Predictors of mental health difficulties and subjective wellbeing in adolescents: A longitudinal study

Abstract Background Mental health and subjective well‐being are of great interest in both health policy and research. There has been considerable debate regarding whether mental health difficulties and subjective wellbeing are two distinct domains or different ends of a single mental health spectrum. This study investigates if predictors of mental health difficulties and subjective wellbeing are the same or different in a large‐scale community‐based sample in the United Kingdom. Methods 13,500 adolescents in year 7 (aged 11–12) and again in year 8 (aged 12–13) completed surveys on emotional strengths and skills, support networks, mental health difficulties and wellbeing. Socio‐demographic factors were gathered from the National Pupil Database. Mental health difficulties and wellbeing scores were standardized to allow comparisons. Results The correlation between mental health difficulties and subjective wellbeing was −0.48, indicating a moderate overlap between the two domains. Some of the predictors (e.g., gender, ethnicity, problem solving, emotion regulation) in year 7 predicted both mental health difficulties and subjective wellbeing in year 8. However, some of the predictors in year 7 only predicted mental health difficulties (e.g., special education needs, empathy) and some only subjective wellbeing (e.g., prosocial behaviour, peer support) in year 8. Conclusion This study provides further evidence for differences in what predicts adolescents’ mental health difficulties and subjective wellbeing. It highlights the importance of not only focusing on preventing or treating symptoms of mental illness but also focusing on improving children’s wellbeing.


