 Mae wedi bod ei wneud hanesu'n didn digwydd i DVD, sy'n ei bwysig ar bethau y bydd yn popethau ffordd penderfwylo ar unig ym Mhom Nwysg Plyg i dddysgu'r ffordd ym Gwydol. Ond eich y gallu'n ddefnyddio'r progect ym ymwyaf芽edau a'r ddigwydd gwyddiant i ddefnyddio'r gwasgau iechydig arno cyflos o'r dda. Davydyd hefoedd yw ei ddwy'n cyfrnân gyffredig folk焔 i'r Unedig neu hefyd i bobl cerddiau ym Mhom Nwysg Plyg ac cross cohorts, trends in mental and physical health using the British birth cohorts. Before his PhD, David completed an MSc in clinical mental health sciences at UCL and worked on research related to mental health provision among young people and in migrant populations. And he's talking about adverse childhood experiences and multiple mental health outcomes for adulthood and analysis of the 1958 British birth cohort. Okay, hello everyone. Thank you for the introductions and thank you for having me. So I will present a body of work on adverse childhood experiences and how they are linked to mental health outcomes throughout adulthood. And this is the analysis of 1958 British birth cohort. And I would also like to thank my co-authors. So adverse childhood experiences typically are defined as traumatic, stressful, psychosocial events, conditions, when tend to co-care, persist over time and outside of child's control. And these tend to include abuse, different forms of abuse, physical, emotional, sexual neglect, but also household dysfunction, for example, mental illness among parents, interpersonal violence, divorce, substance abuse, events like that. So there's a lot of research showing that ACEs, so adverse childhood experiences, are linked to mental health, that are bad for mental health, even several decades later. And despite the large amount of literature, there's still a lot of limitations which I think we can address, especially with our great cohort studies. So for example, ACEs tend to be reported retrospectively rather than prospectively. And this has a lot of biases. A lot of research is cross-sectional rather than longitudinal. Tends to focus on individual mental health outcomes rather than a range of outcomes, which makes a comparison of different outcomes, difficult direct comparisons. Quite often the studies are not adjusted for important confounding, for example family characteristics, because simply this data is not there. And they focus on individual ACEs or some kind of summed up scores rather than exploring both types of exposure. So this research was eventually split into two separate studies, and that's how I'm going to present it. They're both under review, and they can be found online as pre-prints. So I'm not going to try to focus on the findings without borrowing you on the pathological details. So you can read up on those yourself. So study number one looks at how strong is the association between ACEs and a range of mental health outcomes measure between age 16 and 55. And in study number two we looked at how strong the association is between ACEs and trajectories specifically of psychological stress between age 23 and 50. So the idea behind this research was to use this so-called outcome-wide longitudinal design. So it's a term coined by van der Veel and he proposes that when we have the data on several related outcomes and we're interested in a given exposure, we should include those outcomes. We should just have them there in the study because then it's just much more efficient. We are looking at the same exposures. You will look at the same covariates, the same sample, so we can directly compare effect sizes. So when this big meta-analysis are being done, it's just we make life easier of the researchers and of the readers. So that's what I try to do in my work. And I'm using, as I mentioned, I'm using the NCDS data. So the birth cohort of those born in one particular week in 1958, the current samples around 8,000. My ACEs were measured both prospectively and retrospectively. And all the findings, all the estimates are being here are adjusted for the same covariates. So the gender, fathers, occupational, social class, maternal education, birth weight, gestational age, maternal age of birth. And breastfeeding duration. And the range of outcomes I'm going to present includes life satisfaction, quality of life, self-rated health, psychological distress. So that typically includes symptoms of depression, anxiety, any kind of medications taken for mental health problems or visiting mental health specialists. So this is my ACEs. Sorry for the table. But the key here is that prospectively we measured things like separation, divorce, substance misuse, family conflict, death of parent, mental health problems among parents, physical neglect, offending. But prospectively we also included things like abuse, different type of abuse and emotional neglect. So this is just where the type of adversities we were able to study with the data. Okay, I'm going to slow down a little bit here. So this is the key findings of study number one. So this shows the association between number of adversities, both measured prospectively and retrospectively. That's on the Y axis and different outcomes at different ages. I should probably mention here that the outcomes are standardized. So the beta coefficient represents standard deviation so we can then talk about effect size in standard deviations. And as mentioned, a range of outcomes were included. The key message here is that whatever the outcome we are looking at, there are strong effect sizes. The association is pretty strong. Most children who experienced adversities tend to have worse mental health. And for example, if we look at, let's say, having three adversities versus none, we can expect around half a standard deviation of worse mental health on a range of different outcomes on average, which is a pretty