think are the most promising with respect to productivity?

Neil Sebire:

In my opinion, unequivocally, the core layer that has to be in place before anything else is for organisations to have some form of comprehensive electronic record system because it's complete madness for people to be talking about what they want to do with AI and they're perhaps still using paper notes for parts of the hospital. I remember when I was a junior doctor a few years ago, having to spend hours and hours every week either filling in forms or looking up to try to find the patient's notes or to try and find the lab test result, etc. Electronic patient record systems get round a lot of that. They say they have specific savings for many of the tasks that go on in the hospital, particularly the pacing type tasks. And in addition, they then allow you to capture accurately a lot more data about what is happening on a day-to-day basis than you had before.

It now generates reliable data that you can start to use for machine learning, for example, for optimisation. That's the sort of core layer. There is then a lot of interest in technological tools and many places have started to use things like RPA, which is robotic process automation, which essentially is nearly always used in the back end for things like billing, etc, where you can use a computer system to do repetitive tasks such as filling in forms, etc. But I think the areas that are really starting to gain traction and a lot of interest with clinicians is the role of ambient technologies. Things like speech to next for potential time savings, summarisation of notes, et cetera. And as with any other sector, there's a lot of interests in large language models or LLMs, but it's unclear at present exactly whether and what those can be used for in a health care setting. But these have the potential, for example, to be able to rapidly read and summarise documents or to generate structured notes and letters. We and many other organisations are actively exploring these types of technologies at the moment.

Jennifer Dixon:

I mean I think your point about the core EPR is really important and it's clear across the country that there is very different maturity of data across different NHS providers. In terms of the ambient, I wondered if you could say a bit more about this.

Neil Sebire:

These kind of ambient technologies really are tools. I guess the easiest way to explain it is similar to many of you may have a device which listens to what is happening that you can then ask questions of by one of the big tech companies in your house. And the concept would be that you could have for example, a device that can listen to a consultation between a patient and a doctor and can then not just do speech-to-text but can add intelligence onto that. Can recognise that something that's been said is an allergy or can recognise that something is a medication and can then go and look up a code for that medication and provide things back. And certainly that's something that we are trialling at the moment. There is a lot of interest globally in whether or not these types of ambient technologies not only may free up a bit of doctor's time, but actually the real interest is can you improve the quality of care by doing this?

It's very easy to build in, for example, kind of checklists to say, okay, at the end of the consultation has the clinician asked all of the things they needed to ask or have you covered all of the next steps for example. It gives you a basis of adding in all of these quality areas but without the individual doctor or nurse having to kind of actively consider them because you've got a system that will nudge you in the back end. These are still in their infancy. Very few of these are currently in use and approved for clinical use. But it's probably fair to say that this general area is one of the most rapidly growing and has the greatest [inaudible 00:17:20] at the moment from tech companies.

Jennifer Dixon:

And Anita made the really important point about workforce. Can you just say how these kind of trials in Great Ormond Street, particularly with this ambient speech-to-text system you described, how the clinicians and how the staff has responded to this and what sort of support did they need?

Neil Sebire:

In general, the response to most of these types of things from clinicians is very positive. I mean, we found this with the introduction of the EPR system and the introduction of these types of tools. And I think a large part of that is because in general, health care systems and the NHS are quite behind people's daily lives. In other words, when we were still doing things with paper and multiple different logins before we had our comprehensive EPR system, clinicians were delighted at the thought of actually having a simple way now where they can see all of the lab tests and they can just type their notes in and see what the dietitian had written without having to go through reams of notes. And similarly, because people are used to using many of these types of AI ambient tools in their daily life, it's almost a frustration to think, well this is out there, why can't we use it in health care?

Well the reason we just can't just use these things in health care is that health care is considered a high stakes industry. In other words, there's more risks to something being done wrongly if you are ordering a type of blood, for example, than if you were ordering some pasta from another site. There are different requirements for any of these tech systems. And one of the things that has actually been very positive in the last year or two, I would say, is that initially when things were being developed from a tech perspective for health care, people would come and say, here's this tool, isn't it great? It does this. But there was very little real world on the ground testing. But that's now changing quite a lot because what people really want to know with these tools is not just how the test performs in theory, but do they like it? Does it deal with accents? Does it work in a clinic that's quite noisy? All of the practical questions that you can only answer by putting it in a real clinical environment.

In general, I would say that the clinical response is very positive to anything that might reduce the administrative burden on clinical staff.

Jennifer Dixon:

You can see, Anita, can't you in future if the supply of care is being shaped by these new technologies, that's going to have very profound implications for the types of staff that we need in future.

