 So today's webinar again entitled functional support and memory, a perspective analysis of the CLSA comprehensive cohort is being presented by Samantha you and Samantha is a PhD student in epidemiology and public health at the University of Ottawa. She's researching public health effects of folate and vitamin B12 deficiency at a global level and assessing relevant public health policies. She also works as a data analyst at the genetics and genome biology program at the Sick Kids Research Institute. Samantha completed her master's of science and epidemiology and bio stats at the University of Waterloo in 2021, where she studied the association between functional social support and memory among middle aged and older adults in Canada using three year data from the CLSA. Her research interests include epidemiological methods, causal models, evidence synthesis and chronic disease epidemiology. And now I will pass it on to her presenter to get started. Thank you. I hope everybody can see me. Yeah, so without further ado, I'll just go on with my presentation. Hello everyone. Thank you Shirley and Jennifer and the rest of the CLSA team for this wonderful opportunity to share my research today. And hello and thank you to everyone for logging on today. I know that everybody is very busy at the end of the year. I hope you find the presentation today very interesting. I'm not sure how many of you are engaged in research related to social support and or covering the function and or memory. I've also heard from the organizers that some of you may be from the general public not directly conducting research with yourselves. So I did prepare some background sliders with definitions and context and classifications as so and so forth. Let me begin. So this is just an opening slide just to show you the prevalence of dementia, although dementia was not the focus of my study it's a it's a severe form of neuro cognitive disorder. You can see here that the prevalence is has been going up, and it's a little higher among women than men, and the risk of dementia goes up dramatically after the age of 65. And when you look at this severe form of neuro cognitive disorder so dementia, you can easily imagine that this condition requires a lot of support from caregivers and families in many different forms, such as helping with daily routines, reminding medications, providing cognitive information and so forth. So this makes it very challenging, as you can imagine for people with cognitive declines to live through a pandemic as we are today, as in person care and many different forms become very limited and daily routines, which are important for folks with cognitive declines become more destructive and restricted in many ways. And according to research approximately 40% of dementia cases can be prevented or delayed. And, and this is an important part of information for as a context of my research as prevention is one of the key focus areas of Canada dementia strategy, which was published just last year. Now moving on to the main part of my presentation. So I'd like to go over some definitions classifications and some background theoretical information about the elements included in my study. So this is how cognition is defined under the DSM presentation combination is a complex thing that comprises of these six different domains. So learning and memory, executive function, perceptual motor function, language, social cognition, and complex attention. And if we zoom in on memory, particularly, it's typically classified into these four subtypes, episodic memory, semantic memory, implicit memory, and working memory. Episodic memory refers to an ability to actively retrieve personal experiences in personal contexts. And implicit memory is about recalling past experiences without much conscious effort like riding a bike, for instance, semantic memory is holding a structured record of facts, concepts and knowledge, like the things we learn in school and education Working memory is storing and using information for short time periods, recalling the grocery list as we were doing grocery shopping in the supermarket. So these are the examples. And as you can imagine, memory supports many facets of our lives, maintaining relationships, performing job functions, putting things into context and ultimately maintaining functional independence. While many of us think that memory loss is a natural part of aging process, memory impairment that goes beyond a certain point is not part of normal aging. And that is linked to higher risk self neuro qualitative disorders. Episodic memory has been reported as one of the first domains in paired and qualitative decline process and impaired episodic memory would look like being unable to remember your latest family trip, or making many simple mistakes performing the tasks that used to be very familiar to you. And this can lead to feelings of uncertainty, frustration, fear and depression. Memory is another domain that declines substantially with age, and this affects the daily task performances, like reading, writing, making plans, and also increased distraction. So cognitive decline occurs on a continuum from minimal to severe impacting a single domain or multiple domains at the same time. Normal age related cognitive changes are small, and they do not undermine daily functions. And neuro qualitative disorder also does not undermine our ability to perform daily activities, but activities may be performed at sub optimal levels and may require some more effort. Now major neuro qualitative disorder or dementia is characterized by a progressive loss of functional independence. So individuals ultimately lose the ability to perform the basic tasks of daily living. Now that age is a greatest risk factor, one of the greatest risk factors for cognitive decline. But other than that, what are the factors or known to accelerate or prevent cognitive decline. And our research has been done on biological factors as you can see on the slide. Some genes, some hormones, notably apolipoprotein e4 allele have been found to be associated with the onset or progression of cognitive decline. And out of many clinical