 So my name is Wei. I'm one of the new support officers in the South Seas. I've been there for like five years or so. So I'm here to introduce you to the next force because I was in a session. So first up we have a friend of ours who will be talking to us about selective sorting with regards to SEMS community health and to the SECC. Okay so I'm one of those people who use both the LS and the cross-sectional data and during those talks I was thinking I should have really put more of the cross-sectional sorting which I haven't so sorry about that. I'll try and slip it in as we're going through. So I'm a population officer currently lecturing at University of Liverpool and all of these friends have had SEMS using the LS and can contribute a lot of it to that learning curve and that familiarity with those kind of data and that kind of has led to different collaborations such as with Matt as he previously suggested but also experiences using big data overseas which I'll touch on a little bit. So it's kind of good to use for your research and any kind of output you want to produce but also as Matt said it's really useful for your career but it doesn't happen how you apply it. So that's my kind of opening slide to say this is your idea whether it's your PhD, your research project, whatever you start thinking yeah I'm going to get there and then this happens it might be okay or something as well but eventually you will get there so hopefully that's how you'll feel. So in order to be what I'm going to do today is talk about selective sorting so that's the key concepts I was looking into and then talk about the census migrator so what kind of things can you do with these different data types but then really focusing on what I did with the LS rather than the cross-sectional data and then talking more about how those experiences helped me use other data in different contexts so that is what I talk about different journeys so it's my little journey starting off at Leeds then going to Queen Mary here in London via University of Auckland and now I'm looking for where I'm now and all of that has kind of so much to do with the experiences I had using the census migrator and the LS. So this is an overview of my thesis so these are the kind of the methods that I use, the objectives that I have and the data types that I address and so I didn't just use the census migrator, I was perhaps a little foolishly ambitious and also tried to stick in with the Health Service of England which I did and it was good but I'm not going to talk about that today so that was one data set that we're not going to focus on but all of the different objectives that I had could be addressed using the samples of unwise records like other I still call it the stars rather than the census migrator and the ONS longitudinal studies and you can use these data sets to get at different angles of the same picture which is what I'm going to talk about as we go through. Using similar methods whether it's something like standardized illness ratios or reduced regression and looking at how these different measures can be used to illuminate different aspects of ethnic inequality and health which is what I was really interested in under the guise of the concept of selective sorting which is what I will talk more about today. So the focus is really going to be on how I use the LS but I will touch on other aspects as we go forward. So before I talk more about that data and what I did with it, what is selective sorting and why do we care about it? And Matt kind of touched on this a little bit in this idea of selection of facts with international migrants but I was interested in selection of facts going on with internal movers or perhaps also in terms of social mobility. So we know that the population is selectively sorted, this selected sorting idea is different by circumstances and that can be through migration at various different scales so we can think about it in terms of residential mobility or we can think about it in terms of social mobility and typically these processes, these sorting processes are thought of in distinct ways. So you kind of either talk about social mobility or you talk about residential mobility or migration but they're interlinked. People often move house in tandem with some kind of change in their social circumstances so they are related to each other. We know that selective as opportunities for migration or social mobility are varying depending on our own individual level circumstances so your social economic status. Also the area in which you live so the kind of opportunities that are available to you both in terms of education, labour market or the kind of further areas you might move to. So it's also selective on ethnicity and on health. So it's the key things that I was looking at. What I was then interested in is how this process of selective sorting might contribute to health inequalities and whether that's going to operate differently in different ethnic groups. So the core idea is that the movement of differently healthy groups between area types and social classes might influence the health profile of those different areas or of those different social classes and then does that happen differently for different ethnic groups. And the core idea is if this collective sorting could it be contributing to changing health gradients? Is it leading to them to widen? Is it just maintaining them as they are? Or is it narrowing them and constraining them? So that's the idea that I was looking into with these different data. So why census my data? I'm not going to go into this in great detail because a lot of it is coming through all the presentations we're already having. But we know that we have these different data steps and we can look back to 1971. So with the cross sectional data this is from 1991, 2001 and 2011. And this is good and different from the LS because you have much larger sample sizes. You can still look at relevant issues in terms of ethnic inequalities, in terms of migration, in terms of health and that granularity of ethnic detail but you can't do it in a linked way. So you have the full coverage of census variables and if you're identifying migrants all you're doing is using a one year migration indicator. So in that census in that data you have it says is your current address different from one year prior? And within that then you can identify people who have moved to the UK in the last year or have moved within the UK in the last year or who have stayed so haven't moved. And you can have that at 1991 and 2011. So that allows you to look at one aspect of this picture. Then you have the longitudinal data. So this is the one census sample so it's much smaller so you have different benefits but you can actually follow people through time. So you can see that as people move what is it actually doing to health gradients between different Arabic types or social classes rather than just looking at associations