The Future Proofing Study: Design, methods and baseline characteristics of a prospective cohort study of the mental health of Australian adolescents

Abstract Objectives The Future Proofing Study (FPS) was established to examine factors associated with the onset and course of mental health conditions during adolescence. This paper describes the design, methods, and baseline characteristics of the FPS cohort. Methods The FPS is an Australian school‐based prospective cohort study with an embedded cluster‐randomized controlled trial examining the effects of digital prevention programs on mental health. Data sources include self‐report questionnaires, cognitive functioning, linkage to health and education records, and smartphone sensor data. Participants are assessed annually for 5 years. Results The baseline cohort (N = 6388, M = 13.9 years) is broadly representative of the Australian adolescent population. The clinical profile of participants is comparable to other population estimates. Overall, 15.1% of the cohort met the clinical threshold for depression, 18.6% for anxiety, 31.6% for psychological distress, and 4.9% for suicidal ideation. These rates were significantly higher in adolescents who identified as female, gender diverse, sexuality diverse, or Aboriginal and/or Torres Strait Islander (all ps < 0.05). Conclusions This paper provides current and comprehensive data about the status of adolescent mental health in Australia. The FPS cohort is expected to provide significant insights into the risk, protective, and mediating factors associated with development of mental health conditions during adolescence.


