Mental health issues, including depression and anxiety, are some of the leading causes of disease burden among young people worldwide (Kieling et al., 2024; Liu & Kuai, 2025). Experiencing these issues during early adolescence or young adulthood is linked to long-term negative outcomes (e.g., higher risk of recurrence, suicidal thoughts; Clayborne et al., 2019). To address this concern, researchers and healthcare providers are increasingly examining the effectiveness of risk prediction tools to identify youth at risk of developing these conditions.
Risk prediction involves analysing a person’s data to estimate their likelihood of developing a particular condition within a specific timeframe. This approach is commonly used in primary care settings to assess the risk of physical health issues, like cardiovascular disease (Damen et al., 2016). However, there has been a lack of studies examining the views of young people in the UK through semi-structured interviews, which limits understanding of their desire for these tools and any concerns that may prevent them from engaging in such interventions.
The current study conducted by Higson-Sweeney et al. (2026) sought to explore how young people in the UK perceive risk prediction tools and the potential personalised preventive interventions for depression and anxiety that may follow.
Methods
The study sample consisted of participants aged 16–25 years living in the UK, who could either access the internet or travel to King’s College London for an interview. Individuals currently experiencing acute mental health distress were excluded. While lived experience or diagnosis of a mental health condition was not part of the eligibility criteria, this information was collected to contextualise the sample.
Participants were recruited primarily through social media and outreach to relevant organisations, initially using convenience sampling before incorporating purposive sampling to enhance diversity (N=25). The average age was 21.04 years (SD = 2.78). In terms of gender, 60% were female, 36% were male, 4% identified as genderfluid, and 8% identified as transgender/gender diverse. Regarding sexual orientation, 32% were bisexual, 8% were gay/lesbian, 48% were heterosexual, 8% identified as queer, and 4% preferred not to disclose. The ethnic breakdown revealed that 28% of participants identified as Asian, 4% as Black African, 4% as mixed, 52% as White British, and 12% as White Other.
The study followed the Consolidated Criteria for Reporting Qualitative Research (COREQ) (Tong et al., 2007) checklist, and interviews were conducted online or in-person by four trained research assistants. A topic guide developed with a Young People’s Advisory Group (YPAG) informed the interviews, which included questions about participants’ understanding of risk prediction tools and personalised preventive interventions. Two vignettes were used to help them discuss a topic that might be unfamiliar or sensitive. Data were analysed inductively using reflexive thematic analysis (Clarke & Braun, 2017), grounded in critical realism, and emphasised reflexivity, with the lead researcher maintaining a reflexive diary to explore how personal experiences influenced the research process.
Results
Four themes were identified through thematic analysis, with relevant subthemes.
Helpful or harmful: risk prediction as a ‘double-ended sword’
Participants viewed risk prediction models and subsequent risk scores as having both positive and negative implications for youth. They highlighted that heightened self-awareness from risk scores could foster proactive prevention of mental health issues; however, they also cautioned that this awareness might lead to hyper-vigilance and misinterpretation of behaviours as symptoms. The perceptions of risk scores varied significantly, as they could validate feelings for those at high risk, offering reassurance and care, while simultaneously causing distress or uncertainty if the score did not align with a youth’s personal experiences.
If I was identified as low risk, it would like take a weight off my shoulders a little bit and I would maybe feel like I could stretch myself like for example the sleep, the socialisation, I would maybe push myself (…) to like not worry about these things as much.
While being identified as high-risk could encourage youth to seek much-needed support by overcoming barriers like waiting times, there were concerns about the potential discouragement or anxiety the scores might provoke, impacting their willingness to reach out for help.
If I was going through a particularly bad time and I was being labelled as low-risk, I think the issue would then be that I would feel more encouraged to, like, dismiss what I’m feeling.
‘Taken with a grain of salt’: are risk prediction models the way forward?
Participants showed openness to risk prediction models and personalised support but expressed scepticism regarding several concerns. There were ethical reservations about assigning risk scores to youth, as participants feared labelling could exacerbate stigma and worsen mental health. They emphasised the need for sensitive communication and transparency regarding how these scores are determined (“I think it can be quite deterministic (…) you have a 90% chance of developing depression. It’s like, oh, yeah? Well, that’s pretty sure it’s going to happen”).
Participants recognised the complexity of mental health and were doubtful that a single model could account for all influencing factors. They noted that data accuracy heavily relies on reliable sources, with a preference for medical records over self-reports, and advocated for regular updates to risk scores. Concerns about data privacy and ownership were prominent. Participants worried about the sensitivity of the data used in risk assessments, questioning who would access it and the potential misuse of information.
Lastly, recognising practical challenges, participants emphasised the need for careful planning and collaboration to integrate risk prediction models into mental health support systems effectively. Factors like resource availability and professional training were seen as critical for successful implementation.
Individual differences impact personalisation
Participants noted that preventive interventions cannot adopt a one-size-fits-all approach and must consider individual preferences. Their choices included one-on-one counselling, group therapy, mental health apps, and self-help resources, with no consensus on the best format for these interventions. They believed that personalising treatments would enhance effectiveness, contrasting with the current National Health Service (NHS) method of uniform treatment, which some viewed as insufficient (“…where everyone is treated the same way, which isn’t a good thing”).