strong effect size. And probably there's a marginally stronger association for life kind of well-being measures such as life satisfaction or quality of life and measures of depression, anxiety or general psychological distress. And this figure shows a similar story here, however, included binary outcomes. So for example, seeing mental health specialist or nursing mental health specialist within a certain time period. And also whenever it was possible, I binarized the continuous outcomes. So when there were some thresholds available, which would indicate having potentially a diagnosable depression or anxiety, I just cut the variables at that threshold to represent the more kind of severe mental health problems. But it's again the same story as previously. So experiencing adversities linked with between 50% and 150% higher risk of having a mental health problem later in the life. Okay, so in study two, instead of looking at age specific associations, we looked at trajectories over time, over age between age 23 and 50 using the same measure. So in this case, it was Malaise inventory, a measure of general psychological distress. So the idea behind this is that we can see if this association has changed with age. So if adversities having a stronger association or impact on mental health over time or this impact diminishes over time or stays relatively stable. I would say the conclusion here is that it stays relatively stable. So we see differences already at age 23 or potentially earlier if we had the data and they remain equally strong up to age 50. The same conclusion is for individual aces. So again, this is just trajectories, but stratified by type of adversity. And so except of parental death, which didn't have such a harmful effect as others, we see a similar trajectory across all of them. And those adversities specifically were reported prospectively. And on this side, there are retrospective measures. So they tend to have slightly stronger association with mental health. So as you would expect, abuse has a quite strong effect. But interestingly, family conflict, which one would expect having not been a severe experience actually has quite strong association, like even stronger and physical abuse and comparable to psychological abuse. So I think that's an interesting finding. And I think that's the key parental saturation divorce actually didn't have as strong effect as other adversities. So seeing that there is a quite large spread of those individual trajectories across adulthood, we try to identify groups or subgroup of trajectories which were similar. With a mixed effects model, so we identify four subgroups. So the largest one with 75, 74% experience relatively low and stable symptoms over their adulthood. And then three other groups experienced between moderate and high or severe symptoms throughout their adulthood. So we identify those four groups. And then after identifying those groups, we wanted to see how strong is the association between experiencing adversities and being in one of those groups. So for example, we can see here that even having just one adversity is already associated with twice as strong risk of being a high symptoms group compared to the low symptoms, where for example having four or more adversities is associated with almost six times as strong risk in being in high symptoms group compared to low symptoms group. However, still 60% children who experienced two or more adversities and 54% of children who experienced four or more adversities were in the low symptoms group. So despite having experienced quite a number of quite severe adversities in their childhood, those children still didn't develop serious mental health problems. So our question was what we're interested in after seeing what distinguishes those resilient children from the children who after experiencing adversities did develop mental health problems. So how they might differ. And we hypothesize a number of characteristics which might buffer the impact of adversities on mental health. And this included father's occupation, social class, gender, parental involvement. So I think I don't remember exactly, but these include questions like mum and daddy reads to the child. I think they spent time doing activities with the child, this kind of questions. Cognitive ability, consciousness, internalizing, externalizing problems and physical health problems. And even though these characteristics were strongly related to mental health, they did resilient children did not differ according to those from those non resilient children. So we just ran a series of interactions and we didn't find that they made children with adversities. They didn't make their mental health worse. But also it didn't improve mental health of children who already experienced adversities. So it just simply does no effect modification according to them. Which was quite surprising actually. We would expect that at least some of these know cognitive ability or having other mental problems in adolescence would have some kind of effect. But basically they didn't. Okay, so the take home messages from this research. So ACEs are bad for mental health, are very harmful and are persistently harmful from the whole adulthood and are cumulatively harmful. So the higher number of adversities the worse mental health gets, this is of no surprise. And this is true regardless of the type of adversity, how they're measured, if they're measured prospectively or retrospectively, the way mental health is measured. So well-being type of outcome or depression, anxiety or symptoms of both. And they didn't present that but also these findings were quite robust to different analytical decisions to do with missing data, sample definitions, coding of these variables. Or different confounding sets. And however we still have to remember that