research, we now know about the associations between several health conditions, as you can see here on the list, and increased risks of cognitive impairment. And these health conditions are also associated with the lifestyle and psychosocial factors, as you can see below. Lifestyle risk factors, predicting poor cognitive function, or physical inactivity, smoking, excessive alcohol use and fatty diets. Psychosocial risk factors have been relatively less studied, the existing literature largely focuses on depression, stress, and social environment. And social environment is what I've been talking about as it relates to social support and increased social support has been cited as a potential buffer against cognitive loss in older adults. So moving on to social support. Social support refers to the social resources that we can use to help with making decisions, solving problems, and in general maintaining positive experiences in life. We can use different terminologies to describe different aspects of social support, like social engagement, social relations, social integration, social activities, and so forth. But conceptually, social support has structural and functional aspects, structural social support refers to the size of the social networks, like how many people do you know. And the frequency of social context, like how often do you meet with these people for interaction. And on the other hand, functional social support is about the perceived nature of social relationships, things like to have close friends. Do you get practical help print support emotional support when you need them. And are you satisfied with these relationships. So the medical outcomes that he social support surveys a tool that measures functional social support specifically, and it brings it down into four components as you can see the very bottom of the tree. So emotional information support is about having someone that understands you having someone that can give you advice, and that you can show your deepest worries. So the changeable support is about having someone that can help you with chores earns driving your doctor when you're sick affection and support. As you can imagine is having someone who we can share love and affection with and positive social interactions is about having someone to have fun with to do leisurely activities with for positive distractions. What's the type of functional social support do you think is likely most strongly associated with cognitive decline or memory function, or most weekly associated. You can drop your text in the tab box. I'm waiting to see if any chats come in. But this was a question that I asked myself a lot during my own research imagining different people in different scenarios, different environments, and I kept coming to a thought that functional social support maybe an evolving thing. Taking different forms at different stages in life, or even taking different effect on your cognition depending on your circumstances or even personality traits. My slide. Yeah. So actually it's important to take a lifespan approach as our needs and our perceptions change as we live into older ages. And there are some theories that describe how functional social support works throughout the lifespan. So the first two models social convoy model and the social emotional selectivity model. They explain that we maintain a stable level of social relationships throughout our lives, but the makeup of these relationships change with time in age. For example, as we age, we may shed less important, more formal relationships to focus more on more important more rewarding and more intimate relationships. The social specificity model says that our need for specific type of support can be best fulfilled by certain individuals and if that support comes from somebody else, it might not have the same effect. So the second model is also very interesting. It suggests that support and conflict may coexist in close and cohesive relationships, and we experienced love and tension simultaneously in close relationships as opposed to having relationships that are always positive or always a negative. So then how does functional social support affect cognitive health. A majority of the studies reported that higher level of functional social support predicted higher global cognition higher perceived social support was found to be associated with higher cognitive function, both globally and across different domains, and also predicted more decline in cognition as well. Emotional informational support show the strong show the strongest association with higher cognitive performance among older adults. And in terms of memory, high level of functional social support predicted delays in memory decline among older adults conducting research is conducting in many different countries. And a number of potential mechanisms explain this associated between functional social support and cognition. The most relevant one in the context of my research. So functional social support is the first one there stress buffering hypothesis, which explains that chronic stress arising from poor or no social can lead to permanent loss of hippocampal neurons and structural damages to the hippocampus stressors trigger elevation of cortisol levels, and this impairs the cognition, and also stressors can induce structural damages in the hippocampus which is one of the key reasons that regulate cognitive function. So under the stress buffering hypothesis social support may serve as a question against stressors. For instance, you may perceive a potentially stressful situation, a little differently if you know that helpful resources will be available to you. Or if others around you help you think that the problem that you perceive as stressful is actually solvable or acceptable or not as bad as you think. Three other mechanisms that I put here on the slide, but in the interest of time I'd just like to move on to the next slide. If you're interested, you can you can discuss them over the question and answer session, or even look them up yourself at your own time. So there's something of the literature search that I have done. The majority of the