 Welcome everyone, this is the open session, thank you for coming. So Anna is in her final year of a PhD in health economics at the University of Manchester funded by Welcome and the PhD is evaluating the effectiveness of community approaches at protecting and improving health and well-being and reducing inequalities. This paper is evaluating the health and well-being returns of social participation and the role of community asset infrastructure, so fire away. Okay, perfect. Hi, well so yeah that was a really nice introduction, thank you. So I'll just go straight into the theory behind like what is social participation. So, oh I'm not insuming ages, there we go. So social participation, it comes under the umbrella of social capital and this social capital from the Putnam definition is to do with connections with others that enrich people's lives and these connections have two parts which the tangible part which is the membership of community groups which leads into the social participation and the intangible part which is the community trust element. So we focus predominantly on this social participation as something we can measure but there's another concept that builds into both these which is asset mapping. So this is to do with identifying things within the community that can enable people to engage in social participation and social capital and helps build community trust. So these can be physical spaces such as parks, leisure centers, youth clubs or institutions like schools, universities, associations such as the groups or individuals who have knowledge and skills to engage to run these sort of groups and additionally local economies such as and there's local budgets that put direct money into like museums and community groups. So the introduction why we investigate why this is an interesting topic. So social capital has been shown to broadly improve health outcomes. This is found in a recent systematic review, however there's been an issues of working out what part of social capital contributes most to health. Could this be the tangible part which is the community groups or the intangible part which is a part of community trust and there's issues of which one contributes more. Social participation has led to improved health outcomes and additionally reduced mortality. So there's important associations with individuals outcome but there are endogeneity concerns around social participation such as reverse causality, those are better health more likely to engage in social participation but social participation is linked to better health, improving health and also this selection bias. Additionally access to these community assets to facilitate social participation especially those community assets that are close in proximity and people can walk to them. So we aim to estimate what the health and well being returns to social participation and also if these effects vary by probability of participating because we found that there's different socio-economic status and is the predictor of social participation. So the data that we use for this and we use the UK household longitudinal study and we've got the special license for this which meets the LSOA special alliance so that's a certain level of geography that we have that we can identify the individuals within the survey and we measure social participation from the social and interest group membership component. So people are asked within the survey I'm going to about what types of organizations they're part of and here are the types of groups that they can list that they're a member of and we use this as our social participation indicator so if anyone answers yes to any of those groups then we mark them as someone who engages in social participation and if they say no then they're not. So then the outcomes that we look at so we have the health and well being so for the health we just use the general self-assessed health score which we reverse coded such that poor is now one and excellent health is five and then for well being we use the SF 12 mental health component and this is derived from six questions using a six point like a point scale so the questions that these questions are about mental health if they accomplish less and make some work less carefully whether they felt calm and peaceful how much energy that it had and whether they felt depressed and also if this mental health interfered with social interactions then this scale and then from this that's basically turned into a score of from zero to 100 with the mean of 50 and then it's given a standard deviation of 10. So this is our model as I mentioned before we had some endogeneity concerns so we had social participation on either health or well being we have issues of reverse causality so this reverse causality so this led to the idea of using an instrument variable and as we said before community assets were found to be a good predictor of social participation. The community assets data that we collect is on community centres, sports centres, places of worship, youth clubs and assets of community value and we use an indicator there within one kilometer of an individual's LSOA centroid. Just a quick note on these assets of community value these are special ones that individuals nominate within their community with their local authority and they have special look like protection and that they can't be disposed of and then individuals want to keep them that in their community. Of these five community assets we have to geocode this data so we have to get their coordinates of all of these data points some of them are a lot easy to get some of them involved googling which was quite time-consuming but that's why we focused our findings just on Greater Manchester because it allowed us to have a good urban area of the UK but also meant it wasn't too time-consuming getting all the data. So for an example for the community centres how we calculated whether an individual had them within their area that we use this example within Stockport so the person's house is in the middle