Detection of risk for future depression among adolescents: Stakeholder views of acceptability and feasibility in the United Kingdom

Abstract Aim: Depression is one of the most common mental illnesses globally and a leading cause of disability. It is often established by late adolescence and thus identifying which adolescents are most at risk is crucial to enable early intervention to prevent depression onset. We have previously developed a risk calculator to stratify which adolescents are at high risk of developing depression and in this study explore the views of stakeholders to ascertain the acceptability and feasibility of implementing such a tool in the UK. Methods: Semi‐structured interviews were conducted with 60 UK‐based stakeholders (12 healthcare workers, 12 social workers, 12 school workers, 12 policymakers and 12 parents). Interviews were audio‐recorded and transcribed verbatim. Transcripts were analysed drawing on framework analysis techniques in NVivo 12. Results: Six overarching themes were identified: facilitators of acceptability; barriers to acceptability; role of stakeholders in implementing risk screening; feasibility of delivering the risk calculator in practice; barriers to implementation; and policy and system implications of using it in the current UK health and social care climate. The implementation of a depression risk calculator in the UK was seen as largely acceptable and feasible by most respondents. There was a strong emphasis on the utility of schools to implement this risk calculator, although it was recognized that training and support would be essential. Conclusions: Stakeholders were generally positive about utilizing a tool to screen for risk of future depression among adolescents in the UK but raised important concerns which should be taken into account before implementation.


| INTRODUCTION
According to the World Health Organization (WHO), mental health conditions account for 16% of the global disease burden in adolescents aged 10-19 (WHO, 2019). In the United Kingdom (UK), one in 10 young people have some form of diagnosable mental health condition (Mental Health Foundation, 2015), with recent data indicating that levels of depression have increased from 9% in those born in the 1990s to 15% in those born in the early 2000s (Patalay & Gage, 2019). This is particularly concerning as depression has been shown to greatly increase the risk for life-threatening physical disorders (Prince et al., 2007) and suicide (WHO, 2014). It has also been demonstrated that up to half of all mental disorders, including depression, begin by the age of 15, with three-quarters being diagnosed by 18 (Kim-Cohen et al., 2003). As such, the UK Government's latest green paper on young people's mental health made the early intervention and prevention of mental disorders a public health priority to reduce morbidity and premature mortality (DoH, 2017). However, the evidence is mixed regarding whether universal programmes to prevent mental illness in young people are effective or cost-saving (McDaid et al., 2017;Stallard et al., 2012). Therefore, it has been suggested that future research might be better directed at targeted prevention efforts for high-risk children and adolescents (McDaid et al., 2017).
Initial trials of targeted prevention programmes among adolescents at high-risk for depression show promising results (Brent et al., 2012;Calear & Christensen, 2020) and a recent meta-analysis found that these were more effective than universal programmes (Werner-Seidler et al., 2021).
Screening for future risk of developing a disorder is commonly utilized in physical health care, for example screening for obesity in children (US Preventive Services Task Force, 2017). Whilst historically neglected, it is increasingly being implemented with young children to aid prevention of mental illness (Feeney-Kettler et al., 2010), for example, screening for internalizing and externalizing problems in preschool children using the Behavioural and Emotional Screener (BASC-2 BESS) (Kamphaus & Reynolds, 2007). However, there is some scepticism concerning the utility of screening for risk of mental illness as diagnoses are often not as clear cut (Banner, 2013), which could result in large numbers of false positives (Mitchell & Coyne, 2007). Incorrectly labelling a young person as being at risk for a mental illness could be particularly problematic as stigmatizing attitudes about psychiatric disorders are still common in the UK (Curtice & Ormston, 2015) and thus those deemed at high-risk might experience shame, distress, discrimination, and social exclusion. Furthermore, previous research has suggested that existing tools are ineffective at determining future risk of mental illness (Najman et al., 2008), which may compound service providers' reluctance to implement risk screening (McDaid et al., 2017).
Recently, as part of the Identification of Depression in Early Adolescence (IDEA) international consortium (Kieling et al., 2019), we have developed a new model for predicting the future onset of depression among adolescents (Brathwaite et al., 2020(Brathwaite et al., , 2021Rocha et al., 2021). This prediction model, or 'risk calculator,' combines easy-to-obtain social and demographic information about young adolescents to predict their individual risk of developing depression at age 18. The model was originally developed and tested in a Brazilian sample but has been shown to have satisfactory performance in a UK sample . However, given the above-mentioned concerns regarding risk screening for mental illnesses, it is crucial to carefully consider the issues that might arise in using such a tool to screen adolescents for depression risk before attempting to implement it within the UK context. Therefore, this study aimed to investigate the acceptability and feasibility of implementing screening for future risk of depression among adolescents in the UK. Qualitative interviews were conducted with a wide range of key stakeholders likely to be involved in its implementation, including health, social care and education service providers, policymakers, and parents of young people.