studies show positive associations between functional social support and poverty the function in older adults, and this may be explained by multiple theoretical mechanisms. However, there were some gap in the literature as well, limited evidence around whether Asian sex moderated the association between these two variables, and there was also a significant heterogeneity in how social support is defined and measured. Many studies fell short of distinguishing between structural versus functional social support. And different skills were used to measure the functional social support as well so it was difficult to compare across different studies. And also a lot of those studies in the literature were limited in scope as well focusing only on older adults above the age of 65, or focusing just on certain cities or certain provinces or certain states. So CLSA was a wonderful platform to overcome this potential gaps in the literature. So based on the 15 evidence and the gaps in the literature I came up with these three research questions. I was interested in finding out the baseline level of association between functional social support and changes in the score over the years in a community dwelling Canadians aged between 45 and 85. And if this association maintained after controlling for different potential confounders. And if this association was modified by age and sex. This is just a brief slide on how the CLSA cohorts of one and the sampling frame and the inclusion exclusion criteria. Any of you would be familiar with the CLSA framework, as I research platform so I don't want to go too long here. And I just want to come to how I came to my own analytic example from the comprehensive cohort. So the bean slide sample size of the comprehensive cohort was 30,097. And over the course of the first three years, you lost about the percent of that population to three different reasons stated here. And participants who either withdrew or dropped out were generally older, I'm at lower level of education, lower level of income and pro level of self rated health, compared to those who remained in the study until the first wave of the follow up three years later. And the analytical sample if you see the flow chart on the right side. I did a complete analysis. So I filtered everyone who did the regular testing at the data collection site at both time points baseline and follow up. And I only included those who had full, who had complete memory scores at baseline and follow up, and also full complete functional social support scores at baseline and complete values for covariate covariates at baseline. So my final sanitize was 2011 on participants. I'd like to briefly go over how I operationalize the functional social support and memory. So first, I did this. This was the outcome variable. The auditory verbal learning test was the simple memory measure using the CLSA ensure others call it rabble. So it does measure immediate recall and delay recall. And I standardize it into that scores separately for English and French speaking participants. They had different names. And as you can see, I combined these two. We needed recall and delay recall scores into composite levels that score. So I averaged them. As you can see, and this was because they had comparable distributions at both time points. And also because of the literature does not distinguish separately how the delay recall versus immediate recall or vice versa is associated with the crime or memory. They, for the most part, if they did measure both separate recalls, they combined them into a composite score and analyze them as a single entity of memory score. And the final outcome variable was the difference in the memory score between baseline and follow up, calling it a change score here. And this was computed by subtracting the combined memory score at baseline. Moving on to the functional social support variable. The medical outcome study social support survey MOS SSS was used to measure the functional social support in the CLSA cohort. And this is a 19 item stuff administered survey measures overall and for the best of types of functional social support. We have a list of the 19 questions on the slide and how each question is assigned to a subject, except the last item, which was computed to an overall score. So each question, the participants are asked to read the level of support of the scale of five, one to five, one is not at the time to a little over time to all the way to five all the time. So the analysis I dichotomize the functional social support score into high or low with the high being scores for or higher, and the low being less than four to come for the skewed distribution of the score. And other than the predictor and the outcome variables, I included 12 covariates and analysis, all of them informed by the literature to control for potential compounding. The baseline value of these covariates, but their distributions may unchanged throughout the three years. And as per the CLSA recommendations I included the participants age, sex and province of residence is a priority variable in all of the models. And now to this for the statistical analysis, I did complete case analysis. I will discuss it later but the nature of missing it missing this in the outcome of the variables did not appear to be at random or complete at random so complete case analysis was a safer approach. And five sets of multiple linear regressions, each for the functional social support, overall, and subtypes, and for each set of base model included age, sex, province as well as the baseline numbers for as informed by the literature, and the full model include all the covariates. Each full model was then assessed for was certified by agents as well. And each of the full model was assessed for fit using technostics do residual plot and observe versus predictive plots. And I was able to assess the potential impact of missing data, I conducted a by various sensitivity analysis to examine the differences in the distribution of the function social support among people with complete versus missing memory values. And also distribution of the members scores among people with