and as you can see there's a community centre here, here, here, here and here so basically we use a programme a statistical programme that then calculates which ones are closest which will be this one and then whether and then afterwards we then see if that's within one kilometre radius of the individual's LSOA centroid and we use across the five assets community assets. So the next part the methods that we use to answer this we use the marginal treatment effects model as this allows us to both use the instrument variable specification but also allows us to estimate the treatment effects across probability of treated so we can estimate the treatment effects of someone with a high probability of attending participating in social participation or compared to low probability. So the first stage we model social participation which contains our assets whether they're within one kilometre of someone's LSOA centroid and a range of controls which include an individual level household level an area level just and then from that first stage we use the first stage estimates in the second stage to estimate the effects on the outcomes and we use the same set of individual household and area level controls that and in the final stage we estimate these treatment effects across the propensity scores as from the first stage it's slightly more complicated than a traditional instrument variable model but it's because we're able to see if there's heterogeneous effects across the participation levels rather than just looking at the average person in the middle we opted for this so additionally we can estimate the average treatment effects the average treatment effects on treated so there's a high probability of attending and then average treatment effects on untreated which is the low probability and then the local average treatment effects which is the complies which is a traditional IV model so these are the results these are column one refers to the summary statistics and then column two is the first stage results for the model so for youth clubs within Greater Manchester 54.3% of the individuals had a youth centre within one kilometre of their house by LSOA centroid and they had it 46.7% had an asset community value within one kilometre of their house so the second column refers to the average marginal effects on social participation so here those individuals who had a youth club within one kilometre of their house there were 6.3 percentage points more likely to engage in social participation than those who did not similarly for assets community value whereas the rest they either had counter-intuitive effects or they weren't significant so these are the results so the graph here is the marginal treatment effects graph so on the bottom axis it's unobserved resistance to treatment but the easiest way to see it is on the right hand side it's the lower probability of social participation and the left hand side is the high probability of social participation so the treatment effects are then plotted across each like quintile at the bottom so those with high probability of engaging in social participation had positive and significant returns to their health with the average treatment on treated effect showing a 2.2 increase in self-sustained health and then for well-being we find this similar trend with those of high probability of engaging in social participation having positive significant returns to their health whereas low probability showing negative effects similarly the average treatment effects on treated showed a 20 increase in the well-being score compared to non-participants so what we find is that community asset infrastructure is a strong predictor of social participation particularly the youth clubs and the assets of community value and there are significant returns to those who are more likely to engage in social participation however this result actually could lead to widening existing widening of existing health inequalities this is because we find that those of higher socioeconomic status are more likely to engage so these people already have better health outcomes so in order to have actual health gains then we should be empowering individuals of low social participation rates such as those of low socioeconomic status to engage with these policies of nominating these assets to community values and this is one way that health inequalities could be addressed thank you that's brilliant thank you Anna it's really interesting and a really nice example of using the LSOA as well data so the special license data I'm sure you've got stories about accessing that data as well so Nicole Andali is a research fellow in the economics department at the University of Aberdeen prior to this position she held a fellowship at Queen's University Belfast where she also received her PhD in psychology in her current role she's working on a project examining health and variable payment contracts her interests include financial decision making debt advice and stress oh I'm sure you can help us all out over to you Nicole Thank you very much for the introduction and I should apologize in advance as well if your live transcript goes hey why are here because I'm not sure it can manage my accent but I'll do my best to speak clearly so and yes I have quite an interdisciplinary background and the team that I'm representing today is very interdisciplinary as well so we have Keith Bender and Janice Theodosiu who are both in the economics department at the University of Aberdeen and Julia Allen and Daniel Powell who are in the health psychology group also at the University of Aberdeen and so I want to talk today about an empirical analysis from a larger ESRC grant on performance-related pay and performance-related pay or PRP is when your payment contract depends on your performance or your output so and it's most straightforward sense you can maybe think of the delivery drivers if you're ordering food and they can choose to be paid by PRP so that means that they get paid per delivery that they make but there's also jobs