between the health status of movers and non-movers for a double. So again you have the full coverage of census variables but your migrants rather than a one year migration question are now identified by a 10 year migration indicator. So is the address different from 10 years prior? So you're getting different angles of the same story. And what I was interested in is these close co-polls of people. So people who were present at the 1990s in 2001 census or in the 2001 and 2011 census. And this adds in loads of other questions particularly when you're comparing it to existing work for example is a 10 year time period enough to look at these health differences should you be looking at a 20 year time period? What happens if you go back to 1971? All these different issues. So with the LS and it's separate from what we did with the SARS. So with the SARS I was really just looking at associations. So modelling health outcomes for movers compared to non-movers for different ethnic groups. But here there are three main approaches. So you could compare gradients. You could say what are the health gradients by deprivation which is what I was looking at different area types for people who have moved and people who have remained in the same place. If you say okay the population is all allowed to move you're letting them be mobile. What does the deprivation gradient look like for health? But what happens if you then put that population back to where they started? Did you see that that movement meant that the gradient from the most of the least deprived areas widened in quality along the gradient? Did it maintain them or did it constrain them? And you can calculate that by looking at the ratio between the most and least deprived areas. Another way of thinking about this instead of just comparing the gradients is actually looking at the health status of people transitioning. If you move into or out of the least deprived areas. If you move into or out of the most deprived areas what is that doing to the overall health gradient? So it's quite complicated. It's quite a lot of things to be holding on to but if you can hear then so if you have moved as migrants and say this but you also have people whose area type might have changed even if they haven't. And what is that combined influence of people moving between area types or the area type changing on overall health gradients that you see? But both of these scenarios are really just interested in the top and the bottom of the deprivation gradient and that kind of forgets everything that's going on in the middle which is why most of us have desired anyway. So another way to look at it then is calculating the slope index of inequality and the relative index of inequality. Again thinking about what happens if people move and what happens if people don't move. So don't worry about grasping all the kind of intricacies of this. This is just kind of showcasing the kind of different things that you can do with these data. So a few conclusions. So first of all that first analytical approach. So we have for 2001 in 1991 period and the 2001 to 2011 period and you're comparing two gradients. You're comparing health status at the end of that study period to either 2001 or 2011 by where people live in 1991 or 2001 so the start of the study period with the health period as it is where they live at the end of the study period. And you want to know what is the gradient, what is the difference, what is that gradient ratio for when people stay where they are or when people move. So just to kind of make it clear, this dark bar is what happens is the health status by where they end up by their destination deprivation quintile. The lighter bar is where they started. So does the movement appear to widen, narrow or constrain health gradients? In this case it appears that movement allows people to move when they move between deprivation quintiles over the 10-year period. It could contribute to widening health inequalities. You see that the difference between the most deprived and least deprived has widened. So that's one way of looking at it and that is our conclusion A. Healthy gradients by deprivation are steeper when groups move within and between deprivation quintiles then occur when the population is put back into their origin quintile. So movement is an appearance to exaggerate health gradients. But how is that happening? So if we take the example from 1991 to 2001 or 2001 to 2011 and then try and walk through this with the 2001 to 2011 example, but what you have here is you have the standardised illness ratio for different transitions. So people who remain in the least deprived area people who move into the least deprived area people who move out of the least deprived area people who remain in the middle people who move out of the most deprived area people who move into the most deprived area and people who remain in the most deprived area. And the first thing to notice is that the best health is with people who remain in the least deprived area. The worst health is with people who remain in the most deprived area. And that's true for both time periods. But then what we're interested in is what happens for movers and sayers and perhaps barriers change and what's going on with this standardised illness ratio. For this movement to be contributing to widening health inequalities, all that you need to see is that the people moving into the least deprived areas have better health not significantly just better health than the people moving out of the least deprived areas and the people moving into the most deprived areas need to have poorer health than the people moving out of them. You don't need to get significant difference for it to start playing out in this way. All that needs to happen is that you see that difference and you do. With the stayers, what you're seeing is a slightly different scenario and this kind of different pattern here perhaps suggests that this movement or this change in area is the best maintaining existing health gradients. It's not necessarily constraining it. So conclusion B, transitions into and out of Q1 and Q5 so at least the most deprived by movers will be contributing to widening health gradients and movers churning within the least deprived areas are in better health than stayers who remain in the least deprived areas and movers churning within the most deprived areas have poorer health than stayers who remain. So there's still something going on there with movers and how they move within or between those differently deprived areas. So then we see the same thing but with a slow connection to quality. You're just getting the same patterns but you're seeing that actually with the healthy groups across all of the deprivation quintiles matters for this change in health gradients. So just to kind of show you what this meant for future research and give an idea of why it's a benefit to use this data but not just for your current ideas but to kind of