complete social support scores versus missing social support scores. First, I like to just go over the descriptive statistics characteristics of the participants at baseline. This is all weighted. And you can see sex was evenly distributed between male and female, but age was not. You can see that over 70% of the participants was under 65 years and only 11% over 75 years older. In terms of the province, Quebec and British Columbia accounted for the for the largest shares and education over 85% of the participants at baseline had some post secondary education. And as for the annual household income, a little over 43% of the participants had over $100,000 per year. And the distributions of our marital status and living arrangements overlap on this exactly with 75% in marital or common law relationships and living with someone. In majority of the participants were healthy, as you can see, almost 70% of the participants had one or no chronic condition, and over 90% had no functional impairment at all, and only 9% stuff reported as curse smokers. Overall, we get this picture of the participants as younger, so young old adults, less than 165 years of age with high level of education, high level of income and relatively good health. Now let's take a look at their functional social support scores at baseline. The distribution is left skewed. As you can see on the right side of the screen with the median scores of all their variables ranging from 4.33 to 4.70 out of five. And 75% of the participants scored 4.80 higher in each subtype. When categorized into high or low scores, 92% to 95% of the participants reported high functional social support. And these distributions were sustained over three years. And the functional social support scores were comparable across different age and sex strata. And all strata over 90% of the participants reported high scores for each functional social support variable. To the memory scores. These are the standardized unraveled scores in the table you can see the scores for immediate recalls, delayed recall, gravel one, gravel two, and gravel combined the combined memory score. And all three distributions are comparable at each time point. The mean score decreased slightly from baseline to follow up by about 0.25. If the memory scores change over three years, you can see that 57% of the participants experienced a decline in their memory over three years while the rest had their scores increase. And although not shown here, over half of those who inclined, who declined or increased had small increases or small decreases, a change of less than one point, one point. So we have this normal distribution of change score with the mean at negative 0.24. And let me specify agent set we see some differences. First, the change scores were the largest magnitude of negative in the youngest age group. So that's 45 to 54 years, which means that they had the largest decline, compared to other age groups. In contrast, the largest positive was seen in the oldest age group 75 years older, which means they had the biggest improvement in the memory scores over three years. And you can see in the table here that the mean and the median of the change scores actually increase with each older age group. And between the sexes, female participants showed substantially larger magnitude of negative compared to males. Now let's look at regression analysis results in both the base model and the full models, the regression coefficients for functional social support were rather small. So in terms of the direction of the association, the regression coefficients being positive indicates that high versus low baseline functional social support is associated with an increase in memory change score, which means more improvement. But then also the 95% confidence interval contains the note value here so we were not. So this was not very conclusive. And in terms of the magnitude of the association, a larger positive regression coefficient would mean a greater increase in the memory scores over three years. But in my models, the magnitude of the regression predictions were small, and probably because the increases in the change scores, the memory change scores were less than one point for approximately half of the participants in whom the change scores actually increase. So the potential support at the largest and it was the only statistically significant variable, having effect on across all models, and all other sub types at their 95% confidence intervals, including the note value. So starting to find the full models by age and sex. First, looking at the age groups, none of the coefficients were statistically significant at the 5% level, and the 95% confidence interval, as you can see, was the widest for the oldest age group, as this group had a smaller sample size. The tangible support for males at the largest and the only statistically significant effect. And now this is about the impact of the missing data impact of excluding participants with missing data. So comparing the group with missing values showed consistent results. And first, participants with missing memory scores at baseline reported lower scores on all five functional social support variables compared to those who had complete memory scores. So the three or follow up those with missing memory scores at equal or lower baseline functional social support scores as well. These were statistically different differences, statistically significant differences. Continuing participants with missing functional social support scores also reported lower mean memory scores at baseline, but slightly higher mean memory scores at follow up. And the pattern was consistent across all five variables of functional social support. The differences were not statistically significant for the affectionate support tangible support and positive social interactions. Overall support and emotional information support the differences was statistically significant. So just briefly captain the findings, the regression models did not suggest a strong association between functional social support and memory change over three years, among the middle and older, except for