Virtual Festival of Evidence | Prof. Carol Jagger
Whole systems thinking for complexity and fragility: challenges and guidance
Carol Jagger: Because we’re living longer, and we have a larger number of older people, we need to be able to estimate how many people are going to need help with daily activities and who will need care.
Interviewer: What is the strongest piece of evidence you’ve worked on that shows the huge impact modelling can have on healthcare?
Carol Jagger: In terms of the impact of extra life expectancy, modelling how diseases impact on disability. I think that bit of the process had been completely forgotten in the past. We know that life expectancy is increasing by two years every decade, but it’s assumed that those years will be free of disability. But if you start looking at how diseases are changing, that doesn’t seem to be the case. That was a really important piece of information for the government: we’re not ready for ageing because we’re going to have more people with disabilities.
Interviewer: In terms of dementia, you’re now involved in quite an exciting new project. Can you tell us a bit more about that?
Carol Jagger: The MODEM project, which is run by the London School of Economics, will model the costs and consequences of intervention for dementia. I’ll be looking in particular at the impact of different diseases on disability, but the new model will allow me to incorporate specifically the effects of dementia and look at how other conditions alongside dementia will produce the numbers required for certain interventions.
Interviewer: What are the implications of this for the way we make provision for patients with dementia here in the UK at the moment?
Carol Jagger: At the moment we, I think, don’t have very much evidence of what works well. We’ll be doing reviews of this and putting forward best practice on how this will contain or reduce costs. We’ll be incorporating socio-demographic variables, like retirement. I wanted the model to be not just focused on dementia, but to be able to reflect some of these other changes that are taking place. My previous model informed government reviews on ageing, and I want this one to as well.
Carol Jagger: This was the title that Terry and Peter gave me and I will talk a little bit about frailty. But first of all, I am rather different, I think, from many of the rest of you. And I think I probably fit better into the session this afternoon, which is more public health. So, I’m trying to stop the people coming to your front door, basically. My background I’m a Statistician, Mathematician by First Degree, Medical Statistician. But I’ve worked on aging since the early 1980s and so now I have very much a focus on disability and dementia increasingly. [00:01:00] I usually say, I’m an aging researcher personally and professionally.
I’m going to talk a little bit about frailty. I’m interested in frailty because I’m more interested in disability because that’s what drives need for long-term care. And frailty, we know a lot more about disability than we know about frailty and I’m interested in where frailty fits with disability within the disablement process, and I’ll talk a little bit about that. But then, I’ll talk mainly about a macro simulation model that I produced some years ago and that has been feeding some of the government reviews, and then a newer version of it, which will be a micro simulation model which I have just started.
I’m quite interested in other pieces of work and the software you’ve used, and how you put in stuff together [00:02:00] from that point of view because it’s very early days in the architecture of that. But I’ll but describe it. And it’s very much part of a bigger project modelling the cost and consequences of dementia, and I’ll say a bit more about that. And all of this is within the context of increasing life expectancy. So, my other big area is healthy life expectancy and I want my simulation models to be able to tell me something about how healthy life expectancy, disability-free life expectancy changes with respect to life expectancy.
So frailty, there are, basically, two frailty models knocking around the main ones. There’s the Rockwood Deficits model, which are quite like the idea of because [00:03:00] I do think of frailty as being a cumulative collection of health problems, diseases, whatever. I’m just a little uncertain about the way it’s put together. And, from my point of view, disability is included in that because you include the function test, the Activities of Daily Living, which I’ll say more about, in those deficits, counted deficits. And therefore, you can’t say how it fits with disability.
But I quite like the, sort of, idea of it, then, there is there’s the Phenotypic Model the, much more, the clinical syndrome, which takes usually five signs and symptoms. And, we’ve very much developed with the views that frailty was a precursor of [00:04:00] disability. And so, disability is, in a sense separate to that. And that’s being, well they’re trying to develop that within general practice as well as the Deficits Model as well.
