 Welcome everybody to the session on ageing. Very glad to be chairing this meeting. I'm Dr. Martin Hyde from Swansea University, a really interesting line up of speakers. It gives me great pleasure to introduce Oliver, who's a MSc research student from the School of Geography at the University of Leeds and whose research focuses on use of course methods in social geography. So Oliver, please take it away. Hi everyone. And so my name is Oliver Hall. I'm a master's by research student at the University of Leeds. And the focus of my research is to develop a composite index to measure loneliness in older populations in England. So the topic of loneliness as a key social issue in the UK has gained increased attention in recent years. And news outlets have characterized the issue as an epidemic and the UK government became the first government in the world to appoint a minister for loneliness in 2018. It's estimated that 24% of older age groups suffer from loneliness some of the time. And this of course is an age group that's growing with some estimates suggesting that by 2039, a quarter of all Brits will be over the age of 65. High rates of loneliness have been cited as a key risk factor for depression, suicide, heart disease, high blood pressure and dementia. In fact, loneliness has been found to be as great a risk to health as obesity and smoking by some studies. These health risks have further negative economic and social impacts of course through increased strain on public services. In fact, the director for acute care for the NHS says that the increasing costs of caring for isolated elderly patients could quote, ultimately cripple the NHS. It's important then to measure loneliness in order to know where exists to intervene. So current methods of measurement are very qualitative. The government advises the use of a direct survey question, which essentially just asks an individual how often they experience loneliness. They also advise using a second measure known as the three eyes and UCLA scale. It asks people whether they experienced the derivatives of loneliness without ever directly using the term lonely with questions such as, how often do you feel like you lack companionship? There are multiple conceptual flaws of these methods, which have already been addressed in the literature, but significantly, they're not well suited to location and service planning for charities and local authorities. As the surveys that include them are not large enough to be applied to the small area level. As such, the campaign to our loneliness have highlighted current attempts of tackling loneliness as quote, hit and miss. A couple of groups have attempted to construct composite indices that map estimates of loneliness at the small area level, including Age UK, Essex County Council, and one in the literature from Lucian Burns. The Age UK index is quite restrictive, including only four variables, and the report outlining its construction is quite unclear, but the parents statistical inaccuracies as well. The Essex index is far more descriptive, but is only applicable to Essex and uses costly commercial data, making it difficult to reproduce for other charities and local authorities. Whilst the Lucian Burns index doesn't make use of statistical analysis for informing decisions in the construction phrase, and it's also only applicable to London. So this project became an attempt to create a loneliness index that made use of statistical analysis, whose construction process is clearly transparent, uses open data, and is applicable at the small area level for the whole of England. So at first, I throw a literature review highlighting many variables that have been found to be associated with loneliness in older populations, but at this point, it was important to conduct our own research to try and ascertain which of these are independently associated to loneliness. For this, we turn to administrative data, both the English longitudinal study of aging and understanding society were considered as they contained a variable of self-reported loneliness, but it was decided that understanding society had a broader selection of variables relating to social, economic, and environmental factors. Using both would have been possible, but obviously it would have negated the benefit of using multivariate model as confounding factors would have been impossible to identify with the two surveys using different sets of participants. We found the following variables to be related to loneliness within a multivariate model. So first, we found widowhood, which was kind of interesting development. Initially, marital status had been categorized into four responses, being married, single, divorced, and widowed, but it was found that being divorced and being single was confounded with living alone, so the incidence of loneliness in those categories were more likely a result of the fact that those people were more likely to live alone than a direct result of their marital status. Living alone was a fairly straightforward variable and had the largest effect of all the variables included within the model. Being dissatisfied with your income was found to have a greater relationship with loneliness than absolute low household income. An interesting finding that pointed to the importance of perceptions rather than an objective evaluations of an individual's personal circumstances with regard to loneliness. A similar development was found with health variables where self-assessment of general health had a stronger relationship with loneliness than the more objective measures, such as presence of chronic diseases or activities of daily living. Being an ethnic minority, that is being non-white British, was found to be related to loneliness as was a lack of neighborhood cohesion as measured by Buckner's index of neighborhood cohesion, which is our qualitative index included in the survey. Lastly, and somewhat surprisingly, we found smoking to be associated with loneliness. Now, unfortunately, the data from Understanding Society could not be used in the index as the sample size in the over 65 age group was too small to apply at the neighborhood level. Using micro simulation or estimation techniques was explored, but it was ultimately decided that either the appropriate inputs weren't available or they were simply out the scope of a one year research project. So stakeholder consultation was conducted as a means of validation, but also suggestions regarding proxy variables were necessary as we couldn't use Understanding Society. So we consulted with both independent age and the British Red Cross. The variables of widowhood living alone, poor general health, and were aroundly accepted and available from open sources. The variable of dissatisfaction with income was not available at a small area level. However, stakeholders saw no issue with the use of absolute measure of household income to represent the material wealth dimension of loneliness in the index instead, whilst it's also commonly