tangible support. All of the 95% confidence intervals included another value and you couldn't conclude with certainty about the direction of the association. And after controlling the confounders, the magnitude of the association diminished slightly and changeable support was the only variable that had statistically significant effect on memory change in the fully adjusted models. And the evidence for that modification by age and sex was not evident. So the results was a were a little different from what I expected. And so I spent a fair amount of time revisiting the literature and try to get some insights from the previous studies. And I narrowed down the literature to these five studies that were most comparable to my own research, a larger scale on the design, looking specifically at functional social support as a predictor, and studying cognitively healthy adults, who are also middle aged or older. So these studies reported a positive association between functional social support and permission. But then if you read closely, their predictor and the outcome variables are slightly differently defined than mine. So comparison was not so feasible. Meanwhile, few articles reported inverse or no associations between functional social support and cognition. And there were some recommendations provided by these authors, some cited on reciprocity theory, which is simply in receiving social support that you cannot reciprocate due to illness or other limitations may cause stress, anxiety and depressive moods. And this may affect cognitive function negatively. But this is not really relevant to my research because as you understand my participants were healthy without much functional impairment. There's also a little description of distinction between fluid and crystallized intelligence and how this may affect the association, the direction of the association. But we need more research to, to map different cognitive domains and tests to different types of intelligence. So the sex differences, some authors explained that for men, high level of tangible support at baseline may also indicate incident declines in cognitive function, because men typically have less intensive networks of social support. And so if they are experiencing a cognitive crime, they may actively seek support to maintain their function. But then this again may not be relevant to the CLS example, given their characteristics. A valid possible rationale relevant to my research include detecting cognitive changes using your psychological tests would be difficult in non or pre pathological stages of cognitive impairment. So longer follow ups and more market declines in cognitive change would be needed to ascertain the association between these two variables of interest. And also a selection bias in favor of participants who perform better than average on memory test was identified in one study, and this may be relatable to my own research as well. Overall, for the research is needed and currently the evidence is quite limited. And so applying these insights to my own research, the sample using my research was younger, physically healthier than the typical samples studied in this field. And my participants may have needed less functional support, compared to those in other studies. And the analytical sample drawn from the baseline CLS data set may not have been optimal for assessing changes in memory over a three year period. A large proportion of the analytical sample was cognitively healthy, because the CLS is screened out persons with overt signs of cognitive impairment at recruitment. And adding to that the level of commitment required to participate in the CLS a comprehensive cohort testing may have the incentivize older adults with maybe some small signs of cognitive challenges from taking part in the study. So this may have resulted in the recruitment of highly selective subgroup of cognitively healthy older adults. And if you recall the oldest age group in my analysis showed the largest improvement in memory over three years, and the number of people with younger age group was twice as many in the older age group than the younger age groups. And lastly, given the cognitively healthy sample, a three year follow up was unlikely to be long enough to detect changes in memory, other studies with relatively short to medium follow up some people without cognitive impairment also found some needed results. So wrapping up the presentation with strength and limitations of the research. So the scope of the CLS a was a big strength that I was able to benefit from the CLS a covered middle. And older age community dwelling adults from seven out of 10 provinces. It's really well with many other previous studies with limited target populations. And also the rest of the longest between the analysis, the longest original design, helping mitigate the reversal causation bias. In the month of covariates, I was able to control for almost all of the potential confounders cited in the literature as well. And the MOSFFS scale is to measure the function of social support were well validated and reported for higher high level of reliability. And also, I was able to look at memory that distinct outcome, because many, many other research memory is just included as one of the cognitive domains and usually just combined into the public global cognitive score. And these are the limitations. CLS participants for volunteers, and they reported higher level of education income and health, compared to the average person in the same age range in Canada. They also reported very tight and left through distributions of functional social support scored at both time points, which could have reduced the variability needed to detect the differences in memory and change scores across the entire spectrum of the functional social support. Another possible selecting bias may have occurred at a participant who showed over a science of cognitive impairment or screened out at baseline. And occasion of people who are generally older and at lower level of education income and health may also have resulted in some selection bias. And also the proportion of participants with missing