A quick note: this blog post was a difficult balancing act to write. I want to make clear from the outset that a) I think mental health problems are very important, b) I think we should pay attention to them, but also c) that we need to keep a broad perspective. And Esther Yao and colleagues certainly did the latter. They examined whether the saying “if it doesn’t help, it doesn’t hurt” also applies to programmes focusing on mental health at school. What did they find? Perhaps less so than we would hope.
In recent years, there has rightly been much more attention for the mental health of young people. Some possible answers include programmes offered at school, a trend we see worldwide. Sometimes this means mindfulness, but there are also examples of cognitive behavioural therapy, social and emotional learning, positive psychology, or simply more knowledge about mental health. And often these programmes are universal. This means that they are offered to all students in a class or school, regardless of whether those students experience mental health problems. That last point turns out to be important.
Because… what if such a programme helps some students, does little for many students, but could actually make things worse for a small group?
58 studies
In their scoping review in the *Journal of Adolescent Health*, Yao and colleagues looked at exactly this question. The researchers collected 58 studies, published between 2019 and 2024 (so including the COVID period, I immediately thought), on such universal mental health programmes for adolescents. Together, these included more than 36,000 young people in the intervention groups. The programmes varied considerably, with mental health literacy and mindfulness being the most common forms.
Oh, before I continue, this is a scoping review and not a meta-analysis. So the researchers don’t calculate one average effect of these different interventions. They try to map which possible negative consequences can be found in the existing research.
And what did they find? Of the 58 studies, 17, or 29 per cent, reported at least one possible form of harm. In nine of the 58 studies, so 16 per cent, there was a statistically significant deterioration on at least one outcome measure for the group as a whole.
But don’t conclude from this that 29 per cent of young people are harmed by mental health programmes. It means that 29 per cent of the studies found at least one possible negative effect. Moreover, deterioration during an intervention doesn’t necessarily mean that the intervention caused it. Young people can start feeling worse during a school year for all kinds of other reasons. This is precisely why control groups are so important, and the researchers noted, for example, that sometimes the control group deteriorated too! That’s why they deliberately talk about potential harms.
Averages can hide things
Only 21 of the 58 studies separately examined whether particular groups of students responded differently. Six of those 21 studies, or 29 per cent, found significant deterioration in at least one subgroup. A study by Therriault and colleagues provides a particularly striking example. These researchers distinguished between students based on the problems they experienced beforehand. The group with few problems at the start deteriorated on all fourteen outcomes after the intervention. The middle group deteriorated on three of the fourteen outcomes. But here it comes: the young people who had many problems beforehand improved on all fourteen outcomes. Do you see the problem? What is good for one may come at a cost for others.
But be careful, this is of course one study and certainly not a general rule that they found in every study. It does illustrate a problem with such universal interventions. A programme can be useful for young people who need help, while the same programme may add little for young people who are doing well, or may even have unwanted consequences.
The same applies to clinically meaningful deterioration among individual students. Only three studies explicitly examined this, and all three found a group of students who deteriorated in a clinically meaningful way. But again, caution is needed: in only one study was the proportion of students who deteriorated larger in the intervention group than in the control group. In another study, it was actually smaller than in the control group. I repeat: we often still don’t know whether the intervention itself is the cause.
Maybe we aren’t looking carefully enough either
The main purpose of this review is not to cause panic, but Yao and colleagues do point to an important scientific problem with important societal consequences: none of the 58 studies had been designed in advance to investigate possible harms or had prespecified outcome measures for harm.
Think about what is written above. We carry out interventions around the mental health of thousands of young people, but when researching them, we apparently mainly ask whether they work. We ask far less systematically whether they might also have unwanted consequences, even though such consequences may indeed exist.
A nice example, by the way, comes from Belgium. In a study of a mindfulness intervention, the researchers asked participants to describe positive and negative aspects of the intervention. Of the 95 young people in the intervention group, 22 mentioned unpleasant experiences. They described, for example, becoming more aware of negative thoughts and unpleasant physical experiences, or experiencing more stress, anxiety and sadness. Interestingly, the standard quantitative analysis in that study did not show significant deterioration. It’s an old rule, but one that seems to hold here too: what you measure partly determines what you find.
How could such a programme cause harm?
I can be soberingly brief about this: we know even less about that. In the literature, the researchers find several possible mechanisms. For example, an intervention may make young people pay more attention to their thoughts and feelings, something I myself remember from a study on mindfulness. As a result, they may start ruminating more or interpret normal, temporary negative feelings as signs of a mental health problem.
There is also talk of a possible hopelessness effect: what happens when you are given a programme that is supposed to help, but you still feel bad afterwards? There is also a possible spoiling effect, whereby young people who later need help may feel that they have already tried something like this and that it didn’t work anyway.
This is mainly based on the literature, not on what the researchers found in the 58 studies. They did find significant deterioration relatively more often with cognitive behavioural therapy and mindfulness. Again, caution is needed. The studies of these interventions tended to be methodologically stronger, which means they may simply have been better at detecting negative effects. At the same time, the authors point out that these particular interventions make students engage directly with difficult thoughts and feelings, whereas mental health literacy or social and emotional learning, for example, tends to focus more on acquiring knowledge and skills.
So should we stop working on mental health at school?
By now, you probably fully understand my note at the beginning. My answer would be: no. That is a conclusion this research does not support. The review deliberately looks for possible harms and therefore needs to be read alongside research showing positive effects and null effects. The authors themselves mention several limitations. There are, for example, few studies with good long-term measurements. Some of the research took place during the exceptional COVID period. The review also looks only at in-person, universal programmes.
But the opposite conclusion, that such programmes are well-intended and therefore can do little harm, seems equally difficult to maintain after this review.