| INTRODUCTION
The global prevalence of mental health conditions in young people is significant (Shorey et al., 2022). Approximately half of mental health conditions emerge before the age of 18 years, with peak age of onset at 14.5 years (Solmi et al., 2022). Among children and adolescents, depression and anxiety disorders are common, with global prevalence rates estimated to be up to 8% for depression (Shorey et al., 2022) and 6.5% for anxiety disorders (Polancyk et al., 2015). Evidence from Australian population surveys mirror this data, with 12-month prevalence rates for depression and anxiety disorders estimated to be 8% and 7% respectively (Lawrence et al., 2015). There is a significant increase in the rate of mental health conditions and particularly depression with the transition from childhood to adolescence (Costello et al., 2011;Lawrence et al., 2015). Depression is the leading cause of disability in children and adolescents  and both depression and anxiety disorders are major drivers of disease burden in this age group (Klaufus et al., 2022). Poor mental health is also associated with suicidality and self-harm, both of which are significant issues for young people. According to Australian data, suicide is the leading cause of death in people aged 15-44 years (Australian Bureau of Statistics [ABS], 2020), and self-harm rates are increasing, particularly among young women (ABS, 2022a).
While there are effective clinical treatments for depression and anxiety disorders, they are limited in availability and do not meet the current demand, particularly for young people (Lawrence et al., 2015;Slade et al., 2009). This is in part due to the stigma associated with service use, difficulties with access, and equity of care (Lawrence et al., 2015). In order to effectively address the global disease burden of mental ill-health, there is growing recognition that prevention and early intervention approaches delivered before, or at, first onset should form part of the strategy, in addition to treatment approaches (Cuijpers et al., 2021). Research has established that prevention programs reduce symptoms and prevent incident episodes of depression and anxiety (e.g., Stockings et al., 2015;van Zoonen et al., 2014;Werner-Seidler et al., 2021). Early intervention is critical because untreated mental illness with onset during childhood or adolescence is a key predictor of illness severity and recurrence during later adolescence (Neufeld et al., 2017) and into adulthood (Korczak & Goldstein, 2009). Therefore, delivering evidence-based interventions early will not only reduce the duration and severity of the incident episode, but will also reduce the likelihood of future episodes later in life (Mueller et al., 1999).
Research to date has found small but consistent prevention effects for depression and anxiety interventions (Stockings et al., 2015;van Zoonen et al., 2014;Werner-Seidler et al., 2021). These effects are likely to be small, at least in part, because contributing risk factors and trajectories of symptom development are not yet well understood (Forsman et al., 2015;Wittchen et al., 2014). An enhanced understanding of mental health risk and protective factors will lead to the development of more precise, targeted, and effective programs for prevention and early intervention. Prospective cohort studies are a key component to building this understanding.
Existing cohort studies that investigate child and adolescent mental health have provided important insights into understanding mental health risk and protection, but many have been associated with limitations. First, the large-scale cohort studies in this area take a general focus on development and health, and accordingly, assess a broad range of outcomes for which mental health is just one, limiting the breadth, precision of measurement, and contribution of the overall study to mental health specifically (Harris, 2013;Sanson et al., 2002;Scholtens et al., 2015). For example, the Longitudinal Study of Australian Children is a study of 10,000 Australian children that collects a wide range of data including parenting, family relationships, education, and health (Sanson et al., 2002). While there is some measurement of social and emotional outcomes, these are limited in scope and do not provide the rich mental health data needed to identify mental health trajectories. A notable exception is the Adolescent Brain 2 of 19 -WERNER-SEIDLER ET AL. Cognitive Development Study in the USA, involving over 10,000 children aged 9-10 years (Garavan et al., 2018). This study has a dual focus on the neurobiological and psychological development of children through to adulthood, and includes a broad range of measures including neuroimaging, cognitive, biospecimen and individual mental and physical health factors (Karcher & Barch, 2020).
Second, of the cohort studies that do focus specifically on risk factors for adolescent mental illness, sample sizes are relatively small, ranging from 200 to 1200 participants (Beesdo-Baum et al., 2020;De la Torre-Luque et al., 2020;Ellis et al., 2017;Grootendorst-van Mil et al., 2021). This limits the potential to examine risk factors associated with subgroups of adolescents, and may not accommodate for attrition rates, diminishing the potential for appropriately powered longitudinal analyses.
The Future Proofing Study (FPS) aims to overcome these limitations by utilizing large-scale, comprehensive, and long-term data relating to adolescent mental health, thereby advancing knowledge about the factors associated with the onset and developmental trajectories of a broad range of adolescent mental health conditions. The FPS was established as a 5-year prospective cohort study with an embedded cluster-randomized controlled trial (cRCT) conducted in Australian secondary schools. Full details of the trial protocol have been published elsewhere (Werner-Seidler et al., 2020). A separate process evaluation incorporating teacher and school staff perspectives has also been conducted (Beames et al., 2021;Beames et al., submitted).
Future Proofing Study data sources include annual student selfreport questionnaires to assess mental health, wellbeing, sleep, and a wide range of other individual, environmental, and social factors.
These are complemented by smartphone-collected measures that will be examined as potential digital predictors and/or correlates of adolescent mental health and wellbeing. Smartphone-collected measures include cognitive task performance (e.g., working memory, executive function), typing and speech characteristics, ecological momentary assessment (EMA), and passive device sensor data (e.g., Global Positioning System (GPS), accelerometry, gyroscope).
Smartphone-collected data will be monitored over time to explore associations with changes in adolescent mental health. For example, emerging research has shown that changes in activity levels as detected from smartphone sensors are associated with wellbeing, however, this has not yet been examined comprehensively (Muller et al., 2020). Modeling and machine learning approaches will be used to develop reliable and valid indicators for prediction of mental illness onset and disease trajectories (Barnett et al., 2019).
Smartphone-collected data will be examined together with other data sources to determine if superior and more accurate prediction can be achieved by combining data sources, compared to using individual measures. Ultimately, this may lead to data-informed approaches for personalizing and enhancing intervention delivery.
Finally, linkage to government records relating to health and education (e.g., emergency department attendance, hospital data from birth, academic data throughout schooling) will be utilized to provide objective information about risk factors, health outcomes, service utilization, and educational outcomes associated with adolescent mental health. To date, there have not been any mental health specific cohort studies of this scale which incorporate such a broad range of different data sources concurrently into the one study (Beesdo-Baum et al., 2020;Ellis et al., 2017).
The broad questions we will address with this cohort are as follows: (i) What is the prevalence of mental health conditions in an adolescent sample, and how do these change at an individual level over time?
(ii) What individual factors are associated with the development of, and protection against, mental health symptoms and conditions for a range of disorders (e.g., depression, anxiety, behavioral, and eating disorders), self-harm, suicidal ideation, and suicidal behavior, cross-sectionally and over time?
(iii) What social and environmental factors (e.g., socioeconomic status, geographical location, school factors) are associated with the development of, and protection against, mental health symptoms and conditions cross-sectionally and over time?
(iv) Are there identifiable and distinct trajectories of mental illhealth over time, and how can they be distinguished? Do these trajectories have different risk factors, and can these trajectories be predicted from risk factors measured at baseline or early life experience?
(v) Does passively collected smartphone data or cognitive task data correlate with or predict mental illness and wellbeing, crosssectionally and over time?
(vi) What is the relationship between mental health conditions and medical history, health service utilization and academic performance, as measured by linked government records, crosssectionally and over time?
The aim of the embedded cRCT is to investigate whether two digital prevention interventions, delivered during the first 2 years of the study, prevent or reduce depression and other mental health conditions, relative to a control group. Full details can be found in the protocol (Werner-Seidler et al., 2020).
The purpose of the current paper is to provide a description and reference to the design, methods, and baseline characteristics of the FPS cohort.

