Early Onset of Distress Disorders and High-School Dropout: Prospective Evidence From a National Cohort of Australian Adolescents

Early Onset of Distress Disorders and High-School Dropout: Prospective Evidence From a National... Abstract Prior research examining whether depression and anxiety lead to high-school dropout has been limited by a reliance on retrospective reports, the assessment of mental health at a single point in time (often remote from the time of high-school exit), and the omission of important measures of the social and familial environment. The present study addressed these limitations by analyzing 8 waves of longitudinal data from a cohort of Australian adolescents (n = 1,057) in the Household, Income and Labor Dynamics in Australia (HILDA) Survey (2001–2008). Respondents were followed from the age of 15 years through completion of or exit from high school. Discrete-time survival analysis was used to assess whether the early experience of a distress disorder (indicated by scores <50 on the 5-item Mental Health Inventory from the Short Form Health Survey) predicted subsequent high-school dropout, after controlling for household and parental socioeconomic characteristics and for tobacco smoking and alcohol consumption. Adolescents with a prior distress disorder had twice the odds of high-school dropout compared with those without (odds ratio = 1.99, 95% confidence interval: 1.24, 3.17). This association was somewhat attenuated but remained significant in models including tobacco and alcohol consumption (odds ratio = 1.74, 95% confidence interval: 1.74; 1.09, 2.78). These results suggest that improving the mental health of high-school students may promote better educational outcomes. adolescence, education, high-school incompletion, longitudinal, mental health Educational attainment is a critical determinant of adult life opportunities, both at the individual level in terms of work, health, and financial circumstances and in a broader societal context in terms of demand on social welfare entitlements and workforce development (1). Data from across countries in the Organisation for Economic Co-operation and Development (OECD) shows a trend of increasing levels of school retention, including high-school completion (2). Nonetheless, Australian data from the mid-2010s shows that approximately 26% of 19 year olds did not have high-school (or equivalent) qualifications (3, 4). The reasons for high-school incompletion are varied and include poor academic performance, absence from school, and a lack of socioeconomic resources (4). One likely further important antecedent for high-school incompletion is early-onset psychiatric illness (5). Like high-school incompletion, psychiatric illness has been shown to be associated with unemployment, poor job quality, low income, financial hardship, and poor housing (6–9). The mechanisms via which early-onset psychiatric illness might affect school incompletion include the disruption of academic achievement (e.g., via impaired cognitive function), disrupted behavior, and social responses from teachers and parents who prejudge (and inadvertently limit) student ability (10). Identifying the contribution of early-onset psychiatric illness to high-school incompletion is necessary to inform effective, targeted prevention and intervention policies. The present study uses longitudinal, prospective data to investigate the specific contribution of early onset of symptoms of anxiety and depression. We use the term distress disorder (DD) to describe likely clinical levels of depression and (generalized) anxiety symptoms, reflecting a transdiagnostic approach that views depression and anxiety as markers of an underlying internalizing or “distress” factor (11–13). Links between early-onset mental disorders and subsequent high-school incompletion have been investigated predominantly by using retrospective data on age of onset. Research from the United States, Australia, and South Africa has found that early-onset disorders such as anxiety and mood disorders, conduct disorder, and substance-use disorders are associated with high-school incompletion (14–17). The reliance on retrospective data is, however, a major limitation (18, 19), with retrospective data vulnerable to recall bias and imprecise for capturing the timing between disorder onset and high-school dropout. A recent systematic review by Melkevik et al. (19) focused on the association of anxiety and depression with high-school incompletion and demonstrated this overreliance on retrospective data, showing that only 4 of 16 studies employed a prospective design (20–23). Closer examination of these prospective studies also revealed 2 key limitations: 1) long time lags between assessments of symptomatology and school outcomes, and 2) inadequate adjustment for socioeconomic confounders and indicators of comorbid risk-taking/externalizing behaviors. For example, a prospective study conducted by Fletcher et al. (21) used a sibling fixed-effects methodology to control for potentially relevant but unmeasured family and neighborhood characteristics common to both siblings (e.g., family resources, neighborhood crime). Analyses of data from approximately 2,400 adolescents found that greater depression was associated with high-school dropout after adjusting for family and neighborhood indicators; however, the association was no longer significant after controlling for risky behaviors (e.g., substance use). The study used a nationally representative sample of US students in grades 7–12, initially interviewed in 1994, when the Center for Epidemiologic Studies Depression Scale (CES-D) was administered to them. Educational outcomes were assessed 6 years later, when respondents were (on average) 22 years of age. Given that depression is episodic in nature, the long time lag between assessments might have underrepresented episodes of depression. A long time lag also makes it difficult to link experiences of DDs directly with high-school dropout, because it increases the opportunity for cumulative confounding factors, such as adverse family circumstances, to play a substantive role. In another example, Fergusson and Woodward (20) analyzed data from 1,265 adolescents in New Zealand and found that depression was associated with early school exit but not after adjustment for socioeconomic status (maternal education and family socioeconomic status at birth) and indicators of comorbid externalizing problems (conduct disorder, attention deficit hyperactivity disorder (ADHD), smoking, and alcohol use). In this study, the assessment of mental health and educational outcomes was more proximal. Adolescents were categorized as having major depression between the ages of 14 and 16 years (interviewed at ages 15 and 16 years), and educational outcomes were measured at ages 16, 18, and 21 years. While the analysis adjusted for maternal educational achievement and family socioeconomic status at the time of the child’s birth, more recent and time-varying measures of socioeconomic circumstances were not considered. The present study used a broadly representative, national household survey to prospectively follow a large cohort (n = 1,057) of Australian adolescents across their final years of high school (annual assessment) to assess how the experience of a DD (representing clinical levels of depression and anxiety symptoms) might influence high-school completion. Uniquely, the longitudinal household-study design allowed us to overcome many of the methodological limitations of previous research. We controlled for a comprehensive range of potential confounders over time, including household- and parent-level factors for both mothers and fathers. Because we assessed confounders annually, the analyses considered changing socioeconomic and family circumstances, accounting for both episodic and entrenched socioeconomic disadvantage. We also adjusted for smoking and alcohol use. These risky behaviors are known to be associated with truancy and high-school incompletion and are highly associated with externalizing problems and other forms of substance use. METHODS Sample Data were from the first eight waves of the Household, Income and Labor Dynamics in Australia (HILDA) Survey (release 14.0). Details of the survey methods are reported elsewhere (24). Briefly, HILDA is a nationally representative longitudinal household panel survey with a multistage sampling design. Interviews have been conducted annually since 2001 with each household member aged ≥15 years. The survey composition is dynamic, with sample loss arising from death or attrition and new respondents entering the sample each wave (e.g., adolescents within a household are eligible to participate when they reach 15 years of age). The HILDA survey was approved by the Human Research Ethics Committee at the University of Melbourne. A unique subset of data was created that included all 15 and 16 year olds who entered the HILDA survey during the first 4 waves, including their first 5 years of data. Data for all respondents was centered on their first wave of participation (considered year 1 regardless of when they entered the survey). This identified a sample of 1,256 participants aged 15 and 16 years with data at year 1. However, 100 participants had already left school at their first assessment occasion (74% of these were aged 16 years), and a further 99 did not participate in any subsequent waves during the follow-up period, leaving 1,057 respondents. A series of univariate analyses showed that residing in a low-socioeconomic-status area, having no mother in the household, and mother’s low educational attainment were the only factors associated with attrition. Table 1 shows respondent high-school completion status over time. Over the study, 673 respondents completed high school (64%) and 295 (28%) exited high school without graduating. The outcome for 89 respondents (8%) was not resolved (6 were still attending school at the final wave). Overall, from year 2 to year 5 (when outcomes were assessed), excluding occasions on which respondents did not participate, there were 2,483 observations. Table 1. High-School Completion Status Over 5 Time Points, Household, Income and Labor Dynamics in Australia Survey, 2001–2008 Year  Still in High School  Completed High Schoola  Early Exita  Did Not Participate in Wave  1  1,057        2  857  24  121  55  3  540  201  219  97  4  150  567  287  59  5  6  673  295  83  Year  Still in High School  Completed High Schoola  Early Exita  Did Not Participate in Wave  1  1,057        2  857  24  121  55  3  540  201  219  97  4  150  567  287  59  5  6  673  295  83  a Cumulative frequencies. Table 1. High-School Completion Status Over 5 Time Points, Household, Income and Labor Dynamics in Australia Survey, 2001–2008 Year  Still in High School  Completed High Schoola  Early Exita  Did Not Participate in Wave  1  1,057        2  857  24  121  55  3  540  201  219  97  4  150  567  287  59  5  6  673  295  83  Year  Still in High School  Completed High Schoola  Early Exita  Did Not Participate in Wave  1  1,057        2  857  24  121  55  