Insights from longitudinal research in child & adolescent mental health

Matt Kempen
Marketing Manager for ACAMH

Posted on

Professor Barbara Maughan delivered this Keynote lecture ‘Insights from longitudinal research in child & adolescent mental health’ on Friday 22 October 2021 at the ACAMH Awards. Professor Maughan was the recipient of ACAMH’s President’s Medal at the virtual awards.

The ACAMH Awards aim to recognise high-quality work in evidence-based science, both in publication and practice, in the field of child and adolescent mental health.

ACAMH Members can receive a CPD certificate, simply email and let us know the date and time that you watched the recording.


What I’m planning to do in this short talk is try and reflect on the contribution that one particular approach to research, and it’s one that I’ve been closely involved with myself across my career, the longitudinal study, what the contribution of that has been to our understanding of child and adolescent mental health problems over time. I think we’ve had some very interesting, exciting findings recently in that field. I thought it might be interesting to try and put those into a historical context and see where we’ve come from in the past and where we might be able to go to in the future. I have some slides I’d like to share with you, so I’m now going to try and share my screen and then we’ll be able to begin.

And I wanted to begin with this wonderful [inaudible 00:01:20] quotation from a book review that was published in the British Journal of Psychiatry in 1967, which says, as you see, ‘The follow up is the great exposer of truth, the rock on which many fine theories erect and upon which better ones can be built. It is to the psychiatrist what the post mortem is to the physician.’ And the review was of this book by Lee Robins, published in 1966, which was one of the earliest contributions in this field. There had been some follow up studies prior to this. This was one of the landmark contributions that were coming out, a number of them, at around this time, and what Lee had done was follow a US clinic referred sample who had been seen in clinics in the 1920s and 30s. She found their clinic records and she went out and found the individuals, who were then midlife, in their 40s, and she interviewed them to see how they’d got on.

And what the book does is provide us with a picture of their lives and the implications that those largely conduct disorder type problems had for their subsequent development. And the book established a set of parameters, in a way, for findings in this kind of area, which have been replicated in lots of studies, subsequently. Key things that stand out for me would be, first of all, that there are continuities in the underlying form of the behaviour that appears in childhood, what would now be called homotypic continuities. In this area, antisocial children tend often, not always, to become somewhat antisocial adults. Some associated features, associated impairments, that go with that pattern, that they have difficulties in a range of other aspects of their lives, in education, employment, relationships, and so forth. Possibly not so obvious immediately at that point, their physical health was also not very good, and they had increased risk of a range of other mental health related issues, particularly alcohol and substance use. And then the one that I picked out at the bottom, I think really wasn’t anticipated at the point that this study was done, that they had an increased risk of disorders of a very different kind, of internalising disorders.

So this sets us up with a kind of benchmark for saying, Well, will these findings be replicated? And would they hold up… Would similar pictures hold up in relation to other disorders? And if we look to the literature over the following decades, I think the answer by and large to both of those questions is yes. There were replications and extensions to these original findings in different kinds of samples, particularly importantly in representative samples rather than clinical ones, in different eras, in different locations and over different follow up intervals. I was lucky enough to be able to participate in a number of studies like this, linking with, for instance, the 1946 British Birth Cohort representative sample in the UK. We found the pattern of impairments for young people where teachers had identified them as showing difficulties in adolescence, pattern of impairments very similar. We found looking further down the line that not only was their health compromised, they were actually at increased risk of premature mortality. And then in one of the London samples that I was more directly involved with myself, worked with Stephen Scott and with a health economist, Martin Knapp, showed that not only were conduct problems problematic for the individuals, they were problematic for the rest of us too, because they created a very heavy burden on services and high cost.

So a body of work developing there still being developed today, trying to understand what those continuities might be about, and at the same time, other investigators were beginning to do similar related kinds of follow ups in samples with other disorders, ADHD, anxiety, depression, and so forth. And very broadly they have come up with a rather similar pattern of findings. Very typically, there are homotypic continuities in the underlying difficulty for which children were first identified or sought help. There are subsequent problems in social functioning, sometimes of a rather different kind from the ones that antisocial children would show. There’s often associations with physical health, and there’s often also progression to other disorders and to other kinds of disorder. So over time we’re beginning to map in the developmental profile of many of the individual common disorders of childhood in this kind of way, what’s the developmental profile likely to be over time.