members scores increased with a pop from approximately 4% at baseline to about 22% at follow up. And after excluding all other produce participants with missing values in different variables using analysis, the analytical sample was reduced about 40% of the baseline cohort. And one last point there is the absence of number of data. Having a number of data for the functional social support, the MSSF scale or rabble score would have been helpful in interpreting the, the scale scores, and the magnitude of the vision coefficients in light of the benchmark indicating the type of scores that you would expect in an average population. That's a good question. My research found a very weak association between functional social support and memory over three years in the comprehensive cohort of CLSA at a descriptive level, approximately half of the participants had their memory scores increase and we have other half decrease. The tangible support was significantly associated with higher change scores in memory, and the evidence of the modification by sex or age was not clear. We would need longer follow up to be able to detect clear associations between functional social support and memory. So, this is it for the presentation. I haven't heard anything from Charlie so I hope I didn't go over time. No, I think. Thank you so much Samantha that was great. We do have some just first of all just a reminder to post your questions. If you have if you have any I know there's actually been a lot of discussion already. The first few questions went back to I think if you think back to your methods slides. There were some questions about why whether you included non marital relationships. Why you didn't include physical activity or races covariance. To answer the first question first. I included. So, I include the living arrangements, which was living alone versus living with somebody. So, in the middle status, there was common law relationships, metal relationships divorced, separated, we don't. So, I, and I group common law and metal relationships together. So I guess common law would be with was included as a variable. Did I answer the question. Yeah, I think so and then the other parts were why you didn't include physical activity and race. Physical activity and race, race. I did not specifically include the race or ethnicity. I understand that from other research done on the CLSA that languages spoken and ethnicities included and represented in the CLSA were diverse. I wasn't able to see much in the literature preceding my, my research that ethnicity was a factor, a big risk factor, or confounder between functional support and cognition or memory. I did see some researchers don't specifically on the, for instance, Chinese Americans or some Asians in the Asian pop in the Asian countries, or immigrant population in living in the western world, and also some studies in the UK Scandinavian countries as well. But then the race itself didn't really come out as a big risk factor or confounder, not really discussed much in the literature so that's why I excluded it. And I also didn't want to put too many covariates to adjust for because it could also kind of burden the models with overfitting as well. And physical activity. I excluded it because I was a little skeptical of how this could actually be measured there's a lot of controversies around how physical activities can be usually self reported, and it could be kind of subjective. And it can't really cannot really be have a physical level objective and actually measured. So, and also it has some time varying nature as well so depending on when you ask. You know the answer could could vary right. So that would having a time varying confounder and the model would also kind of complicate the analysis as well. And the relationship between physical activity and cognition is already being well established. I didn't really feel the need to input it hope that answer the question. I think that was very comprehensive. Also sort of methods related or analysis related. Why did you not use inverse probability or censoring weights or multiple imputation to consider missing data. About the multiple indication, I did consider including the missing values because we did have a lot of missingness. And it was a concern. Towards the end. But then to be able to impute using a robust model. We need first, we would first should be able to assume that the missing as is at random, or completely at random, ideally, but I wasn't able to come to that conclusion. And, as you have heard, if you did the, you know, when you look at the sensitive analysis, the, the people with missing dollars had war reported lower level of. Missing memory scores reported lower level of function for the support and vice versa. So it wasn't it didn't really seem as missing at random. So I did not think that it was safe to just go ahead and impute the data. So I did not consider using inverse probability that some propensity score right I think it could have been a could have been a good approach to kind of even out the different study samples but it wasn't like we had like an exposure group versus non exposure group. It was a cohort study, but the exposure group. So if I, if you were to think that the exposure group was the group with the lower or higher level of social support, then the distribution was very skewed. Right. So I would be very interested in actually going back and calculating the propensity scores and how that would kind of change the analysis. That's interesting. Thank you. Great. Good question. And then there was just somebody just asked to clarify which social support theoretical framework you use to measure functional support theoretical framework. I mentioned in the slides for the, the, how social support would be associated with cognition or memory and stress preference hypothesis was what I focused on as how that would explain the association between support and cognition. And then related to that. How does emotional support differ from affection support. And what about affirmative support. Emotional emotional informational support was was it was defined in the MSSS scale as having having