3  540  201  219  97  4  150  567  287  59  5  6  673  295  83  a Cumulative frequencies. Finally, while item missingness was minimal, it did lead to some further data reduction. In the main model, incorporating a range of individual and household variables (not considering parental characteristics) resulted in missing data for 68 respondents and 253 observations. Most of this missingness was associated with absence of the lagged distress measure, with 27% of these occasions resulting from failure to complete or return the self-completion questionnaire in which mental health was assessed. This missingness was addressed through multiple imputation methods (25). Measures DDs were assessed at all time points using the 5-item Mental Health Inventory (MHI-5), a subscale from the Short Form Health Survey (SF-36) (26). The MHI-5 has been validated against Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Axis 1 diagnosis; has been recommended for screening for anxiety and depressive disorders; and has been used as a measure of likely anxiety and/or depression in epidemiologic surveys (27, 28). The MHI-5 assesses symptoms of depression and anxiety, as well as positive aspects of mental health, in the previous 4 weeks. Scores range from 0 to 100, with higher scores representing better mental health. The present analyses dichotomized the MHI-5, with scores of less than 50 categorized as having a DD. This cutpoint was chosen on the basis of previous research validating the MHI-5 against diagnosis of anxiety and depression disorders assessed using the Composite International Diagnostic Interview (29). Preliminary analyses using the HILDA data confirmed the validity of the MHI-5 as an indicator of distress in comparison with the Kessler Psychological Distress (K-10) scale and that the 5 MHI items represent a single “distress” factor (see Web Appendix 1, available at https://academic.oup.com/aje). High-school completion status At each year, early school exit was coded as a binary variable, contrasting those who were still at school or had just completed high school (0) with those who had left prior to high-school completion (1). At each year, respondents were considered out of scope if at their previous interview (year) they reported having completed or left school. In Australia, education policy is generally the responsibility of the 6 states and 2 territories, resulting in some differences in how education is structured across different jurisdictions. However, the general progression through school involves a year of preparatory school at age 5 years (kindergarten), followed by 6 years of primary school, and 6 years of high (or secondary) school (ending in 12th grade). Attending school is compulsory until the end of grade 10 (15 or 16 years old), with some minor variations between states (30). Covariates Covariates included basic demographics at baseline (sex, age). Household indictors included a measure of region/remoteness, housing tenure (renting vs. other), residence in a socioeconomically disadvantaged area (lowest 30% of Socio-Economic Index for Areas (SEIFA)), and low level of household income (bottom quintile). A number of indicators of parents’ circumstances were included for mothers and fathers: presence in household (either not in household or nonresponder), high-school incompletion, postschool qualifications, receiving income support payments, employed versus unemployed, and employed (currently or previously) in a low-skill occupation. Binary variables representing adolescents’ current smoking status (yes = 1) and alcohol consumption (at least weekly = 1) were included. “Time in study” (years) was also included. Analyses Descriptive statistics from the baseline sample were initially calculated. A series of discrete-time survival analysis models examined the association of prior (12-month lagged) DD (and lagged covariates) with an indicator of early school exit. Discrete-time survival analysis is appropriate given the annual data collection. Preliminary models over the 4-year follow-up period showed that time to early school exit (or censor) was appropriately represented by a linear term, with each additional year at school decreasing the odds of early exit by 18%. A standard approach to model building was employed with variables with a univariable P ≤ 0.2 included in a preliminary multivariable model. Backward elimination was then used to remove variables that no longer met these criteria without leading to poorer model fit (evaluated by comparing log-likelihood statistics) or altering the coefficient of the key covariate by 15% (model A). Two key potential confounders of the association between DD and early school exit (lagged smoking and alcohol consumption) were added at a final model step (model B). Potential sex differences were evaluated by the inclusion of an interaction term (given that sex was identified as a potentially important mediator by Melkevik et al. (19)). The main analyses were weighted by the responding person sample weights included in each wave of HILDA to adjust for nonresponse/selection (see Watson (31)), and multiple imputation (using chained equations to generate 20 imputed data sets based on 10 cycles) was used to address potential bias due to missing data on covariates. A series of further sensitivity analyses were conducted (see Web Appendix 1) to address potential study limitations: excluding respondents who completed high school in the current year, calculating inverse probability weights to adjust for attrition, using random effects logistic regression models, limiting analysis to respondents who entered the study at age 15 years, and testing the robustness of MHI-5 scale. RESULTS Descriptive analyses The baseline (year 1) characteristics of the analysis sample are presented in Table 2. Overall, adolescents who subsequently left school early were significantly more likely to live in rental housing, to not live with their mother, and to report DD, current smoking, and regular (weekly or more frequent) alcohol consumption (based on χ2 tests of association). Preliminary analysis showed that a model including a measure of time-varying DD was superior to a model comprising only baseline DD (log likelihood χ2 = 7.93, degrees of freedom = 1; P = 0.005). Table 2. Characteristics of the Sample at Baseline, Household, Income and Labor Dynamics in Australia Survey, 2001 Variable  Overall (987 Persons and 2,227 Observations), % a  Completed School or Censored (762 Persons and 1,805 Observations), %b  Early Exit (295 Persons and 722 Observations), %c  Age, years         15  66  65  68   16  34  35  32  Female sex  50  52  45  Low household income  16  14  20  Reside in disadvantaged aread  31  28  40  Housing tenure, renter  20  17  27  No father in household  20  19  23  No mother in household  4  4  5  Father did not complete high school  18  17  20  Mother did not complete high school  37  33  49  Father is current welfare recipient  7  6  11  Mother is current welfare recipient  19  17  27  Distress disorder (MHI-5 <50)  8  7  12  Current smoker  6  3  15  Regular alcoholic consumption (weekly or more frequent)  5  3  12  Variable  Overall (987 Persons and 2,227 Observations), % a  Completed School or Censored (762 Persons and 1,805 Observations), %b  Early Exit (295 Persons and 722 Observations), %c  Age, years         15  66  65  68   16  34  35  32  Female sex  50  52  45  Low household income  16  14  20  Reside in disadvantaged aread  31  28  40  Housing tenure, renter  20  17  27  No father in household  20  19  23  No mother in household  4  4  5  Father did not complete high school  18  17  20  Mother did not complete high school  37  33  49  Father is current welfare recipient  7  6  11  Mother is current welfare recipient  19  17  27  Distress disorder (MHI-5 <50)  8  7  12  Current smoker  6  3  15  Regular alcoholic consumption (weekly or more frequent)  5  3  12  Abbreviation: MHI-5, 5-item Mental Health Inventory. a Observations per person: 2.3 (range, 1–4). b Observations per person: 2.5 (range, 1–4). c Observations per person: 1.6 (range, 1–4). d Lowest 30% on Socio-Economic Index for Areas of relative advantage/disadvantage. Table 2. Characteristics of the Sample at Baseline, Household, Income and Labor Dynamics in Australia Survey, 2001 Variable  Overall (987 Persons and 2,227 Observations), % a  Completed School or Censored (762 Persons and 1,805 Observations), %b  Early Exit (295 Persons and 722 Observations), %c  Age, years         15  66  65  68   16  34  35  32  Female sex  50  52  45  Low household income  16  14  20  Reside in disadvantaged aread  31  28  40  Housing tenure, renter  20  17  27  No father in household  20  19  23  No mother in household  4  4  5  Father did not complete high school  18  17  20  Mother did not complete high school  37  33  49  Father is current welfare recipient  7  6  11  Mother is current welfare recipient  19  17  27  Distress disorder (MHI-5 <50)  8  7  12  Current smoker  6  3  15  Regular alcoholic consumption (weekly or more frequent)  5  3  12  Variable  Overall (987 Persons and 2,227 Observations), % a  Completed School or Censored (762 Persons and 1,805 Observations), %b  Early Exit (295 Persons and 722 Observations), %c  Age, years         15  66  65  68   16  34  35  32  Female sex  50  52  45  Low household income  16  14  20  Reside in disadvantaged aread  31  28  40  Housing tenure, renter  20  17  27  No father in household  20  19  23  No mother in household  4  4  5  Father did not complete high school  18  17  20  Mother did not complete high school  37  33  49  Father is current welfare recipient  7  6  11  Mother is current welfare recipient  19  17  27  Distress disorder (MHI-5 <50)  8  7  12  Current smoker  6  3  15  Regular alcoholic consumption (weekly or more frequent)  5  3  12  Abbreviation: MHI-5, 5-item Mental Health Inventory. a Observations per person: 2.3 (range, 1–4). b Observations per person: 2.5 (range, 1–4). c Observations per person: 1.6 (range, 1–4). d Lowest 30% on Socio-Economic Index for Areas of relative advantage/disadvantage. Discrete-time survival analyses Initial univariable discrete-time survival models showed that all covariates except baseline age and parental employment characteristics (workforce status and occupational skill level) were associated with early school exit with P ≤ 0.2 and were, therefore, included in the early phase of multivariate model building. The first model (model A; Table 3), having deleted covariates no longer associated with the outcome, showed that adolescents who reported DD had twice the odds of early school exit in the following year in comparison with those without DD (odds ratio = 1.99), having controlled for sociodemographic characteristics, economic circumstances, and parental characteristics. Table 3. Discrete-Time Survival Models Examining Dropout Prior to High-School Completion, Household, Income and Labor Dynamics in Australia Survey Sample, 2001–2008a Variable  Model A  Model Bb  OR  95% CI  OR  95% CI  Prior distress disorder (MHI-5 <50)  1.99  1.24, 3.17  1.74  1.09, 2.78  Prior smoking      