Welcome to the Project EDWARD podcast – My role in the safe system – with James Luckhurst. 

The March 2024 episode introduces us to a surgeon, who explains how greater sharing of collision data - and medical information about crash victims - could lead to a better understanding of the causes and the costs – both human and financial – of death and injury on the country’s roads – and improve safety in the process.

Crash and casualty data is routinely collected by the police who attend incidents or are made aware of them. It is then published by the Department for Transport through a system called STATS19.

However, to get a more rounded picture of collisions and harm on the road network this should be better linked to other sources of data, particularly injury data collected in the national medical dataset, Hospital Episode Statistics (HES), and that recorded by the ambulance service and other parties involved in the aftermath of road crashes.

That is one conclusion of a report entitled – Data Linkage in Road Safety – funded by the RAC Foundation and the FIA Road Safety Grant Programme and I’m pleased to welcome its author to tell us more…

My name is Seema Yalamanchili, I'm a general surgeon in London and I'm also a clinical research fellow at Imperial and I'm studying road safety as part of my phd.

Let's talk about what you're trying to do, then, because the headline is that collision data and medical information sharing could lead to a better understanding of the causes and costs of road traffic collisions. What's brought you into this?

I could speak to you about this for days, but in summary, actually the main part of my doctoral studies is trying to look at novel digital sources of data, so crowdsourced data and how that might help with road safety data in real time, data that can predict. But to be able to validate those new types of information, we had to cheque them against something we would consider gold standards. So, in the UK, we're pretty lucky. We've got stats 19 that's collected by the police and available from Department of Transport, telling us about the collisions and causal factors for those collisions. But equally on the hospital side, we've got a number of hospital data sets that can tell us about what's happened to patients once they arrive at hospital. What I really wanted was to be able to say, well, this piece of, let's, for example, talk about traffic navigation app data. There's been alert here saying there's a collision. Is that real or not? How bad is it? So the stats 19 will give me an idea of if it's happened, but in terms of the actual injury outcomes, it's not so great. So there is some denomination that's put down by the police officer completing the form that will say their sort of lay understanding of what's going on, but it's not really a proper medical assessment.

And if we're going to start talking about how we get the number of road fatalities down, that's really important. It's not just about fatalities. We've got a lot of near misses, or, in our world, near misses, but actually that's still severe injury that we should be looking at. And without that information, it's very difficult to say, well, this collision led to this kind of injury outcome, because without that link, as a trauma surgeon type person, I've got less information about what led to this person coming to me with this set of problems. And equally, from the police and transport side, it's a bit harder for you to say, well, actually, that was quite a big deal, because a concealed pelvic injury, for example, might not be apparent to a police officer at the roadside. It's only when they get to hospital we realise how sick they are and they can die. So that linkage is really key and they're starting to do it in many other countries now and we've sort of fallen behind a little bit.

Give me some examples of good practice from overseas, then.

Well, I think, to be fair, it is an area of evolution that's going on. Europe, the states, Australasia, and there's even some sort of countries from the global south that are dipping their toe in. So in terms of gold standard practise, again, I'm not sure anyone's 100% achieved it, but there's definitely people making a really good effort. I'd probably say up there leading things is Sweden. Sweden have got a good system. They try to connect health data with the police record in real time, so they consent the patient in hospital, I understand, to try and make the link to their health record. That's one way of doing it. I'm not saying it is the only way and it's the best way, but that's one way of doing it. Recently, New Zealand have tried a different approach, kind of similar to what I've been trying to do, where they've taken a number of existing data sets within the country and tried to link those. And, yeah, again, in other countries, they've done sort of a more, I guess, a smaller level, local regional thing, where they might have just looked at a small group of people as a pilot type thing. And again, in a lot of places, road safety data isn't necessarily kept on a national level, it's quite often regional. And so that's what people have to work with sometimes.

Just talk through what happens for you when you're called to action as a surgeon, then what information is forthcoming, how accurate is it, and how much are you left to find out for yourself once you are working on a patient?

Yeah, it really varies, and it depends on what the ambulance crew are able to tell us. A lot of them are pretty good. They've done it for a long time. They realise that some of that information for us gives us an idea of the severity of what we can expect in a patient in terms of their injury. So they'll tell us a little bit about, for example, the way the vehicle collided. So a t bone or a head on or somebody bullseyeing the screen, something like that will give us a bit of an idea about the vector of the energy transfer. Speed is very important. Again, that gives you the magnitude of energy transfer, and the bigger that is, the more severe the injuries we can expect and people who've been ejected from cars, that sort of thing. Also very important. And we know certain types of injury, so deceleration, so going from very fast to suddenly stopping, that has a tendency to cause certain injuries which are related to particular organs that have a fixed point in your body, so they can't move in the same way, so they sort of get wrenched off their fixed point.

And we look for those when we either have to scan the patient or if there's no time, we have to take them to theatre to directly have a look visually and fix it surgically.

One of the questions relates to demographic groups and why some of them fare worse than others, even though they might have had similar collision circumstances. Can you comment on that?