| METHODS
This research formed part of a multi-site qualitative study conducted in Nigeria, Nepal, Brazil, and the UK, with the aim of informing global and local policy to improve the early identification and prevention of depression in early adolescence (Wahid et al., 2020). The overall purpose of the multi-country qualitative study was to elicit stakeholders' opinions on the feasibility and acceptability of integrating (1) screening for risk of developing depression in adolescence and (2) preventative interventions into health, social and educational settings in each country. Topic guides, types of respondents, codebook development, coding, and analysis were harmonized across sites. The current analyses focus on the UK qualitative component.

| Participants
A purposeful, convenience sampling method was used to recruit stakeholders from London, UK. We aimed to recruit from the follow- Based on recommendations for identification of meta-themes in multi-sited cross-cultural research, 20-40 respondents per country may be needed for thematic saturation across sites (Hagaman & Wutich, 2017). Therefore, our preliminary sampling goal was 45 participants per country. For each UK-based stakeholder group, we a priori targeted 12 respondents as previous research has suggested that this sample size is sufficient for the identification of themes within homogenous groups (Guest et al., 2006). Because five stakeholder groups were targeted for the UK, our sample target was 60 participants. We adopted this approach as it provided a practical solution for planning the number of participants to be recruited in order to maximize the chances of obtaining sufficient data to achieve thematic saturation. However, we acknowledge that this is a complex and contentious concept (see Sebele-Mpofu, 2020) and by no means guaranteed that we would achieve thematic saturation.

| Procedure
All participants who met the inclusion criteria were provided with an information sheet in advance of their interview and asked to provide written informed consent. Ethical approval was granted by the Psychiatry, Nursing and Midwifery Research Ethics Subcommittee at King's College London. In-depth semi-structured interviews were conducted by AB (research assistant) in person or over the telephone with each participant and lasted for between 45 min and 2 h (averaging 1 h).
The audio recordings of the interviews were transcribed verbatim and anonymised.

| Topic guides
The topic guide covered three main areas: the identification of risk for depression in adolescence, facilitators and barriers of different platforms to identify risk and intervene, and system improvements. It was developed in collaboration with qualitative research teams in each of the participating countries, with country-specific adaptations when needed. For topic guide adaptation, we took an iterative approach, with six preliminary interviews conducted with an a priori deductive interview guide. The resulting data were then analysed with an a priori deductive codebook developed based on existing literature and theory (see Wahid et al., 2020). For example, we drew upon the social ecological model of health (McLeroy et al., 1988) and the world system theory on the social origins of disease (Baer et al., 2013) to guide the development of codes to understand the role of individual, interpersonal, institutional, community and policy factors, and their interrelations, in depression risk and identification in adolescence. During this analysis, the interview guide and codebook were modified to include context-specific questions and inductively created codes for subsequent interviews (see Supplementary Materials and Table S1).
Based on preliminary interviews in these countries, a prototype risk calculator was created for participants, modelled after similar risk calculators for diabetes and using some of the domains in the IDEA risk calculator , During the interview, the mock calculator could be used to demonstrate what was meant by a risk calculator and thus gain more in-depth answers in relation to the implementation of risk screening in practice. Participants were instructed that this was just a mock-up and no data were collected when viewing the form.