Anita Charlesworth:

The most important thing that we need to start to try and think about is, these technologies will do two things. One which Neil has been talking about is what's called augmentation of labour. It will help workers do their job better, they'll do different things, it'll be higher quality, it'll be job enriching. And that's one set of factors and that leads to productivity, but that does change sort of who you need things. There are other set of changes where the technology will replace labour.

Jennifer Dixon:

Yeah.

Anita Charlesworth:

Instead of interacting with the human being, we'll interact with a machine. And this is some of the obvious things like booking and scheduling, etc. If you think with a bank 20 years ago you would've gone in to see somebody in the bank or you would have phoned them, but someone would've done all those tasks. Now you, the customer, do all of it on an app. The extent to which technology will replace labour in health care is a really important issue. That might seem threatening, but I don't think we should see it that way because actually globally the amount of health care services that we need across the world to ensure that people can live long and healthy lives is way outpacing the rate at which we can train and recruit people. And globally we've got a big shortage in the skilled health care workforce, a shortage that it is likely to increase, putting huge pressures on systems across the world.

Actually labour saving technology, it's going to be really important globally for health care. And we are talking about very NHS, but also it should reduce the cost of some of that health care and if we're to achieve the global goals of universal health care coverage, I think that's really important. People may say, but what about the digitally excluded? And the digitally excluded is a really important policy question, but it's also worth thinking about that actually in a lot of the world, the one thing that people do have is a mobile phone. I think the issues about equity and digital are really important and really profound and should be a high priority, but probably a bit more nuanced and complex than often we think of them in some of the public debates about this here.

Jennifer Dixon:

There is a need for speed here, isn't there? What you were saying about having sort of a national push at some of these things, for example, obviously at Great Ormond St you're doing great stuff with the ambient work that you're saying, Neil, but in a sense, do we need sort of 10 or 15 or 20 sites trialling this at the same time?

Neil Sebire:

I think that is the case. There are many centres that are doing all sorts of interesting things in this space, but as you mentioned earlier, there is also quite a large spread, if you like, of underlying capability, whether that's the electric patient record system and also infrastructures to do this kind of thing in health care. In general, you cannot run these kind of tools on the IT infrastructures that most health care organisations have. You need platforms that can deal with these kinds of tools and are designed to do so. And increasingly the way to do this is through cloud. And you'd be surprised at how challenging it is to access cloud for these kind of things across health care organisations. And a large part of this is that many of the roles around data and data science, etc, don't exist as roles. There is no well-established professional career pathway for non-clinical informatitions.

In other words, if you're going to be in an organisation and you are now to say this is fantastic, we'd like to do this in five different hospitals, here's the tool. Simplistically, here's the software, who do I give it to? Whose job is it in that hospital to do this? Because the traditional model has been the old-fashioned kind of word processor model that you install a piece of software, then it's done and then a year later you do an update like you would do on your phone, etc. This is not how these tools work.

Anita Charlesworth:

Professor Diane Coyle gave the last REAL Centre lecture and she's an expert on technology and productivity in the wider economy. And there are a couple of things which I think are really important to think through from her lecture that the NHS really needs to take to heart. One is that you do need to make sure that you've got this systematic capability and core platform across the system. But the second is that actually if you look at the wider economy, all of which have access to the same technology in different sectors, there's the most enormous difference between firms at the rate in which they realise the benefits. And that requires both them to have the workforce with the skills, but also the management with the capability and culture to do this. And then actually it requires a different approach to your organisations. She talks about much less hierarchical, much more empowering local.

To get the benefits of this new technology for the NHS, we need to invest in the core underpinnings that Neil's talked about. We need to invest in the innovation, the vibrancies, have lots of [inaudible] across the system, this new cadre of data scientists, but we also need to change a lot of our approach to management and the way we run our systems. The change agenda here is huge. It needs the capital investment that the Chancellor has put in place, but it's so much more than choosing a couple of products and rolling those out.

Jennifer Dixon:

Yes, exactly. And as Neil says, you don't just sort of cut and paste this into the system, it has to develop over time and get better, particularly if it's a generative AI or a machine learning piece, which means you need iterative, formative evaluation as you go to develop the model. Just to return to what Anita pointed out earlier, which is of course one thing is using technology to augment labour in the way that we've just been describing, but another one is to replace labour. And then there are two areas I think. One is to see what more the patients can do themselves, as you were saying, Anita, as in the banking analogy. But the second one is to help to shift care between acute hospitals and the community, which is where everybody would prefer. Back to Neil, where do you see promising technologies in the patient facing element? Where do you see that going? And secondly, more profoundly, is there anything that you're involved in that you think has the potential to really shift care out into the community safely and better?And is that technological related or is that more to do with the nature of clinical treatment itself?