used in the variable in the literature as well. So ONS household income estimates were used. It was suggested that pouring this language ability was a more appropriate indicator than data on ethnicity. This was in keeping with the literature and only hadn't been investigated as there were not enough observations for this in the survey. And so language ability data was taken from the census. Lack of community cohesion needed a proxy variable as Buckner's index was not available and neither were any of its component parts. Initially we explored the possibility of creating a sub-domain of community cohesion that used statistics such as population turnover. However, there's a general lack of enthusiasm from stakeholders for many of the variables we suggested. Only high rates of hate crime was accepted as a proxy variable. Further, there was some statistical justification for the use of this variable as self-reported loneliness and those who perceive hate crime to be a problem in that area showed a strong correlation within understanding society. Stakeholders questioned the causal link between smoking and oneness. And we agree it's unlikely that smoking is directly related to loneliness and that it was probably representing some other latent variable in the model. However, the purpose of this study is not to identify the true causes of loneliness but rather identify the neighborhoods where loneliness is most likely to exist. So for this purpose, we believe that the inclusion of the variable does add value to the predicted power of the index even if it is acting as a proxy. And finally, stakeholders showed support for the inclusion of a variable relating to informal provision of care which was not included in understanding society data. This is in agreement with the wider literature as well as multiple other charity reports including those from the campaign to end loneliness, age UK and care as UK. And so informal provision of care was also incorporated into the index. Where possible, weights were derived from odd ratios generated in the multivariate logistic regression as this gave us a coefficient representing the importance of the indicator to loneliness within our model. The smoking indicator was then penalized by reducing its weight based on stakeholder opinion. The care provider's indicator was defaulted to no weight based on there being no statistical grounds to apply one which implicitly gave it one of the lowest weights and hate crime indicator was penalized on the course geography which was collected as the geography was exerting too much influence on the index. These variables were then normalized onto a common scale multiplied by their assigned weights and added to each other to create the index. So as you can see, this is the loneliness index mapped in England at local authority level and it does display a clear spatial pattern. Coastal areas appear to be at highest risk. The most at-risk authority in the whole of England was West Somerset, which is just this one here. Although when mapped to the neighborhood level, the most at-risk elements were predominantly along the most southern coast areas around Brighton, just here, Eastbourne, and the most at-risk neighborhood was in Christchurch and Dorset, which is just about there. Another interesting finding was the difference between the more urban areas in the North than in the South. So Greater London and the home counties were a very low risk of loneliness, whilst more urbanized areas in the North, particularly in the old industrial towns, sorry, particularly in the old industrial towns around Barrow, Sunderland, Huddersfield, and Doncaster are a particularly high risk. This is largely driven by the inclusion of the income variable with London and the home counties having higher average incomes than those more northern areas. Through comparing it with the Essex County Council Index, we find remarkable similarities. As mentioned, the Essex Index incorporates many variables but uses costly commercial data and so it's not easily reproducible for charities or local government, whilst our index uses freely available open data. Both industries find similar patterns with tendering being at highest risk and along with Harlow and areas of Basildon, Castle Point and Braintree. Whilst Ottersford appears in the North West, appears to be one of the lowest risk wards. Obviously, some differences is the data, the different data was incorporated and collected also in different years. Notable divergent neighbourhoods on the coast of Rocksford, for example, in the South East as well as in South Molden. But from a planning perspective, the new index is suitable. It shows a high degree of similarity with the more descriptive index and clearly highlights the same wards as being high risk, such as tendering and those of low risk, such as Ottersford. What we've demonstrated here then is that it's possible for charities and local authorities to construct an index that's statistically found closely aligned with current interseason use. However, they are also able to do it using freely available open data that can be implemented at any desired level of geography and significantly can be implemented in the whole of England, which allows for benchmarking comparisons between towns and cities and other local authorities to make use of it as well. Another key benefit of the index is that it allows for comparison with other variables that are often hypothesised as being linked with loneliness. Two key examples here are deprivation and population density, both of which lack academic consensus regarding their relationship with loneliness. Regarding deprivation, the plot on the left shows the relationship between index and multiple deprivation deaths out, so just a widely used index of deprivation in the UK and the loneliness index. On the right is the exact same, except we've removed the income variable from the loneliness index as income and IMD deaths out, obviously highly correlated. Our results suggest that there's only a very loose relationship between deprivation and loneliness. We also find a correlation between an association between low population density and high risk of loneliness, but crucially, we find a relationship between coastal proximity and high loneliness risk. This is a phenomena that's not yet been highlighted in the literature, so it was quite an interesting finding. Analysis conducted on the relationship between population density and coastal proximity found no association, meaning that the two parts are not simply representing the same phenomena and low population density and coastal proximity are both related to high risk of loneliness in England. So thank you very much for watching. I've included my email in case everyone likes to get in touch and likes to welcome any questions. It is my great pleasure to introduce the second of our speaker, Sarah Stockforth, who is a research fellow at the University of Sussex working on the Nuffield Foundation project entitled Ethnic