 So welcome to the first of the parallel sessions on COVID-19 and I'd like to introduce our first speaker Eunice Leung from UCL. Eunice is a recent graduate in population health sciences at UCL with experience working in climate change and sustainability. She has worked with UNICEF and UNDP Malaysia to research climate change awareness among Malaysian youth and she's led a statewide school competition to raise environmental awareness among school children. Recently she completed an internship as a built environment not-for-profit, Future of London, where she researched and wrote about the challenges facing the sector in a post-COVID era and ways to facilitate an inclusive recovery. Eunice is interested in the interaction between people and place and the role of the built environment in promoting health and well-being. Looking towards a career in public health and the built environment, she will be pursuing a master's degree in health and urban development this September at UCL's Botford School of Planning. So over to you Eunice. Hi my name is Eunice. I'm a recent graduate of population health sciences at UCL. I'll be presenting a study I did with my supervisor Terry on how changes in social environments have impacted smoking among young adults during the COVID-19 pandemic. To define what we mean by social environments, social environments encompass the immediate fiscal surroundings and social relationships within which defined groups of people function and interact. Social environments are dynamic and change over time as a result of both internal and external forces. And you know for certain from the pandemic that this has forced the way we live to undergo profound changes. So why is this important? Lockdown in social distancing regulations implemented across the globe have induced a range of impacts on smoking behavior, many of which remain poorly understood. Research by Jackson colleagues who found smoking prevalence among young adults aged 18 to 34 years increased by 25% during the first lockdown alone. Additionally, smoking initiation is shifting from teens to young adults. So in 2020, although more young adults have successfully quit smoking, there are also those who have initiated their first cigarette at the same time. Secondly, as mentioned earlier, these social environments have been massively disrupted as a result of the pandemic. Examples of disruptions in social environments include university closures and online classes, moving from independent housing to back home with parents and even transitioning in and out of lockdowns multiple times throughout the year. House smoking patterns have changed as a result of these varying environments subjected to individuals remain unknown. And lastly, exploring these changes are important in informing tobacco control policies in a post COVID era among an age group that largely contributes to smoking in the UK, ever to achieve the UK government's ambition of a smoke-free England by 2030. In this study, we aim to address how smoking behavior in young adults during the pandemic can be better understood by considering changes in their social environments and social interactions and whether these changes have had different impacts on smoking between young adults based under student status before the outbreak. And to achieve these aims, we examine the association between different COVID-19 restrictions defined through time, as well as changes in the frequency of social interaction to the corresponding risk of smoking. And student status before the outbreak was examined to test associations for the both objectives. The Millennium Core Study was used for this study. So this slide shows all the streets from the MCS that was used as the sample in our analysis. The MCS is a nationally representative sample of babies born across England, Scotland, Wales, and Northern Ireland in 2000 to 2002. And the MCS is established to chart the supercensus growth and development of children in the UK and is designed to be a representative of the UK population, which began with an original sample of 18,818 cohort members. Our study used data from the first strip of the MCS from the 2000 to 2002, the 7th strip in 2018, and most recently the COVID-19 surveys that were conducted across three waves from 2020 to 2021. The first wave was conducted during the height of the first lockdown in May 2020, the second a few months later when restrictions were eased and the next in 2021 during the third national lockdown. And using the MCS as a data set over this period of time provides a really great and relevant time to evaluate the effects of different COVID-19 restrictions and the different social environments experienced by cohort members. Smoking in wave three was only measured in those who participated online and had participated in either waves one or two of the COVID-19 survey. So in keeping with the flow of data collection on smoking across these waves, the analytic sample comprised all participants recruited at baseline, followed at age 17 who participated online in wave three of the COVID-19 survey and also participated in either waves one or two amounting to 2,254 participants. Smoking behavior was measured by cohort member smoking status reported in the COVID-19 survey, psychotomized as smoker versus non-smoker. We had three exposure variables which were time, which are meant to show the differences in COVID-19 regulations and changes in social interactions to account for a social environment. And the latter two variables were measured by the questions in the last seven days, how many days that you meet up in person with any of your family or friends who do not live with you. And in the last seven days on how many days that you give help to people outside of your household affected by coronavirus or the current restrictions. And so helping others include activities such as shopping, collecting medicines, checking in on people and any other voluntary work for community use or other organizations. 11 covariates were selected across three different time points at baseline, at suit seven at the age of 17 and across the COVID-19 race, one to three at the ages of 19 to 20. At baseline, time invariant covariates included cohort members ethnicity, parents socioeconomic status measured by home ownership and parents smoking status. Smoking status binge drinking defined as more than five alcoholic drinks on a single occasion. Psychological distress and long-term illness were included from suit seven at age 17, whereas sex and country of residence were included at age 19 to 20. And time varying covariates at age 19 to 20 included participants' relationship status and whether or not they were living with parents. Student status was included as a modifier and this was based on participants' student status before the pandemic on whether or not they attended university. So using random intercept poison regression models, we examined within person differences in current smoking by levels of COVID-19 restrictions over time and social interactions, controlling for covariates at birth age 17 and ages 19 to 20. And analyses were adjusted using non-response wave to provide it in wave three of the COVID-19 survey to account for attrition. As you can see from the table, these associations were tested by building partially adjusted and fully adjusted models with the three main exposure variables with the different covariates. Time was modeled as a categorical variable using wave one, which is made 2020 as a reference category, and time invariant covariates were included on all of these models. Our main findings included that in this age group, compared with the first national lockdown in May 2020, the easing of COVID-19 restrictions in later months was associated with an increased risk of smoking. So compared with the first wave in May, young adults were 74 percent more likely to smoke in wave two, which was September to October 2020 when lockdown restrictions eased and 44 percent more likely to smoke in wave three, which was February to March 2021 during the third national lockdown. Secondly, a higher frequency of social interactions with people outside the household was associated with a higher risk of smoking. Participants who met people outside their household twice or more times a week were more likely to smoke than those who never left their household to do so, and participants who helped people outside their household however did not have a different risk of smoking compared to those who never left the house to help people. We also tested differences in the associations of three main exposure variables by university student status by adding interaction terms and from the interaction tests we found that the risk of smoking across time points differed by student status. So compared with May 2020, which is denoted here as wave one, there was a larger increase in the risk of smoking in wave two, which was September 2020 among university students than non-university students, and this difference was persistent in wave three. So compared with the first wave, there was a larger increase in the risk of smoking in February to March 2021 among university students than non-university students. There was however no significant difference in associations for both the social interaction variables. There were several limitations in our study, including a low response rate. So the analytical sample for this study was