Effects of a Whole-School Prevention Program Targeting Mental Health and Nonsuicidal Self-Injury in Swedish Adolescents: A Cluster-Randomized Experimental Study with Longitudinal Follow-Up

This study contributes to the still limited body of research investigating the effects of school-based universal prevention targeting NSSI, and aligns with earlier findings showing that addressing NSSI in educational settings does not seem to have iatrogenic effects (Baetens et al., 2020; Baetens et al., 2024; Muehlenkamp et al., 2010). Like Baetens et al. (2024) and Zare et al. (2023), it was found that the current school-based prevention reduced NSSI frequency. Reduction of NSSI frequency in adolescents with NSSI is important as repetitive NSSI has negative long-term outcomes (Daukantaite et al., 2020b).

There is an ongoing debate about whether the frequency of NSSI is an appropriate outcome for treatment and prevention studies. The results suggest that versatility (defined as the number of different methods of self-injury) alone (Turner et al., 2013), or in combination with frequency (Ammerman et al., 2020), might be a better indicator of NSSI severity and hence preferable. Despite this, frequency remains the most commonly used behavioral outcome related to NSSI (Calvo et al., 2022; Witt et al., 2021) and is associated with the risk of suicide attempts (Brausch & Boone, 2015). Concerning changes in NSSI frequency, Muehlenkamp et al. (2023) identified this to be significantly associated with a change in suicidal ideation; however, their study was observational, and the mechanisms of these changes were not investigated. Based on earlier research, lower NSSI frequency thus has the potential to reduce the risk of suicidal behavior (Prinstein et al., 2008) and can be an important step toward NSSI cessation (Kiekens et al., 2017).

The estimated NSSI frequency in the intervention group significantly decreased from baseline to post-intervention and was at post-intervention significantly lower than that in the control group. However, this difference did not persist at the follow-up measurement, as, at that time point, the frequency in the control group was also lower and significantly different from its baseline estimate. This might be explained by the follow-up measurement being carried out just after students returned from their two-month summer holiday, something which reasonably affected participants in both the intervention and control group. There is substantial evidence indicating increased self-reported mental health (Kristjansdottir et al., 2013), lower rates of mental health-related hospital admissions (Radhakrishnan et al., 2023; Slaunwhite et al., 2019), and reduced rates of self-harm (Jack et al., 2023) during school holiday periods. Furthermore, in qualitative interviews, students from the present study sample identified academic stress as a major contributor to mental ill-health (Aspeqvist, Münger, et al., 2024), which would have been absent during the holiday.

We did not, however, detect a change in NSSI onset as an effect of the intervention. The sample size can reasonably explain this, as the participation rate was low, and it is a known fact that many adolescents who engage in NSSI begin at an earlier age (Aspeqvist, Andersson, et al., 2024; Plener et al., 2015). Thus, to detect delay or actual prevention of NSSI, one would need to deploy earlier intervention and study considerably larger sample sizes. For example, Baetens and colleagues (2024) focused on a community of 11–14-year-olds and reported a significant reduction in NSSI frequency after a short universal preventive intervention.

Further, results indicated an effect on Stigma Awareness. There was no significant change in the intervention group, but a significant decrease in the control group and exploratory analyses showed that this effect was driven by a decrease among boys in the control group. The effect size was small; however, small effect sizes can have large impacts when interventions are delivered to large populations (Funder & Ozer, 2019; Rosenthal, 1990; Werner-Seidler et al., 2021). The Stigma Awareness subscale of the PMHSS-R is built from items with statements such as “Most people look down on children who visit a counselor because they have emotional or behavioral problems.” where higher scores indicate agreement with the statement. The designers of the PMHSS-R describe the Awareness subscale as intended to measure awareness of prevailing social stigma and differentiate this from personally held stigmatizing beliefs (McKeague et al., 2015). Stigma surrounding mental health and NSSI is something that is addressed in both the KRAS and YAM interventions that were delivered to adolescents in the current study. As mental health stigmatization is a real and widespread phenomenon of great concern (Thornicroft et al., 2022), a reasonable conclusion is that the intervention group ended up with increased awareness of the stigma associated with mental health problems generally and NSSI specifically, which expressed itself as a more realistic, albeit somewhat darker, view of social stigma in comparison to the control group at follow-up. An earlier qualitative study on participants’ experiences with the intervention (Aspeqvist, Münger, et al., 2024) revealed that adolescents reported a greater understanding and compassion for their peers following the intervention. Since stigma poses a barrier to help-seeking (Barrow & Thomas, 2022) and the disclosure of NSSI (Rosenrot & Lewis, 2018), fostering more understanding environments may benefit those needing social support and mental health services.