Inequalities in Later Life. Her research involves statistical analysis of large-scale and complex survey, data, and prior joining to joining Sussex, Sarah again, a PhD in Sociology from the University of Edinburgh and has worked on a number of research roles on policy think tanks. And I was going to talk today about the persistence of ethnic inequalities and health in later life. So Sarah, please, off you go. Yeah, thank you very much, Martin. So yeah, I'm Sarah. I'm a research fellow at the University of Sussex. And today I'm going to present some findings from a piece of work that I've been doing with colleagues at the University of Manchester and the University of Sussex and it's funded by the Nuffield Foundation. So I hope that's changed slide. Let me know if not. So a bit of background to the whole kind of topic. So ethnic inequalities in health and wellbeing are really quite well documented across the early and mid-life course. So we know, for example, that people from minoritised ethnic groups tend to have worse health outcomes than the white majority or the white British group. And the explanations for inequalities are complex, but often socioeconomic inequalities are used to explain health inequalities. And then there's a link between ethnic and socioeconomic inequalities. So for example, people from some minoritised groups are disproportionately disadvantaged on a number of socioeconomic axes. For example, living in more disadvantaged areas, having poor housing quality, higher rates of unemployment, things like that. We also know that there are direct and indirect effects of racism and discrimination on health outcomes. Directly, for example, through stresses from racism and discrimination, either experienced or anticipation of racism. And indirectly, for example, working through socioeconomic inequalities, so for example, through employment discrimination and through structural inequalities. So we know that ethnic inequalities exist in health in the earlier and mid-life course, but we don't currently know a lot about ethnic inequalities in later life. And the lack of evidence here is really important because a key demographic change over the next few decades in Britain will be the rapidly increasing population of older ethnic minority people. So we need policy and health and social care provisions to have the data and evidence to be able to address the inequalities. So the projects kind of as a whole that I'm working on is using existing data resources to address this data and evidence gap. The specific bit that I want to present to you today is looking at the prevalence of ethnic inequalities in health in later life, the extent to which these exist or persist over time, and then what the respective contributions of socioeconomic inequalities and racism and discrimination are on the ethnic inequalities observed in health. So we really are quite limited by current data availability to investigate ethnic inequalities in later life and over time. So in the UK, we do have some really fantastic longitudinal data resources. So for example, we have the English Longitudinal Study of Aging or ELSA and the Older British Birth Cohort Studies where we can robustly analyze aging in the UK. But really these studies are suitable for analyzing older white British people and not older ethnic minority people due to the sample sizes in the studies. We also have some really great longitudinal resources with specific ethnic minority boost samples, for example, understanding society. The world's largest household panel survey repeats the contacts, the longitudinal design and ethnic minority people are deliberately oversampled throughout. But even here in the latest wave, there are really small subgroups for older ethnic minority people. So we do have this lack of data on older ethnic minority people specifically and especially when we want to look at changes over time. And this link here is a blog done by my colleague, Dermot Capadia, who you will have heard speak earlier, which is a really great insight into this issue. So having said all that, there was no single source to study ethnic inequalities in late life over time. So what we did was we searched for a range of suitable data resources and we looked to harmonize as many different data sources as we could over time. So we used the fourth National Survey of Ethic Minorities from 1993, the Health Survey for England, 1999, the Health Survey for England, 2004, the Citizenship Survey, 2007. And then we used two waves of understanding society to incorporate the original ethnic minority boost sample and then the refresher immigrant and ethnic minority boost sample as well. So we chose each survey because they were nationally representative at whatever point in time they were collected. So that might be of just England, of England and Wales or of all of the UK. All of these surveys have suitable ethnic minority boost samples. So their survey design included deliberate over sampling and their adequate measures for us to investigate ethnic health inequalities and look at this over time. So we have these six data sets and they span more than 20 years and we use the same analytical sample in each data set. So we look at those age 40 and over living in England. So we just focus on England because we have the Health Survey for England and so to make it as comparable as possible, we restrict the others to England only too. So we keep those age 40 and over because we know from previous work we've done using census data that ethnic inequalities in health really start to open up in the 30 to 39 age bracket and inequalities are really well established by the age of 40. So we know that rates of poor health for people from many minoritised ethnic groups are equivalent to poor health rates for white British people who are significantly older and there's evidence of an earlier onset of ill health for people from minoritised groups. So that's why we kind of have the age bracket from 40 over. So in order to examine the prevalence and persistence of ethnic inequalities over time we harmonise these different data sets as closely as possible and we estimate separate cross-sectional models in each survey. So each survey has a different data collection and in the survey design in the way that they ask their questions and provide measures, we can account for individual survey designs and weights using the complex survey adjustments and the appropriate weights that are provided within each data set and we also harmonise the analytical sample as I've just mentioned. In terms of variable measurement I will talk about this in more detail on the next slide but we've harmonised measures as closely as possible. So where exact measures do differ across surveys the underlying concept or construct is consistent and so we harmonise the measures according to what the base measure we have available. Then we undertake sensitivity analysis of the alternative ways that we could harmonise these measures and we're really careful in