restricted in that smoking behavior measured in wave three of the COVID-19 survey was asked only to those who had participated in waves one or two and used the web questionnaire online in wave three, and this reduced the potential sample size available for this study by excluding those who participated over telephone and the newer recruits in wave three. The findings are limited by the number and timing of the waves in which data collection is produced in the COVID-19 study, analyzing more time points spread across other lockdown stages may have yielded different findings due to the ever rapid changes in COVID-19 regulations. We also would ideally have liked to measure the places in which participants were meeting and helping other people. And for example, whether this was done in leisure places that sell alcohol to explore the nuances in the types of social interactions that encourage smoking in this age group. These findings contribute to a growing body of work that highlights the magnitude of fluctuations in the risk of smoking among young adults since the start of the pandemic and the importance of social environments in understanding the mechanisms at play, including across socioeconomic groups. The study supports the need for a continued attention to smoking prevention among young adults and new place-based interventions where they be more susceptible to smoking. The larger changes in the risk of smoking among university students also cause attention to the role of higher education institutions in supporting these smoking prevention efforts. Thank you for listening. Thank you very much for an excellent presentation and because everybody's automatically on mute, I'm clapping on behalf of everybody. Okay, so thank you very much. And we will now turn to Andrew Bryce from the University of Sheffield where he's a research associate. He was part of a research program funded by the Health Foundation looking at the effects of health on labour market outcomes, including the impacts of the COVID-19 pandemic. He's currently researching the Disability Employment Gap in the UK in a project funded by the Nuffield Foundation. So over to you, Andrew. Thank you, Jenny. Can you hear me okay? I'm coming through. Yes, fine. Great. Okay, fantastic. Yeah, so my name's Andy and I'm presenting joint work with Mark Bryan and Jenny Roberts and we're all at the University of Sheffield where I am right now. And the name of this paper is Employment Related COVID-19 Exposure Risk Among Disabled People in the UK. And I want to say thank you to the Health Foundation for funding this work. So first of all, just a summary of the work that we did. So there's a lot of evidence, for example, from the World Health Organization that workplaces can be a fertile territory for the transmission of the SARS-CoV-2 virus which causes COVID-19 and particular workplaces are more risky than others. And what we found in our study was that in the UK, disabled people were significantly less likely to work from home and more likely to be working outside the home during the pandemic. So it seems that disabled people were more at risk of picking up the virus through work than non-disabled people. Not only that, but there were also more highly concentrated in occupations with a high risk of exposure to COVID-19. So this all adds to evidence that disabled people experienced a particularly raw deal during the pandemic and helps to explain why three in five people who died from COVID-19 were disabled. Now at the start of the pandemic, I remember it being said that the virus doesn't discriminate, but actually lots of evidence that has been collected since has really thrown that out the water. The virus very much does discriminate. In fact, you could say that we're all in the same storm, but we are in different boats. Now, when I first put together this slide, it was noted to me that putting a cruise ship here probably wasn't the most appropriate thing, because actually when it comes to the spread of COVID, cruise ships are actually one of the most risky environments for the spread of the disease. And so actually, you're probably going to be much better off marooned on your own on a dinghy in the middle of the ocean in terms of COVID safety than you are on a packed cruise ship. But hopefully the pictures do sort of give you an idea of the analogy here. So clearly when you're on a stormy sea, you'd much prefer to be in a big boat than in a little dinghy. And so it's been very similar with the COVID pandemic. So to give an example of that, there's been a huge amount written about the disproportionate effect of COVID exposure, and indeed COVID deaths among people of Black and minority ethnic groups. But there's been less written about disability and how COVID has a disproportionately impacted disabled people. And this is in spite of this very sobering fact that 60% of COVID deaths up to November 2020 in the UK were among disabled people. Now, of course, you can argue here that that's maybe not surprising because of the profile of disabled people tend to be older on average and more likely to have underlying health conditions. And therefore, if you do get infected, it is more likely to lead to serious illness and death. But the ONS data that looks at this, in fact, even when they account for age and underlying health conditions, which indicate as a clinical vulnerability, disabled people still had higher death rates than non-disabled. And so we're clearly more exposed to the virus in the first place. And so this is sort of the motivation for this study. So let's have a look at the data that we use. Well, we use the Understanding Society. And as you saw in Meno's presentation, we looked at the, we used the COVID waves. So these were additional, initially monthly, up to July 2020, and then bi-monthly, on to March 2021 online surveys conducted with the Understanding Society panel. So we found this to be really rich data to really find out about what people in the UK were doing during the pandemic. So we particularly used data on whether people were employed, the number of hours they worked in the week, and whether they often or always worked at home. And this enabled us to split people up into those who worked at home, those who were on furlough, so they were still employed, but they didn't do any hours. And then the remainder, which is those that would be continually do a positive number of hours, but were not working at home, and therefore were going out to work, and therefore coming into physical contact with others. We also matched in data from the latest wave of the regular Understanding Society survey. So that was wave 10 or 11 to identify the health-related characteristics of people in the sample. So we were able to split off disabled people, and these were defined as those with a long-standing physical or mental impairment, illness or disability, who also, when presented with a list of particular specified functions, for example, mobility, memory, lifting, etc. If they indicated that they had at least one of difficulty with the least one of these functions and had a long-standing physical or mental impairment, then they're identified as disabled. And we feel that this is very close to the Equality Act 2010 definition of disabled people, which is what we still use today in terms of equality legislation. And moreover, for those who were employed, we were able to identify their occupation in 2019 before the pandemic. And then this occupation was then matched to what was called what we called a risk indication factor, which is based on work by Kukuchi and Kirana, who used job quality measures from Onet to identify the extent to which each occupation in the UK classifications, the extent to which they were at risk of spread of COVID. So let's just explore that a little bit more. So essentially, the Onet provides data on a whole range of job characteristics, and these were the two that we particularly focused on. So physical proximity to others, the extent to which how close are you actually in contact with other people while you're at work. So for example, hairdressers were actually the top ranking one in terms of physical proximity to others. And then the other job characteristic was exposure to disease, the extent to which you're exposed to disease in your work. And here, it was a factor at nursing and midwifery professionals who were at the top of that list. And by combining those two characteristics, so those that scored high on both those measures were the most high risk occupations. And indeed, the majority of these top risk occupations were in the health and social care sectors. And a lot of these were key workers who were continuing to go out to the workplace during the pandemic. So let's just look at our results. So the first thing to note is that disabled people were less likely to work from home. So even back in January, February 2020, before lockdown, but arguably, you know, the virus was still circulating around at this time, disabled people, the blue line was significantly less likely to work from home than the non-disabled population. And this gap increased in April 2020 when we went into lockdown. And in fact, we sustained right through the pandemic. And even if we take into account people who were furloughed and therefore were staying at home for that reason, we still find that in most months, disabled people were significantly more likely to be going