where maybe you're paid partial PRP so you have a portion of your salary which is fixed and then a portion of your salary which varies depending on your performance so commissioned based jobs or a lot of jobs in finance for example or partial PRP and because this kind of the prevalence of PRP differs depending on exactly which definition of PRP you use it means that it could be between 10 to 40 percent of workers in Europe or the US but I hope I've kind of convinced you that there's a significant portion of the population who are employed on these contracts um traditionally um employers kind of like these contracts because they're associated with higher productivity and also higher earnings for the employee but actually in the past two decades or so um there's been an emerging literature which has found kind of an association between people on PRP and also poor health outcomes and so if you think about these health outcomes there seems to be roughly three different pathways in which this could happen so the first pathway is very simple and people on PRP contracts are more likely to get into accidents and injuries so um truckers for example who are paid by PRP are more likely to have accidents and you can think of it um if you think about the delivery driver again and if you are incentivized to work harder or work faster then you might be likely to cut corners and maybe hurray so that you can make that next delivery and then um you might be likely to put yourself at risk as well but we see this across a wide range of industries the second pathway is that if you're working in PRP and you're trying to prove to um work harder or work longer hours then you might have a trade-off between working harder and um healthy activities during your leisure time so you might have less time for physical activity for example or cooking and and you might be more likely to engage in activities such as drinking alcohol and drug use as kind of a coping mechanism and for dealing with the kind of hard work during your working hours and then finally a third pathway is simply that PRP is inherently more stressful than fixed pay so we know that if you're working in PRP then you're more likely to have a higher variable income stream which is considered more stressful and although we're very good as humans at bouncing back from kind of brief episodes of stress um if we're suffering with long-term stress and then that can eventually compromise the immune system and make us more susceptible to these other conditions that I've listed here um of course um there are some gaps in the current literature and so one of those gaps is that the people who work in PRP are a self-selected sample so people who choose to work in PRP jobs might have specific unobservable characteristics which also affect their health okay so maybe it's that people who work in PRP are just more tolerant of risk and if you're more tolerant of risk then maybe you're also more likely to engage in healthy behavior or unhealthy behavior which is risky to your health and these two poor health outcomes so that's that's one issue the second issue is that really a lot of this data has looked at self-reported measures of health and there is a lack of physiological measures of health and so it's possible that if you're struggling with your mental health already which some of these people in PRP seem to do then you might also be more likely to rate your physical health as worse and we did actually run an experiment where we examined the difference between PRP and fixed groups and we found that those in PRP have higher levels of cortisol and but of course it's an experiment so it was a very small homogenous group and of course we could only look at cortisol after acute stress so we don't really know anything about the chronic stress so in the current paper what we wanted to do was to address these issues by first of all statistically correcting for some of this self selection and by using instruments and secondly we wanted to look at biological health markers as well as self-reported health just to get a slightly fuller of picture and we used data from wave two of the UK HLS so what's really good about the UK HLS is that they include an item about PRP although only in every other wave and in wave two and three they also include a nurse assessment which has this biological health data and so because of that we were able to use wave two because it included both the PRP measure and the nurse assessment and although of course only on a kind of subset of the sample so I've shown a graph here which kind of shows how we go from a relatively large sample into a much smaller sample using our analysis and we only used those who said that they were employees so nobody used self-employed or studying or retired and we only used those who are between 18 and 65 so the typical labor force and only those who completed the nurse assessment module now within that group there's kind of different estimates or there's differences between the sample that completed the self-report data and then slightly fewer people completed the blood pressure data or gave their blood pressure and even fewer people gave blood so that we have blood marker data for them so in order to kind of maximize the sample that we have for these different groups of measures that we're interested in we split them into three different samples so the samples overlap very much so but they are slightly slightly different just because we wanted to have as many respondents as possible for each of the health measures and we have a range of health variables so first of all we have four measures of self-reported health which I've all coded so that higher values here indicate better health we have the GHQ which is a general mental health over the past four weeks and we included one item where people rate their general health from 1 to 5 and then we have the two the two components of SF-12 so the physical component summary and the mental component summary which really are quality of life measures which look at how