But, in terms of disability, what do I mean by disability? If you are a social scientist, you will take me up on this model because this is very much a biomedical model of the disablement process. But, I would argue that it does include the other factors like the environment, medical care, risk factors that can slow you or make you move more rapidly through this process. [00:05:00]
Disability is really around here the activity restriction, restriction in basic Activities of Daily Living (ADLs) and instrumental activities of daily living more household type tasks, which are the building block of activity restriction. And the freed frailty would seem to be around here, predicting activity restriction later. I was shocked when I looked again at these. [00:06:00]
The Activities of Daily Living scale, these basic self-care activities were developed 50 ago by Katz. Then, Paul Laughton and Elaine Brody developed these instrumental activities of daily living which tapped lesser levels of disability. And it’s these, activities that people really want, older people want to be able maintain. From a quality of life point of view, they are very important for old people. But they are very important because without them, you can’t live independently. What Katz’s noted, certainly for these ADLs down here, the basic personal care activities, was that they followed a particular ordering [00:07:00] but, we don’t really use that ordering.
Quite a lot of studies have shown that there is a definite order of loss to activities. This slide is taken from our new cut of eighty-five plus study which is the cohort of eighty-five-year-olds. The hierarchy fits the other studies that have also looked at it. But we were able, because we had a much greater range of activities we could put these into a hierarchy much better. And we could also look at men and women too, which haven’t been done before.
The only difference is that women seem to lose, have difficulty with heavy housework before men do. But other than that, they seem to lose them, losing meaning have difficulty, with in the same [00:08:00] order. And, the first item that you have difficulty with is cutting your toe nails and, the last one is feeding yourself. So, the ones down here are, very much, the basic personal care ones. So when you get down here, you’re in problems. Now, that was done quite recently by my PhD student.
But back in 2001, I looked at this in a much reduced set of just ADLs in a study in Melton Mowbray, because I was based in Leicestershire there. And this, we looked longitudinally at how people age seventy-five and over had difficulty with activities. And we did this over a period of about eight years. And, it was set in a very large general practice in Melton Mowbray, which at the time, was the largest [00:09:00] general practice in the UK. When the government told GPs that they had to start doing an annual health check with their seventy-five-year-olds, most GPs didn’t do this very routinely. But this general practice did, and it asked us to help them develop the assessment and we put in these Activities of Daily Living.
What it shows is that the first ADL that they had difficulty with was bathing, and the second one with mobility around the home. And if you look the median age of onset of difficulty, there was about a three-year gap between them having difficulty bathing and then having difficulty with getting around the home. And then, in fact, there was [00:10:00] a six-year difference between having difficulty on average having difficulty getting to and from the toilet. But then you’ll notice that after that, it all happens pretty quickly.
That it’s about half a year for the next one, having trouble dressing, and less for the next one. So it all happens rather quickly from that point on, from having difficulty getting to and from the toilet, after that, things start to drop off fairly quickly. And I would say, maybe this is where frailty is. And so that’s just where my thinking is at the moment, I’m going to have a look at that in a bit more detail in our 85 plus study.
The British Geriatrics Society has just developed these guidelines fit for frailty [00:11:00] and are suggesting doing a very short screen, a prism of seven questions. And I’m wondering whether in fact if you, so there’s a very general do you have health problems which limit your activities, which is, it seems a bit, sort of, non-descript. Maybe introducing, you know, do you have trouble with a particular activity, might actually be able to screen people better, so I just put that up as some sorts.
But people who are interested in frailty might want to have a look at that document. So I told you about the disablement process and the models, simulation models, that are being developing very much focus on disease as a precursor [00:12:00] to disability. And in the past, most of the projections of future need for health and social care have tended to assume that the prevalence of disabilities is going to remain constant within age groups. And, in fact, I work with economists at the London School of Economics, the PSSRU unit there.
The output from my models have fed into their models, this is what they were assuming. But my macro simulation models, when it fed into theirs, actually showed that this wasn’t the, basic starting point with the easy option, and because, actually, to be able to have a constant prevalence of disability means that you have to be reducing the prevalence of disease [00:13:00] or the disabling effects of those diseases, or the durations of diseases. And so, you have to still be doing something to do that. But I’ll come on to that later.