out to work. Just over a minute left. Okay. They were more likely to be going out to work during the pandemic. And also, when we look at this risk indication factor, so this is among people who were working outside the home, we see that disabled people were on average in more risky jobs than non-disabled people. So what have we learned from this? Well, the first thing to say is that disabled people actually are much less likely to be employed than non-disabled people. So that is an inequality in itself. And so the UK government wants to close this gap. And this is an adorable aim. However, what we show is that it comes with risks. As Laura said, we're saying earlier, COVID-19 is still with us. And even if we get to the end of this pandemic, there's still a risk of further similar diseases coming through. And so we do need to learn from this and make sure that it's possible for disabled people to get into work, but for it to be a safe and inclusive place for them to work. And I've just noted down here a few ideas that I think would help to address this balance. Thank you very much, Andy. And we'll move on now to Chris Deeming from the University of Strathclyde, where he's a senior lecturer. He has interest in secondary data analysis, subjective wellbeing and social attitudes. He's currently working on a UKRI COVID-19 project, analysing ONS secure research service data. And he's going to be talking to us about coronavirus restrictions and subjective wellbeing. Thank you very much. Okay, so I'm talking about a small project that I'm currently working on, which is funded by the ESRC. And it's involving ONS survey data from the OPN, and that is the Opinions and Lifestyles Survey. And I'm specifically interested in the impact that lockdown laws and restrictions had on people's self-reported or subjective wellbeing. So this is what I'll address in my presentation. Okay, so the motivations for this study really were, in theory at least, I'll come on to the practicalities of actually doing the assessment, but I was interested in the impact of the pandemic on subjective wellbeing and also the duration of the pandemic and also the impact of restrictions on subjective wellbeing. I'm interested in the UK as a whole, but also in the nations of the UK, England, Scotland and Wales. And I'm interested in particularly the impact of school shutdowns, work closures, and stay at home measures in the pandemic, and the furlough as well, the furlough scheme, and particularly the impact of these policies, these public health policies on subjective wellbeing. Can we can we assess the impact of the pandemic and also the public health policies on subjective wellbeing and on different sections of the population? So I'm interested in geography, countries, and also different sections of the population. Okay, so the policy context really are there some UK wide lockdown laws, for example, but there are also variations within the different UK nations and some countries devolved administrations actually making their own lockdown laws. So there is some variation and for people interested in modelling and statistics, variation is interesting because we can look at the data and start to examine it and try to explain the impact of the different variations and the different measures on people's subjective wellbeing. We're certainly interested in things like the timing of the different lockdown laws, the duration, the stringency of the policy responses in the devolved administrations as well, and the University of Oxford has been doing some work on the stringency, scoring the stringency the toughness really perhaps of the lockdown measures. So we have some good scores to model and we have some good data from the OPN survey. In a sense, it creates, the situation creates what we might think of as a natural experiment and then we can look at the impact of these different restrictions that come into play at different times and are lifted at different times in different constituent countries of the UK and things like the job retention scheme as well, looking at the wellbeing of people who were on that scheme. So the opinions and lifestyle survey is a long-standing survey it was established in 2012, a merger of two existing surveys. It's a cross-sectional survey and subject to the usual issues across sectional surveys. It's a relatively small survey in the outset of 1,100 adults. So it's meant to be a representative to a degree of adults in the British population and it's a long-standing survey that has been used by the ONS to look at issues of immediate policy interest. In the COVID, in the pandemic, the survey was increased to a much larger sample size up to 6,000 adults surveyed more or less on a continual basis and social impacts was broadly defined across a range of indicators really in questions, impacts on caring wellbeing and things like that, particularly impacts on wellbeing. This is a concern here. So this was rolled out in March and has been running continuously during the pandemic and continues to do so. For the study, we pulled 50 waves of this survey data. This survey data is accessible via the ONS Secure Research Service. So we pulled the data throughout the 2000s and 2001 which gives us a large sample size of 150,000 respondents and we can use that then to start to look at the pandemic across time, space and under different lockdown restrictions. I use a lot of different regression techniques. I'm not going to report all the results today because the analysis is still ongoing in terms of the modeling and I'm particularly interested in subjective wellbeing and the OPN survey uses standard subjective wellbeing questions across the ONS suite of social surveys but also a lot of the international social surveys use these questions as well. And if we look at the look at some of the indicators, for example, this is people satisfaction scores. What I'm trying to do is set this against a pre-pandemic baseline which represents February 2019 which is the red line and then looking at satisfaction scores on average in the population in this instance throughout the pandemic and then trying to introduce in the statistical modeling certainly different lockdown measures. I'm illustrating this here with the second national lockdown with the UK one to show the impact of these lockdown measures on people's satisfaction within the pandemic against different policies but also against the baseline measure pre-pandemic February 2019. And so I have the four different subjective wellbeing scores from the ONS satisfaction, the worth wellness, how worthwhile is life and against the pre-pandemic baseline and the average scores throughout the pandemic. This is the average data from the 50 merged or pulled OPN datasets. Also happiness scores, we can see the happiness drops in the pandemic as might be expected, the reported happiness and we can see the different impacts of these and restrictions as people's happiness returned close to the pre-pandemic baseline restrictions in England and in Scotland we see increasing and then the impact of the second UK lockdown in early January. So this is happiness scores and then I have anxiety scores as well and the baseline is low anxiety and anxiety understandably in the UK population is raised during the pandemic initially quite significantly high at the start and then dropping as lockdown measures are introduced and we can also see a spike actually which I've marked on the chart which is the peak daily COVID-19 deaths peaking there and we can see the population anxiety does indeed spike around this time as there's concern about the pandemic and the introduction of the lockdown as well which actually reduces anxiety people seem to their anxiety eases a bit as lockdown measures to prevent the spread of COVID are introduced. Yeah all these things are going on so there is a lot we can do with this data on the whole subjective well-being does decline happiness shows a different story here is that there is some return increase in happiness throughout the pandemic whether the other measures are largely negative. There are challenges of using this data because I'm interested in trying to separate the pandemic effect from the policy effect as it were because some of these things are going on and they're integrated and trying to disentangle this is quite complicated when the emergencies are announced at the global level in the UK when the ONS start to collect data there are also differences between the countries but a lot of the countries follow similar patterns in their lockdown laws to a degree so it makes it harder to model on the on the plus side there are some good variables in the data set which we can use more than the global policy responses to look at people's well-being and that's particular variables and questions related to the experience of home working home schooling and being on furlough. There are issues in subgroup analysis which the ONS caution against doing too much or over interpreting we feel we've got a pretty good sample to look at this so we are worried about this of course but we are aware that we can use this data to at least estimate some impacts of policy in the pandemic. The question is there was an online survey in the pandemic so I did some work looking at our representation of our sample in a sense and we do find it is a bit skewed in terms of age in terms of single people filling out the survey and in terms