most children with ACEs do not develop mental health problems. And probably the next step is to understand what distinguishes in this resilient children from those who are not as lucky to and who then go on and have mental health problems over the life course. Okay, I think I have a little bit more time but I will stop here. So thank you a lot for listening. And also thank you for all the great work on the catalogue of mental health measures because it just makes life so much easier. So I really appreciate that. And I'm open to questions. Thank you so much. Latif Akani is a first year PhD student at the University of Strathclyde. His research aims at investigating the health and wellbeing effects of the UK national living wage policy. Latif's research interests are in health, labour and development economics. Prior to commencing his PhD, Latif was a research associate at the Centre for the Study of Economies of Africa, CSEA, a policy think tank based in Nigeria. And he's going to talk to us about income trajectories and health outcomes in the UK exploring the impact of stability and volatility. I'll give you a two minute warning before the end of your 15 minutes. Okay. Thank you. Good afternoon everyone, my name is Latif. The paper is an elementary part of my PhD research and my overall PhD research is to look at evaluating the impact of the national living wage introduced in 2016 in the UK on health outcomes, health and wellbeing outcomes. So, before going deep into the policy evaluation, we decided to look at income over time because the UK has a very peculiar, I mean interesting attribute in terms of the wage policy increasing the wage almost every year since before the national living wage and also since the national living wage has been introduced. The country has a target of trying to meet the median wage. So, just a brief overview of what we're talking about, just the background, the data and the method we use, the results and the sum of all the funds. So, as a bit of background, like I said, the empirical evidence on health impact of income are mixed. They are different arguments in terms of the direction of impact, whether health is the one forcing income low income. For example, if an individual has a poor health, there's every likelihood that such an individual will not be engaged in a very good job and that will affect the income. Also, there are all the studies too that I mean have established that income has implication on health. What is very much inconclusive in empirical studies that increase in income does not necessarily translate into improvement in health. For example, if income goes up, if such an individual does not engage in health enhancing consumption, there are studies that have established that it should deteriorate in health, smoking, alcohol consumption, obesity and what have you, there have been studies that have established it. Also, like I mentioned, the social economic status poverty have very much linked to high mortality and mobility, indicating that low income has implication on health too. So, increasing income does not necessarily translate to health. However, when income is low or when there is poor social economic status, it might translate into poor health. Then there is a report by the Office of National Statistics in 2018 that summarized that people stay home after the attacks doesn't keep up with inflation and this tends to make people feel poorer and it needs to decline and also re-spending. So, all this, what this background is telling us is that despite the wage policy, despite the increase in income over the years, that may not necessarily have translated into improving health. So, what we now try to do is to look at, okay, what is the impact of income over the years, especially prior to introducing the national living needs on health outcomes. So, in addition to this, what we also try to look at is to look at the argument of income gradient and income trajectory. From the permanent income, my purposes of fried mandas, this argument that's okay, rather than the transient income, changing the income every year does not necessarily affect or translate into meaningful improvement in people's health and well-being. Rather, what should be considered the permanent income, the overall cycle of income, which tends to also, I mean, are densified in 2001, one of the key studies we reviewed established this also to show that long-term income is more important for health than just the correct income that individual end. And also that when there is instability in income, income when it is not very much predictable, it could be harmful to health. Then in terms of the empirical approach we also see in literature, there's this argument about cross-sectional versus longitudinal methods. While early studies use cross-sectional approaches, there are some that have also used longitudinal methods trying to address some of the endogeneity bias and confounding issues in the cross-sectional approaches. Another advantage of the longitudinal method that we found in literature is that the reporting bias in said reported health, if you are asking an individual about their health, you are likely going to get some reporting bias. But if you ask the same individual the similar set of questions over time, this could address some of those reporting bias. Then the third other aspect that we also tried to address in the studies, the estimation approach now, in using longitudinal approach, what studies do because of the limitation of using accounting for some effect, like the fixed effect, is to use diacutimus outcome, like poor versus good, yes and no. So recently there are new or improved methods of handling all that outcome. So when the scale of outcome measure is more than two, so can you still account for some of those effects? So just in terms of the