much your everyday is everyday activities limited due to your physical health or due to your mental health and then we have the two groups of biological variables so we have blood pressure which includes systolic blood pressure, diastolic blood pressure and then we created a binary variable where we predicted the probability that you would be above the clinical cutoff for high blood pressure and then we also have two inflammatory markers from the blood samples so we have C-reactive protein and fibrinogen both of which indicate worse health if they are higher or more inflammation and they're both associated with acute stress as well as chronic stress okay so I just thought this was quite interesting if we only look at very simple comparisons of the PRP group and the fixed salary employees and we just compare them on these various health measures what we actually see is that the people in the PRP group have better self-reported general health and physical health C-reactive protein and fibrinogen and so all of us suggesting that actually they are healthier than the fixed salary sample and you might think that this seems like the complete opposite of everything I have said so far and you would be right it's very different to the literature that we've already spoken about and there's a couple of reasons for that so first of all the PRP sample are likely to be much younger they're likely to be higher educated and they're likely to earn more of which are also associated with health so that tells us that it's really important to control for these things in regressions when we run these analyses and then finally we also have this issue with self-selection and previously I talked about risk tolerance but we have self-selection the other way as well which is that if you have a propensity to poor health then you're probably less likely to work in a PRP job where your output determines your income so what we did is we run endogenous treatment models where we estimated a regression for each health outcome and then we included these instruments which predict selecting into PRP but that shouldn't predict the health outcome in question so we included it covariates such as income whether they work in a manual or non-manual work and occupation category age gender education level medical status ethnicity hours work per week and country of residence in the UK and then we also included a couple of health covariates so we included BMI whether they had ever been a smoker and also for the biological markers we included taking prescribed medication because some medications can affect both your blood pressure and these markers in the blood and then finally we have the two instruments so the two instruments are firm size as well as a percentage share of PRP workers in the occupation that we work in and so firm size and predicts working in a PRP job because larger firms are more likely to use PRP but there's no reason for why firm size should predict your health in the UK anyway and then and percentage of share of PRP workers in your occupation if you work in an occupation where it's very common to have PRP contracts then you are more likely to work in PRP but the and but this percentage share shouldn't affect your and your health outcome and so we tested this as well in regressions and indeed we found that they do there are significant predictors of PRP but they are not significant predictors of health once you control for PRP okay so and what I've done here is I've only shown you the marginal effect of working in PRP on the health outcomes I haven't shown you any of the covariates because we would be here all day if that was the case and I've also included the kind of means for the overall sample below so that you can get an idea of what effect size this is and on the on the means so you can see for example that the mean GHQ 12 is 25 out of a possible 36 but the PRP group are likely to have 6.9 units less than the fixed salary group to quite a quite a big effect really and what we see here is that PRP is a significant predictor of worse general health and worse self-reported mental health they have higher systolic blood pressure and higher levels of fibromedium and so it's not that we see a significant effect of PRP across the board but we do see it on a range of variables which are connected to chronic stress interestingly we also see that it predicts less activity limitation due to physical health so there's no apparent reasons for why that would be but it is possible that our sample is doesn't really contain any people with severe mobility issues and it might also be that this is something you know severe impacts on your health it wouldn't be seen until you've worked in PRP for a very long time and so maybe our sample doesn't capture that what we did see also was that there were several covariates which predict these health outcomes and that includes gender as well as whether you work in manual or non-manual employment so what we decided to do was just to look at the same regressions but within these specific sub-samples just to dig a little bit deeper there and what we can see here is that although there's plenty of results which are very similar across the different sub-samples there are a few differences so we see that the effect of PRP leading to better activity limitation due to physical health seems to actually be driven by male workers and also we can see that male workers have and that is a sample where we see the impact of PRP on higher C-reactive protein and fibrinogen whereas actually if we look at the female sample or the non-manual sample we see that they have slightly lower levels of C-reactive protein and again we weren't sure why we see these results and I'm very happy to hear any suggestions that we might have and one thing is that we do predict or we do control for how many hours they work but we don't have any information about kind of their shift schedules or anything like that and so it might be that working in PRP allows for some flexibility which is very much appreciated in certain