of a lower education than is reported in other surveys and our sample are more likely to be home owners so we are aware we have a different sample than might otherwise have got from the face-to-face survey work and this is after the correction of the survey weights within the sample as well so it is an example of the sample that does present issues and this project's got various accreditations and recognitions as well which I'll leave there and there's some references and I'll stop there. Thank you very much and this is a very clever use of natural experiment as you say and it's also as Eunice will know I use the OPN data on these four questions in my session for our population health undergraduates on how to measure well-being. That's great, I'll be interested to see your slides. I prefer the short web myself but we have a debate about it in the class. Actually the ONS did the experimental stuff when David Cameron announced the well-being measures back in 2011 and I got the experimental data and wrote a paper looking at the distributions back in 2014 something like that when it was first. With the students we have a debate about a discussion about what timeframe one should look at and I think that's one of the differences between the ONS questions and the WEM webs but we need to move on now so we now turn to our final talk of this session by Edward Webb from the University of Leeds. Edward is a senior research fellow at the academic unit of health economics at the University of Leeds where he's worked since 2016. His primary research interests are decision-making preferences and valuing health and he's going to talk to us now about long-term health conditions and labour market outcomes during the COVID-19 pandemic. Cool, yeah so I'm Edward from Leeds, I'd like to thank my co-authors. This work is funded by the Nuffield Foundation although these are my views not theirs etc. The motivation behind our study is that there's a large amount of research which shows that people with long-term health conditions often have worse labour market outcomes so that could be for example leaving the labour market earlier could be part-time rather than full-time working or it could be earning less conditional on participation in the labour market and I don't think I need to take a long time to convince people that COVID-19 has obviously massively disrupted working patterns given that we're all still looking at screens rather than actually talking to each other. So the idea is that okay well has there been a disproportionate effect of the pandemic on people who already had some problems with their labour market participation and I think this is quite interesting because there are lots of ways that you might think that this could be a negative disproportional impact because it could be that adapting to new ways of working is just more difficult if you struggle with resilience or something like that. But on the other hand it could be that some aspects of this disruption to working patterns is actually positive for people with long-term conditions so they might find it easier to work from home that might benefit disproportionately from avoiding a long commute. So our general strategy is that we use the understanding society both the main and the COVID-19 surveys which we've heard plenty about already today. So our idea is that we can identify participants with various long-term conditions using the main survey. We can then match these to people without a given condition on baseline variables and by baseline I mean the far off Halcyon days of January and February 2020 and then we can track their labour market outcomes in the nine waves of the COVID-19 survey. So that's all well and good so we have a strategy to identify the causal effect of having a given long-term condition in that we match people on their baseline characteristics but I've already said at the start of the presentation that people with long-term conditions often have poorer labour market outcomes anyway so how do we identify whether well if we just track people over a given 18-month period that we just see worse labour market outcomes develop over that time. So therefore we construct a kind of sort of counterfactual analysis by looking at the relevant outcomes in wave seven to nine of the main survey. So this is the last three waves pre-COVID and there's around about two years on average between people's first and their lowest observation. So the idea behind that is that we can say that okay well if we see an effect of having a long-term condition in the COVID data we can see would you just expect to see an effect of a similar magnitude in pre-COVID data. So how do we identify participants? Well basically we identify people having a given condition if they ever say that they have it prior to March 2020 so these are the specific variables we look at. In understanding society there's a bit of a change as to how you ask the question about long-term conditions in wave 10 onwards specifically distinguished between for example different types of arthritis we don't look at the kind of the sub-categories so we don't look at say rheumatoid versus musculoskeletal or no rheumatoid versus osteoarthritis for example we just stick to the main category. These are the specific long-term conditions we look at so we look at asthma arthritis cancer diabetes liver condition epilepsy in wave 10 there's this category of emotional nervous or psychiatric problem introduced and we combine that with the main mental health related condition asked in waves one to nine of the main survey which is clinical depression. We also there's a group of conditions which we put together and call vascular conditions and there's group of conditions which we put together and call pulmonary conditions and we don't look at hypothyroidism or hypothyroidism essentially because they were apparently going to be quite boring. I should say that an actual you know proper doctor has actually looked through these categorizations so there is at least some medical knowledge going into this. In terms of our matching variables we match on kind of the usual things so age, sex, ethnicity we match on their people's work experiences in January February 2020 for example how often they work from home pre-pandemic we also look at things like whether they identify as a key worker in waves one and two of the COVID survey we match on their job type household size location we also match on the number of co-morbidities that people have. So how do we match people so for each condition we impute their missing baseline variables it's done using the Miss Forest package for R and then we use Mahalan novice distance matching so this is a quite similar to propensity score matching there is certainly a paper out there which argues that it gives you a closer match on all variables I think matching is one of those things where there's actually quite a lot of different methods and everyone has their favorite ones this happens to be the one that we use. We also vary the matching ratio based on how many people we find with the given condition so for people where there's very few people with a given condition we match on a ratio of four to one then where we get over a thousand people with that condition we just match one to one. These are our outcome variables we look at so we look at a binary variable whether people are employed or not we look at the number of hours people work conditional on employment we often we look at how often people work from home conditional on being employed so that's either always hybrids or never we look at whether people are furloughed at any point in the COVID-19 survey so there's various furlough related questions asked in the survey and it's quite complicated to combine them so we just say okay binary variable at any point did you say yes I've been furloughed we also look at earnings conditional on employment household income and only for people who are under 65 and not on universal credit at baseline whether I did any point during the pandemic they start claiming universal credit we use cross-sectional models for furlough and universal credit for other models we use random effects panel models and in these panel models we include a time trend and we also interact that time trend with whether people have a long-term condition or not. So just a very quick summary of our results so the amount of people with the conditions vary a few minutes left cool and which we get over 5000 down to epilepsy where there's only 130 people this is a very broad summary of our results so basically here for the panel models I'm showing you whether the main effect of having a condition and the interactive interaction of having a condition and the time trend is statistically significant and if it's significant what area it goes in and you can see that there's there are interesting patterns here and similarities between some conditions and some differences as well as well so we can see that and if we compare that to the counterfactual COVID-19 analysis basically we find very little there so there's very little when we look at the the main survey data just before COVID-19 so it looks like there are various effects of having a long-term condition in terms of worse labor market outcomes and we reckon that this is probably due to COVID-19 there's some similarities between conditions especially for most for people with most conditions they have