3.01  1.88, 4.80  Prior alcohol consumption      1.52  0.92, 2.50  Female sex  0.68  0.51, 0.91  0.70  0.52, 0.94  Time, years  0.97  0.83, 1.15  0.95  0.80, 1.13  Rental housing  1.79  1.27, 2.54  1.63  1.15, 2.32  Father status (referent: present)           Not present in household  1.56  1.08, 2.25  1.47  1.01, 2.14   Nonresponder  1.48  0.69, 3.12  1.49  0.67, 3.04  Mother left school early  1.77  1.28, 2.43  1.69  1.23, 2.33  Father is current welfare recipient  1.30  0.73, 2.31  1.34  0.75, 2.41  Variable  Model A  Model Bb  OR  95% CI  OR  95% CI  Prior distress disorder (MHI-5 <50)  1.99  1.24, 3.17  1.74  1.09, 2.78  Prior smoking      3.01  1.88, 4.80  Prior alcohol consumption      1.52  0.92, 2.50  Female sex  0.68  0.51, 0.91  0.70  0.52, 0.94  Time, years  0.97  0.83, 1.15  0.95  0.80, 1.13  Rental housing  1.79  1.27, 2.54  1.63  1.15, 2.32  Father status (referent: present)           Not present in household  1.56  1.08, 2.25  1.47  1.01, 2.14   Nonresponder  1.48  0.69, 3.12  1.49  0.67, 3.04  Mother left school early  1.77  1.28, 2.43  1.69  1.23, 2.33  Father is current welfare recipient  1.30  0.73, 2.31  1.34  0.75, 2.41  Abbreviations: CI, confidence interval; MHI-5, 5-item Mental Health Inventory; OR, odds ratio. a Analyses weighted by the responding person sample weights included in each wave of the Household, Income and Labor Dynamics in Australia survey. Analyses addressed missing data through multiple imputation methods (see Web Appendix 1 for details). b Model B extends model A by incorporating measures of smoking status and regular alcohol consumption. Table 3. Discrete-Time Survival Models Examining Dropout Prior to High-School Completion, Household, Income and Labor Dynamics in Australia Survey Sample, 2001–2008a Variable  Model A  Model Bb  OR  95% CI  OR  95% CI  Prior distress disorder (MHI-5 <50)  1.99  1.24, 3.17  1.74  1.09, 2.78  Prior smoking      3.01  1.88, 4.80  Prior alcohol consumption      1.52  0.92, 2.50  Female sex  0.68  0.51, 0.91  0.70  0.52, 0.94  Time, years  0.97  0.83, 1.15  0.95  0.80, 1.13  Rental housing  1.79  1.27, 2.54  1.63  1.15, 2.32  Father status (referent: present)           Not present in household  1.56  1.08, 2.25  1.47  1.01, 2.14   Nonresponder  1.48  0.69, 3.12  1.49  0.67, 3.04  Mother left school early  1.77  1.28, 2.43  1.69  1.23, 2.33  Father is current welfare recipient  1.30  0.73, 2.31  1.34  0.75, 2.41  Variable  Model A  Model Bb  OR  95% CI  OR  95% CI  Prior distress disorder (MHI-5 <50)  1.99  1.24, 3.17  1.74  1.09, 2.78  Prior smoking      3.01  1.88, 4.80  Prior alcohol consumption      1.52  0.92, 2.50  Female sex  0.68  0.51, 0.91  0.70  0.52, 0.94  Time, years  0.97  0.83, 1.15  0.95  0.80, 1.13  Rental housing  1.79  1.27, 2.54  1.63  1.15, 2.32  Father status (referent: present)           Not present in household  1.56  1.08, 2.25  1.47  1.01, 2.14   Nonresponder  1.48  0.69, 3.12  1.49  0.67, 3.04  Mother left school early  1.77  1.28, 2.43  1.69  1.23, 2.33  Father is current welfare recipient  1.30  0.73, 2.31  1.34  0.75, 2.41  Abbreviations: CI, confidence interval; MHI-5, 5-item Mental Health Inventory; OR, odds ratio. a Analyses weighted by the responding person sample weights included in each wave of the Household, Income and Labor Dynamics in Australia survey. Analyses addressed missing data through multiple imputation methods (see Web Appendix 1 for details). b Model B extends model A by incorporating measures of smoking status and regular alcohol consumption. The final model (model B) incorporated the measures of adolescent substance use (lagged smoking status and alcohol consumption). While smoking was associated with increased odds of early school exit, DD remained significant, with respondents with DD having 74% greater odds of leaving school early the following year compared with those without DD. Being male, living in rental housing, having no father in the household and having a mother who left school early were also characteristics associated with early school exit. On the basis of this model, we consider the likelihood of early school exit for an adolescent boy, who does not smoke or regularly drink alcohol, who lives in rental accommodation with his father (who receives income support payments), and whose mother did not complete high school. While a boy in these circumstances without DD had a likelihood of 22% (95% confidence interval: 11.5, 33.3) of early school exit, the likelihood of early school exit for an adolescent in the same circumstances but with DD was 34% (95% confidence interval: 17.8, 49.8). Sensitivity analyses confirmed the robustness of the results (see Web Appendix 1). The inclusion of a term for interaction between sex and DD did not improve model fit (log likelihood χ2 = 2.59, degrees of freedom = 1; P = 0.107), although coefficients were greater for girls than for boys. An initial model that did not adjust for nonresponse, attrition, and missing data (see Web Table 1) produced similar odds ratios for the association between DD and early school exit, as did models using the continuous MHI scale (see Web Table 2) and models testing the MHI items for depression and anxiety separately (see Web Table 3). The exclusion of adolescents who completed high school in the target year (see Web Table 4) and random effects models (see Web Table 5) produced similar (or somewhat stronger) odds ratios for the association between DD and early school exit. Restricting the analysis to those initially aged 15 years produced slightly weaker coefficients (see Web Table 6). Supplementary analyses S4–S6 outline the details of various models aimed at reducing bias (see Web Tables 7–10). DISCUSSION The findings from this prospective study provide strong evidence that adolescents experiencing a DD (depression/anxiety symptoms) are at greater risk of high-school dropout than are those who are not. The relationship between DD and subsequent high-school dropout remained significant after taking account of a broad range of socioeconomic circumstances (including household- and parent-level factors) as well as adolescents’ alcohol and tobacco use. In the context of previous research, while several retrospective studies have similarly pointed to a strong link between early-onset mental health disorders and high-school incompletion (14–17), the findings of prospective research focused on anxiety and depression have varied depending on the lag between interview dates and the sociodemographic and health behavior covariates (19–21). The present study represents a significant advance on previous research in that the models controlled for household and parental socioeconomic circumstances (as they vary over time), both of which are strongly related to high-school completion. The present study is also unusual in its application of multiple waves of data, spaced annually. The repeated measurement of DD and other covariates over time enabled consideration of the proximal association between experiences of DD in adolescence and subsequent high-school dropout. We showed, for example, that a measure of lagged DD was a significantly better predictor of high-school dropout than a time-invariant measure of DD at age 15 years, demonstrating the importance of multiple-wave data on mental health as opposed to baseline data alone. The importance of anxiety and depression for subsequent high-school incompletion was evident in the finding that the association endures after taking account of tobacco and alcohol use, which are each strongly and independently associated with high-school dropout. (We note that this inclusion may represent an overcorrection because substance use could be a response to the experience of anxiety and depression symptoms). This study adds weight to the hypothesis that early-onset DD disrupts the opportunity for adolescents to successfully attain important social and economic milestones, such as high-school completion. However, there are other important indicators of educational disruption, such as school absence (due to poor psychological health or truancy), poor student engagement, and poor academic achievement (32–34). High-school dropout likely represents the endpoint in a series of disruptions in relational and academic growth that are influenced by, and in turn influence, anxiety and depression symptoms. While high-school incompletion is the endpoint in this investigation, it is important to consider postschool transitions such as entering vocational training, university education, and employment. Despite disrupted schooling, some individuals will find new pathways to further education and employment, while others will be unable to do so (5, 35). Given that roughly half of all mental disorders have their onset during adolescence (mid-teens) (36), there is great scope to more robustly delineate the impacts of early DD on educational and socioeconomic opportunities and to generate new data on the personal, social, and economic benefits of early intervention. A number of limitations should be noted. Diagnostic measures of mental health (i.e., mental disorders) were not available. However, the MHI-5 score is a valid indicator of depression and anxiety in the general community (29). The HILDA survey included no information on the school environment, academic achievement, or school absence, nor did it include psychometric measures of externalizing or conduct behaviors. Australian data has shown that high-school completion is strongly associated with academic performance (4) and with externalizing behaviors (20). It may be that DDs are a consequence of these underlying factors. However, we adjusted for a range of sociodemographic and substance-use measures that would also likely be associated with these academic and behavioral factors. This study also did not consider different school settings, such as contrasting private or public schools. The present results showed a relatively high rate of high-school noncompletion (28%), but this is consistent with census and other data on high-school completion published by the Australian Bureau of Statistics and other reputable sources (37, 38). Because the HILDA survey is restricted to respondents aged 15 years or older, we did not capture those who left school very early. However, analysis restricted to 15 year olds eliminated most of this missingness and produced the same pattern of results. It is also important to recognize that, with time and perhaps the resolution of their DD, some in their late teens or early 20s without high-school qualifications may return to education. The present study, by design, did not consider return to study or longer-term outcomes. This is an important direction for further research. Similarly, while not significant in this study, sex differences in the association between DD and school dropout warrant further consideration. Despite these limitations, this study provides unique longitudinal data on the mental health and social circumstances of adolescents and their parents/households. The range of sensitivity tests conducted (applying different methods, measures, weights and multiple