our language, in our interpretation about what our harmonised data can tell us and what it can't tell us just kind of by the nature of it. So we use two main health outcomes throughout this piece. We use limiting long-term illness and we use self-rated health. So the wording of these two health measures does change in the different surveys and this is not uncommon. Even in the census, the wording of these questions changes every census year and we aggregate the measures into binary measures of yes and no for reporting a limiting long-term illness and a binary measure of having fair or poor self-rated health compared with anything better than fair. So we're using logistical regression models throughout this work. Our main measure on the right-hand side of the equation as it were is ethnicity and we have the white majority or white British group as our reference group throughout. Ethnicity is self-reported with predefined categories and we restrict our analyses to the main minoritised ethnic groups in the UK because they have the greatest sample sizes and because they're most comparable across the surveys. So that's the Irish, Black African, Black Caribbean, Indian, Pakistani, Bangladeshi and Chinese groups. There is a caveat here that Black African people were not sampled in the Fourth National Survey or the Health Survey from the 1999 but they are sampled in all the other linked surveys. So when we come to harmonise the rest of the variables, some are more straightforward than others. So age and sex are comparable in all the surveys and they're quite straightforward. Some required a little extra work. So for example, we use NSSEC for a measure of social class which is the National Statistics Socioeconomic Classification. This is the measure used by the Office of National Statistics. But NSSEC wasn't available kind of in a ready-made measure in all of the surveys. So we had to construct it using the underlying information such as what's called SOC codes or the occupation codes and information on the employment status. So in some surveys, we had to construct it ourselves and then in others, it was readily deposited. Some required even more work than that. So for example, income was collected differently in the different surveys. So for most surveys, it was collected as income brackets. So response was shown a sheet of paper and they had to point or mention which bracket they came under. In understanding society, it's different. It is a continuous measure. So in that case, we decided to use a relative measure of income, split the measure into quintiles and then use this rather than an absolute or raw measure to try and mitigate this issue. And then finally, the measures for racism and discrimination firstly weren't available in all surveys. So they were only available in the fourth national survey of ethnic minorities, the Citizenship Survey and understanding society. And there are no agreed or standard ways of measuring racism and discrimination within a survey context. So the original questions in the surveys are similar but they're not identical and they broadly map onto a distinction between experienced physical or verbal attacks compared with measures of fear of harassment or anticipation of racial harassment. So that's kind of how we conceptualize those in the models as well. Okay, so what we find is that ethnic inequalities exist in both health outcomes. So in limited long-term illness and in self-rated health for those aged over 40, in particular, we find that Pakistani and Bangladeshi respondents are the worst affected by poor health for both health outcomes and black Caribbean and Indian respondents also have significantly worse self-rated health than white British respondents. And these trends are observed in all survey years and we also find that age is significant and women tend to have worse health than men which is also consistent with previous work. So when we additionally adjust for our measures of socioeconomic position and racism and discrimination where they're available, we also find associations with poorer health. So this graph here plots the changes in predictive probabilities of fair or poor self-rated health compared with the white British group. So that's represented by the dashed line at zero. And we plot them separately by survey year and for each model. So model one is the base model. It includes ethnicity, age, age squared and sex. Model two then adjusts for socioeconomic position which is income, education and social class. And then model three, where it exists. It's not in all surveys. Additionally adjusts for racism and discrimination. So what we can see is that first of all in model one, we see these differences in inequalities compared with the white British or white majority group. And then we see that when we add in socioeconomic position, the inequalities are reduced slightly for most groups and it looks to be more substantially reduced for Pakistani and Bangladeshi groups in particular. And then when we add in racism and discrimination, there are smaller changes in the probabilities of reporting fair or poor self-rated health where these measures are available. So we're finding that ethnic inequalities are partially explained by contemporaneous measures of socioeconomic position and experiences of racism. But these measures are collected at one point in time. And so we can't really get an accumulation of disadvantage and its effects on health outcomes in later life even though we can continue to observe ethnic health inequalities in older ages. So our results are associational and not causal and we can only indirectly look at the effects of accumulation of disadvantage. So it supports the theoretical approach but currently there's not suitable longitudinal data in the UK to directly test the role of the accumulation of disadvantage. So finally, just kind of reflect on the methodological conclusions. So our work is reacting to the data and evidence gaps in the UK for older ethnic minority people in particular. And so by harmonizing a variety of data resources, we found a partial solution to this data problem. We can track, we can observe the associations all the time and across surveys. And by harmonizing measures, we can kind of see what the effects of socioeconomic position and racism are on the inequalities that we observe in later life. But the approach we present here really just needs to be reinforced with robust data collection. We think that future work in data collection needs to better represent older ethnic minority people and particularly in longitudinal surveys. Relatedly to that, we think that data collection needs to include longitudinal or life course measures of socioeconomic inequalities and racism and discrimination so that we can better understand and hopefully directly test the impact of the accumulation of disadvantage on the outcomes that we can observe in later life. So just quickly some acknowledgments to the project team that's Dr. Lye the Carriers at the University