| Design
The FPS is a prospective cohort study with an embedded cRCT (the trial protocol has been published elsewhere: Werner-Seidler et al., 2020). Study variables were assessed at baseline and will be collected annually for 5 years. Ethics approvals were obtained from

| Participants
All adolescents enrolled in Year 8 at participating schools were invited to take part. To participate, students required a smartphone with iOS or Android operating system and an active phone number.

| Measures
See Table 1

| Participant school and individual characteristics
Participants' school characteristics were imputed from a publicly available database curated by the Australian Curriculum, Assessment and Reporting Authority (ACARA), and included state, remoteness (e.g., major cities, inner regional, outer regional; ascertained through the Accessibility and Remoteness Index of Australia, ARIA+; Australian Institute of Health and Welfare [AIHW], 2004), school sector, school size, and Index of Community Socio-Educational Advantage (ICSEA: an index applied to all Australian schools, calculated from parent and community socio-demographic data; ACARA, 2020). Participants' individual characteristics were assessed via self-report questionnaires and included age, sex at birth, gender identity, sexuality, Aboriginal and/or Torres Strait Islander identity, country of birth, language spoken most at home, household makeup, perceived socioeconomic status, and mental health and disability diagnoses.

| Depression
The Patient Health Questionnaire for Adolescents (PHQ-A;Johnson et al., 2002) is an adaptation of the PHQ-9, a nine-item depression severity screening tool based on Diagnostic and Statistical Manual of Mental Disorders-Fourth Edition criteria (range 0-27; higher score indicates greater depression). A threshold of ≥15, reflecting moderately severe symptoms, was used to determine caseness. The internal consistency of the PHQ-A in the current study was excellent (Cronbach's α = 0.88).

| Anxiety
The Children's Anxiety Scale Short-Form (CAS-8) is an eight-item measure of anxiety, based on the Spence children's anxiety scale (CAS; Spence, 1998). The CAS-8 incorporates questions assessing generalized anxiety and social anxiety. The CAS-8 has demonstrated good reliability and provides population-level, standardized norms, with a range of 0-24 (higher score indicates greater anxiety; Spence et al., 2003;Spence et al., 2014). A threshold of ≥14 was used to determine caseness. The internal consistency of the CAS-8 was high in the current study (α = 0.88). An additional 7items from the CAS were administered to allow for calculation of the CAS Social Phobia and Generalized Anxiety subscales (each with total scores ranging 0-18).

| Psychological distress
The Distress Questionnaire-5 (DQ5; Batterham et al., 2016) is a fiveitem screening tool assessing psychological distress. The scale has strong psychometric properties, a range of 5-25 (higher score indicates greater distress), and a threshold of ≥14 as the clinical cutoff. The internal consistency of the DQ5 in the current study was good (α = 0.88).

| Suicidal ideation
The Suicidal Ideation Attributes Scale (SIDAS) is a measure of suicidal ideation severity in the past month. It has high internal consistency and good convergent validity (Van Spijker et al., 2014). The score range is 0-50, with a higher score indicating higher suicidal ideation.
A threshold of ≥21 was used to determine clinical levels of suicidal ideation. Internal consistency of the SIDAS was good in the current study (α = 0.78).
T A B L E 1 Summary of primary, secondary, and additional outcome measures, potential mediating and risk factors, and data collection timepoints     Follow-up assessments are conducted primarily at school to minimize loss at follow-up, and are scheduled for 6-weeks, 6-months, 12-months, 24-months, 36-months, 48-months, and 60-months postbaseline. At the 6-week assessment, all students received a $20 gift card to reimburse personal data costs. Linkage with government records relating to health and educational outcomes will be conducted in 2023 and 2026.

| Statistical analysis
Data quality was routinely monitored by the study Data Manager. ) for continuous outcomes where 0.01 is small, 0.06 is medium and 0.14 is large, and phi (φ) for categorical outcomes where 0.10 is small, 0.30 is medium and 0.50 is large (Cohen, 1988;Keppel, 1991).