imputation) demonstrate the robustness of the findings. In conclusion, the present study followed a large cohort of Australian adolescents across the final years of high school. Experiencing clinical levels of depression and anxiety symptoms during early adolescence was associated with increased risk of high-school dropout. The findings support calls for mental illness prevention and early intervention programs targeting adolescents and the school environment. In the Australian setting, such services and programs are becoming common (e.g., Headspace, beyondblue schools program, and MindMatters). The evaluation of these programs should consider social and economic outcomes, such as reduction in high-school dropout. The present findings highlight the need to provide and evaluate mental health support services for young people who have left the education system to enhance their chances of returning to education or gaining meaningful employment. There may be broad benefits of such service and policy interventions. ACKNOWLEDGMENTS Author affiliations: Centre for Mental Health, Melbourne School of Population and Global Health, University of Melbourne, Melbourne, Victoria, Australia (Peter Butterworth); Melbourne Institute of Applied Economic and Social Research, University of Melbourne, Melbourne, Victoria, Australia (Peter Butterworth); and National Centre for Epidemiology and Population Health, Research School of Population Health, College of Medicine, Australian National University, Canberra, Australia (Liana S. Leach). The Household, Income and Labour Dynamics in Australia Project was initiated and is funded by the Australian Government Department of Social Services and is managed by the Melbourne Institute of Applied Economic and Social Research. This work was supported by Australian Research Council Future Fellowship (grant FT13101444), National Health and Medical Research Council Early Career Fellowship (grant 1035803), and a research grant from Australian Rotary Health. The findings and views reported in this paper are those of the authors and should not be attributed to either the Australian Department of Social Services or the Melbourne Institute. There have been no involvements that might raise the question of bias in the work reported or in the conclusions, implications, or opinions stated. Conflict of interest: none declared. Abbreviations DD distress disorder HILDA Household, Income and Labor Dynamics in Australia Survey MHI-5 5-item Mental Health Inventory REFERENCES 1 Buchmann C, Hannum E. Education and stratification in developing countries: a review of theories and research. Annu Rev Sociol . 2001; 27: 77– 102. Google Scholar CrossRef Search ADS   2 Organisation for Economic Cooperation and Development (OECD). Education at a Glance 2015: OECD Indicators. Paris, France: OECD Publishing; 2015. http://www.oecd.org/edu/education-at-a-glance-2015.htm. Published November 24, 2015. Accessed October 5, 2017. 3 Australian Institute of Health and Welfare (AIHW). Australia’s welfare 2015. (Catalogue no. AUS189). Canberra, Australia: AIHW; 2015. https://www.aihw.gov.au/reports/australias-welfare/australias-welfare-2015. Released August 20, 2015. Accessed October 5, 2017. 4 Lamb S, Jackson J, Walstab A, et al.  . Educational opportunity in Australia 2015: Who succeeds and who misses out. Melbourne, Australia: Centre for International Research on Education Systems, for the Mitchell Institute; 2015. http://www.mitchellinstitute.org.au/reports/educational-opportunity-in-australia-2015-who-succeeds-and-who-misses-out/. Published October 26, 2015. Accessed October 5, 2017. 5 Freudenberg N, Ruglis J. Reframing school dropout as a public health issue. Prev Chronic Dis . 2007; 4( 4): A107. Google Scholar PubMed  6 Butterworth P, Olesen SC, Leach LS. The role of hardship in the association between socio-economic position and depression. Aust N Z J Psychiatry . 2012; 46( 4): 364– 373. Google Scholar CrossRef Search ADS PubMed  7 Evans GW, Wells NM, Moch A. Housing and mental health: a review of the evidence and a methodological and conceptual critique. J Soc Issues . 2003; 59( 3): 475– 500. Google Scholar CrossRef Search ADS   8 Ezzy D. Unemployment and mental health: a critical review. Soc Sci Med . 1993; 37( 1): 41– 52. Google Scholar CrossRef Search ADS PubMed  9 Stansfeld S, Candy B. Psychosocial work environment and mental health—a meta-analytic review. Scand J Work Environ Health . 2006; 32( 6): 443– 462. Google Scholar CrossRef Search ADS PubMed  10 Mcleod JD, Fettes DL. Trajectories of failure: the educational careers of children with mental health problems. AJS . 2007; 113( 3): 653– 701. Google Scholar PubMed  11 Krueger RF, Caspi A, Moffitt TE, et al.  . The structure and stability of common mental disorders (DSM-III-R): a longitudinal-epidemiological study. J Abnorm Psychol . 1998; 107( 2): 216– 227. Google Scholar CrossRef Search ADS PubMed  12 Krueger RF, Eaton NR. Transdiagnostic factors of mental disorders. World Psychiatry . 2015; 14( 1): 27– 29. Google Scholar CrossRef Search ADS PubMed  13 Watson D. Differentiating the mood and anxiety disorders: a quadripartite model. Annu Rev Clin Psychol . 2009; 5: 221– 247. Google Scholar CrossRef Search ADS PubMed  14 Breslau J, Lane M, Sampson N, et al.  . Mental disorders and subsequent educational attainment in a US national sample. J Psychiatr Res . 2008; 42( 9): 708– 716. Google Scholar CrossRef Search ADS PubMed  15 Kessler RC, Foster CL, Saunders WB, et al.  . Social consequences of psychiatric disorders, I: educational attainment. Am J Psychiatry . 1995; 152( 7): 1026– 1032. Google Scholar CrossRef Search ADS PubMed  16 Leach LS, Butterworth P. The effect of early onset common mental disorders on educational attainment in Australia. Psychiatry Res . 2012; 199( 1): 51– 57. Google Scholar CrossRef Search ADS PubMed  17 Myer L, Stein DJ, Jackson PB, et al.  . Impact of common mental disorders during childhood and adolescence on secondary school completion. S Afr Med J . 2009; 99( 5 Pt 2): 354– 356. Google Scholar PubMed  18 Esch P, Bocquet V, Pull C, et al.  . The downward spiral of mental disorders and educational attainment: a systematic review on early school leaving. BMC psychiatry . 2014; 14: 237. Google Scholar CrossRef Search ADS PubMed  19 Melkevik O, Nilsen W, Evensen M, et al.  . Internalizing disorders as risk factors for early school leaving: a systematic review. Adolesc Res Rev . 2016; 1( 3): 245– 255. Google Scholar CrossRef Search ADS   20 Fergusson DM, Woodward LJ. Mental health, educational, and social role outcomes of adolescents with depression. Arch Gen Psychiatry . 2002; 59( 3): 225– 231. Google Scholar CrossRef Search ADS PubMed  21 Fletcher JM. Adolescent depression and educational attainment: results using sibling fixed effects. Health Econ . 2010; 19( 7): 855– 871. Google Scholar CrossRef Search ADS PubMed  22 Miech RA, Caspi A, Moffitt TE, et al.  . Low socioeconomic status and mental disorders: a longitudinal study of selection and causation during young adulthood. AJS . 1999; 104( 4): 1096– 1131. 23 Stoep AV, Weiss NS, Kuo ES, et al.  . What proportion of failure to complete secondary school in the US population is attributable to adolescent psychiatric disorder? J Behav Health Serv Res . 2003; 30( 1): 119– 124. Google Scholar CrossRef Search ADS PubMed  24 Wooden M, Watson N. The Household, Income and Labour Dynamics in Australia (HILDA) Survey: Wave 1 Survey Methodology. Melbourne, Australia: Melbourne Institute of Applied and Economic Research; 2002. (HILDA project technical paper series, no. 1/02). http://melbourneinstitute.unimelb.edu.au/assets/documents/hilda-bibliography/hilda-technical-papers/htec102.pdf. Accessed October 5, 2017. 25 Little R, Rubin D. Multiple Imputation for Nonresponse in Surveys . New York, NY: John Wiley & Sons; 1987. 26 Ware JE Jr, Gandek B. The SF-36 Health Survey: development and use in mental health research and the IQOLA Project. Int J Ment Health . 1994; 23( 2): 49– 73. Google Scholar CrossRef Search ADS   27 Cuijpers P, Smits N, Donker T, et al.  . Screening for mood and anxiety disorders with the five-item, the three-item, and the two-item Mental Health Inventory. Psychiatry Res . 2009; 168( 3): 250– 255. Google Scholar CrossRef Search ADS PubMed  28 Rumpf HJ, Meyer C, Hapke U, et al.  . Screening for mental health: validity of the MHI-5 using DSM-IV Axis I psychiatric disorders as gold standard. Psychiatry Res . 2001; 105( 3): 243– 253. Google Scholar CrossRef Search ADS PubMed  29 Gill SC, Butterworth P, Rodgers B, et al.  . Validity of the mental health component scale of the 12-item Short-Form Health Survey (MCS-12) as measure of common mental disorders in the general population. Psychiatry Res . 2007; 152( 1): 63– 71. Google Scholar CrossRef Search ADS PubMed  30 Cobb-Clark DA, Ribar D. Financial stress, family conflict, and youths’ successful transition to adult roles. Bonn, Germany: Institute for the Study of Labour; 2009. (Discussion paper series no. 4618). http://ftp.iza.org/dp4618.pdf. Accessed October 5, 2017. 31 Watson N. Longitudinal and Cross-sectional Weighting Methodology for the HILDA Survey. Melbourne, Australia: Melbourne Institute of Applied and Economic Research; 2012. (HILDA project technical paper series, no. 2/12). http://melbourneinstitute.unimelb.edu.au/assets/documents/hilda-bibliography/hilda-technical-papers/htec212.pdf. Accessed October 5, 2017. 32 Fröjd SA, Nissinen ES, Pelkonen MU, et al.  . Depression and school performance in middle adolescent boys and girls. J Adolesc . 2008; 31( 4): 485– 498. Google Scholar CrossRef Search ADS PubMed  33 Plüddemann A, et al.  . The prevalence of risk for mental health problems among high school students in the Western Cape Province, South Africa. S Afr J Psychol . 2014; 44( 1): 30– 35. 34 Wang MT, Peck SC. Adolescent educational success and mental health vary across school engagement profiles. Dev Psychol . 2013; 49( 7): 1266– 1276. Google Scholar CrossRef Search ADS PubMed  35 Waghorn G, Chant D, Lloyd C, et al.  . Earning and learning among Australian community residents with psychiatric disorders. Psychiatry Res . 2011; 186( 1): 109– 116. Google Scholar CrossRef Search ADS PubMed  36 Kessler RC, Amminger GP, Aguilar-Gaxiola S, et al.  . Age of onset of mental disorders: a review of recent literature. Curr Opin Psychiatry . 2007; 20( 4): 359– 364. Google Scholar CrossRef Search ADS PubMed  37 Australian Bureau of Statistics. Education and Work; 2014. (no. 6227.0). http://www.abs.gov.au/AUSSTATS/abs@.nsf/DetailsPage/6227.0May%202014?OpenDocument. Released December 12, 2014. Accessed October 5, 2017. 38 Australian Bureau of Statistics. Schools, Australia; 2015. (no. 4221.0). http://www.abs.gov.au/AUSSTATS/abs@.nsf/Lookup/4221.0Main+Features392015?OpenDocument. Released March 18, 2016. Accessed October 5, 2017. © The Author(s) 2018. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png American Journal of Epidemiology Oxford University Press

Early Onset of Distress Disorders and High-School Dropout: Prospective Evidence From a National Cohort of Australian Adolescents

Loading next page...