sub-samples we know from other studies that women might appreciate flexibility in the workplace more due to other factors and so maybe that cancels out some of the effect of PRP on health. Okay so to conclude we have some evidence for PRP workers suffering from worse mental health and biomarkers and specifically when we look at those related to chronic stress we do see some affections as well and male workers seem to have better quality of life regarding physical health and we see some of the opposite effects in specifically female and non-manual worker samples. There are some limitations to our study so we only have a very broad measure of PRP and we don't know whether they're working in full PRP or impartial PRP and we would expect to see different effects so people who are working in full PRP are likely to be more stressed and we can't tease that apart and we also don't have any information about risk preference or personality traits although we do include smoking as kind of a proxy for risk and we do think that self-selection should control for some of that. However we do think that we have some results that suggest that using PRP contracts can have a detrimental effect on the employed population and this is something that employers should take note of this off because even though PRP leads to higher productivity we also know that work-related stress leads to high absences in the workplace so this might actually cancel out that positive effect for employers. Okay thank you very much I've included a link here to our website if you would like to read more about our project and I'm very happy to take any questions if there are any. So thank you very much and we'll move on to our final speaker who is Mel Der Lang. Do you want to share your presentation Mel and I will introduce you. So Mel Der Lang is a Wellcome Trust funded PhD student in the MRC Integrative Epidemiology Unit at the University of Bristol. She has a Master's of Research in Health and Wellbeing from Bristol University. Mel's current PhD research looks at the effects of the spring and autumn clock changes on cardiovascular outcomes and depressive symptoms in the UK. Her wider research interests include exploring the effects of modifiable risk factors for disease such as sleep, nutrition and physical activity. Sounds all sounds fascinating so over to you Mel. Okay hello everyone so today I'm going to talk to you about some research that I did for my master's dissertation which looked at the time windows that people eat within each day and whether this is related to metabolic health. So research is increasingly showing that when we eat in addition to what and how much we eat is important for our weight and health. Time restricted eating is a form of fasting that involves limiting food take to a certain time period each day usually 12 hours or less without changing the quantity or quality of your diet. In general fasting is believed to be beneficial for health because it causes the body to switch from burning glucose to burning fat for energy and gives the body time to rest or repair itself. An addition time restricted eating is thought to be beneficial for metabolic health because it works for the body's own natural circadian rhythm. Animal studies strongly suggest that time restricted eating is beneficial for metabolic health however the evidence from human studies is less convincing. So for example studies of Ramadan fasting suggest that time restricted eating is beneficial for healthy individuals but not necessarily other subgroups. Meanwhile small scale clinical trials suggest that whilst adherence to time restricted regimes is high they offer little metabolic benefit over normal eating patterns. Other researchers looked at what eating windows people naturally eat within in the real world and by daily eating window we mean the time between when somebody first starts eating or drinking in the morning and when they finish eating or drinking in the evening and studies have shown that generally most people eat for a period of over 12 hours a day so could potentially benefit from time restricted eating. However the number of studies conducted is small and it's difficult to generalize the results because of small sample sizes. In addition there's been little research looking at the link between daily eating window and metabolic health and there's been little research done in the UK. This means that the true potential of time restricted eating to improve the metabolic health of UK adults is unclear. So the aim of this research was to evaluate the potential use of time restricted eating as a dietary intervention to improve the metabolic health of UK adults. More specifically we wanted to use a large representative sample of UK adults to identify what length of daily eating window people eat within in the UK. We wanted to identify the socio-demographics and health behaviors of those with the longer daily eating window and we also wanted to explore whether there's a relationship between metabolic health and length of daily eating window and then finally we wanted to formulate recommendations as to whether time restricted eating could be used to improve the health of UK adults and if so what daily eating window could be feasible. To do this we conducted a secondary data analysis of the UK National Diet and Nutrition Survey. The NDNS is a cross-sectional survey conducted every year to examine new nutritional intake and eating habits of the UK population aged one and a half and over. It collects data from a representative sample of the UK population. Data collection takes place in two different stages. In stage one participants are visited by an interviewer. They complete a computer-assisted personal interview. They have their height and weight measured. They complete a physical activity questionnaire and they complete a four-day food diary which I'll talk a little bit more about in a second. For those who complete three or four days of the food diary are then visited by a nurse. They have their fasting blood sample taken, their