INTRODUCTION
Adolescence is a critical period for young people's mental health and wellbeing (Dahl et al., 2018;Patton et al., 2016). The majority of mental health disorders show first onset before the age of 25 (Jones, 2013).
Furthermore, adolescents perceive lower levels of life satisfaction and experience emotional distress more frequently and with greater intensity than younger children or older adults (Arnett, 2008). Understanding the predictors of mental health disorders in childhood can provide insight into what sets changes in mental health in motion.
Recent research estimates that one in eight children and young people experience mental health problems in the UK (NHS Digital, 2020).
Mental health disorders experienced in adolescence can cause a high burden for both the society and the individuals and can lead to This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. significant impairments in various areas of life such as family life and professional life as well as quality of life (Copeland et al., 2015).
There are two main distinct yet overlapping perspectives on wellbeing: one focuses on hedonic wellbeing and the other on eudaimonic wellbeing (Lambert et al., 2015;Ryan & Deci, 2001). According to the hedonic philosophical tradition, subjective wellbeing is a person's evaluation and feelings of their own life (Diener, 1984) and refers to the levels of positive affect, low levels of negative affect, and a high degree of overall life satisfaction (Diener & Lucas, 1999). On the other hand, the eudaimonic perspective refers to wellbeing as distinct from happiness and states that it occurs when individuals' life activities are congruent with deeply held values, which are holistically or fully engaged (Waterman, 1993). Some regard wellbeing as requiring both hedonic and eudaimonic components; that is, the combination of feeling good and functioning well (Huppert & So, 2013). For example, Seligman (2011) proposed the PERMA model to describe wellbeing, which is the acronym for positive emotion, engagement, relationships, meaning, and accomplishment.
Some consider wellbeing and mental ill-health to be two ends of the same spectrum of mental health. However, World Health Organization defined health as "a state of complete physical, mental and social wellbeing and not merely the absence of disease or infirmity" (World Health Organization, 1948, p. 1). Moreover, there has been identification of individuals who experience good wellbeing in the presence of symptoms of mental ill-health (Kinderman et al., 2015;Sharpe et al., 2016) and many individuals otherwise free of mental disorder do not feel healthy or function well (Keyes, 2005). The Complete State Model of Mental Health (Keyes & Lopez, 2002) combines the two spectrums. It states that an individual may have high mental illness symptoms and low wellbeing "floundering" which is defined as complete mental illness, or they may have high mental illness symptoms but high wellbeing "struggling" which is defined as incomplete mental illness. On the other hand, they may have low mental illness symptoms and low wellbeing "languishing" which is described as incomplete mental health, or they may have low mental illness symptoms and high wellbeing "flourishing" which is defined as complete mental health. In line with diagnostic criteria for a mental disorder not requiring all the symptoms to be present, the operational definitions of flourishing (Huppert & So, 2013;Keyes, 2002) may not require all the features of positive feeling and functioning to be present which would allow for an overlap between symptoms of mental disorder and features of flourishing (Huppert, 2014).
Depending on the theoretical framework, each researcher prioritises different components and develops a measurement scale (Huppert, 2014 focuses more on the psychological functioning rather than the subjective feeling states (Ng Fat et al., 2017).
The relationship between mental health difficulties and subjective wellbeing has been investigated in adult populations. Twin studies have shown evidence of some shared variance between mental illness and wellbeing.  have shown that for internalising psychopathology (prevalence of clinical levels of major depression, generalized anxiety or panic disorder in the previous year), genetics played a larger role than environment in the shared associations between wellbeing and internalising psychopathology. Moreover,  found that the strongest risk factors for low mental wellbeing were genetic factors that impact on both internalising psychopathology and externalising psychopathology and environmental factors that impact on externalising psychopathology (defined as alcohol-related problems and smoking). Similarly, Bartels et al. (2013) showed that the overlap between psychopathology and wellbeing was mainly genetic. Hofgaard et al. (2021) investigated the contributors to illbeing and wellbeing within a twin study design. They have identified that the genetic correlation between promotion of wellbeing and prevention of illbeing, despite stress, was 0.43 and the heritability was 0.24 and 0.30 respectively. Different factors were common to both prevention of illbeing and promotion of wellbeing (finding meaning in daily activities, a perception of good health, experiencing positive affect, being optimistic, not feeling lonely, and believing in one's capacity to handle difficult demands), while others were found to represent unique effects on either prevention of illbeing (being physically active and trusting others) or promotion of wellbeing (feeling satisfied and secure in close relationships). Studies that investigate the association between mental health difficulties and wellbeing are fewer across the adolescent populations. A study by Patalay and Fitzsimons (2016), investigated a wide range of individual, family, social and wider environmental predictors of mental health difficulties and wellbeing for 11-year-olds using a large, nationally representative sample (Millennium Cohort Study, MCS). Their results showed that alongside some factors that predicted both outcomes (e.g., sex, being from a single parent family, arguing with parents, sibling bullying and peer problems), many factors were uniquely associated with either designed to improve mental health should also consider having elements to improve wellbeing children's mental illness (e.g., ethnicity, cognitive ability, special education needs, communication difficulties and chronic illness) or subjective wellbeing (e.g., being overweight, parent problems, being bullied by peers and having a safe neighbourhood). While this study provides further evidence for mental health problems and wellbeing as distinct constructs, some methodological shortcomings limit conclusions. Notably, although subjective wellbeing was reported by the adolescents themselves, mental health difficulties were parent reported. Hence, the uniquely associated factors for mental health difficulties and wellbeing might have been due to different reporter perspectives. A further study by Rees (2018) investigated the predictiveness of early childhood (age 5) family and socio-economic factors on children's subjective wellbeing and emotional and behavioural difficulties at the age of 11. Similar to the study by Patalay and Fitzsimons (2016), their results showed predictors that are uniquely associated with subjective wellbeing and emotional and behavioural difficulties. However, this study also relied on the MCS and, therefore, shared the same limitation of different reporters as the Patalay and Fitzsimons' study (Patalay & Fitzsimons, 2016).
Current study draws on a recent, large-scale community-based self-report survey of adolescents to explore the correlates of mental health difficulties (emotional and behavioural difficulties) and subjective wellbeing. Specifically, child and family demographic factors, child reported emotional strengths and skills and support networks in year 7 were used to predict child-reported mental health difficulties and subjective wellbeing in Year 8.