data and the method now, because we are also trying to focus on longitudinal approach to capture our income over time and affect health and well-being. So we use the understanding society survey and we consider 10 waves because of the large data sets. Over 40,000 households are considered, and between 2009 and 2019 is where we consider. Although when we adjusted the data for similarity of periods, because usually every wave is conducted over 24 months, so what we then do is to make sure every individual income and health is related to the same period. Before we have data for 2011 to 2018, then another advantage of this data is that it has a very detailed measurement of people's income, household income, individual income, and different measures of health and well-being. So we consider this, I mean, five major outcomes are the general health, the general health questionnaire to do as a project for mental health. Then we also consider the long-term disability or chronic health measure, then satisfaction with leisure time and life satisfaction. How does income affect these five outcomes? So for the baseline model, like I said in the background, we use the fixed-effect order logic model. So what this model is, rather than the binary fixed-effect model, it accommodates other outcomes variables. For example, the general health is asking people how do they feel in terms of their general health, and there are five options between poor, fair, good, and excellent, about five categories. For mental health, for example, the case-ness, the GHQ12 has about 12 plus, at least about 13. So with all these different skills, rather than do what other studies do, trying to now categorise into poor versus good, we try to, I mean, use the variables the way they are collected, rather than altering them to not bias whatever result or whatever outcome we'll be getting. So for the baseline model, we consider the impact of the household income. We also consider some covariates like the age, the occupation of household age, the number of individuals employed in the household, and other factors. So for the baseline model that is under that, then we now also move forward to look at the income trajectory. Okay, stability in income, volatility in income, these are different concepts. Apart from the income itself, how stable household income is, or an individual income is, how does it influence their health, and when income is volatile too, how does it influence their health. So overall, we consider about four different variables there. So rather than using the household income now, we consider the stability in household income, which we measured as the average of the household income over time. Then volatility in income also we use the standard deviation of the act percentage of income. Then the last two models, or the last two variable models you consider, one is on the duration of income scale. That okay, when a household spends so much time as a low enough, it tends to have different impact compared to an household that ends above the median income. So apart from the baseline model, we consider the household disposable income. So the order for category of models we estimated, look at different concept of income, and how each of these concepts could influence health differently. For the main results, so the baseline model that I highlighted is just looking at the impact, apart from the other covariates, I only reported the main variable of interest there, which are the income measures. So this is looking at the impact of the household disposable income over time after we have accounted for inflation and equalised it on the mental health, general health, longstanding disability, satisfaction with leisure and overall life. So in summary, what we observe here is that increase in income or current income over time as a positive implication or a positive impact on health, both in terms of improvement in general health, improvement in mental health. People tend to report improvement in health as their income increases over time. Although within the result does not show any meaningful means significant relationship for the income units. So for average income now, what we use to measure income stability. So we observe that when average income over time, so here what we did is to use a cross-sectional approach. So we have reached their income over the nine years period, 2011 to 2018. And we look at the impact of the average over the past nine years on the current status. That also shows that there's a positive impact in terms of the average. Then volatility, which we measure as a standard deviation is negative and significant. Then we now look at, for every year we compute the median income. That okay, what is the median, overall median income for every year. We now consider for every household are they below or above. So based on the number of years they fall below the median income. We use that as a project to look at the low income. Spell, and that shows a negative impact that has also continued to fall below median income. It tends to lead to, I mean, decline in the prediction of the health. So while for those that are both consistently over the years, they tend to have positive impact. So based on these results and giving our overall interest, which is to look at the impact of the living wage. So we then break our results down, our estimation down into looking at, okay, high enough versus low enough. So we use the 2018 data. We now divide household into different income content. So we have the bottom 20% and the top 20% in us. So this chart here is just showing the head distribution of the low earners and the high earners. So for example, the gray area is showing that people in the bottom 20% tend to report more of poor versus fair health compared to people in the top 20% which is in the unshaded to red