 
/lp/ou_press/early-onset-of-distress-disorders-and-high-school-dropout-prospective-x4b6z0xqQH
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.
ISSN
0002-9262
eISSN
1476-6256
D.O.I.
10.1093/aje/kwx353
Publisher site
See Article on Publisher Site

Abstract

Abstract Prior research examining whether depression and anxiety lead to high-school dropout has been limited by a reliance on retrospective reports, the assessment of mental health at a single point in time (often remote from the time of high-school exit), and the omission of important measures of the social and familial environment. The present study addressed these limitations by analyzing 8 waves of longitudinal data from a cohort of Australian adolescents (n = 1,057) in the Household, Income and Labor Dynamics in Australia (HILDA) Survey (2001–2008). Respondents were followed from the age of 15 years through completion of or exit from high school. Discrete-time survival analysis was used to assess whether the early experience of a distress disorder (indicated by scores <50 on the 5-item Mental Health Inventory from the Short Form Health Survey) predicted subsequent high-school dropout, after controlling for household and parental socioeconomic characteristics and for tobacco smoking and alcohol consumption. Adolescents with a prior distress disorder had twice the odds of high-school dropout compared with those without (odds ratio = 1.99, 95% confidence interval: 1.24, 3.17). This association was somewhat attenuated but remained significant in models including tobacco and alcohol consumption (odds ratio = 1.74, 95% confidence interval: 1.74; 1.09, 2.78). These results suggest that improving the mental health of high-school students may promote better educational outcomes. adolescence, education, high-school incompletion, longitudinal, mental health Educational attainment is a critical determinant of adult life opportunities, both at the individual level in terms of work, health, and financial circumstances and in a broader societal context in terms of demand on social welfare entitlements and workforce development (1). Data from across countries in the Organisation for Economic Co-operation and Development (OECD) shows a trend of increasing levels of school retention, including high-school completion (2). Nonetheless, Australian data from the mid-2010s shows that approximately 26% of 19 year olds did not have high-school (or equivalent) qualifications (3, 4). The reasons for high-school incompletion are varied and include poor academic performance, absence from school, and a lack of socioeconomic resources (4). One likely further important antecedent for high-school incompletion is early-onset psychiatric illness (5). Like high-school incompletion, psychiatric illness has been shown to be associated with unemployment, poor job quality, low income, financial hardship, and poor housing (6–9). The mechanisms via which early-onset psychiatric illness might affect school incompletion include the disruption of academic achievement (e.g., via impaired cognitive function), disrupted behavior, and social responses from teachers and parents who prejudge (and inadvertently limit) student ability (10). Identifying the contribution of early-onset psychiatric illness to high-school incompletion is necessary to inform effective, targeted prevention and intervention policies. The present study uses longitudinal, prospective data to investigate the specific contribution of early onset of symptoms of anxiety and depression. We use the term distress disorder (DD) to describe likely clinical levels of depression and (generalized) anxiety symptoms, reflecting a transdiagnostic approach that views depression and anxiety as markers of an underlying internalizing or “distress” factor (11–13). Links between early-onset mental disorders and subsequent high-school incompletion have been investigated predominantly by using retrospective data on age of onset. Research from the United States, Australia, and South Africa has found that early-onset disorders such as anxiety and mood disorders, conduct disorder, and substance-use disorders are associated with high-school incompletion (14–17). The reliance on retrospective data is, however, a major limitation (18, 19), with retrospective data vulnerable to recall bias and imprecise for capturing the timing between disorder onset and high-school dropout. A recent systematic review by Melkevik et al. (19) focused on the association of anxiety and depression with high-school incompletion and demonstrated this overreliance on retrospective data, showing that only 4 of 16 studies employed a prospective design (20–23). Closer examination of these prospective studies also revealed 2 key limitations: 1) long time lags between assessments of symptomatology and school outcomes, and 2) inadequate adjustment for socioeconomic confounders and indicators of comorbid risk-taking/externalizing behaviors. For example, a prospective study conducted by Fletcher et al. (21) used a sibling fixed-effects methodology to control for potentially relevant but unmeasured family and neighborhood characteristics common to both siblings (e.g., family resources, neighborhood crime). Analyses of data from approximately 2,400 adolescents found that greater depression was associated with high-school dropout after adjusting for family and neighborhood indicators; however, the association was no longer significant after controlling for risky behaviors (e.g., substance use). The study used a nationally representative sample of US students in grades 7–12, initially interviewed in 1994, when the Center for Epidemiologic Studies Depression Scale (CES-D) was administered to them. Educational outcomes were assessed 6 years later, when respondents were (on average) 22 years of age. Given that depression is episodic in nature, the long time lag between assessments might have underrepresented episodes of depression. A long time lag also makes it difficult to link experiences of DDs directly with high-school dropout, because it increases the opportunity for cumulative confounding factors, such as adverse family circumstances, to play a substantive role. In another example, Fergusson and Woodward (20) analyzed data from 1,265 adolescents in New Zealand and found that depression was associated with early school exit but not after adjustment for socioeconomic status (maternal education and family socioeconomic status at birth) and indicators of comorbid externalizing problems (conduct disorder, attention deficit hyperactivity disorder (ADHD), smoking, and alcohol use). In this study, the assessment of mental health and educational outcomes was more proximal. Adolescents were categorized as having major depression between the ages of 14 and 16 years (interviewed at ages 15 and 16 years), and educational outcomes were measured at ages 16, 18, and 21 years. While the analysis adjusted for maternal educational achievement and family socioeconomic status at the time of the child’s birth, more recent and time-varying measures of socioeconomic circumstances were not considered. The present study used a broadly representative, national household survey to prospectively follow a large cohort (n = 1,057) of Australian adolescents across their final years of high school (annual assessment) to assess how the experience of a DD (representing clinical levels of depression and anxiety symptoms) might influence high-school completion. Uniquely, the longitudinal household-study design allowed us to overcome many of the methodological limitations of previous research. We controlled for a comprehensive range of potential confounders over time, including household- and parent-level factors for both mothers and fathers. Because we assessed confounders annually, the analyses considered changing socioeconomic and family circumstances, accounting for both episodic and entrenched socioeconomic disadvantage. We also adjusted for smoking and alcohol use. These risky behaviors are known to be associated with truancy and high-school incompletion and are highly associated with externalizing problems and other forms of substance use. METHODS Sample Data were from the first eight waves of the Household, Income and Labor Dynamics in Australia (HILDA) Survey (release 14.0). Details of the survey methods are reported elsewhere (24). Briefly, HILDA is a nationally representative longitudinal household panel survey with a multistage sampling design. Interviews have been conducted annually since 2001 with each household member aged ≥15 years. The survey composition is dynamic, with sample loss arising from death or attrition and new respondents entering the sample each wave (e.g., adolescents within a household are eligible to participate when they reach 15 years of age). The HILDA survey was approved by the Human Research Ethics Committee at the University of Melbourne. A unique subset of data was created that included all 15 and 16 year olds who entered the HILDA survey during the first 4 waves, including their first 5 years of data. Data for all respondents was centered on their first wave of participation (considered year 1 regardless of when they entered the survey). This identified a sample of 1,256 participants aged 15 and 16 years with data at year 1. However, 100 participants had already left school at their first assessment occasion (74% of these were aged 16 years), and a further 99 did not participate in any subsequent waves during the follow-up period, leaving 1,057 respondents. A series of univariate analyses showed that residing in a low-socioeconomic-status area, having no mother in the household, and mother’s low educational attainment were the only factors associated with attrition. Table 1 shows respondent high-school completion status over time. Over the study, 673 respondents completed high school (64%) and 295 (28%) exited high school without graduating. The outcome for 89 respondents (8%) was not resolved (6 were still attending school at the final wave). Overall, from year 2 to year 5 (when outcomes were assessed), excluding occasions on which respondents did not participate, there were 2,483 observations. Table 1. High-School Completion Status Over 5 Time Points, Household, Income and Labor Dynamics in Australia Survey, 2001–2008 Year  Still in High School  Completed High Schoola  Early Exita  Did Not Participate in Wave  1  1,057        2  857  24  121  55  3  540  201  219  97  4  150  567  287  59  5  6  673  295  83  Year  Still in High School  Completed High Schoola  Early Exita  Did Not Participate in Wave  1  1,057        2  857  24  121  55  3  540  201  219  97  4  150  567  287  59  5  6  673  295  83  a Cumulative frequencies. Table 1. High-School Completion Status Over 5 Time Points, Household, Income and Labor Dynamics in Australia Survey, 2001–2008 Year  Still in High School  Completed High Schoola  Early Exita  Did Not Participate in Wave  1  1,057        2  857  24  121  55  3  540  201  219  97  4  150  567  287  59  5  6  673  295  83  Year  Still in High School  Completed High Schoola  Early Exita  Did Not Participate in