a lower likelihood of employment there's also some differences for example we only found a significant increased likelihood of being furloughed for people with pulmonary conditions and exploring that they also people with pulmonary conditions also had the biggest reduction in working hours so we can think there might be various reasons for that so it could be that employers are driving this so they saw people as most at risk and therefore were least likely to ask people to come in or it could be that people with pulmonary conditions were much more cautious there's some good news so there's no we don't really see much of a reduction in earnings conditional unemployment but it could just be that it takes time to change people contracts so it might just take time for the the effect of having a long-term condition during the pandemic to feed through there's also little difference observed in working from home patterns so we reckon that that could be because a lot of the changes were legally mandated and we would need to look in future have evolves this is my last slide so people with long-term conditions it looks like they need increased support due to being in the due to COVID-19 so increased support especially with staying in the labour market and it doesn't look from our data like the outcomes are equalizing by September 2021 so in future it'd be useful to see whether effects persist and also I think it's quite important to investigate the causes behind different labour market outcomes so we don't also know whether these changes are necessarily worse for well-being because maybe actually people who left the labour market with a long-term condition might that might actually beneficial for them and so yep that's the end of my presentation thanks for listening hello everybody I'd like to start by introducing Deborah Chilequa who we're delighted to have a third year medical student from Hull York Medical School joining us and hopefully a start of a very long and distinguished career using health studies so Deborah's conducting research that explores socio-demographic inequalities in relation to cervical screening uptake in the UK she has completed a degree in biomedical science and contributed to research focusing on developing therapeutic interventions for vaginal and urinary tract infections so over to you Deborah and so and hello my name is Deborah Chilequa and the title of my study is a cross-sectional analysis of ethnic inequalities in cervical screening uptake in the UK using understanding society so this is just an information on cervical cancer and why are we focusing on it so globally cervical cancer is the fourth most common cancer in women so it does affect a large population of women sorry and then to combat its incidence the UK developed a programme where the human population of virus vaccination was introduced to reduce the incidence of cervical cancer unfortunately the vaccine does not prevent or cervical cancer so it is very important that women still go routine cervical screening and this is because the 10 years survival rate of cervical cancer is only at 51% and this means that 49% of women actually die within 10 years of being diagnosed and this is rather unfortunate because it's approximated that 500 deaths are prevented each year for a relatively simple procedure which is cervical screening so essentially this is just a cervical screening flowchart so you can follow along with what exactly happens women who are 25 to 64 years old are invited for cervical screening the patient cervix is examined through this lovely piece of equipment on the right called a speculum and the pap smear is taken and sent to the histopathology lab and one of three results will be given either you'll have an infection either it's normal or it is abnormal which then further examinations of the women's cervix will be undergone despite the relatively simple procedures and cervical screening is at an all-time low and it's actually a 19 year low in England and that is the same across all age groups social economic inequalities as always thought to be a major contributor to low cervical screening uptake which with low depravity in socioeconomic sector being associated with a lower uptake and more cervical cancer mortality so essentially this graph is showing this is a study that shows that cervical cancer screening is the lowest in the lowest quintile of income deprivation in England and unfortunately you cannot see my mouse but what is what if you can kind of just make out that in the first bar in the first bar of 2007 the least deprived areas had a higher percentage coverage and in the most deprived areas so the quintile and the lowest quintile of income deprivation in the most deprived areas had a lowest percentage of cervical screening coverage and this was the same from 2007 all the way to 2012 so low income deprivation in England is associated with low cervical screening attendance so what we do know is socioeconomic inequalities is one factor against that has been extensively studied that affects low attendance and cervical screening less attention however has been given to ethnic inequalities as a contributor of cervical screening so like do does and women's ethnicity alone affect how often she attends and if she attends for cervical screening this is in part due to poor recording of ethnicity in primary care and some studies have attempted to examine the association between ethnicities and ethnicity and cervical screening there was a study for example that found that white British women were more likely to attend cervical screening compared to minority ethnic women the problem with the study however was it was unable to disaggregate women beyond white British so it was white British women in comparison to the rest of the ethnic minorities in the UK and so that brings me on to thank you our next slide which is where we now talk about what we wanted to do so we wanted to look at ethnic groups in those minorities and see if there's a difference in cervical screening uptake within them and so we use the national understanding society data set which allowed us to compare a wider population of ethnic groups so it brings me to a question which is the study was aimed to look at whether ethnicity is an independent predictor of cervical screening uptake in the UK and we controlled for socio-demographic factors which I will show in the next slide thank you so as mentioned we use the national representative survey understanding society and sampled 12,006 minority women age 25 to 64 living in the UK we did this and we used a weighted and unweighted sample in the next steps we did cross tabulations to examine the relationship between socio-demographic variables on the right and cervical screening uptake socio-demographic variables and considered where age ethnicity religion educational status economic activity IMD score access to a car or van the number of GP visits and what makes them 12 months whether they're born in outside of the UK having English as their first language their social economic activity long-standing illness or disability and their region and then after this the significant variables were then put into multivariate analysis using logistic regression to determine if ethnicity was in fact an independent predictor of cervical screening uptake and next slide please so the results of the step one the cross tabulation analysis were proven fruitful because we were able to see that ethnicity was a significantly related to whether a woman attended cervical screening and more with other socio-demographic variables and this was the case of both weighted and unweighted data I will first show the weighted data the table was quite small I do apologize it's getting everything into one table so in the on the left are all the significant variables and the significant socio-demographic variables and you can see that they're all significant in a screening uptake religion educational status economic status socio-economic classification access to a car van number of GP visits being born outside of the UK and having English as his first language for our study though what was very important for us was that ethnicity like I said was an important contributor and what was interesting in fact as you can see in the red circle Asian women of Asian ethnicity were actually about a much lower percentage 17.6 comparison to white women which was 27.7 and just this is just to say that this was the same for the weighted data as well the only difference was that English being as a first language and being born in the UK was not a significant variable in the weighted but Asian women were also the least likely ethnicity to attend cervical screening and so in our next step was the logistic regression and where we took the socio-demographic variables and wanted to see which were the independent contributors to cervical screening uptake and importantly as you can see Asian women was in fact an independent variable as to whether women did in fact take attend cervical screening sorry having access to a car van was also significant in determining whether a woman went to cervical screening and also the number of GP visits that a woman has was also very as it was also a contributor sorry to where there's a woman went through cervical screening so this is just essentially a summary of the results so after as I said adjusting for the confounding variables