We don’t really know whether disability rates are improving in the UK. And, in fact, we had two pieces of evidence which were rather opposite from the cognitive function and aging study. The first study and I’ll say more about Cephas later, the first study, which began in 1991/92 also had a repeat, it was a longitudinal study. So it couldn’t really look at changes in cohorts. But, for the Cambridge cohort, they introduced a new sixty-five to sixty-nine cohort in 1997. [00:14:00] We could look at the change in disability for the newer cohorts just in Cambridge. And that showed that the prevalence of disability had gone up over time.
Whereas another study that I was involved with in Gloucester, which was an older age group, seem to suggest that the prevalence of disability had gone down. I have now done the analysis for the two cognitive functions on aging study. So that’s all the centres with a twenty year age gap. And we published, last year that the prevalence of dementia had gone down. And, Peter might say more about that this afternoon because that’s part of the public health modelling.
But in fact, although the [00:15:00] prevalence of moderate and severe disability seems to have remained the same, the prevalence of mild disability has gone up. So this is a good reflection of what’s gone on. tha It does look like the age specific prevalence of disability has not remained the same over time.
Diseases is at the start of most conceptual models of disablement process. And in fact, what sort of diseases are we talking about? Well, from Cephas one, we looked at the disabling effect of different diseases, the population attributable risk of disability. And you can see that the main [00:16:00] cause of disability in old age is arthritis. But there’s some more that you maybe you wouldn’t think of, eyesight problems, but also dementia is another disabling disease as well as being fatal.
The interest in including disease into the simulation is really because there has been substantial reduction in mortality from different diseases. We know about the increase in obesity and that impacts on diseases. And now there’s a reduction in the prevalence of dementia. So we know at least, the change in that too. And there are multiple diseases which affect disability and so I [00:17:00] didn’t want to come up really with just a model that had a single disease. And also, the different risk factors such as obesity impact different diseases. And you could say that treatment might impact different diseases as well. The same treatment might impact on different diseases too. So you want to be able to really encompass all of that complexity.
And we also know that multi morbidity, multiple diseases, increase with age. And this again was from Cephas. But I’ve got another slide, in a minute from the eighty-five plus, so you can see quite clearly that the proportion of people with no diseases back in 1991 reduced across age. [00:18:00] The proportion with three or more diseases increased and these were self-reports. These were in Cephas self-report of doctor diagnosed disease. But in our eighty-five plus study, with a cohort of 85 year olds, they were the same single birth cohort born in 1921. And here, we collected information about diseases from the GP records so it wasn’t self-report. Different biases, nothing is ever perfect. But you notice that when we created the disease count, it doesn’t start at zero.
Not one of our eighty-five- year-olds had nothing of the eighteen diseases and conditions that we considered in our disease count, none of nobody, had nothing. [00:19:00] On average, men had four diseases women had five. And around thirty percent had six or more. And these are the ones that are coming through your door, probably. And I would say that single disease based health care delivery doesn’t fit terribly well with this. And you’ve been talking mostly about in-patients, but if you think about out-patients, some of these older people could spend a whole week going to different out-patient clinics. And we have to think of a better way of doing that. Because it really, I don’t think it’s sustainable the way it is.
Now the macro simulation model, it was developed as part of a project called MAC2030 modelling aging populations to 2030. I said before [00:20:00] it improved on single disease models because of this fact that there’s a lot of multi morbidity in old age, and that risk factors in treatments might affect more than one disease. When you are developing scenarios you want to be able to reflect that in your modelling. And it was based on the two year transitions to disability and death within the cognitive the first cognitive function in aging study so now it’s a bit old because it’s based on 1991 or two baseline data and so that’s why the new models will be developed.
What it did was it produced projections of the number of elder people with disability and the disability was defined by the PSSRU models so it was [00:21:00] level to require social care, it involved activities basic activities of daily living. But it also outputted, projected the number of old people overtime with this different conditions and I included seven main conditions, no I included more than seven but I only did scenarios based on these five, arthritis, coronary heart disease, stroke, diabetes and dementia.