cell. Same similar thing for the GHQ, which we used to call the mental health. It shows that people that are earning better tend to report less distress in their mental health compared to those that are in the low 20%. Also with leisure time, this for leisure time, those that are earning big majority seems to be less satisfied with their leisure compared to those that are earning low. So this is just a way of showing some of the characteristics of different households and their income distribution. So based on these characteristics, we then partition our analysis into two to look at low household earners versus high household earners. And our result seems to be consistent also except that for the average income rather than have a significant result as we have before. It shows that for low earners, even when their income is stable because the income is low, it doesn't tend to predict improvements in their high significantly compared to the high earners. For volatile income, they tend to be negative and significant especially for the general health and the mental health. While similar thing is not, for example, the high earners showing that although income have implication on health, but when you say income, you need to first look at the threshold of the income before you can begin to discuss whether it's significant or not. For the sake of time, we also did something here to look at before and after 2016 because 2016 was when the nationality was introduced. So, and our result also showed that post 2016, there seems to be more relationship or more significant relationship between income and the different measures we use. This table here is just trying to show to us that by the time we go to the policy identification and trying to capture those beneficiary because they really have implication. So in summary, our result shows that both current and average income are significant predictors of improvement in health. And we also observed that when household income is stable, they tend to be more likely of improvement in, I mean, reporting improvement in their health while volatile income tend to predict poor health. Then finally, our result also suggested that when income is increasing, there seems to be less contentment especially from high earners with their leisure time. People tend to be less satisfied with their leisure time when their income is increasing. This tends to suggest that people tend to work more or people tend to want to take up more time at work than leisure, which have implication on their leisure time. One of the limitation is that in our partitioning and our post and pre-2016 analysis, we didn't really capture the true impact of the policy, like I said, but we are only trying to look at the significance of the magnitude of impact increases or decreases. Thank you. So, we'll now move on to Maria Christodulu, Christodulu. And if you'd like to start sharing your screen. Maria is a postdoctoral researcher in bio-demography in the Department of Statistics at the University of Oxford. Her training is in evolutionary biology and statistics, and she's going to be talking about erosion of representativeness in a cohort study, which links very neatly to the question that Tammy was asking. Yes, I thought that was an excellent link, actually. So, good afternoon. I would like to talk to you a little bit about some of the work we've been doing on the 1958 British Birth Co-Hort, which we've heard a little bit about from David earlier, which is, I don't know if you can see my slides changing, is everything okay? Excellent. Thank you. So, the overall piece of work I am involved in is on ageing and understanding the underlying trajectories behind ageing, human ageing. But I'm very interested in reproductive ageing, specifically how women age. The reason I find reproductive ageing particularly good to work with is, A, it's to me inherently interesting. B, the fertility window is actually very clearly marked. We have events at the beginning and at the end, and we can actually see a whole trajectory of a woman from her menarchy, which is the time of the first period, up until menopause. And there is already some understanding on both the genetic influence of timings, but also the socioeconomic impact that may be influencing those timings. It's also a shorter window than lifespan and ageing. And it's a little bit less morbid because you're not waiting for people to die to be able to include them in your study, which is, it's an added bonus. Now, fertility itself is hidden. We only observe the results of fertility. We see pregnancies, we see live birds, we see unsuccessful pregnancies. And also we are observing the fertility of two individuals through the outcome of a pregnancy. And again, we have very strong links between the environment and sociological impact. And this is when we started considering how we wanted to look into these events. And the 1958 British birth cohort is essentially the perfect data set for us. Initially designed to study perinatal mortality to set up to see essentially early deaths for children. It included all births that occurred in the first week of March 1958. That was over 17,000 individuals. And the idea of including absolutely everybody who was born in that specific week gives us a very good idea of all sorts of ways of life, parts of the UK. All these factors are just included within the study. We've had up until age 16, there were also replenishing sweeps, which included people who were born outside the UK but moved into the UK by age 16, which means that after that age they were living within the country. And then since then they have been followed and study has changed hands a few times. But a vast trove of incredible variables have been collected over actually more than nine sweeps, but nine were the release ones that we have studied. And in 