Dondena Seminar Series
When Allan Hill, of both the Harvard School of Public Health and the University of Southampton, spoke at Dondena this spring, he brought a message advocating a new and better approach to studying the demography of health.
We’ve got to develop better measures, he says, and we’ve got to come to terms with the biology. The way forward will not be easy. We will need partnerships because social science has what biology needs, and vice versa. Because of this, it may be that only young researchers exposed to the new biology will be equipped to get us there.
Give me your three-minute elevator speech for people who are going to miss your talk today. What are they missing?
Two main points. First, we’ve made a really poor job of studying the demography of health. If you compare the professionalism and time and years of work that’s gone into the analysis of mortality by cause, age, sex, social class and so on, standardized measures and so forth, the study of morbidity is really at a very early stage. And so when it comes to looking at whether you’ve got this compression of morbidity, in other words, as people live longer, are they living longer sicker lives, or are they going to live longer healthier lives, it’s very difficult to come up with a clear-cut answer to that question. And that’s why there’s an enormous debate in the literature because the measures aren’t quite right.
And the second is, in terms of the way ahead, social scientists are really behind when it comes to thinking about incorporating biomarkers and other physical measurements into their analyses. So that when we think about bloods and all of the things that are being collected by biological scientists, social scientists have no idea about the complexity of using these data, not only because biological processes themselves are really complicated. Just because you have a genetic marker for BRAC1 or BRAC2, giving you high-risk for breast cancer, doesn’t mean to say you’re necessarily going to develop breast cancer, depending on your exposures and other things. But also because of the size of the data. In some of the cohorts that are in European biomarker data, we’re talking terabytes of data, it’s not like 5 extra variables on the end of a file, and there’s really no formal education of social scientists yet in how to do this kind of thing.
This is interesting because Mark Hayward was here last week and he was telling a similar story.
He was a student of mine, actually, back at the London School when I was there.
Let’s go back to the measures. What are the measures, but what should they be?
Well, I think a lot of it revolves around self-reported health, and there's a sort of skepticism amongst the medical fraternity about whether people can report their health in any accurate way. In the UK, the best predictor of attendance at a GP surgery is the question “Is your health better, worse, or about the same (as it was last year)” or just the open-ended question “How would you rate your health? Excellent, good, all right, poor and so on.” But of course when you look at the reasons for those visits to GP’s offices, I think a large fraction are for non-medical reasons. People are going for bereavement, because of depression associated with unemployment, seeking an honest broker when they’ve gotten into some situation with the government where welfare benefits are being cut, something of this kind.
Really? They go to their GP for that?
Because it’s free, and because it’s open to all, and GPs will hand people on to other professionals. And the more we learn in cohort studies, when you ask people about their health and their future health, their assessments of their own health condition are very good predictors of what’s going to happen in the health domain five years into the future.
The problem with a lot of these health questions is that they’re very non-specific. They’re not going to tell you about breast cancer or something very detailed, but they are going to tell you about overall health, and so the question is, are people reporting on health, or is it something called “well-being”, broadly defined? And so what I think we need to think about is, is health really a subset of well-being? Well-being would be income, your conditions of living, do you own your own house, and so forth, as well as your health. Or, is one causing the other in a very complicated way?
There are more questions, more questionnaires, about health in the world than almost people, because the mistake that we make is that we always start with a new instrument. Everybody who starts a study invents their own questions, their own questionnaires. And we’re losing the opportunity to have international comparability. There have been some attempts such as the World Mental Health Survey, with trying to standardize questions, trying to think really clearly about depression in its various forms, from serious to mild, and how you quantify that. But when you think about some particular conditions like bipolar depression and suicidal tendencies, would require much more specific questions to get at what’s going on. But there’s increasing agreement, particularly in the mental health area where there aren’t biometric tests for illness, for mental disorders. You rely on reports.
We know something about stress markers on the DNA, Dr. Ros Wright at Children’s in Boston and others done some work on children in Kuwait who have been under the Iraqi occupation. A very nice natural experiment with kids who were in Kuwait, and those who took off to London or Geneva during the occupation. And still the health patterns of those children and adults are divergent today, even though the Iraqi occupation was quite mild compared to what’s going on in Syria today, for example. But unfortunately there are markers on the DNA, which suggests that it might be inheritable, some of these effects of stress, and so it’s beginning to be part of clinical practice now, where instead of relying particularly on children, because children express stress in different ways, some of them become more violent, some might become very docile and compliant. And to distinguish those that really need clinical care, it’s really important to have these biomarker tests. It’s a huge field that the social science community really hasn’t ventured into in any systematic way.
Do you think it’s because they don’t see the connection or because it’s unclear how to get the biological training that they need to be able to do the research?