| RESULTS
A total of 200 schools consented to participate and were randomized. However, as baseline data collection coincided with multiple COVID-19 lockdowns, a total of 66 schools withdrew from the study (see Figure 1 for reasons). There were 134 schools enrolled in the final sample, of which 77 (57.5%) were government schools and 57 (42.5%) were non-government schools. From these 134 schools, 20,533 Year 8 students were invited to participate.
Consent was obtained from 7577 parents, and baseline data were collected from a total of 6388 students. There were 3266 students randomized to the intervention arm, and 3122 students randomized to the control arm. Table 2 shows participant school and individual characteristics. Table 3 shows a comparison of the key demographics from our cohort with youth population estimates in Australia.

| Participant school and individual characteristics
Study participants were in Year 8 (M = 13.9 years; SD = 0.51).
Most students (85.6%) attended schools located in NSW, the same state as the study team. Approximately three quarters of students attended schools in major cities and 24% attended schools located in inner and outer regional areas, reflecting the spread of the Australian population (AIHW, 2021). Approximately half of participating students attended government secondary schools (50.9%), which is slightly lower than the proportion of students attending government secondary schools in the Australian population (58.9%; ACARA, 2021).
Schools attended by study participants had a mean number of 952 students and 76 full-time equivalent teaching staff, with a student to teaching staff ratio of 12.5 students to one teacher, similar to the national average (ABS, 2022b). Participants attended schools with a median ICSEA of 1048, which is slightly higher than the Australian population median of 1000 (ACARA, 2020).

For anxiety disorders, Generalized Anxiety Disorder and Social
Anxiety Disorder were most common (7.8% and 6%, respectively).
A diagnosis of Attention Deficit Hyperactivity Disorder was reported by 6.2% of the cohort. The most common disabilities were visual impairment (5.9%), autism (2.7%) and learning disability (2.1%; see Table 2).

| Participant symptoms on clinical measures, wellbeing and quality of life
Participant symptom data are presented in Participant data were collected from 2019 to 2022 so we conducted ANOVAs on the outcomes reported in Table 4 with baseline year entered as a fixed factor. The overall tests were significant, suggesting higher levels of mental health symptoms in later years.
However, the effect sizes were all η p 2 <0.003 demonstrating negligible effects as a function of baseline year.
The proportion of participants meeting clinical thresholds on primary and secondary measures (depression, anxiety, psychological distress, suicidal ideation) are presented in Table 5. Data are presented overall and broken down by school remoteness, school ICSEA, gender identity, gender diversity, sexuality diversity, Aboriginal and/ or Torres Strait Islander identity, and English as a second language.
Participants who attended school in regional areas were more likely to meet the clinical threshold for depression (p = 0.046, φ = 0.03) and anxiety (p = 0.002, φ = 0.04) relative to participants who attended school in major cities but did not significantly differ in symptom profile for psychological distress (p = 0.41) or suicidal ideation

| Additional variables
See Table 6 for descriptive data on self-harm, suicidal behavior, psychotic-like symptoms, alcohol and substance use. Data are presented overall and separated by gender identity (male, female, nonbinary).