2002 there was a biomedical survey, which was used to also get some collections that was used for DNA work. As a study it is extraordinary in its ability to give outside information about all aspects of life. And because of its sampling strategy, every birth in that one week, it has been considered a representative study for the UK population. And by looking at, for example, the general characteristics in terms of BMI or socioeconomic background, early in life versus age 40, there have been studies that have demonstrated that it has remained in its broad strokes quite representative. But for me representativeness only makes sense in the context of the question you're asking. Something can only be representative based on what it is you're looking for. Something can't be representative of a whole population for all aspects. Some things it will be perfectly spot on and for others it won't be. So as we were looking at fertility, I wanted to see how well everything would match the actual national statistics we have. We are actually extremely lucky in our specific field that we have national statistics to compare against and to understand how the cohort behaves against other people of their age group. So the first thing I wanted to get my head around is how many people stayed and how many people left. And how many of them were male and how many were female and how many actually stayed to be genotyped or at least up until the wave when they were genotyped. And I'm a visual person so I started drawing it out and doing mind maps about who comes in and who goes out and who has stayed up until and we do see an attrition which is unsurprising and has been studied extensively. But the idea is ok, people have dropped out but if it's a random sort of drop out that we have then it won't affect the overall character of the cohort we collected. Without in mind I went to the Office of National Statistics and National Records Scotland. We are surprisingly lucky to have both these agencies making this pretty available. And for the Office of National Statistics for that particular age group I could get the maternities and I could get which are all births that have occurred to women for specific years. And I could go to National Records Scotland and do the same. And then I could use the collection of pregnancy history variables in the cohort to create exact trajectories for the fertility of the women in the cohort and then see how many pregnancies occurred every single year for each woman in the cohort. After collecting the data it's a bit more challenging in a sense that for the particular age group we're looking into England and Wales release the fertility numbers in terms of aggregates so we have them by age group. So women in the age group of 20 to 24 for example for that year and we don't have specific data for it. But for Scotland we also have exact counts of the age of the mother at a specific year so we could recreate all the women who were born in 1958 how they compare to the women who were born that particular week in 1958 and have remained in the cohort. So the first thing we did is we tried to find a way to disaggregate the data to go from England and Wales and remove from the actual group data and create an actual curve for women who were all born in 1958 how many maternities per 1,000 women per year. And we started with Scotland because with Scotland we had both the aggregates and the actual counts. It's a smaller count and they collect the data in a different way. So we tested various methods and we were lucky to find a method from the Max Planck Institute, the quadratic programming spline that seemed to match perfectly the Scottish trends. The Scottish trends are in dark blue if you can see and the estimate from the aggregate data from the splines is the yellow one you can see and it follows it very closely. So we were quite happy with that but again an estimation method is just that, an estimation method. It's not going to be magic. So I went into the cohort and I also did the national statistics so I used the spline method to create a curve for the national statistics for all women born in 1958 and then I looked at the women who have given me a complete history, complete cases as we heard from Latif earlier. So complete cases up until after the end of their fertility and also to see women who had dropped out before the biomedical suite, before their mid-40s. And we found two interesting things. The first one was for the women who remained, I also looked at terminations because terminations are recorded in the United Kingdom under the 1967 abortion act and genome eternities. I will not be lying to you if I told you that I like that we have lower reported terminations in the cohort than national statistics surprised me. It did not because I can imagine many reasons as to why this could have happened. We could have a social deserability bias. It could be the way we asked the question. If you speak into an interview you may be less willing to be upfront about parts of your history but it does surprise me that we seem to be having different numbers in the cohort for maternities than we do in the national records. We seem to have fewer actual maternities from both the people who remained in the cohort until after their 40s and from those who have dropped out earlier, although those who have dropped out earlier are actually a little bit closer to the national records, especially for the age group between the 20s to 30s. And we're not entirely sure why this is happening. So what we are currently doing is we're looking into the 1970 cohort to see if we're seeing similar patterns or if this is just something unusual that happened in that particular group. Now returning to the original questions, we also wanted to see how dropouts may have affected essentially what we're trying to look