| Data analysis
Data analysis was conducted in coordination with parallel studies conducted in Nigeria, Nepal, and Brazil (Wahid et al., 2020). Here we focus on the results from the UK site. A systematic and progressive method, based on Framework Analysis (Smith & Firth, 2011), utilizing a combination of deductive and inductive approaches (employing Grounded Theory techniques; Strauss & Corbin, 1990), was used to understand the cultural and contextual considerations for each site.
The protocol for the data collection and analysis was published in advance (Wahid et al., 2020).
During data collection, debriefing forms and analytic memos were completed by the interviewer to maintain a contextualized, reflective account of the interviews and to assist in development of inductive codes, whilst also noting methodological decisions in relation to the use of the topic guides in practice. Following preliminary coding, the constant comparison method was used to refine themes until thematic saturation was reached (Strauss & Corbin, 1990). The interview transcripts were analysed by the interviewer (AB) using NVivo-12 software and consensus agreement was reached with HLF.

| Sample characteristics
Of the 60 participants interviewed, there were 12 healthcare workers, 12 social care workers, 12 policymakers (eight from healthcare or social work; four education-related), 12 school workers, and 12 parents/caregivers. Fifty-one provided demographic information (see  facilitators of acceptability; and barriers to acceptability. Table 1 provides full details of these themes and subthemes (plus five additional subthemes common only in some stakeholder groups), the frequency of occurrence in each stakeholder group, and illustrative quotes.