| DISCUSSION
The FPS cohort at baseline provides the most comprehensive, up-todate Australian information about mental health during the early adolescent years since the last national survey, conducted almost a decade ago (Lawrence et al., 2015). Participant characteristics reported in this paper indicate that the FPS cohort is comparable to the Australian adolescent population in relation to gender, remoteness, Aboriginal and/or Torres Strait Islander identity, and LGBTQA + identity (see Table 3 for reference samples). To the best of our knowledge, this cohort provides the most comprehensive gender diversity prevalence data for adolescents under the age of 15 in Australia, something that has been omitted from previous studies, and similarly, furthers the field by providing detailed, large-scale prevalence data about sexuality diversity for adolescents under 15 (Fisher et al., 2019;Mission Australia, 2021;VAHI, 2020).
The FPS cohort includes a relatively high proportion of Australian-born adolescents who report speaking mostly English at home. It is well-documented that culturally and linguistically diverse populations are underrepresented in health research for a range of reasons (Hughson et al., 2016;Smith et al., 2018). In the case of mental health research, these reasons may include language and communication barriers, as well as the stigma associated with accessing mental health services (Wohler & Dantas, 2017). The FPS cohort also includes more students who attend non-government schools with a higher median ICSEA relative to students in the overall Australian population. Despite our best attempts to attract a representative sample by advertising widely, approaching all school types, and producing recruitment materials in multiple languages and formats, there is likely to be some degree of sampling bias due to the opt-in nature of the study. Government schools represent 57.5% of the schools participating in the study, however, the opt-in consent rate was higher from parents at non-government schools. Having resources available at schools to devote to study recruitment, as well as household factors and parental availability, likely contributed to these differential consent rates, which is consistent with perspectives provided by school staff on this issue (Beames et al., submitted).
Related to this point is that smartphone ownership was necessary for study inclusion, raising the possibility of sampling bias. At study commencement, estimates suggest 90% of Australian teenagers owned a smartphone (Nielsen, 2019), providing some parameters around the extent of this sampling bias risk.
Key findings from our initial baseline analyses showed that female and nonbinary adolescents had elevated symptom levels across all clinical measures relative to their male counterparts with effect sizes ranging from small (e.g., externalizing symptoms) to large (e.g., depression, anxiety, distress), with nonbinary participants faring worse than their female peers. Participants attending school in regional areas were more likely to reach clinical thresholds for depression and anxiety, and those attending less advantaged schools were more likely to report symptoms above clinical thresholds for depression, anxiety, and psychological distress, although the size of T A B L E 5 Proportion of cohort meeting clinical cut-offs for depression, anxiety, psychological distress, and suicidal ideation overall and by demographic characteristics The main strengths of the data reported in this study include the comprehensiveness of the self-report information across a range of mental health and demographic domains, the representativeness of the sample, and the reporting of current prevalence estimates for common mental health symptoms and suicidality among Australian adolescents. There were significant mental health symptom levels reported in this cohort, with higher rates for some population groups (e.g., gender and sexuality diverse adolescents). The FPS will continue T A B L E 6 Self-harm, suicidal behavior, psychotic-like symptoms, alcohol and substance use in overall cohort and in male, female, and nonbinary participants Note: Data are n and %. To protect the identities of individuals, values less than 5 are censored and reported as '<5'. Data are presented by gender identity.
-15 of 19 to monitor prevalence levels in these high-risk groups, as well as the whole population, and will provide evidence about how mental health morbidity changes over time in adolescents.
The FPS cohort will advance knowledge by leveraging a broad range of data sources (e.g., self-report, cognitive task performance, passively collected sensor data, government health and education records) to create a rich and detailed picture of the development of adolescent mental health conditions. Moreover, the large FPS sample will provide sound statistical power for future studies to examine predictors and mediators of mental health outcomes, and the ability to use advanced statistical approaches (e.g., latent class analysis, growth curve modeling) to identify subgroups of adolescents and individual trajectories of mental health and wellbeing. Data from the FPS cohort are likely to inform the targets of prevention and early intervention strategies well into the future, providing focus for more personalized and precision-based approaches to mental health. While not the focus of the current paper, the embedded cRCT is one of the largest of its kind and will be able to answer questions about the effectiveness of digital prevention approaches, at scale (Werner-Seidler et al., 2020).
There are several limitations associated with the FPS that need to be considered. First, using an opt-in school, parent, and participant consent process likely led to some sampling biases which included a slightly more advantaged and less linguistically diverse sample relative to the general Australian population. Nonetheless, the study still reflects the Australian population with respect to gender, remoteness, and other diversity characteristics. Relatedly, adolescents who were not in formal schooling were not included in this cohort study.
Second, by coincidence, baseline data were collected during 2019-2022. From 2020 to 2021, many regions of Australia were subject to varying levels of COVID-19 related lockdowns and restrictions. It is within this context that the baseline data need to be considered.
While a potential limitation given this broader context, this timeframe provides a unique opportunity to assess how young people fare over the long term after such significant disruption. Finally, the embedded cRCT may be a limitation of the cohort study if intervention effects on mental health outcomes are detected. If this is the case, the control arm (n = 3122) will comprise the sample to answer our core cohort study research questions. Although a limitation to the pure cohort study, if effects of the interventions are detected, there will be an opportunity to examine potential interaction of risk and protective factors on intervention effects.
The FPS cohort is expected to provide a major contribution to understanding different aspects of development, experience, and behavior as risk and/or protective factors for mental ill-health and wellbeing during the adolescent years.

ACKNOWLEDGMENTS
We would like to offer our heartfelt thanks to the wonderful young people, their families and school communities who are participating in this study. This study would not have been possible without the enthusiasm, dedication and commitment from the wellbeing staff from our partner schools, including deputy principals, year advisors, school counselors, and student support officers among many others.
A study of this scale also would not have been possible without the