and ethnicity was an independent predictor and specifically that Asian women were much less likely 32% likely actually due to the odds ratio for attending severance cervical screening but interestingly there were other factors that directly directly affected cervical screening uptake and women were more likely to go for cervical screening if they had access to a car or they had visited the GP a specific number of times and I think this was from six to ten visits in 12 months then they were more likely to have done cervical screening but however Asian as an ethnicity is quite a broad ethnicity and so we wanted to aggregate look a little bit deeper into what groups in between those ethnicities were also being could have been affected by how often they went to cervical screening and so this table shows with the unweighted data that again as we can as we knew before Asian women were the lowest to uptake for cervical screening we can actually even see further that it was actually Bangladeshi Asian women who were even less likely at 11.7% and followed by Pakistani Asian women and then followed by Indian Asian women and we also split up white ethnicity into British Irish or other whites and black and African into African and Caribbean sorry and mixed and so but we also there was also another variable that we wanted to look at and break down even further which was religion and religion is quite highly associated with ethnicity thank you and so no religion and religion in our cross tabulations were significantly contribute country that was a significant contribute contribute to sorry to whether women went through cervical screening uptake and when we looked at the different categories within that religion itself within religion itself we could find that Muslim women were actually much less a lower percentage of cervical screening attendance than other any other major religions in the UK and so we thought to combine ethnicity with Muslim religion to see if we could add further depth into understanding the type of women who significantly do not attend cervical screening and that brings us to the final table and the sensitivity analysis and to no surprise for us we actually found that Bangladeshi women had the lowest percentage uptake which was 10.7 percent and this was followed by Pakistani Muslims which was 14.8 percent and then African Muslims which had 19 percent and then two minutes other thank you other Muslims and Indian Muslims as well so what does that mean exactly so I think our study supports previous research that ethnicity is in fact a significant factor in predicting cervical screening uptake after adjusting for socio-economic deprivation and other socio-demographic factors in highlighting ethnic equalities and cervical screening it shows the importance of ensuring uptake of screening which is all parts of the population so that a woman who has the same likelihood of being having health care for cervical cancer and it's not impacted by what ethnicity she is essentially this has implications for future research and practice in terms of identifying uptake amongst Asian women and a certain studies have shown that language barriers prevent a lack of awareness in women I think a lot of women in this study were said that if they didn't they were receiving letters to attend cervical screening but they were unable to read the letters so they didn't really know that the cervical screening was something that they needed to attend and psychosocial barriers that may be in relation to ethnicity and religion and some studies were also showing that immigrant women thought that their health was not considered a priority in their home country so now when they came over to the UK their health and going for regular checkups again was still not something that they thought that they should take priority over and so a review of interventions to improve cervical screening uptake could be of benefit to ethnic minority women and specifically maybe Asian women as our studies shown so a limitation of this study was that it was unable to be conducted longitudinally due to the use of cervical screening questions only being asked in wave 10 so hopefully if there's more of this of cervical screening questions asked we can follow along and see what happens in the next coming years with cervical screening so thank you very much for allowing me to present thank you very much I now like to introduce Harrison Smalley from the University of Nottingham this research was conducted as a master's research project at the University of Nottingham by Harrison Smalley under the supervision of Professor Kim Edwards Harrison recently completed his medical degree at the University of Leicester and he has strong interest in public health and sport and exercise medicine he's going to talk to us about understanding the burden of chronic back pain so I'm going to talk about our recent spatial micro simulation study of chronic back pain and ward level in England so first give you a bit of a background about chronic back pain so chronic back pain is back pain that's present for three months or more it's sometimes split into lower or lumbar and thoracic back pain it's a very prevalent condition affecting around 14 to 20 percent of the population it's very expensive as well these figures are quite old now but it's been estimated to carry a direct healthcare cost of £1.5 billion annually and the total cost of the economy through loss of productivity of over 10 billion annually while national estimates give useful information on disease they are a simplification of the true picture so national estimates misrepresent what's happening at a small area level such as ward level by averaging out areas of high and low prevalence leaving areas with an exceptionally high prevalence of a disease to go and notice and that can hinder equitable health resourcing and the implementation of more effective targeted interventions understanding how and why chronic back pain prevalence varies across England could carry substantial benefit for public health planning so this takes us to our aims so our primary aim was to create a validated simulated data set of chronic back pain prevalence at ward level across England using a spatial micro simulation model this data set could then be mapped with their aim to analyze the simulated data set trying to determine why the spatial pattern of chronic back pain prevalence is the way it is particularly focusing on the influence of physical activity finally we aim to simulate the effect of policies to increase physical activity on chronic back pain using what if analysis so in summary we we wanted to know what the current situation is why it is the way it is and what could be done to change it yeah so on to the methods so firstly to understand what's on the screen there and what we want to know what is spatial micro simulation so spatial micro simulation is a technique that can be used for various things it takes a survey file for example the health survey for England that has national level data on an outcome of interest from a representative sample but lacks data specified by small area it then also takes a census file that has demographic data on each small area but no outcome of interest it then matches people from the survey to each small area on the basis of each area's demographics so spatial micro simulation can be used for various things so small area estimation use the small area prediction as well and also policy simulation so in this study we used a spatial micro simulation program called SIMO BST which is named after its original use but has validated usefulness and simulating of the health variables including musculoskeletal disease so firstly HPC data sets containing physical activity were combined with 2011 census data to give a geographically specific physical activity data set and stage two then combined this with health survey data set containing chronic back pain to leave us with a final geocoded data set combined in both chronic back pain and physical activity and these data sets were combined on the basis of constraint variables which are variables that were chosen as best predicting physical activity and chronic back pain on the basis of our logistic regression analysis and internal validation of simulation outputs so I won't go into detail in this presentation on the constraint selection and model validation process but the output of the final simulation was then mapped and analyzed using spatial order correlation and geographically weight progression and finally what if analysis was performed by altering individuals moderate to vigorous physical activity levels in the input health survey data sets and then repeating the simulations so I'll touch on these methods of analysis a bit more and we go through the results but yeah so primarily we were looking to simulate a map chronic back pain prevalence across England so this is the mapped output of our final simulation and darker color indicates high prevalence in this you can begin to see a pattern of high prevalence on these coasts and in the southwest with relatively low prevalence in the southern central area so next we took this output and analyzed for spatial auto correlation so here we have our local Moran's eye lies a cluster map so this map shows areas of significantly high or low prevalence in the context of their neighboring wards so for example high high in the dark