Wave  1  1,057        2  857  24  121  55  3  540  201  219  97  4  150  567  287  59  5  6  673  295  83  a Cumulative frequencies. Finally, while item missingness was minimal, it did lead to some further data reduction. In the main model, incorporating a range of individual and household variables (not considering parental characteristics) resulted in missing data for 68 respondents and 253 observations. Most of this missingness was associated with absence of the lagged distress measure, with 27% of these occasions resulting from failure to complete or return the self-completion questionnaire in which mental health was assessed. This missingness was addressed through multiple imputation methods (25). Measures DDs were assessed at all time points using the 5-item Mental Health Inventory (MHI-5), a subscale from the Short Form Health Survey (SF-36) (26). The MHI-5 has been validated against Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Axis 1 diagnosis; has been recommended for screening for anxiety and depressive disorders; and has been used as a measure of likely anxiety and/or depression in epidemiologic surveys (27, 28). The MHI-5 assesses symptoms of depression and anxiety, as well as positive aspects of mental health, in the previous 4 weeks. Scores range from 0 to 100, with higher scores representing better mental health. The present analyses dichotomized the MHI-5, with scores of less than 50 categorized as having a DD. This cutpoint was chosen on the basis of previous research validating the MHI-5 against diagnosis of anxiety and depression disorders assessed using the Composite International Diagnostic Interview (29). Preliminary analyses using the HILDA data confirmed the validity of the MHI-5 as an indicator of distress in comparison with the Kessler Psychological Distress (K-10) scale and that the 5 MHI items represent a single “distress” factor (see Web Appendix 1, available at https://academic.oup.com/aje). High-school completion status At each year, early school exit was coded as a binary variable, contrasting those who were still at school or had just completed high school (0) with those who had left prior to high-school completion (1). At each year, respondents were considered out of scope if at their previous interview (year) they reported having completed or left school. In Australia, education policy is generally the responsibility of the 6 states and 2 territories, resulting in some differences in how education is structured across different jurisdictions. However, the general progression through school involves a year of preparatory school at age 5 years (kindergarten), followed by 6 years of primary school, and 6 years of high (or secondary) school (ending in 12th grade). Attending school is compulsory until the end of grade 10 (15 or 16 years old), with some minor variations between states (30). Covariates Covariates included basic demographics at baseline (sex, age). Household indictors included a measure of region/remoteness, housing tenure (renting vs. other), residence in a socioeconomically disadvantaged area (lowest 30% of Socio-Economic Index for Areas (SEIFA)), and low level of household income (bottom quintile). A number of indicators of parents’ circumstances were included for mothers and fathers: presence in household (either not in household or nonresponder), high-school incompletion, postschool qualifications, receiving income support payments, employed versus unemployed, and employed (currently or previously) in a low-skill occupation. Binary variables representing adolescents’ current smoking status (yes = 1) and alcohol consumption (at least weekly = 1) were included. “Time in study” (years) was also included. Analyses Descriptive statistics from the baseline sample were initially calculated. A series of discrete-time survival analysis models examined the association of prior (12-month lagged) DD (and lagged covariates) with an indicator of early school exit. Discrete-time survival analysis is appropriate given the annual data collection. Preliminary models over the 4-year follow-up period showed that time to early school exit (or censor) was appropriately represented by a linear term, with each additional year at school decreasing the odds of early exit by 18%. A standard approach to model building was employed with variables with a univariable P ≤ 0.2 included in a preliminary multivariable model. Backward elimination was then used to remove variables that no longer met these criteria without leading to poorer model fit (evaluated by comparing log-likelihood statistics) or altering the coefficient of the key covariate by 15% (model A). Two key potential confounders of the association between DD and early school exit (lagged smoking and alcohol consumption) were added at a final model step (model B). Potential sex differences were evaluated by the inclusion of an interaction term (given that sex was identified as a potentially important mediator by Melkevik et al. (19)). The main analyses were weighted by the responding person sample weights included in each wave of HILDA to adjust for nonresponse/selection (see Watson (31)), and multiple imputation (using chained equations to generate 20 imputed data sets based on 10 cycles) was used to address potential bias due to missing data on covariates. A series of further sensitivity analyses were conducted (see Web Appendix 1) to address potential study limitations: excluding respondents who completed high school in the current year, calculating inverse probability weights to adjust for attrition, using random effects logistic regression models, limiting analysis to respondents who entered the study at age 15 years, and testing the robustness of MHI-5 scale. RESULTS Descriptive analyses The baseline (year 1) characteristics of the analysis sample are presented in Table 2. Overall, adolescents who subsequently left school early were significantly more likely to live in rental housing, to not live with their mother, and to report DD, current smoking, and regular (weekly or more frequent) alcohol consumption (based on χ2 tests of association). Preliminary analysis showed that a model including a measure of time-varying DD was superior to a model comprising only baseline DD (log likelihood χ2 = 7.93, degrees of freedom = 1; P = 0.005). Table 2. Characteristics of the Sample at Baseline, Household, Income and Labor Dynamics in Australia Survey, 2001 Variable  Overall (987 Persons and 2,227 Observations), % a  Completed School or Censored (762 Persons and 1,805 Observations), %b  Early Exit (295 Persons and 722 Observations), %c  Age, years         15  66  65  68   16  34  35  32  Female sex  50  52  45  Low household income  16  14  20  Reside in disadvantaged aread  31  28  40  Housing tenure, renter  20  17  27  No father in household  20  19  23  No mother in household  4  4  5  Father did not complete high school  18  17  20  Mother did not complete high school  37  33  49  Father is current welfare recipient  7  6  11  Mother is current welfare recipient  19  17  27  Distress disorder (MHI-5 <50)  8  7  12  Current smoker  6  3  15  Regular alcoholic consumption (weekly or more frequent)  5  3  12  Variable  Overall (987 Persons and 2,227 Observations), % a  Completed School or Censored (762 Persons and 1,805 Observations), %b  Early Exit (295 Persons and 722 Observations), %c  Age, years         15  66  65  68   16  34  35  32  Female sex  50  52  45  Low household income  16  14  20  Reside in disadvantaged aread  31  28  40  Housing tenure, renter  20  17  27  No father in household  20  19  23  No mother in household  4  4  5  Father did not complete high school  18  17  20  Mother did not complete high school  37  33  49  Father is current welfare recipient  7  6  11  Mother is current welfare recipient  19  17  27  Distress disorder (MHI-5 <50)  8  7  12  Current smoker  6  3  15  Regular alcoholic consumption (weekly or more frequent)  5  3  12  Abbreviation: MHI-5, 5-item Mental Health Inventory. a Observations per person: 2.3 (range, 1–4). b Observations per person: 2.5 (range, 1–4). c Observations per person: 1.6 (range, 1–4). d Lowest 30% on Socio-Economic Index for Areas of relative advantage/disadvantage. Table 2. Characteristics of the Sample at Baseline, Household, Income and Labor Dynamics in Australia Survey, 2001 Variable  Overall (987 Persons and 2,227 Observations), % a  Completed School or Censored (762 Persons and 1,805 Observations), %b  Early Exit (295 Persons and 722 Observations), %c  Age, years         15  66  65  68   16  34  35  32  Female sex  50  52  45  Low household income  16  14  20  Reside in disadvantaged aread  31  28  40  Housing tenure, renter  20  17  27  No father in household  20  19  23  No mother in household  4  4  5  Father did not complete high school  18  17  20  Mother did not complete high school  37  33  49  Father is current welfare recipient  7  6  11  Mother is current welfare recipient  19  17  27  Distress disorder (MHI-5 <50)  8  7  12  Current smoker  6  3  15  Regular alcoholic consumption (weekly or more frequent)  5  3  12  Variable  Overall (987 Persons and 2,227 Observations), % a  Completed School or Censored (762 Persons and 1,805 Observations), %b  Early Exit (295 Persons and 722 Observations), %c  Age, years         15  66  65  68   16  34  35  32  Female sex  50  52  45  Low household income  16  14  20  Reside in disadvantaged aread  31  28  40  Housing tenure, renter  20  17  27  No father in household  20  19  23  No mother in household  4  4  5  Father did not complete high school  18  17  20  Mother did not complete high school  37  33  49  Father is current welfare recipient  7  6  11  Mother is current welfare recipient  19  17  27  Distress disorder (MHI-5 <50)  8  7  12  Current smoker  6  3  15  Regular alcoholic consumption (weekly or more frequent)  5  3  12  Abbreviation: MHI-5, 5-item Mental Health Inventory. a Observations per person: 2.3 (range, 1–4). b Observations per person: 2.5 (range, 1–4). c Observations per person: 1.6 (range, 1–4). d Lowest 30% on Socio-Economic Index for Areas of relative advantage/disadvantage. Discrete-time survival analyses Initial univariable discrete-time survival models showed that all covariates except baseline age and parental employment characteristics (workforce status and occupational skill level) were associated with early school exit with P ≤ 0.2 and were, therefore, included in the early phase of multivariate model building. The first model (model A; Table 3), having deleted covariates no longer associated with the outcome, showed that adolescents who reported DD had twice the odds of early school exit in the following year in comparison with those without DD (odds ratio = 1.99), having controlled for sociodemographic characteristics, economic circumstances, and parental characteristics. Table 3. Discrete-Time Survival Models Examining Dropout Prior to High-School Completion, Household, Income and Labor Dynamics in Australia Survey Sample, 2001–2008a Variable  Model A  Model Bb  OR  95% CI  OR  95% CI  Prior distress disorder (MHI-5 <50)  1.99  