The Health Foundation is funding an exciting new research programme that uses rigorous methods to draw out how good health impacts on key areas of our lives – and on society and the economy more widely. It will help make the case that good health is an asset that should be invested in.

Dr Laura Howe from the MRC Integrative Epidemiology Unit at the University of Bristol explains how her team are using longitudinal, intergenerational data to make sense of how our health, and that of our parents, impacts on our education, work, and social relationships. 

What is your project exploring?

Good health is an asset, but it’s something we often don’t notice until we don’t have it and we’re ill. 

There’s already a lot of evidence about how our social and economic circumstances affect our opportunities to be healthy. There’s far less evidence around how our health can shape our social and economic outcomes, and impact on society and the economy more broadly.

Our project is looking at a long list of mental and physical health conditions and seeing how they affect education, work, and social relationships. We’re also looking at whether there are times in a person’s life when good health is particularly important for their school, work, or social life. And finally, we are studying which aspects of a parent’s health affect their children’s school, work and social life.  We want to build the evidence base around how people’s health acts as a determinant of a healthy society and prosperous economy. 

What are the challenges?

It’s really complicated because everything we’re interested in is related to a multitude of other factors. For example, if we want to study the effect of depression on education, we know that teenagers who are depressed are more likely to come from a low-income household, have physical health problems, and so on. All of these factors could also be linked to lower educational attainment. If we can measure these factors, and we can measure them perfectly, we can calculate the effect of depression on education accurately. But we know that this is unlikely to be the case, meaning that standard methods give biased answers. Researchers call this ‘confounding’. 

Another problem is that performing badly in school could lead to depression – so even if depression did not cause lower exam results, the link in the opposite direction would mean that it looked like it did. This is called ‘reverse causality’. 

How are you addressing these challenges in your research? 

A lot of traditional research uses randomised controlled trials. In these trials, people are randomly assigned to two groups, and one group is given an intervention. Because the groups are randomly assigned they should be similar in all other ways, so there would be no confounding. This would show that it’s the intervention that is causing the result, and not any other factors.

For the questions we are interested in, this is neither possible nor ethical! We can’t create randomly assigned groups, and give one group a health condition such as depression in order to study the impact this had on education. And we can’t use a group of people who have depression, because we know there’s likely to be confounding factors.

But what we can do is use a method called Mendelian randomisation. This method allows us to use genetic markers across the population as an instrument to help us understand how health conditions affect education. We’re not really interested in the contribution that genetics makes to outcomes such as education – we’re using genetic markers as a tool to help us create two randomly assigned groups.

In our study, we take genetic variants known to be related to the health conditions we are interested in. Genetic variants are fixed at conception and, as in a randomised controlled trial, if we group people based on their genetic risk for depression, the groups should be similar in all other respects. This removes the problem of confounding. Reverse causality is also not a problem, because your exam results cannot change your genes. 

Of course, not everyone with a genetic risk factor will go on to develop depression as there are so many other factors that influence people’s health. But we would expect a higher number of people with a genetic risk factor to go on to develop the condition. So, if people with high and low genetic risk for depression differ in their exam results, we can be fairly confident that this is because depression has a causal effect on education. 

Are there any challenges in using this approach? 

Yes, lots! One challenge is in interpretation. Your genetics generally affect your level of health across your whole life, so the effects we see from Mendelian randomisation reflect this ‘lifetime average health’, and do not necessarily translate to the effect we could expect to see for a specific intervention targeted at a particular age group. There are also lots of assumptions, which we evaluate as best we can. We can also only apply these methods to health conditions where we have good knowledge of the underlying genetics, so some conditions that are likely to have big effects on work (eg back pain) are difficult to study with Mendelian randomisation. 

What are your biggest hopes for your work?

The ultimate goal is to influence policy and drive investment in health and in its determinants by providing a social and economic rationale for doing so. 

I’m particularly excited to discover more about the intergenerational relationships: how parents’ health affects all sorts of outcomes for their children. We have access to data on genetics, health and other factors for whole families at multiple time points across their lives, so it’s got the potential to be really powerful.

How is this type of research helping to gain new insights in other areas of health? 

There’s so much great work going on using these approaches to understand the causes of disease. We are also using Mendelian randomisation to build a deeper understanding of the social determinants of health, including understanding how and why education affects heart disease, Alzheimer’s disease and other aspects of health. 

Mendelian randomisation is also great for myth busting. For example, lots of studies using traditional methods show that drinking small amounts of alcohol is better for our heart than abstaining from alcohol completely. But when we use genetic variants that affect alcohol metabolism, we can see that small amounts of alcohol probably are harmful to health. The higher risk of heart disease (and other health conditions) in non-drinkers is probably driven at least partially by a ‘sick quitter’ effect (ie people stop drinking alcohol as their health declines), and because the people who drink moderately are the healthy, wealthy groups in society. 

It’s such a fascinating time to be involved in this type of research. With bigger datasets we’ve been able to find more and more genetic variants for so many different traits. There’s just so much more to discover. 

Social and economic consequences of health status: causal inference methods and longitudinal, intergenerational data is led by the University of Bristol, in partnership with Public Health Wales, the University of Bath and Cardiff University.