of Sussex and Dr. Dermak Pajian, Professor James Nashu at the University of Manchester. And this work's been funded by the Noffield Foundation. Yes, that's me. I'd like very much to introduce Awa Dunwin-Aminu from Robert Gordon University. He's a doctoral researcher at the School of Nursing, Midwifery, Paramedic Practice at Robert Gordon University in Scotland. He works as an early stage researcher within the Marie Curie Actions Innovative Training Network for Euroagism. He holds a BDS in Dentistry from the University of Lagos in Nigeria and an MSc in Gerontology from the University of Southampton. He's a member of the Alumni Network for Commonwealth Scholars, awesome. And his research interests are in aging, health and wellbeing, life course research, frailty, genetic, geriatric, oral health and health policy. And in his doctoral thesis, he's looking at the risk factors associated with frailty in the context of AIDS discrimination using data from ELSA. I would like to hand over to you. You have 15 minutes. Thank you very much. Thank you so much, Martin. And hi, everyone. My name is Bertrand Amin to ask, introduced. I'm going to be presenting a topic on social isolation and loneliness. We looked at association with perceived age discrimination in the perspective analysis of ELSA. These studies funded by the European Union were right in 2020. And I'm working as part of the innovative training network called EuroAgesm on that in Maricuri Fellowship at the Robert Gordon University. And I'm also doing my doctoral study in the same university. The EuroAgesm network actually is focused on examining ageism across different fields including health, IT in the media and social media and in the mass media. And every other sphere, looking at providing scientific data to combat ageism faced by other individuals across Europe. I'll probably be talking more about that later or share a slide to give a better view of that. But of course, if you are interested in ageism research, you can look up our web at EuroAgesm.eu. My study today is focused on social isolation and loneliness. Of course, we all know that social relationship is a very key factor for human wellbeing and health and can be maintained through social participation. This is partly determined by a social network of a person and can be objectively measured through the frequency of social contacts. Social isolation is a reflection of lack of frequent social contacts with members of the same social network and loneliness relates to dissatisfaction with the frequency or quality of social contacts. In some studies or in the literature, sometimes loneliness has been used interchangeably to mean subjective social isolation because the social isolation that measures social contact is really deemed as objective measures of social isolation. So sometimes loneliness is referred to as subjective social isolation. So social isolation, the challenges of social isolation of course can never be overemphasized. It has been linked to all course mortality among older adults. Social isolation and loneliness have been found to be associated with medical conditions like cardiovascular problems and heightened inflammatory biomechanical disease conditions. And of course, social isolation and loneliness have been linked to reducing quality of life of individuals. And just like previous speakers have mentioned, a little bit of the determinants of social isolation or loneliness specifically. Chronic illnesses that really limit activities for older individuals have been deemed to be one of the determinants of social isolation. Age as a factor because of other issues that probably relate with age because of course, as we age, individuals are likely to live alone. So there will be a change in living arrangement. There will be change in marital status as well. So as a result of that, these have been found to be correlated with loneliness and social isolation. Gender as well, which may definitely not probably be related to gender specifically as it is, but of course, because of other issues that becomes prominent in later life due to inequalities between both genders, social economic deprivation, most specifically, which we are looking at in this particular story, passive discrimination. It's been looked at in the literature mostly focused on racism and other form of discrimination, such as racism or sexism. But in this study, we are looking at a different dimension because we found that there's been really little or a paucity of study that looked at the dimension of age. The population of older individuals aged 65 years and over is expected to increase from in the next one decade, markedly, and in the UK, people aged 65 years and over makes up about 18% of the entire population. And this is set to increase even further in the next few years. These demographic changes are sometimes perceived negatively by this society leading to an increased risk of discrimination towards older individuals. So my study basically fitting into the bigger project, which is the Euro-Ageism project, it's focusing on ageism, but most specifically the aspect of age discrimination because ageism and compasses discrimination, prejudice and stereotype and it can be a micro level which is usually referred to as self-directed ageism, which is probably the expression of negative attitude to ageing by individuals and it could be a meso level or micro level. The meso level is at institutionalized level and the macro level is at society level. My study will be looking at it mostly from the macro level, which is the discrimination faced by older individuals in the society. Previous study has shown that 29% of older individuals aged 65 years plus in the UK reported age discrimination and this discriminatory experience occurred during everyday events and social activities such as use of GM or groceries or even visit to the hospital. So this is going to be very interesting for healthcare delivery. So the objectives of this particular study is to look at the prevalence of social isolation and loneliness among older adults aged 65 years and over and to examine if age discrimination in any way influences social isolation and loneliness. Of course, like previous speakers also mentioned, there was this intention to look at social isolation and loneliness not at the community level only, but also in the institutionalized dwellings for older individuals like sheltered homes or care homes. But unfortunately, because of our design was focused on doing a prospective study, we were very limited in the type of data we could use. So it was only else that we found that as collected age discrimination data and I think we'll probably discuss more about this maybe in the question session. It was supposed to be, we were planning to have a comparison with the developed nations to see if there was going to be differences in the prevalence of age discrimination and the way social isolation and loneliness is reported. But unfortunately, there was no longitudinal