1.24, 3.17  1.74  1.09, 2.78  Prior smoking      3.01  1.88, 4.80  Prior alcohol consumption      1.52  0.92, 2.50  Female sex  0.68  0.51, 0.91  0.70  0.52, 0.94  Time, years  0.97  0.83, 1.15  0.95  0.80, 1.13  Rental housing  1.79  1.27, 2.54  1.63  1.15, 2.32  Father status (referent: present)           Not present in household  1.56  1.08, 2.25  1.47  1.01, 2.14   Nonresponder  1.48  0.69, 3.12  1.49  0.67, 3.04  Mother left school early  1.77  1.28, 2.43  1.69  1.23, 2.33  Father is current welfare recipient  1.30  0.73, 2.31  1.34  0.75, 2.41  Variable  Model A  Model Bb  OR  95% CI  OR  95% CI  Prior distress disorder (MHI-5 <50)  1.99  1.24, 3.17  1.74  1.09, 2.78  Prior smoking      3.01  1.88, 4.80  Prior alcohol consumption      1.52  0.92, 2.50  Female sex  0.68  0.51, 0.91  0.70  0.52, 0.94  Time, years  0.97  0.83, 1.15  0.95  0.80, 1.13  Rental housing  1.79  1.27, 2.54  1.63  1.15, 2.32  Father status (referent: present)           Not present in household  1.56  1.08, 2.25  1.47  1.01, 2.14   Nonresponder  1.48  0.69, 3.12  1.49  0.67, 3.04  Mother left school early  1.77  1.28, 2.43  1.69  1.23, 2.33  Father is current welfare recipient  1.30  0.73, 2.31  1.34  0.75, 2.41  Abbreviations: CI, confidence interval; MHI-5, 5-item Mental Health Inventory; OR, odds ratio. a Analyses weighted by the responding person sample weights included in each wave of the Household, Income and Labor Dynamics in Australia survey. Analyses addressed missing data through multiple imputation methods (see Web Appendix 1 for details). b Model B extends model A by incorporating measures of smoking status and regular alcohol consumption. Table 3. Discrete-Time Survival Models Examining Dropout Prior to High-School Completion, Household, Income and Labor Dynamics in Australia Survey Sample, 2001–2008a Variable  Model A  Model Bb  OR  95% CI  OR  95% CI  Prior distress disorder (MHI-5 <50)  1.99  1.24, 3.17  1.74  1.09, 2.78  Prior smoking      3.01  1.88, 4.80  Prior alcohol consumption      1.52  0.92, 2.50  Female sex  0.68  0.51, 0.91  0.70  0.52, 0.94  Time, years  0.97  0.83, 1.15  0.95  0.80, 1.13  Rental housing  1.79  1.27, 2.54  1.63  1.15, 2.32  Father status (referent: present)           Not present in household  1.56  1.08, 2.25  1.47  1.01, 2.14   Nonresponder  1.48  0.69, 3.12  1.49  0.67, 3.04  Mother left school early  1.77  1.28, 2.43  1.69  1.23, 2.33  Father is current welfare recipient  1.30  0.73, 2.31  1.34  0.75, 2.41  Variable  Model A  Model Bb  OR  95% CI  OR  95% CI  Prior distress disorder (MHI-5 <50)  1.99  1.24, 3.17  1.74  1.09, 2.78  Prior smoking      3.01  1.88, 4.80  Prior alcohol consumption      1.52  0.92, 2.50  Female sex  0.68  0.51, 0.91  0.70  0.52, 0.94  Time, years  0.97  0.83, 1.15  0.95  0.80, 1.13  Rental housing  1.79  1.27, 2.54  1.63  1.15, 2.32  Father status (referent: present)           Not present in household  1.56  1.08, 2.25  1.47  1.01, 2.14   Nonresponder  1.48  0.69, 3.12  1.49  0.67, 3.04  Mother left school early  1.77  1.28, 2.43  1.69  1.23, 2.33  Father is current welfare recipient  1.30  0.73, 2.31  1.34  0.75, 2.41  Abbreviations: CI, confidence interval; MHI-5, 5-item Mental Health Inventory; OR, odds ratio. a Analyses weighted by the responding person sample weights included in each wave of the Household, Income and Labor Dynamics in Australia survey. Analyses addressed missing data through multiple imputation methods (see Web Appendix 1 for details). b Model B extends model A by incorporating measures of smoking status and regular alcohol consumption. The final model (model B) incorporated the measures of adolescent substance use (lagged smoking status and alcohol consumption). While smoking was associated with increased odds of early school exit, DD remained significant, with respondents with DD having 74% greater odds of leaving school early the following year compared with those without DD. Being male, living in rental housing, having no father in the household and having a mother who left school early were also characteristics associated with early school exit. On the basis of this model, we consider the likelihood of early school exit for an adolescent boy, who does not smoke or regularly drink alcohol, who lives in rental accommodation with his father (who receives income support payments), and whose mother did not complete high school. While a boy in these circumstances without DD had a likelihood of 22% (95% confidence interval: 11.5, 33.3) of early school exit, the likelihood of early school exit for an adolescent in the same circumstances but with DD was 34% (95% confidence interval: 17.8, 49.8). Sensitivity analyses confirmed the robustness of the results (see Web Appendix 1). The inclusion of a term for interaction between sex and DD did not improve model fit (log likelihood χ2 = 2.59, degrees of freedom = 1; P = 0.107), although coefficients were greater for girls than for boys. An initial model that did not adjust for nonresponse, attrition, and missing data (see Web Table 1) produced similar odds ratios for the association between DD and early school exit, as did models using the continuous MHI scale (see Web Table 2) and models testing the MHI items for depression and anxiety separately (see Web Table 3). The exclusion of adolescents who completed high school in the target year (see Web Table 4) and random effects models (see Web Table 5) produced similar (or somewhat stronger) odds ratios for the association between DD and early school exit. Restricting the analysis to those initially aged 15 years produced slightly weaker coefficients (see Web Table 6). Supplementary analyses S4–S6 outline the details of various models aimed at reducing bias (see Web Tables 7–10). DISCUSSION The findings from this prospective study provide strong evidence that adolescents experiencing a DD (depression/anxiety symptoms) are at greater risk of high-school dropout than are those who are not. The relationship between DD and subsequent high-school dropout remained significant after taking account of a broad range of socioeconomic circumstances (including household- and parent-level factors) as well as adolescents’ alcohol and tobacco use. In the context of previous research, while several retrospective studies have similarly pointed to a strong link between early-onset mental health disorders and high-school incompletion (14–17), the findings of prospective research focused on anxiety and depression have varied depending on the lag between interview dates and the sociodemographic and health behavior covariates (19–21). The present study represents a significant advance on previous research in that the models controlled for household and parental socioeconomic circumstances (as they vary over time), both of which are strongly related to high-school completion. The present study is also unusual in its application of multiple waves of data, spaced annually. The repeated measurement of DD and other covariates over time enabled consideration of the proximal association between experiences of DD in adolescence and subsequent high-school dropout. We showed, for example, that a measure of lagged DD was a significantly better predictor of high-school dropout than a time-invariant measure of DD at age 15 years, demonstrating the importance of multiple-wave data on mental health as opposed to baseline data alone. The importance of anxiety and depression for subsequent high-school incompletion was evident in the finding that the association endures after taking account of tobacco and alcohol use, which are each strongly and independently associated with high-school dropout. (We note that this inclusion may represent an overcorrection because substance use could be a response to the experience of anxiety and depression symptoms). This study adds weight to the hypothesis that early-onset DD disrupts the opportunity for adolescents to successfully attain important social and economic milestones, such as high-school completion. However, there are other important indicators of educational disruption, such as school absence (due to poor psychological health or truancy), poor student engagement, and poor academic achievement (32–34). High-school dropout likely represents the endpoint in a series of disruptions in relational and academic growth that are influenced by, and in turn influence, anxiety and depression symptoms. While high-school incompletion is the endpoint in this investigation, it is important to consider postschool transitions such as entering vocational training, university education, and employment. Despite disrupted schooling, some individuals will find new pathways to further education and employment, while others will be unable to do so (5, 35). Given that roughly half of all mental disorders have their onset during adolescence (mid-teens) (36), there is great scope to more robustly delineate the impacts of early DD on educational and socioeconomic opportunities and to generate new data on the personal, social, and economic benefits of early intervention. A number of limitations should be noted. Diagnostic measures of mental health (i.e., mental disorders) were not available. However, the MHI-5 score is a valid indicator of depression and anxiety in the general community (29). The HILDA survey included no information on the school environment, academic achievement, or school absence, nor did it include psychometric measures of externalizing or conduct behaviors. Australian data has shown that high-school completion is strongly associated with academic performance (4) and with externalizing behaviors (20). It may be that DDs are a consequence of these underlying factors. However, we adjusted for a range of sociodemographic and substance-use measures that would also likely be associated with these academic and behavioral factors. This study also did not consider different school settings, such as contrasting private or public schools. The present results showed a relatively high rate of high-school noncompletion (28%), but this is consistent with census and other data on high-school completion published by the Australian Bureau of Statistics and other reputable sources (37, 38). Because the HILDA survey is restricted to respondents aged 15 years or older, we did not capture those who left school very early. However, analysis restricted to 15 year olds eliminated most of this missingness and produced the same pattern of results. It is also important to recognize that, with time and perhaps the resolution of their DD, some in their late teens or early 20s without high-school qualifications may return to education. The present study, by design, did not consider return to study or longer-term outcomes. This is an important direction for further research. Similarly, while not significant in this study, sex differences in the association between DD and school dropout warrant further consideration. Despite these limitations, this study provides unique longitudinal data on the mental health and social circumstances of adolescents and their parents/households. The range of sensitivity tests conducted (applying