Because of my other interests in healthy life expectancy it also outputted disability free life expectancy and years with disability because that’s important that we talk about a compression of disability and I [00:22:00] will explain what that means in a minute but in fact there is no evidence to suggest yet we have got it, this Lancet paper will in fact, tell you that we haven’t got it.
The projections originally the very first model that I built just had men and women together but the later model for map 2030 separated men and women out. Of course because life expectancy for women is higher than men and disability is higher in women than men as well so there are considerable differences and there are differences in some of these diseases as well. Though in the scenario she probably wouldn’t make any gender differences and the projections have been used already quite a lot in the House of Lords report ready for aging [00:23:00] I was asked to produce some new estimates for them and they were used originally in the Dilnot Commission that was one of the things that they were funded for.
Just to explain what I mean by compression of disability and expansion of disability, disability free life expectancy often presented like this life expectancy expected number of remaining years lived at a particular age is the full length of the bar so here you would life expectancy 16 years this might be typically be well it’s just a bit underestimate because I have just made up numbers to make them easy but they might be at age 70 or something like that.
And 12 years that re spent free of disability and four years are spent [00:24:00] with disability and if you look 20 years on life expectancy has gone up to 20 years so it has gone up four years well it’s going up two years every decade so that’s why I made those that’s fit with what is happening. If we are compressing disability what would happen is that the years free of disability would increase faster than the years with disability so you would be compressing disability in a shorter space towards the end of life.
In an expansion of disability then the extra years are more likely to be years with disability then without disability and in fact we are showing that to be the case certainly with any disability. [00:25:00] The cognitive function in aging study was originally in six centres but we never used Liverpool because it was a slightly design and it started a bit earlier so we tend to use the five centres: Newcastle, Gwyneth, Nottingham, Ely which is Cambridgeshire and Oxford.
It’s a stratified random sample of people aged 65 and over the general practice formed the sampling frame and the equal number 65 to 74 and 75 and over so we plenty that were very old even in the days that they were not so many of them. Very importantly for looking at total population it included those in institutions there were 13,000 baseline in 1991/2 [00:26:00] and with the two year follow up and we included 11 diseases in the model along with the sociodemographic variables, education being one of them.
Though we didn’t in any of the scenarios change education which we could have done but we didn’t and I now realise maybe we should have done because certainly the preference of dementia has gone down and part of that has been the increase in education in the newer cohorts and death information they were flagged at the national death registry this is an old slide it not called that anymore.
So the main elements of this model SimPop we used CFast to estimate the two year transitions to disability and death, conditional on this range of diseases [00:27:00] and then had a projection stage that applied those transition rates to age the population and then we had to adjust to calibrate because it used the CFast mortality rates, we had to calibrate it to the Government Actuaries population projections and that was not very easy to do and so it’s still based on the 2006 population projections because that was a major operation to readjust it to the later ones.
We also found within CFast when we did the projections that the prevalence of diabetes was underestimated compared to the health survey for England so we recalibrated that as well to the health survey of England but we checked some of the other diseases as well for the health survey of England they were fine but diabetes [00:28:00] wasn’t. So in terms of operationalising the disease scenarios there were basically three parameters that you could play with so you could play with the prevalence of disease to reflect changes in the cohorts or the risk factors.
You could play with the disabling effect of the disease which was really reflecting and any changes in treatment or in severity of disease for instance, the treatments for hypotension for vascular disease and then mortality from the disease again to reflect possible changes in treatments or severity and we used systematic reviews, we did systematic reviews for those five conditions [00:29:00] to look at what evidence there was for changes in the treatments or prevalence or incidence or disabling effects of those diseases.
We got pretty good evidence for treatments that affect mortality because for clinical trials mortality is a common endpoint but disability is rarely an endpoint, it was hard to see what effects treatments might have on disability but effects on mortality were relatively easy and even with stroke trials it was a bit infuriating because they don’t now but they used to have combined outcomes of disability and death so you couldn’t pull out the effect on disability and the effect on mortality which is a bit frustrating. [00:30:00] We looked at a number of different scenarios but they were three main scenarios that we developed at part of map 2030 for all our models.
