Background Due to the problems unyielding nature, key bodies such as the Scottish Government have endorsed whole school approaches to improve adolescent mental health. Developing effective interventions in complex social environments can benefit from novel investigative methods and tools based on complexity theory. Whilst standard tools find it difficult to understand the consequences of different environments, using computational techniques such as agent-based models (ABMs) make it possible to understand the dynamic consequences of interventions. Therefore, dynamic co-produced models can help to identify interventions which best account for the complexities of the environment in which the intervention takes place.

Methods We ran workshops with youth advisors, staff, and students in two Scottish schools, asking about feelings and behaviours in various situations. Based on these workshops and the scientific literature, our ABM was developed to include positive and negative interactions, and different relationships to environments. Having identified teacher location as a key intervention point to improve adolescent mental health, we investigated the effect of teacher movement during breaktimes on student stress by varying the probability that the teachers move around the school from zero to one. We further assessed how the effects of teacher movement varied according to two versions of the model. These differed in terms of the data used to inform agent behaviours and included either 1) rules based on the scientific literature, or 2) rules based on the workshop data. Mean student stress score with 95% uncertainty intervals were computed for each model version and teacher movement probability.

Results Key workshop observations include classroom decoration, and loud, busy corridors. Teacher location had a non-linear impact in both versions of the model, with stress peaking at movement probabilities of 0.25 (literature) and 0.10 (workshop) before declining. While the literature-based model suggested unsupervised breaktimes led to lowest stress, the workshop-based model suggested improvements with more frequent opportunities for teachers to intervene and reduce stressful interactions within corridors.

Discussion The sharing of perspectives was a key qualitative outcome for this method, and encouraged better understanding between different age groups, and between students and teachers. Asking about the situation provided information which helped uniquely develop the model and engaged discussion of difficult concepts such as expectation violation. Upon investigation, each perspective provided differing suggestions for best intervention, showing that co-produced dynamic models are required to understand and reduce issues around adolescent mental health.