different methods, measures, weights and multiple imputation) demonstrate the robustness of the findings. In conclusion, the present study followed a large cohort of Australian adolescents across the final years of high school. Experiencing clinical levels of depression and anxiety symptoms during early adolescence was associated with increased risk of high-school dropout. The findings support calls for mental illness prevention and early intervention programs targeting adolescents and the school environment. In the Australian setting, such services and programs are becoming common (e.g., Headspace, beyondblue schools program, and MindMatters). The evaluation of these programs should consider social and economic outcomes, such as reduction in high-school dropout. The present findings highlight the need to provide and evaluate mental health support services for young people who have left the education system to enhance their chances of returning to education or gaining meaningful employment. There may be broad benefits of such service and policy interventions. ACKNOWLEDGMENTS Author affiliations: Centre for Mental Health, Melbourne School of Population and Global Health, University of Melbourne, Melbourne, Victoria, Australia (Peter Butterworth); Melbourne Institute of Applied Economic and Social Research, University of Melbourne, Melbourne, Victoria, Australia (Peter Butterworth); and National Centre for Epidemiology and Population Health, Research School of Population Health, College of Medicine, Australian National University, Canberra, Australia (Liana S. Leach). The Household, Income and Labour Dynamics in Australia Project was initiated and is funded by the Australian Government Department of Social Services and is managed by the Melbourne Institute of Applied Economic and Social Research. This work was supported by Australian Research Council Future Fellowship (grant FT13101444), National Health and Medical Research Council Early Career Fellowship (grant 1035803), and a research grant from Australian Rotary Health. The findings and views reported in this paper are those of the authors and should not be attributed to either the Australian Department of Social Services or the Melbourne Institute. There have been no involvements that might raise the question of bias in the work reported or in the conclusions, implications, or opinions stated. Conflict of interest: none declared. Abbreviations DD distress disorder HILDA Household, Income and Labor Dynamics in Australia Survey MHI-5 5-item Mental Health Inventory REFERENCES 1 Buchmann C, Hannum E. Education and stratification in developing countries: a review of theories and research. Annu Rev Sociol . 2001; 27: 77– 102. Google Scholar CrossRef Search ADS   2 Organisation for Economic Cooperation and Development (OECD). Education at a Glance 2015: OECD Indicators. Paris, France: OECD Publishing; 2015. http://www.oecd.org/edu/education-at-a-glance-2015.htm. Published November 24, 2015. Accessed October 5, 2017. 3 Australian Institute of Health and Welfare (AIHW). Australia’s welfare 2015. (Catalogue no. AUS189). Canberra, Australia: AIHW; 2015. https://www.aihw.gov.au/reports/australias-welfare/australias-welfare-2015. Released August 20, 2015. Accessed October 5, 2017. 4 Lamb S, Jackson J, Walstab A, et al.  . Educational opportunity in Australia 2015: Who succeeds and who misses out. Melbourne, Australia: Centre for International Research on Education Systems, for the Mitchell Institute; 2015. http://www.mitchellinstitute.org.au/reports/educational-opportunity-in-australia-2015-who-succeeds-and-who-misses-out/. Published October 26, 2015. Accessed October 5, 2017. 5 Freudenberg N, Ruglis J. Reframing school dropout as a public health issue. Prev Chronic Dis . 2007; 4( 4): A107. Google Scholar PubMed  6 Butterworth P, Olesen SC, Leach LS. The role of hardship in the association between socio-economic position and depression. Aust N Z J Psychiatry . 2012; 46( 4): 364– 373. Google Scholar CrossRef Search ADS PubMed  7 Evans GW, Wells NM, Moch A. Housing and mental health: a review of the evidence and a methodological and conceptual critique. J Soc Issues . 2003; 59( 3): 475– 500. Google Scholar CrossRef Search ADS   8 Ezzy D. Unemployment and mental health: a critical review. Soc Sci Med . 1993; 37( 1): 41– 52. Google Scholar CrossRef Search ADS PubMed  9 Stansfeld S, Candy B. Psychosocial work environment and mental health—a meta-analytic review. Scand J Work Environ Health . 2006; 32( 6): 443– 462. Google Scholar CrossRef Search ADS PubMed  10 Mcleod JD, Fettes DL. Trajectories of failure: the educational careers of children with mental health problems. AJS . 2007; 113( 3): 653– 701. Google Scholar PubMed  11 Krueger RF, Caspi A, Moffitt TE, et al.  . The structure and stability of common mental disorders (DSM-III-R): a longitudinal-epidemiological study. J Abnorm Psychol . 1998; 107( 2): 216– 227. Google Scholar CrossRef Search ADS PubMed  12 Krueger RF, Eaton NR. Transdiagnostic factors of mental disorders. World Psychiatry . 2015; 14( 1): 27– 29. Google Scholar CrossRef Search ADS PubMed  13 Watson D. Differentiating the mood and anxiety disorders: a quadripartite model. Annu Rev Clin Psychol . 2009; 5: 221– 247. Google Scholar CrossRef Search ADS PubMed  14 Breslau J, Lane M, Sampson N, et al.  . Mental disorders and subsequent educational attainment in a US national sample. J Psychiatr Res . 2008; 42( 9): 708– 716. Google Scholar CrossRef Search ADS PubMed  15 Kessler RC, Foster CL, Saunders WB, et al.  . Social consequences of psychiatric disorders, I: educational attainment. Am J Psychiatry . 1995; 152( 7): 1026– 1032. Google Scholar CrossRef Search ADS PubMed  16 Leach LS, Butterworth P. The effect of early onset common mental disorders on educational attainment in Australia. Psychiatry Res . 2012; 199( 1): 51– 57. Google Scholar CrossRef Search ADS PubMed  17 Myer L, Stein DJ, Jackson PB, et al.  . Impact of common mental disorders during childhood and adolescence on secondary school completion. S Afr Med J . 2009; 99( 5 Pt 2): 354– 356. Google Scholar PubMed  18 Esch P, Bocquet V, Pull C, et al.  . The downward spiral of mental disorders and educational attainment: a systematic review on early school leaving. BMC psychiatry . 2014; 14: 237. Google Scholar CrossRef Search ADS PubMed  19 Melkevik O, Nilsen W, Evensen M, et al.  . Internalizing disorders as risk factors for early school leaving: a systematic review. Adolesc Res Rev . 2016; 1( 3): 245– 255. Google Scholar CrossRef Search ADS   20 Fergusson DM, Woodward LJ. Mental health, educational, and social role outcomes of adolescents with depression. Arch Gen Psychiatry . 2002; 59( 3): 225– 231. Google Scholar CrossRef Search ADS PubMed  21 Fletcher JM. Adolescent depression and educational attainment: results using sibling fixed effects. Health Econ . 2010; 19( 7): 855– 871. Google Scholar CrossRef Search ADS PubMed  22 Miech RA, Caspi A, Moffitt TE, et al.  . Low socioeconomic status and mental disorders: a longitudinal study of selection and causation during young adulthood. AJS . 1999; 104( 4): 1096– 1131. 23 Stoep AV, Weiss NS, Kuo ES, et al.  . What proportion of failure to complete secondary school in the US population is attributable to adolescent psychiatric disorder? J Behav Health Serv Res . 2003; 30( 1): 119– 124. Google Scholar CrossRef Search ADS PubMed  24 Wooden M, Watson N. The Household, Income and Labour Dynamics in Australia (HILDA) Survey: Wave 1 Survey Methodology. Melbourne, Australia: Melbourne Institute of Applied and Economic Research; 2002. (HILDA project technical paper series, no. 1/02). http://melbourneinstitute.unimelb.edu.au/assets/documents/hilda-bibliography/hilda-technical-papers/htec102.pdf. Accessed October 5, 2017. 25 Little R, Rubin D. Multiple Imputation for Nonresponse in Surveys . New York, NY: John Wiley & Sons; 1987. 26 Ware JE Jr, Gandek B. The SF-36 Health Survey: development and use in mental health research and the IQOLA Project. Int J Ment Health . 1994; 23( 2): 49– 73. Google Scholar CrossRef Search ADS   27 Cuijpers P, Smits N, Donker T, et al.  . Screening for mood and anxiety disorders with the five-item, the three-item, and the two-item Mental Health Inventory. Psychiatry Res . 2009; 168( 3): 250– 255. Google Scholar CrossRef Search ADS PubMed  28 Rumpf HJ, Meyer C, Hapke U, et al.  . Screening for mental health: validity of the MHI-5 using DSM-IV Axis I psychiatric disorders as gold standard. Psychiatry Res . 2001; 105( 3): 243– 253. Google Scholar CrossRef Search ADS PubMed  29 Gill SC, Butterworth P, Rodgers B, et al.  . Validity of the mental health component scale of the 12-item Short-Form Health Survey (MCS-12) as measure of common mental disorders in the general population. Psychiatry Res . 2007; 152( 1): 63– 71. Google Scholar CrossRef Search ADS PubMed  30 Cobb-Clark DA, Ribar D. Financial stress, family conflict, and youths’ successful transition to adult roles. Bonn, Germany: Institute for the Study of Labour; 2009. (Discussion paper series no. 4618). http://ftp.iza.org/dp4618.pdf. Accessed October 5, 2017. 31 Watson N. Longitudinal and Cross-sectional Weighting Methodology for the HILDA Survey. Melbourne, Australia: Melbourne Institute of Applied and Economic Research; 2012. (HILDA project technical paper series, no. 2/12). http://melbourneinstitute.unimelb.edu.au/assets/documents/hilda-bibliography/hilda-technical-papers/htec212.pdf. Accessed October 5, 2017. 32 Fröjd SA, Nissinen ES, Pelkonen MU, et al.  . Depression and school performance in middle adolescent boys and girls. J Adolesc . 2008; 31( 4): 485– 498. Google Scholar CrossRef Search ADS PubMed  33 Plüddemann A, et al.  . The prevalence of risk for mental health problems among high school students in the Western Cape Province, South Africa. S Afr J Psychol . 2014; 44( 1): 30– 35. 34 Wang MT, Peck SC. Adolescent educational success and mental health vary across school engagement profiles. Dev Psychol . 2013; 49( 7): 1266– 1276. Google Scholar CrossRef Search ADS PubMed  35 Waghorn G, Chant D, Lloyd C, et al.  . Earning and learning among Australian community residents with psychiatric disorders. Psychiatry Res . 2011; 186( 1): 109– 116. Google Scholar CrossRef Search ADS PubMed  36 Kessler RC, Amminger GP, Aguilar-Gaxiola S, et al.  . Age of onset of mental disorders: a review of recent literature. Curr Opin Psychiatry . 2007; 20( 4): 359– 364. Google Scholar CrossRef Search ADS PubMed  37 Australian Bureau of Statistics. Education and Work; 2014. (no. 6227.0). http://www.abs.gov.au/AUSSTATS/abs@.nsf/DetailsPage/6227.0May%202014?OpenDocument. Released December 12, 2014. Accessed October 5, 2017. 38 Australian Bureau of Statistics. Schools, Australia; 2015. (no. 4221.0). http://www.abs.gov.au/AUSSTATS/abs@.nsf/Lookup/4221.0Main+Features392015?OpenDocument. Released March 18, 2016. Accessed October 5, 2017. © The Author(s) 2018. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)

Journal

American Journal of EpidemiologyOxford University Press

Published: Nov 8, 2017

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off