red indicates a high prevalence ward amongst other high wards so a cluster of high prevalence whilst high low indicates a high prevalence ward in a relatively low prevalence region so a spatial outlier and then the same follows for low prevalence wards in blue so low loads loads around by low wards and though high is an outlier low war so looking at this you can you can now really see those clusters of high prevalence along the east coast and in the southwest heritature as well on the Welsh borders another notable cluster of high prevalence there's also a relatively large area containing clusters of high prevalence where the boards of South Yorkshire, Derbyshire and Nottinghamshire mean low prevalence clusters can be seen in the south especially in and around London as well as cities of the midlands and north okay so next for this project we're particularly interested in the influence of physical activity or specifically physical inactivity on chronic back pain prevalence it being a possible target for interventions to reduce chronic back pain so firstly if you just compare side by side the maps of chronic back pain prevalence and fitness inactivity prevalence you can see that there's a very similar spatial pattern there we analyzed the association between physical inactivity and chronic back pain using geographical weight of progression so I've only displayed the geographical weight of progression results in a very brief form here but to just summarize what they showed so initial univariate geographical weight of progression found a strong positive correlation between physical inactivity and chronic back pain prevalence at war level this relationship was largely explained in the multivariate GWR model by confounders and these confounders were the proportion of residents that are over 60 in low skilled jobs female obese smokers white or black people and disabled people as well so next we went on to our final aim which was simulating the counterfactual scenarios for increases in physical activity levels so to do this we used a previously validated method which was altering physical activity values of participants in the input health survey data set and then re-running the simulation so we did this for increases of moderate vigorous physical activity of 50 minutes 30 minutes and 60 minutes and we found a detectable reduction in chronic back pain prevalence for increases of 30 minutes and 60 minutes but found no detectable change for the 15 minute increase so what what can we take from all this so chronic back pain prevalence varies at war level across england and generally we've seen that there are clusters of high prevalence predominantly in coastal areas low prevalence in cities an area level physical inactivity is highly positively correlated with chronic back pain and that's largely explained by confounders policies to reduce physical inactivity were likely even so in a significant but relatively small reduction in chronic back pain prevalence yeah so just finally briefly discuss a bit about the strengths and limitations of this study and possible future directions so the real strength of this project was the use of the spatial micro simulation methodology this allowed us to simulate our other variables of interest besides the outcome variable at small area level whilst maintaining the relationship between variables at the micro level we also use two high quality data sources in that of the 2011 census and the four years of the health survey for england that we use and touch on a few limitations firstly the health survey uses self-reported physical activity data which is inferior to using a more objective measure like accelerometer data but obviously this is quite unrealistic of an aim for the requirements of the data set for use in this study and secondly is the term back pain so we gained chronic back pain data from the 2017 health survey and but the health survey for england doesn't differentiate between lower and thoracic back pain so that's why our outcome was just back pain so the problem with that is it affects the comparability with wider epidemiological literature which tends to focus on lower back pain so that's kind of a consideration for future health surveys chronic pain questions finally this study is limited by the use of a static spatial marker simulation model so it means that the time period of the simulation is dictated by the years of the data sets we use to construct it so this isn't a current day estimation of chronic back pain prevalence in terms of future work there's a lot of avenues to explore to imagine a few work could be done to simulate chronic back pain at a finer spatial scale for specific areas of interest so even at ward level you've got that averaging out that I mentioned earlier at the national level so with an award there might be pockets of extremely high prevalence that we're currently unaware of we focused on the influence of physical and activity on chronic back pain but following on from this work it would now be easier to simulate scenarios for changes in other predictors of chronic back pain such as obesity so that would be something to consider and finally a dynamic or pseudo dynamic model could be constructed to overcome the issues that we've got with this out-of-date estimate well thank you very much and we move on now to our final talk our last two speakers Joanna Semlien is an academic psychologist based at the University of East Anglia with expertise in LGBT plus health inequalities and Jane Skinner is a lecturer in medical statistics at the University of East Anglia she specializes in the analysis of large observational data sets so over to Joanna and Jane can you see the slides yes we can yes so what we're bringing to the party is we're a double act and our slides worked so on that note we're um yes we're already a success so wonderful thank you so much for um that introduction and I shall um talk for a bit then I'll hand over to Jane and then Jane's going to hand back to me which um hopefully will all function completely seamlessly I feel like I'm speaking to an audience that are fully aware of the opening bullet point but you know I think it's worth stating that we really need to be monitoring health inequalities it's an important part of public health policy and that sort of underpins a huge amount of work that everybody here today is undertaking specifically though I also want to refer to sexual orientation identity and it's and the need for recording that from there are multiple reasons why I need to keep repeating this because I've been probably speaking to this bullet point for about six or seven years and I still find myself not needing to delete it from my slides because obviously we need to comply with equal opportunities legislation and everyone is of course setting out and intending to do that but the recording of sexual orientation identity is really not happening as standard um at all and um in order for us to comply with equal opportunities legislation but also to allow us to monitor sexual minority health we really need to be recording that um but actually um we don't really have the kind of ongoing data to look at but luckily we have national health surveys um so that makes us very happy here's a little bit of background a little sort of intro for you around um why why are we looking at sexual minority populations why is this important well there's been a growing uh evidence base around health disparities in this group over the last uh certainly growing over the last decade starting with there was a very significant systematic view looking at mental health which was published in 2008 um showing that a significant increased risk um certainly around suicidality around depression anxiety but also substance misuse um we looked started then but that was very US centric so then we managed to um start to analyze data from the health the UK health surveys so initially we've looked at smoking and found that lgb young adults are more likely to smoke and that was using l site which is now um next steps we can also see that we've got so we have mental health we have risk behaviors we also know that at least 80 percent of uh uh this data is about 10 years old now but you know just to give you a kind of line in the sand discriminations and harassment is really significant in this population and there is a theoretical kind of underpinning um called minority stress theory which is and um this this posits really that there's both an internal and an external manifestation of that received prejudice victimization social stigma and that that is what's underlying the health differences that we're observing and this could and and offers a perfectly um reasonable explanation for for what is um what what may be what may be going on um the other sort of background really that I want to present is that um the lgb and oh I'm going to come to the oh in a minute explain that to you but lgb oh health inequalities research as I said there is a kind of there is a uh a growing evidence base but uh it's extremely impoverished in terms of its quality um lots and lots of poor quality papers lots of uh small community samples being used snowball sampling as a technique um questionnaires that haven't been validated and um sadly huge amounts of repetition which the population constantly take part in surveys which really um I find quite frustrating but anyway so we've then got