Participants
This study investigated data that were collected between 2016/17-2017/18 from children who were part of a programme called Head-Start. Started in 2016, HeadStart is a 6-year, £67.4 million National Lottery funded programme set up by The National Lottery Community Fund. HeadStart aims to explore and test new ways to improve the mental health and wellbeing of young people aged 10 to 16 and prevent serious mental health issues from developing. To do this, six local authority led HeadStart partnerships (Blackpool, Cornwall, Hull, Kent, Newham and Wolverhampton) are working with local young people, schools, families, charities, community and public services to design and try out new interventions aiming to promote young people's mental health, wellbeing, and resilience. Schools have been selected by these partnerships; in most cases driven by perceived need in these schools but also by the potential for schools to successfully engage with the programme. The analyses reported are based on 13,500 students who completed annual surveys (The Wellbeing Measurement Framework, WMF) in year 7 (aged 11-12) and again in year 8 (aged 12-13). The socio-demographic characteristics of children were extracted through a data linkage with the National Pupil Database (NPD). The sample was not drawn to be representative of all school children in England; it was based on local areas that were part of the HeadStart programme (Big Lottery Fund, 2016). Compared to the national average, the study sample had a slightly higher proportion of children/young people from deprived backgrounds based on free school meal (FSM) eligibility (study sample: 16.04%, national average: 12.9%). The study sample had much lower proportion of children with special education need (SEN) support (study sample: 11.59%, national average: 14.4%), slightly higher proportion of White children/young people (study sample: 77.3%, national average: 75.2%) and females (study sample 53.6%, national average: 49.3%).

Main outcomes
At year 8 (aged 12-13), mental health difficulties was measured with the aggregated Emotional Symptoms and Conduct Problems subscales of the Strengths and Difficulties Questionnaire (SDQ; Goodman et al., 1998) which is a widely used questionnaire of psychopathology symptoms in the United Kingdom since the late 1990s.
Higher scores indicate greater symptoms.

Support networks
Family connection, school connection, peer support, community connection, participation in community life and participation in home and school life were measured using the SRS (Lereya et al., 2016).
Higher scores on the SRS reflect better outcomes.

Procedure
Every year, children in participating schools completed surveys using a secure online system during a usual school day. The online system was designed to be easy to read and child friendly. Ethical approval was granted by the UCL ethics committee (reference: 8097/003) and every year, consent for participation in the research was sought from parents prior to, and from children and young people at the outset of, the survey sessions. Parental opt-outs were received by post, phone or email and child assent was recorded via computer at the beginning of survey sessions.

Statistical analysis
Multiple imputation with chained equations (n = 20) were used to impute missing values (using the mi package in Stata 15). To ensure plausibility of the missing at random assumption, our imputation model included all the predictors and outcomes across the 3 years of HeadStart data collection (2016/17 to 2018/19). Overall, missing cells were at 34.7% of the total, with missingness varying from 1.7 for mental illness to 9·8% for wellbeing (for more information see Table 1).
To investigate whether the predictors of mental health difficulties and subjective wellbeing are the same or different, regression analyses were conducted predicting standardized scores (z-scores) for mental

<0.0001
Note: Each column represents a new regression model. All models were tested using multilevel linear regression.     Table 2 for the regression coefficients). Figure 1 displays the regression coefficients from respective regressions when the model is first added in to predict standardized scores of mental illness and subjective wellbeing.

DISCUSSION
There is a growing interest in the study of subjective wellbeing of children and adolescents (Holder, 2012;Patalay & Fitzsimons, 2016, 2018Rees, 2018). The current study investigated which child and family demographic factors, emotional strengths and skills, and support networks measured in year 7 predicted adolescents mental health difficulties and subjective wellbeing in year 8.