Subsequent decades also have seen the development of some rather different kinds of longitudinal studies, so new population based child cohorts that undertook repeated assessments. Many of the early studies just had one dipstick follow up period. Now we were getting studies that planned to do repeated assessments, and they planned not only to focus on one disorder but the full spectrum of the common disorders of childhood, so the Christchurch and Dunedin studies in New Zealand, New Zealand has been a wonderful contributor in this area, Great Smoky Mountains Study in the US are just some of the most obvious examples. There’s others in many other parts of the world. And what those studies allowed was follow-ups across specified developmental periods, so looking for instance, from childhood to adolescence or from adolescence to early adulthood.

And evidence from all of these sources, as well as retrospective studies of adults, asked backwards, when did your difficulties start, all cumulated together to come to a kind of general conclusion, which I think had not been anticipated prior to perhaps the turn of this century and more recent years, that not only are childhood problems important in of themselves and in childhood, they actually appear to be the roots of much psychiatric disorder across the life course. Many adult psychiatric disorders, some people would estimate as much as 75%, actually have their roots in childhood difficulties. So, by contrast, for instance with the situation for chronic physical health problems, which tend to accumulate at the end of the life course, it seems that risk for in the mental health domain is evident much earlier in development.

Okay, so that’s a kind of background scenario, what’s been happening in the last few years, and probably the most important thing in that domain that’s enabled further development in terms of our ideas and our understanding has been that some of those prospective cohorts with repeated assessments have matured into their middle adult years, and this has enabled us to have what you might call a lifespan perspective on psychopathology, to see how things unfold over a long period of time. And the first thing that that has enabled us to see in a range of different cohort studies is that mental health problems are actually very common, and indeed in the population as a whole they’re probably normative. Across a range of these longitudinal cohort studies, we see that cumulative prevalence rates, when you assess people repeatedly over time, are between 60% to 80% or even above that. Many of us are going to experience difficulties like this. Very important message from a public health point of view, clearly.

And the little graphic I have at the bottom of the page there shows data from one of these studies from the Dunedin study in New Zealand, and it shows across eight assessments, three of which have been collapsed together to give us six assessment points here, it shows the proportion of the sample who had met criteria for disorder when they were assessed at none of these occasions, one, two, three, four, five, six, and you can see the most common pattern is actually to meet criteria on one or two occasions. Over 40% of this representative sample of about 1,000 people met criteria just once or twice. It looks as though it may be the case that a number of mental health problems are relatively self-limiting. They occur once or maybe twice. They don’t necessarily signal a long history of difficulties. The individuals who are in the red bars, they do have much more persistent problems. Their difficulty is more likely to start early in development. And then there’s an interesting group, the pale blue bar in the graphic, who don’t meet criteria at all over repeated assessments. The investigators here called those people with enduring mental health, and I’ll come back and share a little bit more about them later on.

These studies with repeated assessments also make it possible to look at longitudinal patterns of disorder across the life course. The Dunedin sample has now followed their participants another time, so we have nine assessments taking them up to age 45 years. And the investigators in this study have grouped their participants into three groups in terms of the initial disorder they presented with or were assessed with, internalising, externalising and thought disorder problems, and they find that if you track those over time, you do find the homotypic continuity that the prior investigators were finding, but you also find major evidence of more complex patterns across the life course. So, for instance, for individuals who initially were identified as showing internalising problems, if you track them later on down the line, 70% of them, huge majority, will over time show disorder not just another internalising disorder, but a disorder from another family, and only 14% show a single disorder, and that’s actually many of those 14% are people who only show disorder at all on a single occasion. Same kind of pattern for initially externalising difficulties.

And then for individuals whose first disorder is a thought disorder, nearly all of them are going to show difficulties in a different domain. Though the investigators themselves, Avshalom Caspi and his colleagues, conclude, in contrast to assumptions of diagnosis, specific research and clinical protocols, we found evidence that virtually no one gets and keeps one pure disorder type. Exceedingly important implications. They identified in this paper a huge range of other features, which are extremely interesting and informative. They said there are three, what they call life course parameters, of mental health problems that tend to converge in the same individuals, younger age at onset, more years of duration of disorder, which is not simply a function of the younger age and more diverse types of comorbid disorder patterns.