data available that collected questions on age discrimination. So we were really left with ELSA. And for ELSA as well, we were hoping that we will be able to compare data for community dwelling in the older individuals and those in care homes. But over the course of the analysis, only 15 participants moved into care homes from ELSA. So it was really challenging to do that. So we only stick to analyzing data for people who are resident in the community. ELSA is a panel survey which most of us probably are aware of individuals age 50 years and over and the data has been collected since 2002. The most recent since it's the night wave. So we started from wave five of ELSA which was the wave where age discrimination was included in the data collection. And that's for now is the only wave that age discrimination questions were included in ELSA. So other measures we have is the social isolation measure apart from the age discrimination measure. The social isolation measure was derived from the marital status of the participants. So it really measures social context in that context in marital status being member of an association and at least maintaining modally contact with children, family, and friend. And this ranges from zero to five and loneliness was accessed using the University of California three item revised loneliness scale which ranges from one to nine. The data was analyzed using the RStudio. The perspective analysis was done using the generalized estimating equation and the outcome variables were decadamized. Social isolation and loneliness were determined by for social isolation, those who had a score of two and above and loneliness those who have a score of five and above following previous study because the social isolation and loneliness data was actually positively skewed which is another challenge which I've mentioned earlier because most of the participants probably felt positive about their level of interaction with others and really don't want to identify with being lonely or isolated. So this was one of the challenges with the data. We also introduced some covariates, age, gender, longstanding illness from previous studies that have identified the determinants of social isolation and loneliness. We also did sensitivity analysis to examine if similar results will be obtained if we use the variables in the continuous form. From the data, we linked the data using the respondents unique ID over the period. So it was about eight years of follow-up analysis. 2,385 responses were analyzed and the prevalence of social isolation was 32% and loneliness was about 29%. And age discrimination was the 8%. And I must say that the auto I quoted earlier that I'm sorry about the noise is the aircraft, yeah. The auto I quoted earlier that mentioned that age discrimination prevalence in the UK was 29% actually also analyzed elsewhere of course because they analyzed everybody above the age of 50. So it was much lower compared to our study that looked at 65 years and over. We did age categories for our study and yes, I've mentioned that. So for the baseline, our baseline analysis was done using just the data collected after we five and then the future analysis was done from wave six to nine. So we have a baseline status for the social isolation and loneliness and future status for social isolation and loneliness. Our study found that there was a reduced risk of social isolation among individuals who reported age discrimination. And this was also consistent in the future status of social isolation. And for loneliness, we found that the risk of loneliness was higher among those who reported age discrimination and this was consistent as well for future status of loneliness among the participants. This is interesting findings in short because of time I'm trying to manage the time. Our theory going into the data was that individuals who experience age discrimination are likely to withdraw from societal engagement or participation and our expectation would be that age discrimination probably will lead to social isolation. But interestingly, our findings suggest otherwise. So findings from these studies suggest that those who reported age discrimination were not likely to be socially isolated or report socially isolation in contrast to previous study. The previous study in question and et al looked at social isolation among individuals who reported discrimination as a result of age. But among these participants are people who already have pre-existing conditions like HIV. So we felt that this might have influenced their data in a way because of the stigma that comes with the condition that their participants actually presented with. But in our data we found that individuals who reported age discrimination were not likely to report to be socially isolated. And the result of these studies suggest that older adults may have experienced age discrimination through their social contacts and I like the detrimental effect of age discrimination. So it is very likely that those who reported age discrimination were people who were more socially engaging and this may have been the instances where they experienced age discrimination. The finding from this study also showed that the risk of social isolation and loneliness was significantly higher amongst the oldest, old individuals. And this may be because these individuals are more likely to be living in long-term conditions. They are more likely to be living lonely. And that may be the reason why they are more at risk of social isolation and loneliness. The strengths of this study lies in the study design because it's one of, we were not able to find any study that have done prospective analysis of social isolation and loneliness using age discrimination as the explanatory variable. And the limitation relates to the data attrition, the self-reported data and risk of recall bias. And one-time collection of age discrimination data in the ELSA were hoping that the future ELSA data collection would include age discrimination. Future direction would include the effect of COVID-19 on social isolation and loneliness. We know that physical isolation was one of the measures used in addressing COVID in reducing the risk of infection spreading. So we wanted to see how much this affects social isolation and loneliness among other individuals and how this is related to age discrimination. Our recommendations would be that policy and awareness and international support should be used to combat ageism. And there's a need for social interventions to address age discrimination alongside other factors that influence social isolation. These are my references and thank you for listening. It gives me great pleasure to introduce our final speaker, Sharon Cadigan, who, as we said, is going to talk about herpes zoster shingles. She's a research fellow in epidemiology and electrical health records with an electrical health records group at the Department for Non-Communities Epidemiology at London School of Hygiene on topical