a confident having a very close best friend that you can share your fears. You can show your crisis with it again, you can trust that you can get sound advice and support from versus affection support affection support is more of a question that anyone to do have someone to love to share hugs to share effects to share affection with so slightly different, but I get, but I also get the question because I think a lot of it would kind of overlap, because in between spouses you would get emotional support as well, but some literature actually goes further and specify these and say emotional information and our support comes best from friends, compared to family members and affection support comes best from spouses. Yeah, I suppose these are difficult concepts to to tease out from, from one another. So hopefully that answered Carolyn's question. The next question is not sure if you missed it but did you see the moderation effect of age and sex. The evidence was not clear. Yeah by age or sex. Yeah. And you also said the social support component is negatively skewed. How did you address this. Oh, the distribution of functional social support score at baseline was skewed. So, we reported high level of social support, a functional social support at baseline and at follow up as well. So I dichotomize the the variable into high and low to kind of a come for the skewed nature of the distributions. We might have already touched on this question but could you please share which theoretical framework supports your classification of social support as structural and functional support. And what about quality of relationships. Yeah, that's a very interesting question. First part, the theoretical, the theoretical background to classic lines, social support as functional versus structural. I'm not really aware of a specific theory that speaks to that but going back to 19 minutes. Well initially starting from 1979 I think from Dr. Berkman's, Lisa Berkman's research. That's when association between social support in general and cognition in many different forms started to be researched and in the initial years. So my focus was on the structural aspect of social support. How many people do you know how big is your network size, how often or do you talk to them, and how many hours you spend, and so forth. So, so the structural aspect would have to focus in the early years of the research in this field. But later on, I remember reading an article about calling it a second wave. And then people started to question the quality of the relationship is more important than the quantity, and how you perceive that relationship is more pertinent to when comes to social issues or how it impacts cognition in general. So, I think there was a gradual transition, as people were getting more aware of different components and aspects of social support. They were not aware of a specific period of that. But there was that kind of like a divide when people kind of transitioned and about the quality of the relationship. That's another construct I think that's also been used in measuring social support. So, actually there are a lot more than just functional social support there's also. You could also mean positive or negative, right. So do you also do do you also have like negative aspects of the relationships, do you get critiques or criticism, or do you have tension and conflict a lot in the relationships. So that's usually when, when I read about the quality of the relationships of the center literature, it's usually isn't negative or positive. But in this MLSS as a scale, we didn't update that scale doesn't measure that. So that was one of the limitations as to the scale. Great. Well, we're coming towards the end of the webinar just a reminder. If you have any questions now is your last opportunity so there's a couple of quick questions left. And the next one is going to be simple but did you have missing data at baseline. Yes, I had missing data at baseline and covariates functional social support and and the memory variables. I had missing this in all of those at baseline as well and they were student from the analysis especially because the predictor variable and the covariates I only use the baseline. And, and I just see Caroline responded to your, your response and so I'll just let maybe you read that after. Yeah, so I think that pretty one more question okay last one we'll get in and how did you deal with that missing data at baseline. We talked about how you dealt with that at the follow up. How did you deal with the baseline missing data. So, my research was not like done separately at baseline and follow up. I narrow down my analytical sample to those who had complete values that all the variables that I needed. Right, so people who had missing that missing this and the baseline variables that I said I use they were excluded, as well as those who had missing members scores that follow up. That brings us to the, we're almost at one o'clock so thank you again to all our presenters I hope, I hope we answered all your questions. If not, you can always follow up via email to the CLSA or to Samantha directly. We really appreciate your participation in the webinar series. I'd like to remind everyone that's interested that the next deadline for data access applications is January 12 of 2022. Please visit the CLSA website under data access to review all the available data as well as the additional details about the application process. I'd also like to remind everyone to complete their anonymous survey upon exiting the zoom session. You should get that automatically. In terms of the upcoming CLSA webinar. Join us in the New Year when we'll resume for the webinar series in January. We're still figuring out our lineup so you can monitor the CLSA website for that information. And finally remember that the CLSA promotes this webinar series using the hashtag CLSA webinar. So we invite you to follow us on Twitter at at CLSA underscore ELCB. So, for attending our webinars today but also the past year and we really hope you have a safe and relaxing holiday season over the next several weeks. Thank you so much everyone, and thank you.