The first one was just population aging alone, so that was that the diseases the prevalence of diseases was going to remain as they are and all we are going to see is the effect of the larger numbers of older people coming through and that mortality rates are going to continue to fall according to the projections the government projections. Then the second scenario was that the current trends in health continue and this was the negative one as opposed to the population aging alone. The preference of the key diseases would increase because predominantly [00:31:00] of obesity, the onset of disability would be increased and mortality would be reduced so that seemed to be what was going on.
At the time and then improving population health which really was very much, people and the health service took a very positive approach to improving health and the treatments were available for everybody to whom they would be applicable which maybe was not the case in this one. The one thing we couldn’t do which is important in cohorts that are coming through, is model ethnicity and the effect of [00:32:00] increasing the ethnic population in terms of the older population because CFast didn’t have
It was in centres where they were certainly 91, very few ethnic minorities but of course ethnic minorities have particularly high levels of coronary heart disease and diabetes and obesity so that might increase but we have no data at all on disability in ethnic minorities and very little on anything else really. To give you some results so this is just the population aging alone, what we call the central health scenario and what it showed is that we were modelling [00:33:00] not surprisingly with a name like map 2030, we were modelling up to 2030.
When we started our projections from 2010 the total population with disability will increase by 83% most of this driven by the increasing numbers of those who are 85 and over who are the fastest growing section of the population and will remain so over that time scale so the numbers with disability in that age group will more than double. In terms of numbers with key diseases again the big one is arthritis and this maybe fuelling the increase in mild disability that we found between CFast one and CFast two, that I have to explore [00:34:00] in more detail but obesity would also do that too.
Numbers with arthritis would increase by 50%, with stroke would increase by just under 50% and numbers with dementia would increase by 61%. I haven’t actually looked at that against the numbers with the prevalence of dementia have gone down and I must do that. In terms of life expectancy and disability free life expectancy it really wasn’t good news, life expectancy would increase by 2.8 years this is women’s similar figures for men.
Disability free life expectancy would increase it would increase by 1.6 years but it wouldn’t be all those 2.8 [00:35:00] years and so years with disability would also increase so you would get an expansion of disability and I played around with how can you get a compression? To get a compression you have to reduce the disabling effect of all of those key diseases by 50% to get your reduction in years with disability so that’s minus which is a heck of a lot I would say. So we are not going to be there too quickly, you could potentially almost get there by reducing the prevalence of those diseases but you would only really get there at age 85 and part of [00:36:00] that is because at older ages there is more disability so you have got further to move it if you see what I mean.
If you do the both of them then you get…in fact it is called a relative compression because you are actually increasing the proportion of life free of disability as opposed to reducing it but you are not at an absolute compression. So there are a lot of limitations of this model but it was like a starting point really to start thinking about how diseases and disability might link together and previously [00:37:00] people had just assumed, well disability will reduce by about 10% or well it won’t increase, it will remain the same or it will reduce because we are compressing disability and this is what some of the government models were using without any evidence at all.
That we were in fact compressing disability it was just an assumption now we are living longer and healthier now so we must be but in fact the evidence doesn’t seem to be there at all. It’s based on some quite old data now but whatever happens you need the institution population if you are really going to look at disability free life expectancy because you need that for the total population because we have changed the definition of who goes into the institutional care a lot over [00:38:00] time and are still doing that, then if you ignore that population you get biased results because you are comparing a population that had more people in institutional care that you are not looking at, with a population that has got less but that you are not looking at. So there are more disabled people out there in the community because they are not going into institutional care.
What you couldn’t look at in this institutional model is the changes in multi-morbidity, you couldn’t see how the number of diseases that people had changed over time and that was on thing I was interested in looking at. Which is why am going to do a different model now and it was complex as I said to update with the new GAD projections. As a statistician what bothered me [00:39:00] was that there was no measure of uncertainty within the estimates and deterministic models don’t trouble some people but they trouble me.