It’s really complicated, and it’s moving so fast. Now the medical education is so different. We wouldn’t recognize the biogenetics taught today. Harvard Medical School, after the first year, they’re not out of the cell. They don’t even get to organs, it’s a very reductionist sort of education. And there’s this whole vocabulary about SNIPS and genome-wide associations (GWAS) and all of these kinds of initials, that unless you invest heavily or else work with a colleague who knows what they’re doing in that field, it’s really hard to catch up. It’s a field that’s expanding at the speed of light. For people of my age, there’s no point in starting now.
And so what do we do?
Well, I think form partnerships. I think what universities like Bocconi and Southampton and others should be doing is arranging training courses for social scientists, to put on the table what we know about the value of biomarkers. We know a lot, you know, the long-term relationships between nutrition, height, and prosperity. Tall people earn more. So what are the mechanisms there? In the past, height would be a marker for strength. Taller and stronger people produce more and work longer hours. But now it’s motor skills and intellectual development. And we’re seeing in the developing world quite a lot of children growing up who, because of their nutritional challenges earlier on, are never going to be as productive as people being brought up in a better environment.
So there’s these kinds of issues that social scientists know something about, because we’ve been looking at this, but without actually getting into the biology of it. On the other side of the biological sciences, they don’t use big samples. If you say to them I’ve got 3,200 women under surveillance in Accra they say I only want 10, with their medical and social histories, and then we’d be able to look at their bloods and histories and genetic composition and so forth. So there’s a mismatch in terms of the levels of analysis, but also in terms of the mechanisms.
What is important from the biological sciences point of view is that, with all of these gene-environment relationships we’re beginning to understand, there’s a probabilistic feature to these relationships. It’s not entirely deterministic, if this then that. It’s not a clear-cut 1-to-1, so we could find people with markers, we could find people with all sorts of traits, physical traits, but it doesn’t necessarily translate into the condition.
And so there really is a point of intersection, and I think there’s been talk at NIH in the U.S. of training courses for people from both sides, but until people start collaborating on particular problems, and maybe around certain conditions, and actually getting to learn limits of each other’s resources, it’s going to be hard to get them on the same page. But I think the link between some of the biological characteristics of, for example, height and of course obesity is a very important trend, and productivity and earnings, then the intergenerational transmissibility of those things, those are the center of what demography ought to be doing.
In terms of anything else that ought to be going on in the social sciences. What else would you love to be seeing a lot more of?
If you see what’s going on in something like the economics profession, there’s been a tremendous amount of methodological innovation to try and get at a causal effect in more a clean-cut way than with instrumental variables analysis, trying to look at cohorts rather than just simply cross-sectional studies, but I think that the current interest in trials is something that can be replicated in some of the other social sciences.
We do essentially what we call in public health descriptive epidemiology very well. We’re very good at looking at associations. But I think it’s going to take a real trial to demonstrate some of these effects. And many of them are stronger than we would have anticipated from the association studies.
So there have been some new trials – such as conditional cash transfer experiments—giving women money to keep their children in school. And so you just give them the money. You don’t actually say it’s for school fees, or you don’t pay for school fees directly, but you say that this is one of the things we think is worth while. And when you look downstream at what happens as a result of those transfers, the investment more than pays for itself. There are huge returns to education, particularly for girls. And so when you experiment, as Michael Kramer has done in Kenya, with these types of interventions, or even things like for children who are lagging in big classrooms, randomizing them to extra care, extra reading practice, very often just by older children who can read, so you bring up the tail of the class and again the impact of those kinds of interventions are really, really strong.
So you begin to get a sense that the policy actually has some basis in fact, because when you go with associational findings to ministers and people who are ready to legislate, they will ask, where in the world has this worked? What’s the guarantee that it will work here? And second, what’s the local evidence that this is worthwhile doing? And I think that with trials data, and we’re not just talking about clinical trials, but with real experiments, and there are many natural experiments that could be exploited, it doesn’t necessarily need to be that you set up a trial from scratch, but the idea of putting into practice what you propose to do in an experimental way, is something that social scientists have been slow to pick up. We tend to look at after the fact, what was the effect.
So I think that somehow catching up with the policy environment, being ready with the trial data, before politicians make a mess of things, is something that we could be much more interventionist about. But to be able to stay up with the pace of legislation, parliaments really move quite quickly to change the rules and move the goal posts, and if you haven’t done the work more or less real time, you’re not part of the debate.
You leave here tomorrow and what’s the next terrific academic thing on your calendar that you’re excited about?
The most exciting thing would be to get the 3rd wave of the Women’s Health Study in Accra funded, because we’ve been seeing these women since 2003, and we’re starting now with some interventions, and the one we’re starting with is a proposal to address the obesity epidemic in a slightly indirect way. We find that going up to people and saying “You’re fat and you need to do something about it” isn’t really very well received. So the idea is that you approach them with a diabetes prevention program, and you say “Well, we would really like to intervene now so that you don’t develop diabetes later in life.” If you look at an extreme population like the Kuwaitis, the diabetes rates are sky-high. We want to do a pilot to really start on the intervention. But of course the interventions are identical to those dealing with obesity.
Last updated 10 December 2016 - 05:39:11
Dondena - Università Bocconi