a field that's uh really it's an unfunded research field um lots of unpublished um papers it's an under research population and actually a poorly funded research topic and I can speak from experience it has been a very US centric field and still is really the vast majority of publications are from the United States and um one of the issues you have even when you're trying to understand and look at the data and the the knowledge that's out there is that there's a huge variability in the way in which sexual orientation identity is measured sometimes it's attraction sometimes behavior sometimes the self-identified category and also lots of different data um day papers present different forms of aggregation and disaggregation across the different sexual orientation categories that makes for a complex field and it's not very helpful in that variability so from a UK perspective the ONS standardized um created a standardized question 2009 which was a fantastic breakthrough in terms of the UK situation this is the question that still stands and it um began to be incorporated in the UK national health surveys which was a fantastic game changer in terms of us being able to look at UK the UK picture just very briefly on other you noticed in the previous question one of the options was the word was um was the category other this is important and significant because it is an option other than lesbian gay and bisexual but also other than heterosexual and of course there is an option to um say that one doesn't know or refuse that's a verbal option that's available so it is a meaningful category that people are choosing they're actively choosing that so back in 2016 um when I looked at common mental disorder in this population I included the group other and I did the same when we also went ahead and analyzed looked at BMI in this group and those are the two papers um that are referenced just there and both of those we included other and found um an association with higher risk of poor mental health and unhealthy weight compared to um the comparative groups in those analyses that is a meaningful category but it's very heterogeneous it is a you know arguably a limitation in and of itself to conclude much from it but I still intend and have continued and will today be presenting data on that group because it may well include people who are not comfortable with those categories um and or may have a gender identity other than the male female options that are provided in the surveys at the moment there may be age differences in that as well but younger people are choosing other but nevertheless um it is definitely a category of um I think of of importance so I basically covered what I was going to say in this slide in the previous slide but I'll just draw attention to the fact that we've used a range of data sets using IPD and Jane's going to talk about that next so I'm not going to preempt her um very briefly um I've added this to the slides that are online because I did some of these slides as latest yesterday and didn't include this but I just want to point out that the data set that we used was a pool data set of 100,000 and 100,503 people drawn from 14 UK surveys and I want to give you um before the aim which is the one way around but I just added it the participants um the numbers of people and the percentages that identified in our categories that we have analyzed today these are lower than some of the categories we're now seeing in the data that's being provided by ONS um from the APMS but I think this is because we are using older data sets and Jane and then I perhaps repeat we'll talk about that later I'm going to hand over to Jane thank you okay so um thanks very much for joining us today um as you will have gathered the um aim that we're doing here is to examine the association between the categories that Jane described the LGBTQ sexual orientation identity and we're looking at type 2 diabetes among adults in the UK so um these reflect the data that we were able to um draw down from the UK data service so a few years ago um they stopped making sexual the I think they still ask the question but they stopped making the sexual orientation identity variable available so these are somewhat older studies so the approach we took was individual participant data meta-analysis with logistic regression to examine this association and unlike most meta-analyses we're not extracting data from publications we're using the individual person level data and we are able to identify 14 health surveys that collected data on these lgbo categories and type 2 diabetes and one of the advantages of this approach is that we can estimate associations for smaller groups such as lgbo categories for which the original studies couldn't do because they were underpowered so um the way this approach works is that we calculated summary statistics for each study and then pull those estimates and we used a random effects model because there's a range of years and places and study populations so it's reasonable to expect heterogeneity um so for each study we calculated OZ ratios and their standard errors and these were pulled to produce an estimate of the average effect size for the studies um in Prelimp Joe was talking earlier about how you aggregate the groups we found in preliminary analyses that there weren't different effects for men and women so we analyzed gay men and lesbian women together and adjusted for sex so um we also um for this part for this um analysis did two main sets of analysis one we've called minimally adjusted and one we've called additionally adjusted so the minimally adjusted analyses are very minimally adjusted they just included age and sex as covariates and the additionally adjusted analyses also included smoking drinking BMI category ethnic minority status and education status um we also considered marital uh cohabiting status consumption of five a day physical activity and waste measurements and too few studies had the last two and the first two didn't seem to be associated with the outcome um this is just a single slide and i'm going to move on to the results next but there's quite a lot of sort of hidden work behind this of harmonizing different studies to produce um the same variable and you tend to lose complexity when you have when you do this so for example smoking was are you a common smoker yes no um uh education status was do you have a degree that sort of thing okay so um this is a uh forest plot of our analysis um i expect most of you are familiar with that so i'll just very brief if you're not i'll very briefly say that um each each study contributing study is a row and the plot shows the odds ratio and its associated confidence interval estimated from that study and the diamond at the bottom shows the overall estimate of risk so which was for this analysis it's gay um and men and lesbian women um compared to heterosexuals adjusted for age and sex so um the um line coming down one is obviously the line of no effect and you can see um even though the individual um studies might not be statistically significant when you combine them we do get a statistically significant result which shows that gay men and lesbian women are um have higher odds of um developing of having of sorry of having type two diabetes and um when we compare that to the um additionally adjusted model which includes all the covariates that i mentioned on the previous slide or the last slide but one you can see that the effect remains so it's suggested that it's not explained by these these covariates so we get a similar result for um bisexual people they are at um their show increased um odds compared to um heterosexual people of developing type two diabetes and again that remains um we're moving on to the additionally two minutes oh sorry i've given you a bit of extra because of the slide okay sorry okay so um and similarly increased risk for the other group that joe mentioned um and but that doesn't remain after adjustment so um all three groups showed increased risk with um minimal adjustment and the effect remained for lesbian gay people bisexual people um after additional adjustment for additional factors so this is the first study using this data to do this analysis um one proviso is that um if you're very eagle-eyed you might have noticed that there weren't 14 studies in all of the forest plots so because of the small numbers some of the studies had no diabetes cases for some of the sexual orientation groups which meant that they couldn't be included in the analysis um which would bias the estimates upwards okay so moving over to joe very very briefly um three points um firstly a cry out that sexual orientation needs to continue to be included as standard in all health surveys um my third point which i'm going to say second is that actually i'd like us to be monitoring sexual orientation as standard in all clinical settings to allow us to target health interventions but also that we need access to this data so we need to be able to um analyze data beyond around we've got i think it runs out around 2015 we can no longer access um that data so um at the moment we have a real gap um in in what we're able to uh look at so that's a kind of call for um something to help us to be able to access um that data but that's that's our findings that's um our sort of final points and that gives you probably no time for questions but you can always email us