waist and hip measurements are taken and their blood pressure is measured. This is an example of the food diary used in the NDNS. So as you can see as well as information on what people are eating, there's information on the time of each eating occasion in hours and minutes. So participants are asked to continue their normal eating habits and record the information as they go along rather than relying on memory and they complete the diaries on three or four continuous days with the start day randomized to ensure a representation of all days of the week. In our study we included all participants who taken part in any of the years one to nine of the NDNS. We only included adults aged 19 or over who have completed three or four days of the food diaries. This gave us a total sample of 6,802 participants of those around three quarters were visited by the nurse and just over half provided a blood sample. Our exposure variables were our measures of metabolic health so BMI, waist-hip ratio, systolic and diastolic blood pressure, cholesterol, triglycerides, C-reactive protein, HVA1C and glucose concentrations. We had a length of daily eating window as the outcome variable because we wanted to understand the associations in the context of how much daily eating windows change. We also included a variety of socio-demographic and health behavior variables in our analysis so age, gender, ethnicity, employment, education and also behavior variables like physical activity, sleep, smoking and alcohol consumption. In terms of data analysis it's worth noting that the NDNS data are fully anonymized and they are waited for selection and non-response bias at each stage of the data collection process. In terms of statistical analysis we conducted a simple linear regression analysis to examine the relationship between socio-demographics and health behaviors and length of day eating window and then we conducted a multiple linear regression analysis adjusting for confounders to look at the relationship between metabolic health and length of daily eating window. Moving on to our results, in our study the main daily eating window was 13 hours and 33 minutes and 78% of participants had the day eating window of over 12 hours. Our simple linear regression analysis showed a number of characteristics associated with having a longer daily eating window. These included being older, white and male, being degree educated, being employed, not smoking, having a higher calorie intake, having a higher proportion of your calories coming from sugar, getting less sleep, being more physically active, drinking alcohol more frequently and having a higher fruit and vegetable intake. Our multiple linear regression analysis showed that after adjusting for confounders, BMI or waist tip ratio and LDL cholesterol, the so-called bad cholesterol, were negatively associated with the length of daily eating window and we also found that HDL cholesterol was positively associated with the length of daily eating window. So these results were the opposite of what we expected because they suggest that worse metabolic health is associated with a shorter daily eating window. But in keeping with what we did expect, CRP, which is a measure of inflammation, was positively associated with the length of daily eating window. It's worth noting here that the effect sizes for the metabolic markers we found were too small to be meaningful in the real world. And also we found no association between daily eating window and our other measures of metabolic health, so HBA1c, glucose, blood pressure and triglycerides. The strengths of this study include the large nationally representative sample, as well as the wide variety of nutrition, health and socioeconomic variables available in the NDNS. The limitations include the fact that people tend to underreport in food diaries so they may accidentally or deliberately forget that glass of wine or snack that they had before bed and so their actual daily eating window may be longer than that recorded. Also, participants only recorded one time point for each eating occasion, not a start and a finish time. So again, their actual daily eating window may be different from what they report. The cross-sectional design of the study means that we only have a snapshot of people's eating patterns and metabolic health. It means we only have associations, not causal relationships, because there may be residual confounding. And also having daily eating windows, the outcome variable made it quite difficult to interpret whether the results are clinically meaningful. So it'd be useful to look at the relationship in the other direction, so having daily eating window as the exposure variable, because there are differences in metabolic markers such as weight that are recognised to be clinically significant. So the results of our study suggest the time restricted eating could be a feasible intervention for UK adults because having a daily eating window of 12 hours would only require a short reduction in daily eating window for most people. We also found distinct sociodemographic and health behaviours associated with a longer daily eating window. So these people could be identified and targeted by time restricted eating interventions. However, we found inconsistent associations between metabolic health and daily eating window and the effect sizes were too small to be meaningful. On this basis, it's not possible to recommend time restricted eating to improve the metabolic health of UK adults. Future research in this area could use longitudinal data to look at people's long-term eating habits and their long-term metabolic health. It would also be useful to look at the timing of the daily eating window kind of earlier in the day versus later in the day and whether this is related to metabolic health. Future studies could also use more objective, accurate and passive measures of food timing, as Laura Johnson mentioned this morning. So risk-borne accelerometers, wearable cameras, and continuous glucose monitors. Thank you for listening.