medicine. And is currently working on a range of projects related to herpes zoster dementia, cognitive decline using a variety of data sources, including health care for England, about which you will talk today. And the dominantly inherited Alzheimer network data, which sounds fascinating, and linked to UK electronic health data records. So please, Sharon, take it away. Great, thank you very much. As you said, my name is Sharon. I'm an epidemiologist working with the electronic health records group at LSHDM. And I'm going to talk to you today about some research that we've recently completed looking at the prevalence of and the factors associated with herpes zoster in England. So it's a cross-sectional analysis of the health survey for England. So as I said, I'm generally electronic health records. So this is me moving totally in a different direction using survey data. So in today's presentation, I'm going to start... Sorry, I'm trying to change my slide. Sorry. So I'm going to start by giving a quick overview. So I'm going to talk about some of the background of our study and the aim of our study. I'm going to give a very brief overview of health survey for England, because Chloe actually went through this in the very first session this morning if people were there. And then going to move on to the methods of results, strengths, weaknesses, and conclusion of our study. So just to give a little bit of background. So herpes zoster, or as most of us know it as, shingles, is caused by the reactivation of Verceloposaster virus, which is actually more commonly known as chickenpox. So chickenpox actually affects 95% of the population. However, a much smaller proportion of people will actually be affected by shingles in the range of 25 to 35%. So after initial infection, Verceloposaster establishes a lifelong latency in a sensory layer. And then the reactivation of the zoster occurs with declining cell-mediated immunity and affects around 25 to 35% of people. And it can go on to cause some serious complications, generally in older people or people with some comorbidities. The instance of shingles also appears to be rising. And as I said, it's not necessarily an older condition anymore. So it is rising and often in younger people. So the risk of shingles, sorry, the risk of shingles increases significantly with age and immunosuppression. So that's what we already know. However, relatively little is known about other risk factors for zoster. And importantly, a lot of the evidence to date actually comes from electronic health records or medical record data. And many cases of shingles actually don't present to medical care. So for example, in younger people or in milder cases and using or solely relying on administrative data may actually miss these cases. So I'm usually a huge advocate for using electronic health record data and I do love electronic health record data. But I suppose it's just important to note that this is one of the valuable sources of survey data and why it's very useful. So that leads me to the aims of this study. So firstly to describe the lifetime prevalence of self-reported herpes zoster by age and gender. And secondly to investigate other potential risk factors associated with self-reported herpes zoster. So very briefly, what is health survey for England? Well, health survey for England as many of you may already know is an annual cross-sectional survey of a new nationally representative sample of the English population. Is commissioned by NHS Digital and it is conducted by Natsen and University College London. And each year it essentially provides the governments with data on key indicators for health. And each year it actually includes a different population and it also looks specifically at a different topic on top of the core survey each year. So in 2015 for this study we actually requested and funded some additional questions on shingles which is why we have this study today. So how is the data collected for health survey for England? Well, it's face-to-face interviews with self-completion questionnaire it's computer assisted personal interviewing. And the topics that it cover are things like general health, social care and lifestyle behaviors. And also the interviewers also take weight and height measurements. Then this is followed by a nurse visit for a clinical examination for consenting participants. And in that examination, things like weight circumference, blood pressure, urine, blood, saliva samples are taken. In terms of the response rate, so I've just taken a specific one for 2015 here as this is what we're looking at. So there was a 57% response rate for adults overall and there was an 85% of adults within responding households. So generally each year to health survey for England surveys are approximately 8,000 adults. So in terms of methods for this specific study, so our sample looked at 8,022 out of 8,034 adults over the age of 16 who completed the questions on shingles. So it's about 99.9% of them basically completed our shingles questions. As I said, the data sources health survey for England in 2015 and our outcome variable is self-reported healthy disaster. So in terms of our outcome variables, so participants were told that shingles is a painful blistering rash caused by the same virus that called the chicken box. Have you ever had shingles? And if they answered yes, what year did you have shingles and what age were you when you had shingles? So in terms of our covariates, so here we looked at this study covered some covariates around the area of socio-demographic variables, health behaviors and also some clinical conditions. So the socio-demographic variables included gender, age, ethnicity, household size and wellbeing using the WEMVS score. The health behavior variables then included smoking status, alcohol consumption, BMI. So BMI was calculated using the direct height and weight measures that were taken by the interviewer and where these were missing, where people may have refused to have those measurements taken. We supplemented this with self-reported height and weight just for completeness of the data. And then finally physical activity. Then we also looked at some clinical conditions. So we had diabetes, respiratory disease, digestive disorders, genital urinary disorders and mental health conditions. So moving on to the statistical analysis. So all of the statistical analysis was completed using Stata version 16. Probability weights were applied to the data using survey data commands to account for the complex survey sampling strategy. And lifetime prevalence of self-reported herpes zoster was described by age, gender and study sample characteristics and was compared using chi-square test. We then went on and looked at multivariate regression models to compute the adjusted odds ratios accounting for our potential confounders from the previous slide. And then we also taught about potential effect modifiers and looked at age and gender as a sensitivity analysis by adding interaction