into, which is fertility events. Fatility is not a snapshot in a woman's life. It's a whole story. Let's start with the timings of her first pregnancy, of her first birth specifically. And we notice that the two groups we had, the people who dropped out before 40 and the people who stayed, had slightly different profiles in terms of the timing of the first birth. We also notice, however, that factors such as socioeconomic background and education are slightly different between the people who persisted and the people who dropped out. So, for example, we had delayed fertility, which was associated with education, which wasn't a surprising finding. And finally, age at menachee, which I expected to be closely linked with socioeconomic factors but nothing else, didn't actually show any particular differences, but I do feel that in this case, the actual resolution of the data that we collected for that particular event was too wide. We're collecting only data for menachee at age 14 or 15 or 13, which may not have the correct resolution. We need to actually understand the differences between them. I seem to have talked very quickly. So, if there are any questions. Thank you. I'd like to introduce Maitri Khurana, who's also at UCL. Maitri studied psychology at Harriet Watts University at the Dubai campus and then went on to do her MSc at UCL in clinical mental health sciences. The paper she is presenting is based on her master's dissertation, and she's currently working as an assistant psychologist at the Lighthouse Arabia Centre for Well-being in Dubai. So, over to you, Maitri. Thank you for that. I hope everyone can hear me well. Thank you, Jenny, for the introduction. So, my name is Maitri, and I will be talking about my research on the association between sensory impairments and suicidal ideation in attempt, which is a cross-sectional analysis of nationally representative English household data. I'm just going to get started. Okay, so suicide is, of course, a global public health problem. And as you can see from this graph, it's actually one of the leading causes of death worldwide. Identifying risk factors for suicide is important because it is considered to be a preventable cause of death. And one such risk factor could be the presence of sensory impairments. Now, individuals with sensory impairments, particularly visual and hearing impairments, tend to report a poorer quality of life and mental health. And despite both of these visual and hearing impairments having well-established associations with mental health disorders such as depression, the current evidence base regarding their association with suicidality is rather sparse and inconsistent with most of the, with a lot of the research saying, some of the research saying that there is an association while others saying that there isn't. A lot of the current research also comes from the elderly population. So we decided to kind of bridge the gap and try to assess for the association in the general population. So our theoretical explanation for this comes from the Integrated Motivational Volitional Model which possess that triggering events, which in this case would be the presence of sensory impairments and the communication social difficulties that they bring, could lead to, could create conditions for suicidal thoughts and where things like defeat and humiliation come in, this could then create conditions for suicidal, this could add to the suicidal ideation and where motivational factors such as sort of belongingness and perceived burdensomeness comes in, that could again add to the suicidal ideation and then access to means and exposure to another suicide could then lead to suicidal attempts. So we aim to assess if there was an association between sensory impairments and suicidal attempt and ideation in the general population. So we analyzed data from the fourth adult psychiatric morbidity survey which was conducted in 2014. The sample consisted of English adults aged 16 and over living in private households and the final sample set consisted of 7,546 individuals. So for our exposures in the APMS survey there were two questions that assessed visual impairment. They assessed near-sightedness and far-sightedness. So the questions were with your glasses or contact lenses if you wear any. Do you have any difficulty seeing ordinary newsprint at arm's length and with your glasses or contact lenses if you wear any? Do you have any difficulty clearly seeing the face of someone across a room that is four metres or 12 feet away? We combined both of these to create a binary variable representing having a visual impairment or not. And then for our hearing impairment there was a question within the APMS that assessed do you have any difficulty hearing or use hearing aid and we used this binary variable as it was. We also created a general sensory impairment variable where we assessed having a sensory impairment or not so that could be of either kind, visual or hearing. And we had a dual sensory impairment variable as well so that was having both visual and hearing impairment at the same time. Our outcome measures, we had two outcome measures. The first one was suicidal ideation which from the APMS survey was have you ever thought of taking your life even though you would not actually do it? And suicidal attempt which was have you ever made an attempt to take your life by taking an overdose of tablets or in some other way? Both of these were measured within the past year. We also had five predetermined clinical and socio-demographic variables that we considered as covariates in our models. These were gender, age, socio-economic status and diabetes. Just to note that we did not consider depression and anxiety as part of our final models and the reason we did this is because they're likely to be on the causal pathway between