Neil Sebire:

Okay. That's a big question, but that's fine. I mean I think more generally there are kind of three broad areas that I think are going to be hugely affected in the coming decades. One is around clinical decision support. It's around how you make clinicians make better decisions. I'm going to use myself as an example. I'm a diagnostic pathologist. What I essentially do is I look down a microscope at some tissue and I look at essentially a picture of that tissue and I say, I've seen this before. This is this type of disease. And in theory a system should be able to recognise those pictures at least as well or better than me, but we haven't trained it to do so yet. There will be certain tasks within health care itself that will be more or less replaced. And then there's really the bit that is only just starting, which is I think what you're alluding to here, which is really the patient decision support systems.

In other words, what will be the types of tools that will help those. Now some of these will be modifications of these generic tools we've been talking about like speech to text, etc. But the machine learning parts of monitoring patients out of hospital I think is probably going to be the first area where we are really going to see this. And there are various trials going on all around the country in its widest sense hospital at home, which really just means are there things that you would have had to come into a centre away from your house to have monitored that we could now do in some virtual way. And we still have not solved quite how to fit all of this into the care pathways and how these things will work different patient groups.

Jennifer Dixon:

Diagnostic tests particularly and ongoing monitoring.

Neil Sebire:

I think they're going to be the area that will be done first. But of course there are various trials going on for monitoring patient's activity at home in, for example, dementia that have shown quite promising results for very early prediction of superimposed urinary tract infection. Not based on anything to do with the urine, but based on how many times the patient is getting up to go to the toilet or how many times the patient is waking up in the night, for example. There's a whole range of additional types of things that if you like before wouldn't have been classed as ‘health care’, but may now be.

Jennifer Dixon:

And what about the direct AI to patient decision support? That's a long way away I guess, but that conceivably will come.

Neil Sebire:

I think generally that's quite a long way away for clinical care. I guess the best example of this in use at the moment are some of the more modern diabetes tools, where there may be an insulin monitor and pump that can directly interact with a patient through an app and warn them of an event. That's probably the first area where there is really a direct relationship now with the patient and the device, so to speak, rather than the patient and the doctor.

Anita Charlesworth:

From an economic point of view, I think this is one of the important things that we really do need to think about. Historically what technology tends to do in health care is often make the individual procedure cheaper, but expand the number of people who are going to get the care. It's possible for technology to make health care more productive but also more expensive simultaneously, because you do more of it. And of course the whole point of markets is to invent things that then more of us use and to expand that market. And a lot of the dynamics obviously in health care technology are coming from firms who are looking at how to expand the market as well. And the direct patient issue is really important here because obviously if what that does is to provide information to people, but they don't know what to do with that information and so therefore their response is to go straight into the health care system and then the health care system has to work out what to do with that information. And in most cases the right answer to that is to do nothing, what we've done is swamped our system with a whole lot of noise and low value activity. How to avoid that problem and make sure that this innovation is targeted at the things that will genuinely benefit outcomes and improve the sustainability of our system.

Jennifer Dixon:

Let's have a last question. Thank you both very much. There's likes to be a new government after the next election. What two or three things would you be advising a new secretary of state on how to improve productivity in the NHS?

Anita Charlesworth:

The place I would start is in capital investment. We've invested increasingly in the workforce but not matched that with an investment in capital. And I would make sure that we are using that to ensure that across the NHS we have the core foundation interoperable IT systems that will be critical to whatever direction technology takes us.

Jennifer Dixon:

And Neil.

Neil Sebire:

Well, for me this is very easy and I promise you that Anita and I did not confer before you asked this question, but I fully agree because if there is to be benefit at scale from AI and technology, please do not spend money on AI. The way to get value from AI is to invest and ensure that we have the underlying infrastructure to allow us to leverage whatever tools are developed. That means ensuring that hospitals have appropriate electronic patient record systems, infrastructure, networking, data engineering, cloud services, etc. We do not need lots and lots of products being bought. What we need is underlying capability to allow you to then build upon. That has to be the focus, not getting caught up in a shiny thing that is very specific.

Jennifer Dixon:

We must leave it there. Thank you very much to Anita Charlesworth and Neil Sebire for their insights today. Join us again next month and until then, many thanks to the team for producing this podcast. That's Leo and Sean at the Foundation and Paddy and team at Malt. And it's goodbye from me, Jennifer Dixon. Until next time.