But the strengths were that because this was a very large cohort we could look at low prevalence diseases so we could look at diabetes, it wasn’t just diseases like arthritis that has a high prevalence we could include multiple diseases and we could look simulate but rather artificially I guess that the respect of joint risk factors for instance obesity. This was the first projections of disability free life of expectancy that had been done. As part of a new project I’m going to be developing a dynamic micro simulation model, this project [00:40:00] is called Modem and it’s got a very snazzy website that you might want to look at.
Modem stands for, modelling the outcome and cost impacts of interventions for dementia and it’s led by the London School of Economics, Martin Knapp and my model will be a particular work package and will feed into the interventions and costs work package amongst others. Other work packages will feed into mine particularly the ones doing the systematic reviews will feed evidence into mine. So still at the early stages of this, I’ve built the architecture of it in terms in my head and on paper, but nothing else yet. It’s very much modelled on an Australian model called [00:41:00] Dynoptasim, which came out of NATSIM, the National Simulation Unit in Canberra, at Australia National University in Canberra.
So, it’s modelled very much on that, what it will do is very similar aims to the macro simulation model that I had. But the outputs will be, somewhat different because we’ll be using a different outcome measure slightly to which will have a better feel for need for care. But will also be able to pick out multiple diseases, and how they’re tied in with disabilities, specifically [00:42:00], for the care packages.
This was very much within the framework of a dementia project. But I didn’t just want to produce a simulation model that looked at dementia. I wanted it to be able to be more general. And so, that’s why it’s modelled on the older one. We’ll be able to look at disease burden in terms of multi morbidity. As I said, that will be very important because when they’re looking at the care packages for people with dementia, the care packages will need to know what other conditions they’ve got, and to cost those too. And most importantly for me, this model will allow me to get measures of uncertainty around my estimates too.
It’ll be based on the new Cephas cohort. But I’m going to pull in, as far as [00:43:00] I can those aged fifty to sixty-four from the English longitudinal study of aging. And that’ll need quite a bit of data manipulation, because I’ll need to get variables to be the same across the two studies and they’re not all. The scenarios for risk factor reduction and treatment will be informed by the reviews being undertaken in the other work packages. Other the projects funded under the same call, this is a big NIHR ESRC call. And there were a number of projects funded. I’ll be able to pull in some other evidence from those two, then we’ll look at known trends in some of the disease risk factors. Smoking and obesity will be part of the model explicitly this time and not just [00:44:00] implicitly and changes in sociodemographic variables as well. As I said, we’ve decided that disability is going to be measured by something called, interval of need, which is again, IADL and ADL based, but put together rather differently.
And just to explain what these are, it puts people into four categories. And in fact, this was a piece of work that I’ve cited a piece of paper where I’ve used it in our eighty-five plus study. But it was done originally by a guy called Bernard Isaacs who was a geriatrician back in the mid-1970s’ and haven’t really been taken up that much. A few of us have used it, but it seems to be quite a sensible way of putting together ADLs and IADLs in a way that, encapsulates need for care a lot better than just saying [00:45:00] somebody’s got a problem with two or more ADL, which doesn’t tell you terribly much.
So people are put into one of four categories, and their critical interval need require twenty-four hour care, if they are very cognitively impaired, based on mini mental state score. They can’t get to and from the toilet on their own they’re incontinent and need help dressing. So you can pull in other variables as well. They need help getting to and from the toilet. So those are the sort of basic things; something where you can’t predict when somebody’s going to need help.
Short interval needs are, needing help at regular times every day. So they need daily help, but they it’s to get people up, to [00:46:00] put people to bed, to give them a meal. You don’t need somebody all the time. Long interval need, needs help less than daily, typically with the instrumental household activities needs help shopping, laundering, whatever. And then the rest are independent. We’ll be using this as the outcome. But I’m not quite decided how I’ll operationalise that and whether I’ll do it by the single ADL variables and then put them together or what.