F I G U R E 1
Regression coefficients predicting mental illness and subjective wellbeing 6 of 10 - The results showed that from the child and family demographic factors, being a boy was associated with lower mental health difficulties and higher subjective wellbeing across all the models. This is consistent with literature suggesting that girls are more likely to display depressive symptoms across most of the life span, beginning at some point in adolescence (Piccinelli & Wilkinson, 2000;Yoon et al., 2022) which might be due to earlier pubertal timing (Ge et al., 2001) and negative coping styles such as rumination (Nolen-Hoeksema & Girgus, 1994). Being Asian or being Black, compared to being White, was associated with fewer mental health difficulties and better subjective wellbeing in the current study, which is in line with previous research (A. Goodman et al., 2008;Maynard & Harding, 2010). Consistent with research indicating that poverty is a potent predictor of both mental illness and poor wellbeing (Bradley & Corwyn, 2002), our results indicated that free school meal (FSM) eligibility, which is linked to low income, was associated with higher mental health difficulties and lower subjective wellbeing across all models.
Similar to previous findings, goals and aspirations (Deci & Ryan, 2000) and high problem-solving skills were associated with low mental health difficulties and higher subjective wellbeing in the current analyses. Previous literature has suggested that problem solving skills is a major intrapersonal and interpersonal process that leads to quality relationships and an enhanced quality of life (Wallander et al., 2001). Similarly, and consistent with the findings presented here, emotion regulation has been linked with both mental health difficulties and subjective wellbeing (Kraiss et al., 2020).
Indeed, trainings in social problem solving and emotion regulation are commonly used in remedial programs in schools and in the treatment of a wide variety of psychiatric disorders (Frey et al., 2000). Our results showed that perceived stress has been associated with both mental health difficulties and wellbeing. A previous systematic review has shown that stress is related to poorer quality of life by exacerbating various aspects of physical and mental health for university students (Ribeiro et al., 2017). On the other hand, the current analysis showed that high empathy was only associated with higher mental health difficulties and high prosocial behaviour was only associated with high subjective wellbeing. Previous research has shown that high levels of empathy may increase the vulnerability to mental health difficulties (O'Connor et al., 2007).
With regards to prosocial behaviour, it has been long thought that prosocial behaviours affect the wellbeing of the helper as well as the help recipient. Indeed, previous research did identify positive effects of prosocial behaviour on the wellbeing of the helper (Weinstein & Ryan, 2010).
From the factors associated with support networks, only family connection was associated with both mental health difficulties and wellbeing in the current study. Previous research has shown that parent-child relationship quality has been associated with mental health outcomes (Bukowski et al., 2006;Williams et al., 2009). Furthermore, according to the attachment theory, interactions within the family environment can enable or thwart the development of capabilities to build positive relationships elsewhere (Bowlby, 2005) which may lead to better mental health and wellbeing (Moore et al., 2018). Contrary to a previous finding (Patalay & Fitzsimons, 2016), and despite being moderately correlated to peer support and participation in home and school life, school connection was only related to mental health difficulties in the current analyses. The reason behind the difference may be due to the school connection measure used in this study being different than the one used in Patalay and Fitzsimons study. The findings from this study showed that having someone who cares about the individual, listens, and believes in them was associated with lower mental health difficulties. This is in line with the attachment theory suggesting that having a secure emotional connection to key individuals providing a base for psychological and social development (Berkman & Glass, 2000;Bowlby, 2005). Previous research also has shown that those with good school connectedness had the lowest risk of depressive symptoms (Bond et al., 2007). However, our findings indicated that high levels of peer support, community

CONCLUSIONS
In conclusion, the present study addresses the methodological limi-

CONFLICT OF INTEREST
The authors have declared that they have no competing or potential conflicts of interest.

ETHICS STATEMENT
Ethical approval was granted UCL ethics committee (reference: 8097/003) and every year, consent for participation in the research was sought from parents prior to, and from children and young people at the outset of, the survey sessions. Parental opt-outs were received by post, phone or email and child assent was recorded via computer at the beginning of survey sessions.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.