sensory impairments and suicidal ideation and attempt. So for our statistical analysis we used multi-level logistic regression models to describe the association between each type of impairment and suicidal attempt as well as suicidal ideation. We used complete case analysis where we utilized data where they had complete information on all of our exposure and outcome variables. We conducted two sensitivity analyses where we were trying to assess if missing data could be playing a role so we used best and worst case scenarios. We conducted two post-hoc analyses. One was to assess if depression and anxiety could actually be playing a role in this relationship. We added CISAR scores to our final models so that's scores based on the revised clinical interview schedule and we added this to our model just to assess if depression and anxiety could be playing a role. We also conducted post-hoc analyses where we added each of our covariates in turn to see what effect they were having on our model. So as you can see from the results in each of the cases those with sensory impairments had higher odds of having thought about suicide in the past year and this was the case for whether we looked at just the general sensory impairment, dual sensory impairment, visual or hearing and this was the case for a suicidal attempt where we got much higher odds of having attempted suicide. So our main findings were that those with sensory impairments had greater odds of having thought of an attempted suicide in the past year. When we added each of our covariates in turn we found that age was a strong negative confounder in that a lot of our unadjusted models were underestimating this relationship and there also appeared to be some evidence of contribution of depression and anxiety in that when we added Cesar scores a lot of our associations did get attenuated. So some of the strengths of the study is of course that it does contribute to a very limited amount of literature that exists on this topic. We also used nationally representative data, which means our study is more likely to be generalizable. Our findings were adjusted for predetermined socio demographic and clinical covariates and were also robust to our sensitivity analyses that we used assimilating any biases that could be introduced by missing data. We also need to acknowledge some of the limitations. Firstly of course this is a cross-sectional data set and we looked at a lot of longitudinal data before I started talking about my topic and of course this means that we can't rule out things like reverse causation and this is particularly important when considering something like suicidal attempt because past suicidal attempt can lead to future suicidal attempt. So this is something we need to consider when looking at this study. Also we need to acknowledge that sensory impaired population is quite heterogeneous especially in terms of how they react to their sensory impairment. For example there is some research that suggests that having a sensory impairment for a longer time such as having a congenital sensory impairment could actually have those people could actually have a better quality of life just because they can adjust to the disability better than someone who has maybe developed it more recently. We also need to consider that certain settings that have populations with higher suicidal ideation and attempt could not be considered such as inpatient units and prisons these populations could not be explored. Also because the APMS survey looks at private house people living independently in private households it may actually be that we may have missed out people with a higher degree of sensory impairments because they are unable to live in private households independently. So that could potentially be a limitation. We also wanted to acknowledge that the measure of hearing impairment that we chose included people who use hearing aid and while this obviously means that their hearing is corrected to a certain extent we decided to be inclusive in our definition of hearing impairment because there are still studies to show that even though your hearing is corrected just wearing a hearing aid adds to the stigma associated with hearing impairment and the stigma of disability so we decided to include them anyway. We also also all of the APMS survey is self-reported measures so of course there is the likelihood of social desirability bias and okay so some of the interpretations and implications that come about from this study is of course both of these impairments involve communication difficulties that means that access to mental health care will be limited. So individuals and professionals who work closely with these populations for example GPs, ophthalmologist, autolaring colleges even community audiology services or opticians if they are warned about these results and if there is a training that associates with trying to identify symptoms of suicidality then they could then refer them to further mental health care as well. We also need to consider the factors that may be involved in this association of which loneliness may be particularly important to consider in future research because obviously having a sensory impairment can create a lot of feelings of social isolation and loneliness so that could be a potential mediator. Other factors to consider include stigma of disability, locus of control self-perception, self-esteem also of course we need to look at this in longitudinal research just to identify what the temporal nature of this relationship is. So yeah our findings found strong evidence to support a cross-sectional association between sensory impairments and societal attempt and ideation in view of the cross-sectional nature of the data we need further longitudinal research in order to explore the temporal relationship. Thank you.