give you that learning curve and as Matt said once you've got to group to BLS anything else seems relatively okay and that is definitely true. You do have a lot of support from the Celsius team but it's not easy and try not understand how it fits together, means that once you've done it with these data you can go to New Zealand, you can be there for five weeks and you can get three papers out of it so it's just like that's good and you'll be able to do that having got to grips with these different data sources. And actually using these data which is big data and linking administrative records it's just so much easier to understand why they matter so much easier to start applying different models so I took the framework that I've done with BLS and started looking at this but instead of with General Health looking at this for residential mobility over a much richer time period and looking specifically at Cardiovascular disease but different ethnic groups in New Zealand so Maro Pacific, New Zealand and Europeaners and you get the same kind of stories which is always a good thing you'll see consistency in different data sets but you'll see playing out in different ways and you can do different analyses having got that grounding in the longitudinal data that you might have used in the LS. So this here is using projection analysis and glossary analysis to start thinking okay in the LS you can say 91, 2001 and 2011 you have three time points. In these later I had 36 quarter so you can analyse residential mobility and relationship with health outcomes in a very different way. This isn't really showing much detail this is just a highlight that it's beneficial to use the LS and you get different experience and you can start using things applying your same frameworks in different contexts such as New Zealand or perhaps you could go to Scotland and other islands and do the same things and you get that experience and the teams that you collaborate with that is a real benefit. Oh and I have to put that up. No, not long. Can you say especially between 2011 the importance between any way So what this is trying to look at is the idea that there are lots of reasons why inequalities change. There's lots of mechanisms going on that are driving changing health gradients and looking at this was just this is the kind of this was a relatively under explored area the ways that movement between area types contributes to these changing health gradients. There is quite a lot of literature on it and it hasn't been looked at specifically in terms of ethnicity and given that different ethnic groups have really different socio-economic experiences live in very different areas have very different kind of historical trajectories in the UK let alone their own individual trajectories so it's the fact the idea that could this relatively under explored issue explain things differently for different ethnic groups and what I found was that these processes did tend to play out across all ethnic groups I didn't go into showing you for every single ethnic group that I did because there isn't time and the patterns were largely the same but then there were some interesting things for example the Indian within the Indian group there was a very very high amount of inequality and the contribution of these movements between area types and social classes at the same time had an important impact Yeah so the particularly broken down by ethnic group and started to try and think okay how are these different groups moving where ethnic groups have settled in the UK kind of shapes their access to the labour market housing and education sector so that will then have an effect on how and where people move so you see and saw in the UK and in New Zealand certain ethnic groups are much more likely to be churning within the most defined areas and this has a really important implication for the household so if people are constrained within a certain area but also having to be more mobile you have to start looking more detail of what that nature of that movement is and what is the effect of that When we look at things like the inequalities and how the training of the labour group which are based on small areas how much should we be thinking about this different possible explanation I think quite a lot for certain groups I think the kind of the positive moves are generally across bigger scales that are captured in small areas but it's these people who are churning and moving a lot in kind of precarious housing sector perhaps in precarious employment they're the ones that are suffering I think that policies when you're thinking about how to address inequalities should be area based but also really focusing on those at risk groups who are vulnerable to moving a lot and I think that that small scale is important but it depends on the area so it's not going to be true everywhere if you've got someone with a very concentrated private rental sector this is probably more likely an issue if you've just got a social rented or an occupied sector it's not going to have the same impact The last area I'm sure this one So if I'm just going to spend this for a few weeks at a quintile of preparation and then you have to snapshot all the time what you have here is these are the time points that we've got and these are seven clusters of people moving in different ways through deprivation so for example are people moving into more deprived areas or are they kind of going like that across the deprivation spectrum are they starting off kind of less deprived and then moving down into most deprived so it's instead of saying that when you're which is what can do there less all you can really do is say this is the difference in deprivation between two quite large time points which is still very informative as I found but here you can say what characterizes more detailed deprivation for these people who are churning are they churning around quintiles four and five for example and what does that do for their final health risk so that's what this is showing this is the identified seven trajectories of moons and what I've done here I guess my question is how do you interpret the wings so the if it takes too long then I'm going to cut you afterwards yeah thanks do you have all the important spot work intentions do you take account of the fact that areas people might not move in an area but the deprivation changes over time yeah so we have people who were movers and deprivation and they kind of move between different deprivation quintiles but then also stayers so this lot pale bar here is people whose areas change during that time period and so they haven't moved but their area has changed and the time scale at which you look at that really matters as you would expect sort of like lag effects and all that kind of thing but here you do see different patterns to the movers but there's still an overall impact on health and quality that gets stronger the longer the time period is thank you