| Observed themes
All respondents, regardless of group, discussed the utility of the risk calculator at a similar frequency but were also concerned about the potential negative consequences of labelling, how accurate the tool was likely to be, and ethical concerns about data usage and lack of available services for those identified as at-risk. These included, for example, the potential for human error in the administration of the risk tool, as well as the application of population-level adverse childhood experience items at an individual level (Kelly-Irving & Delpierre, 2019), and latterly, the potential for identifying adolescents at risk without having the resources to then implement effective prevention work. However, despite the reservations of some service providers about the accuracy of the risk calculator, most respondents were of the opinion that something is required to identify those at risk, and that this tool is a step in the right direction.
At least a quarter of each group felt more progressive attitudes among young people would increase acceptability of the tool, though some flagged the need for more stigma reduction among parents and greater education generally about what being 'at-risk' means to facilitate acceptability. For instance, concerns were voiced about the potential for being given an 'at-risk' label resulting in a fatalistic attitude to the development of problems in the future and thus that being identified as being 'at-risk' would bring similar levels of stigma as a psychiatric diagnosis would. As such, some stakeholders drew similarities between the risk calculator and genetic testing, such as negative consequences on the psychological well-being of individuals identified as at-risk of a future disorder (Botkin et al., 2015). Respondents in our study indicated the risk calculator may serve as a defeatist selffulfilling prophecy for some young people and were distrustful of making definitive predictions in mental health contexts.
All groups, except healthcare and social work policymakers, felt that sensitive delivery of the tool would be important for its acceptability, meaning that it was presented in such a way as to avoid blaming the young person or their parents for the results, while parents were less concerned about transparency of the tool than the other groups. Transparency was discussed by professionals as requiring the tool to be delivered with an explicit understanding of what might or might not happen next, and the extent of confidentiality regarding any disclosures resulting from the tool. Several healthcare and school workers expressed reluctance about asking some of abuse-orientated questions in the risk calculator to children, due to the potential safeguarding repercussions they would be responsible for initiating, whereas this was less of a concern for social workers. ing; delivery of the risk calculator; barriers to its implementation; and policy and system implications resulting from its implementation. Full details of these themes and subthemes (plus 12 additional subthemes common only in some stakeholder groups), the frequency of F I G U R E 1 Thematic map depicting the main themes across all stakeholder groups for the acceptability of implementing risk screening for future depression among adolescents in the UK T A B L E 1 Themes observed in interviews with healthcare workers (H) (n = 12), social workers (S) (n = 12), related policymakers (H&S Pm) (n = 8), school workers (Sch) (n = 12), related school policymakers (Sch Pm) (n = 4), and parents/caregivers (P) (n = 12) regarding the acceptability of screening for future risk of depression in the UK Option to opt out Some respondents especially school workers) felt there would always be some who will not approve of this risk calculator and that this may be best addressed by ensuring the right to withdraw consent in a similar manner to sex education. tends to be more the concern of practitioners in asking these questions, or in giving the kind of inventories for them to complete. Because they have been sort of worried about the children getting upset about it. But to be honest about it, I think that more of this is in the minds of the practitioner than in the young person themselves." UK-064 [H&S Pm] "…if they would disclose anything about maltreatment…I would worry a bit myself in terms of, well, how that reflects on me as a parent" UK-011 [P] Fallibility of risk screening Some respondents were concerned about the accuracy of the risk calculator, which may pose a barrier to its acceptability as this undermines its utility. Parents in particular were concerned that it was very brief and was particularly focused on environmental risk factors, as opposed to family history or personality traits.
occurrence in each stakeholder group, and illustrative quotes are provided in Table 2.
Stakeholders from all groups consistently identified schools as being the most feasible and accessible platform from which to deliver screening to the largest number of young people. Notably, there were broad similarities between all respondents in terms of the barriers to feasibility, with most discussing a lack of resources (i.e., money for training, and staff time to devote to its implementation) given chronic understaffing in existing services. Moreover, respondents were concerned that schools are increasingly being used as a place to take on more pastoral care than the system is currently competent to manage and that teachers are ill-equipped to deal with mental health concerns.
One way in which stakeholders discussed overcoming these barriers was to deliver risk screening via an app, which was deemed to be the most accessible and cost-effective way of identifying children at risk. Frontline service providers were keen that the use of apps was professionally led, whilst policymakers were keen that there was F I G U R E 2 Thematic map depicting the main themes across all stakeholder groups for the feasibility of implementing risk screening for future depression among adolescents in the UK T A B L E 2 Themes observed in interviews with healthcare workers (H) (n = 12), social workers (S) (n = 12), related policymakers (H&S Pm) (n = 8), school workers (Sch) (n = 12), related school policymakers (Sch Pm) (n = 4), and parents/caregivers (P) (n = 12) regarding the feasibility of implementing screening for future risk of depression in the UK Respondents talked about parents being well placed to identify risk of developing depression, providing collateral information to professionals, but equally that they may be part of the problem and so are not as well placed as professionals to identify risk. Parents also acknowledged that those without personal experience of depression might be less able to identify their child was at risk.  (83) Sch 4 (100) Sch Pm 10 (83) P "But I'm wondering if it's something that you stumbled across and then sort of came out as being high risk of depression, and you do not quite have the confidence, or the skill set to follow up on any of these. I'm being overcautious, but…I suppose some of the questions, yes, I think for teens, I think…it ought to be sort of done in a kind of safe space with someone who they can then kind of go to rather than you doing it in your bedroom and coming out with this risk score and then not feeling quite sure what to do with it." UK-055 [P] Self-directed Respondents felt the risk screening could be feasibly completed by young people themselves, and the anonymity via use of technology could encourage honest answers from young people. However, they also raised concerns regarding confidentiality and how, when and where the information divulged would be shared.