And what we seem to be seeing here is something that’s reminiscent of work that’s being done in other areas, some from longitudinal studies, but also from cross sectional and from genetic studies, too, that there may be a broad underlying vulnerability to mental health problems, often referred to as the p, for psychopathology, the p factor by analogy, with g for intelligence. And clearly, if these findings are true and this is one study at the moment, but it seems very likely that it will be echoed by other studies, then it clearly has implications for all sorts of thinking in all sorts of areas in this field, implications for aetiology and for aetiological research. Maybe we shouldn’t be looking for specificity here if the longitudinal patterns of disorder are not tremendously specific, maybe we should be expecting that risk factors won’t be specific either, and that seems to be the case in some research already anyway.

Maybe there are implications for intervention, too. Maybe we should be thinking not only we clearly do have to respond to the initial difficulties that children present with, but maybe we should be trying to work on more trans- diagnostic interventions on the assumption that many children who present in clinics, for example, are likely to show other patterns of difficulties over time. And maybe we should be trying to work on strategies that would enable them to build stronger, enduring patterns of mental health. That, of course, is more challenging because up to now, the longitudinal studies that I’ve been talking about have tended to focus on mental ill health, mental health difficulties, and not strengths.

But there are a few pointers towards what we might be expecting there. The little graphic I showed you before I alerted you to that group in the pale blue, it was just 17% of the sample, but nonetheless a chunk of individuals who, after repeated assessments, had not met criteria for disorder. How did they differ from the people who were vulnerable to disorder on one or two occasions? And the investigators highlighted a few features that they were able to pick out. First of all, that pale blue group had a number of what appeared to be advantageous personality traits. In particular, they weren’t strongly negatively emotional in childhood. They weren’t socially isolated; they had good friendship networks. They didn’t have much of a family history of psychiatric problems, which might hint at the genetic piece in this story, and they had higher levels of self-control in childhood.

Now very Interestingly, just recently, another study has been published, this time using UK data and looking earlier in the life course, using the UK Millennium Cohort study, large sample of young people born around the turn for century, where again, the investigators have repeated information this time from SDQ scores, repeated five times between ages 3 and 14 years, which enables them to identify a pale blue group of individuals in this study, too, so this is children who don’t meet threshold criteria on the SDQ at any of these points. These children, somewhat brighter than their peers, and certainly didn’t have high levels of special educational needs. They also had good emotion regulation skills so very reminiscent of the Dunedin findings. They were cooperative with other children and with adults, and they enjoyed school. So just some very initial straws in the wind as to the kinds of features that we should perhaps be thinking to try and build strategies to bolster if we’re going to be able to help individuals play to improving their mental health over time.

How are we going to do this? My argument would be that longitudinal studies are going to make important contributions in the future, as I think they have done in the past. And it turns out that, although I’ve just highlighted very few studies here, in fact there’s a hugely rich tradition, particularly in the UK as a matter of fact, of longitudinal studies that can be helpful in this way. My colleague Louise Arseneault at the SGDP has developed this wonderful instrument, the Catalogue of Mental Health Measures. I’ve been a bit involved in this project, which is really lovely, which is designed to provide easy access to information about the mental health measures that are available in British cohort and longitudinal studies to maximise the uptake of these wonderful data and to facilitate mental health research. If you haven’t explored the catalogue as yet, I would really encourage you to do so.

And I’m just going to finish by giving you a little hint of the richness of the data that are available. Each of the lines on this graphic relates to one study. So there are 46 studies, longitudinal studies, on the Catalogue at the moment, and the ages, if you can see them in very tiny figures right across the top, starting at age zero. So all of the pink bar studies at the top of the graphic are birth cohort studies. This is an area where I think we probably excel in the rest of the world in having information from the very earliest stages in children’s development. And as you can see, some of the longitudinal lines here carry on for a long time. We have birth cohort studies that go from very early development up to the age 70s. We have other pink studies, and these are narrow band age groups that have been followed over time, beginning in adolescence and also some that begin very much in older age. And then the green bars are related to studies where the age at recruitment is broader in the first place. It might be a wide range of adolescents or it might be from adolescence into middle adulthood, whatever, but also where samples attract over time.

There are wonderful resources available here to help us explore and take forward some of those questions that I have outlined that prior studies have highlighted for us. I think the next generation of longitudinal research is going to be extremely exciting, and I’m looking forward to seeing the findings that come up from that. Thank you so much.

Add a comment

Your email address will not be published. Required fields are marked *