The current study found no significant differences between intervention and control conditions on help-seeking, social support, self-criticism, and emotion regulation, even though the importance of help-seeking, for example, is emphasized throughout YAM. These null results do not confirm earlier research on school-based interventions (Baetens et al., 2024; Cipriano et al., 2021), which found effects on attitudes toward help-seeking and emotion regulation, for example. YAM has previously shown effects on help-seeking (Lindow et al., 2020). Due to the low participation rate in the current study, it might be underpowered to document effects. In qualitative evaluations of the intervention (Aspeqvist, Münger, et al., 2024), participants reported being aware of the importance of help-seeking message, which was, however, not reflected in the quantitative measures included here. Interestingly, a “help-seeking vs. autonomy dilemma” was found, where participants reported being aware of the need to seek help from adults in matters related to mental health, and, at the same time, expressing a need to balance this against their wish for autonomy and coping for themselves. This implies that help-seeking may be a more complex construct for this age group, which needs to be considered when interpreting the results. As Foulkes and Stapley (2022) pointed out, listening to adolescents’ perspectives and suggestions for improvements and combining qualitative and quantitative measures in evaluating preventive interventions has the potential for a more nuanced understanding of results.

Explorative analyses yielded significant interaction effects when gender and baseline NSSI history were included in the linear mixed effects models. Thus, the whole-school preventive intervention appeared to have effects that were differentiated by these factors. Regarding stigma awareness, it became clear that the overall effects were specific to and more pronounced in male participants. This has potential importance, as an environment with less stigmatization might increase help-seeking among boys, a group that is otherwise less likely to seek help (Barrow & Thomas, 2022; Haavik et al., 2019).

In the intervention group, there was also a positive effect on bullying specific to girls, indicating that they experienced less fear of others, being made fun of, or being bullied. A lower level of bullying and negative social interaction can have a large impact, as bullying is linked to a range of adverse mental health outcomes longitudinally (Lereya et al., 2015) and also, as negative social interaction has been linked to increased NSSI urges (Haliczer & Dixon-Gordon, 2023).

Further, there was a reduction in norms on not talking to others about suicidality among boys. There are gender differences in suicide among adolescents and young adults. Suicide attempts are more common in females, but suicide rates are higher in males. (Miranda-Mendizabal et al., 2019). Breaking the silence among boys is one step forward in suicide prevention. As described above, the YAM program aimed at reducing suicidality and was effective in a large randomized controlled trial (Wasserman et al., 2015).

Effects found to be differentiated by NSSI in the present study are difficult to interpret. Participants in the intervention group without a history of NSSI seem to have experienced a decrease in social support from friends as measured with MSPSS. Estimates suggested a potential increase in social support among participants with NSSI in the intervention group, but this difference was not statistically significant. On the KIDSCREEN Peers and Social Support subscale, reports from participants with a history of NSSI point to an increase in their friendships as an effect of the intervention. Social support is a protective factor for NSSI (McEvoy et al., 2023) and is important in NSSI cessation (Andersson et al., 2024; Simundic et al., 2024). Further, estimates point to an improvement in the school situation (as measured by the KIDSCREEN School Environment subscale) for those with NSSI in the control group. A possible explanation for this is that some change took place locally at one of the schools between measurements. On the one side, the overall aim of universal interventions is to achieve population-level effects. However, the fact that group-specific effects emerged can also be interpreted as support for the suggestion that preventive interventions may not be best investigated solely in terms of effects on whole sample means but by potentially looking further into effects on specific subgroups (Birrell et al., 2025; Greenberg & Abenavoli, 2016; Supplee et al., 2013). However, caution needs to be taken in interpreting these results due to the explorative nature of the analysis. Future research could further investigate the nature of subgroup-specific effects on universal prevention targeting NSSI.