So the methods, the baseline characteristics of the individuals will be of three types, social demographic, I’m now deciding to put in retirement status in there as well because there’s a lot of interest in [00:47:00] whether delaying retirement is actually good for your health. Lifestyle behaviours, smoking, physical activity, want to be able to get obesity, social engagements important for dementia, not least. And then diseases, I’m going to try and keep the same that I had before. I’m pulling in respiratory diseases as well, I’ll probably have a “any other” category as well, which I didn’t have last time.
It’s going to be a discrete time simulation with monthly transition probabilities that what Dynoptasim had which will better account it will allow a more close approximation to continue with time. And we’ll be able to validate quite a lot of this [00:48:00] against a health survey for England although remembering that the health survey for England doesn’t include people in institutions. And at the moment, I’m going to programme this in Sass, it’s a statistical programme but there’s a more programmingly type element of it, which I used for the macro simulation model. That’s what Dynoptasim is in and open to other ideas. Ok, anything that, sort of, minimises the work and means you can get stuff off quicker would be good.
So I’m interested in any of the views on that but I’m not, by any means, an expert in simulation modelling. I’m not aware of any of the other things that you use. And so, just to say thank you to all the other people that have been a part of this particularly our funders, the ESRC [00:49:00] and the NDA which was the funder on the MAC2030 project. And MRC who funds Cephas, Axa who fund me. The Centre of Excellence in Population Aging Research which introduced me to Dynoptasim in some ways too, because I’m an international collaborator on their project.
Ok. Thank you.
Interviewer: We got ten minutes for questions.
Q&A: Can I just ask about ethnicity?
Carol Jagger: Yes.
Q&A: Are you going to have a chance to include that this time around?
Carol Jagger: Well, possibly not, because Elsa doesn’t have much in a way of ethnic minorities either. Cephas two might have more but the [00:50:00] three centres that carried over, well ones Nottingham, Nottingham might have a bit but they’re not big centres of ethnic minorities really.
Q&A: If you’re going to have distributions you probably could apply some small pilot study to say how much things like these…
Carol Jagger: And that’s in a sense what we did in the other one as well.
Q&A: I was really fascinated looking at the progression slides, I certainly will start to include toenail assessments is my first order of business. Did you look, because obviously I see a lot of people who are much younger than that group [Unintelligible 00:50:51] and did you look in your code at all at either the impact of the probabilities on the age of onset [00:51:00] versus people who are pretty well [Unintelligible 00:50:51] although I take your point that there’s nobody really exists above the age of 85 and the second question was really if you do have comorbidity, can it change the order of which the frailty process occurs, how robust is that as a model?
Carol Jagger: The hierarchy is very robust age because that’s one of the things that I did in Melton, was I looked at different age groups and the hierarchies that is the same across the age group. Comorbidity wise, we don’t know but I suspect what it will do is not change the order but speed. You through and that’s what I want to look at [00:52:00]
Q&A: The way it’s modelled [Unintelligible 00:52:05] one of the big problems that I have with conception of these types of models, every time you added a new factor you exponentially increase the number of potential states that an individual can be in and curse the dimensionality. How would you deal with that, given that you’re going to have seven, is it possible to see the stages?
Carol Jagger: Well dynoptasim seems to have done that, I mean they’ve got a model that is up and running, I’m doing this as an add-on to theirs. Not everybody will have those seven diseases and yes, I’ll have to be careful. It might be that I have to pull out some of those diseases and just put them into another category. I’m keen to still be able to [00:53:00] capture them but maybe not by name, if you see what I mean.
Q&A: [Unintelligible 00:53:07] this is quite a medical model of disability as we said at the beginning and I was struck by your slide that you had to change disability by 50% in order to impress disability if I’ve got that right. But in a sense, if we intervene with aids and adaptations, if you change the world around the patient, then disability has a different meaning. Because we are intervening we are making life better for people with disabilities. So, to what extent, [00:54:00] is it still the same?