terms into the multivariate model. So this table is just given a very, very, very brief overview of the characteristics of the participants by their shingles status. I'm not gonna go into much detail here, but basically what you can see is, for example, among those with shingles almost 60% were women compared to 40% with men. And also shingles were more, they were more likely to be in older people and also in those from white ethnic backgrounds. So moving on to prevalence, the overall weighted lifetime prevalence of shingles in this study was found to be 11.5%. And this farce plot shows the lifetime prevalence of self-reported shingles by sample characteristics. So again, I'm not gonna go into too much detail here, but you can see that the prevalence was higher among women. So it was 12.6 versus 10.3 among men. And also, again, you can see differences there in the age groups as they get older, you can see differences in the ethnic white group. Moving on, you can see, you can see here among X, regular X occasional smokers, for example, slightly higher than never in current, 14.9%. And again, when we look at the clinical conditions, we can see, for example, that among respiratory disease, digestive disorders and genitourinary disorders, we can see that there was a slightly higher prevalence than some of the other ones. So moving on to the results from our agent sex adjusted and our fully adjusted multivariate logistic regression models. So here you can see that this table presents the results of both the agent sex adjusted and the fully adjusted models. And after adjusting for a range of socioeconomic and clinical factors, we can see that gender, so we see gender, age, ethnicity, they all appear to have an association with self-reported herpes zaster. So age was a strong predictor of herpes zaster risk and with significantly increased odds among older age groups. And also the odds of having had herpes zaster was 21% higher, we can see here in females compared to males. And also people from white ethnic backgrounds had more than twice the odds of having had shingles compared to those from non-white ethnic backgrounds. So we can see none of the health behaviors here were shown to be associated with herpes zaster. Again, obviously we'll speak under limitations but important to note that you are based in this on recall as well, self-report. So moving on, we can see here that among the clinical conditions, so we can see that the risk of herpes zaster was increased by 51% here and people who reported having digestive disorders. So moving, so I will come back to some conclusions but moving on to the strengths of the study. So the study utilizes data from a nationally representative survey which contains some detailed information on socio-demographic and lifestyle factors. And it also provides self-reported data on herpes zaster which is more sensitive than relying on information from medical records. As I said, a lot of people actually don't present to their GP with shingles unless it's really, really bothering them. And there is quite a lot of cases of mild shingles. And also self-report has been shown to be valid in some contexts as well. So it is valid, you typically shingles in terms of the symptoms. It's quite easy to know that you have shingles if you haven't, thankfully I haven't had it myself but apparently it's very easy to know. In terms of limitations, so obviously the cross-sectional, some limitations do exist. Cross-sectional study design may not provide an accurate picture of health at diagnosis in particular for clinical conditions. So basically we cannot assess temporality and that may result in potential for reverse causality. Also because we use survey data to measure lifestyle risk factors, non-responsive misclassification of risk factors may exist. And here longitudinal studies can be utilized to measure detailed lifestyle factors such as physical activity for example and follow-up for instance, herpes zaster diagnosis. Also the data collected on the clinical risk factors and the medication, in particular immunosuppressive therapies was very limited compared to other data sources. So for example, that is probably one of the benefits of the electronic health records over survey data as you do get quite complete medication information. And then also we did not have information on the herpes zaster vaccine. However, it is important to note that this was introduced in the UK in 2013 and actually only for a very limited subset of people age 70 years and above and with a catch-up campaign for those age 79. So it's unlikely that it actually affected the results in younger adults. And in addition, as with all observational studies there is some possible unmeasured bias as a limitation. So herpes zaster is primarily diagnosed clinically using self-report outcome which may lead to potential overestimation. However, self-reported herpes zaster has previously shown to be accurate compared with physician diagnosis and also herpes zaster verification questionnaire has been compared. And in that study actually, there was one study that did look at that specifically and they actually found that there was a 90, I think it was a 98 or 99% agreement. So just to make a couple of conclusions on the study. So this study found the overall weighted lifetime prevalence of shingles to be 11.5%. Age, gender, ethnicity and the gest of disorders may be risk factors for herpes zaster among a national representative sample of adults in England. Maybe just give a little bit of kind of discussion around those. So for example, take gender, women may be more likely to attend their GP with symptoms with shingle symptoms. And it's possible that hormonal or biological differences between the genders could also play a role. In terms of the digestive disorders. So the digestive disorders were associated with higher odds of herpes zaster in this study. And one potential explanation for this could actually be related to immunosuppression, which is a well-established risk factor for herpes zaster. So for example, people with inflammatory bowel disease, they routinely take immunosuppressant medications such as cortisol and inhibitors. And similarly, you get the same with respiratory conditions. So respiratory conditions in this study, in the age and sex adjusted, they were significant and they did just lose their significance in the fully adjusted models. So just some things to think about in terms of explaining these results. So these potential risk factors, they should be explored in future longitudinal studies for confirmation and also to investigate the possible mechanisms. And finally, I just want to acknowledge my co-authors. So Charlotte Warren-Gash from LSHDM, Judy Brewer from UCL, Andrew Hayward from UCL and Jenny Mendel from UCL. And also the study was funded by NIH or UCL Biomedical Research Centre and Charity Fast Track Grant. And that's it. So I know I'm a very fast speaker, so hopefully you were able to pick up some of that. And thank you.