Well, it's a very complex answer to a relatively straightforward question, because there's lots of reasons why this might happen. But if I give you the examples, for example, cyclists, the cyclists, as many people in the road safety space know, fare worse proportionately for the number of people who are actually on the road cycling. There's a couple of reasons for that. One, they are considered vulnerable road users. They're kind of exposed, like pedestrians are, so they will take a lot of energy transfer if they are hit. But the other thing that tends to happen is we do see a disproportionate number of women who die from cycling injuries when cyclists are involved with a bigger vehicle. So, for example, a bus or a heavy goods vehicle. And again, it could be anatomical, they're smaller. But one of the reasons we think it might be happening is just the driver behaviour, the way they behave around these larger vehicles, particularly at certain types of turns, might be more hesitant, so they're not able to get out of the way of those vehicles. Potentially, the male cyclist might. Again, this is a generalisation and it's just a working theory, which is yet to be proven, but like I said, we see a disproportionate number of women in the outcomes, but potentially men are a little more aggressive getting off that starting line with the lights and get out of the way of that blind spot and may not get hit in the same way, so that's a possibility.

But ultimately, yeah, we see a lot of them getting crushed pelvis, massive bleeding, and that can kill, and in some cases they lose limbs, that sort of thing. So that's one example. Another is, again, theoretical, and it's a sort of thing that we might want to be able to look at if we had linked data. But certain age groups may not fare well. For example, head injuries in the elderly while they are elderly, so that is one contributing factor. But it is also possible that they're not necessarily triaged to the right services in the right timely way that they might do if they were a different demographic. And again, it might be a regional thing. Some places you might have closer proximity to a trauma centre that can offer the right care very quickly. In some places you can't, there's a longer transfer time. So lots of potential answers to that question, but the only way to actually give you something definitive and evidence based is to have that linked data, to be able to say, well, this type of collision, this type of person we know as a trend, leads to this thing, and we know what they got in hospital, we know how long it took to get them there, we know they did or didn't get x or y treatment, and that might be why they're getting this outcome.

Well, two questions to help round off our conversation, Seema, which I could go on for days. Just an observation before those questions. I live in Powys, which is a Welsh county. It's a massive county. There is no accident and emergency hospital in this county. There you go. The choice of which centre I might go to is just grateful to get to wherever that will take us, I suppose. Anyway, two questions. In an ideal world, just give us scope out what success could look like in terms of better linkage of data. And secondly, who needs to be involved again?

I mean, there's what I really want and I'm quite ambitious in some ways in that I think we should be linking national data sets, but I also think if we want to stay ahead of the game, not just be on catch up, we should start to look towards integrating some of these novel sources of data, telemetrics, traffic, navigation apps, et cetera. But before we can even do that, we've got to do what the first barrier ahead of it is, as I've mentioned in the report, is that we don't have a routinely available transport police linked to health outcome national data set and where it has been done locally, again, it's not on a routine basis. So what I'd really like is for there to be linkage between stats 19 collected by the police and run by Department of Transport linked to ambulance. The reason this is important is because the ambulance data tells a lot about the initial emergency treatment from a medical point of view, not just from police and fire. And it has a second important thing in that it helps us to link to the hospital because people don't necessarily go to their local hospital, because we triage according to the severity of injury in most places, having very clear information that a patient's gone to x place or y place is really important for linkage.

And then also, eventually that will also give us a bit of clinical data from the ambulance guys, which is very important as well. And then we've got the hospital national data set, which is great and has everyone in it, but it doesn't tell us all that much about injury. It's sort of a little bit, it says some things, but it's relatively basic from an injury perspective. But we do have a national trauma registry which we should be linking to as well. And that has far more granularity in terms of the injuries that we've picked up, the interventions that we've done and the outcome. So if we could get those, because all of them are now national data sets, that's the first thing. When I was starting this work, the ambulance data was very much regional, but now we can link nationally if we want to, but it's now about the mechanisms and getting the permissions and trying to do this in a joined up national way, that would be my goal. So doing this, lots of the stakeholders coming around the table, and that's really got to be people from the health sphere. So ambulance trauma specialists on the health side, the health side also includes our record keepers, so NHS, digital, on the police and transport side, obviously Department for Transport police and any other stakeholders really, who have an interest in supporting road safety and want to help advocate.

Because although I've mentioned the data players, really just from a mechanistic operations point of view, one of the really important things about getting this off the ground is we've got to demonstrate, which I did on a sort of local level for my work, but we've got to demonstrate that there is patient and public support for this, because to get through the GDPR of it all, which is extensive, the thing that has to be demonstrated is that this is in the public benefit, which it clearly is for those of us who are interested in road safety. But we've got to show that coming from a wider group as well.

So it's all about smart use of all the data that's available Seema, I guess?

There’s so much more on top of that. So it depends on how you want to interpret smart use. I would say on a very basic level, it's just about getting it to make more sense. And so to do that, the first step has got to be not using it in a siloed way, it's got to be when you do linkage, your insights are just far greater. But we're also living in an age where the tools are so much more powerful. The growth in machine learning and AI applications for this is there's so many things we can do. We could start picking out patterns, we could have done that in a traditional statistical manner. But there's so much more that we can achieve with that. Now that we have those tools available in the wider sort of research world, they can now be applied to this. I think a lot of researchers at the moment spend their time trying to do these linkages, or getting round these linkages, or working with aggregated data, and they're really sort of handicapped. But with this, they can just get on and do the work they set out to do.

I did not set out to do this as part of my PhD at all. I just wanted to have a linked data set that I could work with. So if I can help bring that for the next generation of road safety, it's just. That would be great.

This edition of the project Edward podcast featured Seema Yalamanchili, a general surgeon and clinical research fellow at the Imperial College London Institute of Global Health Innovation. The producer was Peter Baker. And I'm James Luckhurst. Thanks for listening.