The current study adds to the small body of research documenting effects of preventive interventions targeting NSSI. The results show that whole-school intervention of the type investigated here can have an effect on NSSI frequency in an adolescent population. Reduction in NSSI frequency was recently reported in another intervention combined with KRAS (Baetens et al., 2024). While YAM has documented effects related to suicide attempts and ideation (Wasserman et al., 2015), it has not previously been part of an intervention targeting NSSI. This study confirms earlier research indicating that it is possible to influence adolescents’ perceptions of stigma (Lanfredi et al., 2019). Finally, this study acknowledges the potential importance of examining effects in specific subgroups (Laurenzi et al., 2023).

A key strength of this study is its randomized controlled design, which enables conclusions about the specific effects of the intervention. Although a larger sample size would have been ideal, the current sample is sufficient to statistically verify a range of experimental effects. The study was carried out in a real-world setting—ordinary Swedish secondary schools—and this enhances the generalizability of the results to the broader population. Another advantage is the extensive survey battery, which allowed for analysis of effects across various relevant constructs.

However, there are some limitations to the study that need to be mentioned. One limitation is the participation rate, which was 25.3% of eligible adolescents. This was partly affected by the ethical requirement for both caregivers to provide active, informed consent. Moreover, the study was initiated during the last phase of the COVID-19 pandemic, which also affected recruitment; there were no school closures in Swedish compulsory (primary and lower secondary) schools (Lindblad et al., 2021), but social distancing affected procedures during the recruitment phase. While we are not aware of systematic effects on participation, the study sample may differ in some respects from the total eligible population. There could be a potential bias in adolescents who participated in the study, as caregivers with an interest in the topic and research in general may have been more likely to give consent. Regrettably, since publicly available statistics on the participating schools are limited and differ from the sample data regarding question types, this prevents an analysis of selection bias. However, eligible participants and their caregivers were unaware whether their school was assigned to the intervention or control group (the control group received the same interventions at a later stage). Thus, there is no reason to suspect different participation patterns between the groups. The achieved attrition rates of 24.4% at post-measurement and 13.5% at follow were relatively low and, while not optimal, limit the potential effects of attrition bias. Demographic data in Table 1 show that regarding sex, living with both parents, and country of origin, attrition seems not to have caused differences between the full (baseline) sample and the study sample (participants with data from all three measurements). Further, while Cronbach’s alpha indicated acceptable internal consistency reliability, the low OmegaT for the social support from family subscale indicates that the subscale might be reflecting more than one single construct in the study sample. Previous research points to challenges with retrospectively self-reporting behavior like NSSI (Daukantaite et al., 2020a), seemingly affecting the present sample as reported lifetime NSSI prevalence decreases across measurements. There was no measurement of cognitions and attitudes about NSSI before and after the intervention, which is a limitation. This would have allowed for more participant responses on NSSI-related variables, including those with no NSSI experience. Another limitation is that student attendance for each session was not recorded, nor were adherence ratings for fidelity to the methods used. As the study was cluster randomized with few clusters (i.e., schools), changes at the school level might have influenced the findings. No such changes were, however, brought to the attention of study staff. Moreover, as described above, seasonal effects have been documented on various mental health constructs, and as the follow-up measurement was carried out shortly after the summer holiday, such effects might have influenced the data. Lastly, as the study had a single treatment arm with all components of the whole-school intervention, study data do not allow for differentiating the effects of the components.

Proactively addressing mental health in educational practice can help promote healthy adolescent development, and further research is needed on whole-school prevention strategies targeting NSSI and mental health. Ideally, samples should be considerably larger, as this would allow for detecting smaller effects and further investigation into what predicts beneficial responses for whom. Whole-school prevention studies should preferably be conducted with modular approaches, as researchers could determine which components contribute to the intended outcomes. Further research is also encouraged on gender-specific effects and the underlying mechanisms behind potential differences in outcomes, as needs may be met to varying degrees across genders by universal preventive interventions.