‘You still need like a person in the process’: the importance of human involvement
This theme highlights the essential role of human oversight in both the development and implementation of risk prediction models and interventions. Participants emphasised the need to engage stakeholders, particularly youth and their networks, in creating these tools. Educating users about the tools, the data they use, and the benefits of prevention is crucial for fostering trust and encouraging engagement.
Human involvement is vital during the delivery phase as well. While participants accept algorithms for risk scoring, they believe healthcare professionals should validate these results to account for individual nuances. They also highlighted the importance of personal interaction to discuss treatment options, expressing concern over impersonal communications and advocating for discussions that respect individual privacy and needs.
I see those models as more of early identification or again risk identification, and then you take it further with the doctor or something. I wouldn’t personally be confident with a risk prediction model fully diagnosing me.
Conclusions
In conclusion, young people show cautious openness toward risk prediction tools and personalised prevention for depression and anxiety. However, successful implementation requires addressing concerns surrounding accuracy, privacy, ethics, feasibility, and potential harms.
Strengths and limitations
A key strength of Higson-Sweeney et al. (2026) is its qualitative design, which provides detailed insight into young people’s attitudes through semi-structured interviews. Including 25 participants aged 16–25 enabled researchers to explore complex concerns about accuracy, privacy, ethics, personal choice and human involvement. The use of reflexive thematic analysis was also appropriate for identifying patterns and differences across participants’ accounts, while the researchers’ use of reflexivity and involvement of a Young Person Advisory Group strengthened the credibility and relevance of the findings. Recruitment was guided by information power rather than relying solely on a predetermined sample size, which is appropriate for qualitative research and suggests that the sample was considered sufficient to address the research question.
However, the study has limitations. The convenience sample was UK-based and relatively small, limiting the generalisability of findings to wider or culturally diverse youth populations. Additionally, participants tended to interpret risk prediction as a simple risk score or label, which may have influenced their responses and perceptions of prediction tools. Finally, the authors noted that fraudulent participants were identified in the sign-up survey due to inconsistent IP addresses and information. To address this issue, they conducted visual identity checks and reviewed interview recordings for authenticity. As a result, one participant was withdrawn from the study due to reliability concerns.
A further limitation is that participants were asked to consider hypothetical risk prediction tools rather than experiencing an actual risk assessment or personalised intervention, meaning their views may not fully reflect how they would respond in practice. Furthermore, presenting risk prediction primarily in terms of a risk score or label may have shaped participants’ interpretations of the technology.
Implications for practice
In clinical settings, the findings suggest that risk prediction tools are most useful when embedded within a wider, human-led care pathway rather than treated as standalone assessments. Clinicians could use prediction results as a prompt for further conversation, exploring the young person’s current circumstances, concerns and protective factors before making decisions about support. A practical implication is the need for clear protocols governing what happens after a high-risk prediction, including timely clinical review, communication with the young person and access to appropriate preventive support.
Services could also routinely monitor whether predictions lead to beneficial outcomes or unintended consequences, such as increased anxiety, stigma or disengagement. This would allow tools to be evaluated not only for predictive accuracy but also for their real-world clinical impact. Training clinicians in communicating uncertainty may further help young people interpret predictions without viewing them as fixed outcomes.
Statement of interests
Anamarija Veic has no conflicting interests to declare. She acknowledges the use of AI to assist in revising the text to enhance clarity and correctness. This tool has helped with grammar checks and rewriting long sentences to increase the overall quality of the blog.
Edited by
Dr Dafni Katsampa.
Links
Primary paper
Nina Higson-Sweeney, Anna Peycheva, Josefien Breedvelt (2026). ‘It is just a prediction; it’s, like, not fact’: youth attitudes towards risk prediction tools and personalised preventive interventions for depression and anxiety. BMJ Mental Health, 29(1), e302327.
Other references
Clarke, V., & Braun, V. (2017). Thematic analysis. The journal of positive psychology, 12(3), 297-298.
Clayborne, Z. M., Varin, M., & Colman, I. (2019). Systematic review and meta-analysis: adolescent depression and long-term psychosocial outcomes. Journal of the American Academy of Child & Adolescent Psychiatry, 58(1), 72-79.
Damen, Johanna AAG, Lotty Hooft, Ewoud Schuit, Thomas PA Debray, Gary S. Collins, Ioanna Tzoulaki, Camille M. Lassale et al. “Prediction models for cardiovascular disease risk in the general population: systematic review.” BMJ 353 (2016).
Kieling, C., Buchweitz, C., Caye, A., Silvani, J., Ameis, S. H., Brunoni, A. R., … & Szatmari, P. (2024). Worldwide prevalence and disability from mental disorders across childhood and adolescence: evidence from the global burden of disease study. JAMA Psychiatry, 81(4), 347-356.
Liu, Z., & Kuai, M. (2025). The global burden of depression in adolescents and young adults, 1990–2021: Systematic analysis of the global burden of disease study. BMC Psychiatry, 25(1), 767.
Tong, A., Sainsbury, P., & Craig, J. (2007). Consolidated criteria for reporting qualitative research (COREQ): a 32-item checklist for interviews and focus groups. International journal for quality in health care, 19(6), 349-357.