| DISCUSSION
To our knowledge, this is the first study to investigate, with a range of key stakeholders, the acceptability and feasibility of implementing screening for future risk of depression among adolescents within the UK context. We found that the majority of stakeholders interviewed in this study suggested that risk screening for developing depression is both acceptable and feasible in the UK, with many citing its utility for filling a gap in service provision that currently fails to prevent young people's mental health from deteriorating to a crisis-point.
However, a number of caveats and important issues for consideration were also raised regarding the implementation of such screening.
Stigma was a concern impacting acceptability of risk screening for developing depression. Being labelled as being 'at-risk' of a mental illness in adolescence can be both beneficial in order to facilitate access to support, as well as harmful as a result of the negative reactions of their peers (Welsh & Brown, 2013). This age group may be particularly at risk of stigmatization from peers (Crosnoe et al., 2008). These views are consistent with recent criticisms that diagnosing individual-level vulnerabilities from probabilistic, population-based tools is inherently problematic and may serve to exacerbate inequalities rather than serve those who are most vulnerable to future problems (Kelly- Irving & Delpierre, 2019).
Schools were highlighted as being the most feasible setting to implement risk screening. However, despite schools being widely thought of as the most accessible place to support the majority of children (Hanley et al., 2017), only 2% of secondary schools currently have a published policy relating to mental health (DfE, 2018). Indeed, a report by the Mental Health Foundation (2018) found that 73% of teachers felt mental health was not given sufficient priority during their training. This echoes the concern raised by many respondents that more investment would be needed to skill-up staff to implement depression risk screening and ensure clear pathways were in place to access preventive interventions.
The idea that apps should be delivered under the supervision of a professional until efficacy and safety is better established is consistent with previous research which suggests that professionals are cautious when referring individuals to online resources because of concerns regarding their effectiveness (Sinclair et al., 2013). At least in the prepandemic era, there was limited evidence concerning the efficacy of smartphone apps in children and young people's mental health care (Hollis et al., 2017) and this potential approach for risk screening thus requires careful investigation.

| Limitations and future research
It is important to note that within the qualitative interviews, some respondents mistook the nature of the risk calculator to be something that was being used to identify current symptoms of depression rather than those at high risk of developing depression in the future, which may have impacted upon the findings obtained. However, the iterative nature of the interview topic guide, and the use of a prototype risk calculator to demonstrate its potential use in practice to respondents may have improved the depth and specificity of our Research Respondents, largely those working in health and social care, explained that the best way to drive policy change is to fund and engage in research to begin to understand the scale of the problem, and the utility of screening and prevention for risk of developing depression.
6 (50) H 3 (25) S 5 (63) H&S Pm 1 (8) Sch 0 (0) Sch Pm 2 (17) P "…this could be one off research definitely, but if we incorporate it into the system, you know, we need to show kind of figures possibly. I think that would help, that if you nip it in the bud, this is how much you are saving the NHS or recouping resources later on…I think that's how they count is not it?" UK-029 [P] findings. The participants in this study were, however, a relatively homogenous group (predominantly white and female) who are not representative of the wider population in the UK, thus limiting the generalizability of the findings. Unfortunately, we were unable to recruit adolescents into our study due to ethical restrictions within the UK. Future research would benefit from investigating the attitudes of adolescents themselves towards risk screening as these may be affected by generational shifts in attitudes towards mental health as discussed by the respondents in this study. It may also be useful to investigate the effectiveness of stigma reduction campaigns and the next steps required to improve de-stigmatization of mental health issues to increase the wider acceptability of such a risk screening tool.

| CONCLUSION
Risk screening for future development of depression in adolescence was seen by most respondents as largely acceptable and feasible in the UK. This may be because of widespread anti-stigma campaigns, such as the Time to Change campaign (a social movement led by the UK-based charities 'Mind' and 'Rethink Mental Illness'), and a growing recognition of the issues facing young people, placing prevention nearer the top of the public, political and clinical agendas. Although the opinions of stakeholders were largely positive regarding the risk calculator, a lack of resources was identified as a major barrier to its feasibility in practice. This is reflective of wider concerns within mental health care in the UK, where CAMHS are recognized as reaching only 35% of those in need (CQC, 2018), and even fewer of those who may require preventative support. Clear referral pathways and accessible services would need to be in place before implementation of risk screening for future depression among adolescents could be considered. Subsequently, research will be required to develop a physical or online screening tool based on our predictive model and test its implementation potential in school settings in the UK.