Carol Jagger: I think this tries to address some of the public health side of the lifestyle factors if you like and the reduction in smoking, alcohol, obesity that everybody is going on about. Yes, I completely agree that the introduction of aids earlier could help that. I almost think that’s something separate in terms of the hierarchy, and in fact that hierarchy, there’s a professor of practice within Newcastle who’s got very excited about that because of the new social care act, he’s talking about trying to [00:55:00] re-enable people a bit earlier and therefore slow down the earlier part of it so that if necessary people go down faster at the end but over a shorter period of time.
I may be able to capture that, if I did look at the separate AGLs, IAGLs, but then you get into too many variables, so I’m not sure.
Q&A: what has helped to, lots of things obviously, but if you were thinking about aids and adaptations and treatment, then, does anybody know which is the most cost-effective so that you could… How many years of disability gain would you…By introducing particular intervention, [00:56:00] which are most cost effective?
Carol Jagger: No we don’t, that’s the problem, but I think we could if we started using some of that, that’s what Peter my colleague wants to look at.
Q&A:About confirmation in terms of the actual functional ability, which is mainly affected by the patients the other of course is, who you’ve got in your immediate family, so there is something about family structure and close relationships in terms of functional disability that means that we need social care or care whatever. That’s another…
Carol Jagger: That’s why I chose difficulty rather than, needs help with, because needs help with, is more dependent upon whether you’ve got someone in your household to give you help and therefore you may be disabled at a lesser level but have help, [00:57:00] than somebody who lives on their own school and has to cope on their own, as it were. Difficulties, less biased if you like by living arrangements.
Q&A: A very interesting talk, I was slightly puzzled by counting intuitive differences between genders inter-sociability, females live longer and spend more of their time being disabled. Did you have any idea from your model? What are the factors that actually affect the difference?
Carol Jagger: Every study worldwide shows that women have more disability than men, it’s just standard. No, there was nothing from the model than women tend to have disabling diseases and men have fatal diseases, [00:58:00] men don’t have the staying power basically, dare I say. The other factor, that I haven’t really looked at but which I suspect, is that there’s a threshold of strength to do some of these ADLs and that women reach that threshold before men do. We know hungret strengths less for women than men, and hungret strength affects ability to do ADLs. Maybe women, just because they’re not as strong reach that threshold faster than men do.
Q&A: [00:59:00] One of the presentation things, how robust do you feel that the homework agency have chosen to use, as indication for this model care drives risk for this base, for example I didn’t see cancer in the model. You would assume that a fair amount of disability may be associated with the cancer development over time, can you comment at all about how robust the model is with the core reasons you’ve chosen?
Carol Jagger: Cancer is not a great disabling disease, it’s a fatal disease, maybe that’s changed a bit over time, but in fact I suspect not enough to put it into the top five. The other studies that have looked at effective diseases on disability, cancer has come out very low, it’s fatal rather than disabling.
Interviewer: [01:00:00] I think perhaps we should, given that it’s 11:10 and we’re slightly behind schedule, I’d just like to, well actually two things, firstly, thank you so much, Carol, that was really, really interesting and I think the real lesson for us, to whom simulation means obviously a slightly different than it does to you in terms of, the engagement that you’ve got with policy makers, the use that the models have been put to. I always cite your work as an example of successful implementation, whereas the work that I do, you really struggle to come up with examples and there’s an interesting lesson when we come to reflect on evidence and the use of models by policy makers, for changing things.
But its public health, epidemiology and this simulation modelling that you do, is a real shining light in terms of gaining acceptance with people who don’t understand [01:01:00] what’s under the bonnet but they are quite happy to use your results to inform policy and yet people like me struggle, that was one thing. The other thing was just to say to anybody who’s new, that’s just arrived today wasn’t here earlier on we are collecting people’s views on that first question, that evidence means different things to different people. If you have any personal thoughts on what evidence means to you, there’s a board there that says, Confidence, Trusted, Wanted.
If you could just write what your background is and what evidence means to you, what a model would have to do, to convince you that it was trustworthy and that you could believe the model results. [01:02:00]
It was a really terrific presentation. Thanks a lot.