Trajectories of Mental Health over 16 Years amongst Young Adult Women: The Australian Longitudinal Study on Women's Health

Acknowledgement: The Australian Longitudinal Study on Women’s Health is conducted by a team of researchers from the Universities of Queensland and Newcastle, and funded by the Australian Government Department of Health. This analysis was funded by an Australian Research Council Discovery Project Grant to the third author (DP120100167).

Mental health problems appear to be most common during adolescence and early adulthood (e.g., Hamdi & Iacono, 2014; Australian Bureau of Statistics, 2008). Longitudinal studies of general populations in a number of countries consistently identify groups of individuals who continue to experience chronic or recurrent mental health problems throughout the adult life span (e.g., Campbell, Matestic, von Stauffenberg, Mohan, & Kirchner, 2007; Colman, Ploubidis, Wadsworth, Jones, & Croudace, 2007), a finding that has led to a theoretical and empirical focus on adolescence and early adulthood as a crucial time in the development of chronic mental health problems.

This article addresses longitudinal trajectories of mental health problems among a national cohort of young Australian women. Issues concerning mental health in Australia are similar to those in other developed democracies. A recent national survey shows that approximately 45% of Australians will meet criteria for a mental disorder at some stage in their lives, with 20% meeting criteria in any 12-month period. Women are more likely than men to be diagnosed with mental disorders (although men have higher rates of substance abuse), and rates peak in early adulthood and decline with age. Being unpartnered, unemployed, with limited education, or otherwise socially disadvantaged is associated with greater risk of mental disorder, as are smoking, physical inactivity, obesity, and physical illness or disability (Australian Bureau of Statistics, 2008).

In this article, we take a broad population-health approach to mental health. Although we do not question the value of intensive, smaller-scale, research and intervention with individuals with clinical psychiatric disorders, we adopt the classic epidemiological approach of Rose (1985), focusing on relative risks of incidence within populations rather than on individual cases. Rose (1985) is particularly associated with the view that a focus on the large number of people with mild to moderate problems leads to theories and interventions that are more able to produce significant change at a population level, than will a focus on the relatively smaller group with severe problems. This approach is consistent, for example, with the World Health Organisation, whose publications and interventions focus on population-level mental health, and whose model of mental health focuses primarily on broad social and economic determinants, the role of mental health in social engagement and social capital, and the relationships between mental health, physical health, and the capacity to contribute positively to family, social and economic life (Herrman, Saxena, & Moodie, 2005).

Many psychological theorists assume that mental health problems that emerge in adolescence or early adulthood will tend to persist. For example, Schulenberg, Sameroff, and Cicchetti (2004) have argued that the transition between adolescence and early adulthood is critical to lifecourse mental health. Their perspective is that the stresses associated with the major simultaneous changes in social roles, family relationships, and material conditions that are normative at this time may either disrupt previous trajectories of successful adjustment, or contribute to exacerbations of existing but subthreshold psychological problems. On the other hand, theoretical and empirical research on psychological resilience (e.g., Bonanno, 2005; Howard, Galambos, & Krahn, 2010) suggests that at least some individuals have the capacity to achieve positive mental health despite major traumatic events during this life stage.

Although poor mental health in early adulthood has been shown to be a strong individual predictor of persistent or recurrent mental health problems, population-based research supports the view that this is not inevitable: Up to 50% of people experiencing a first episode of major depression will not experience a recurrence (e.g., Eaton et al., 2008). At the population level, Colman et al. (2007) followed a cohort of 4,627 participants in the 1946 British Birth Cohort study from age 13 to age 53 and identified six distinct trajectories of mild to moderate depressive symptoms, including 6% of the entire sample who experienced a single episode of depression in adolescence with no recurrence.

Numerous longitudinal studies have examined patterns of both clinical and subclinical indicators of poor mental health over time (e.g., Galambos, Barker, & Krahn, 2006; Howard et al., 2010; Kuchibhatla, Fillenbaum, Hybels, & Blazer, 2012; Rhebergen et al., 2011; Tanner et al., 2007). However, most longitudinal studies focus on describing and classifying patterns of change in mental health and do not assess predictors, correlates, or outcomes, limiting our ability to understand and predict the course of poor mental health within and between populations over time, or to identify groups at risk.

Research suggests that young people with recurring poor mental health differ from those with nonrecurring poor mental health and from those with ongoing good mental health (Hammen, Brennan, Keenan-Miller, & Herr, 2008). Comparison of those with and without recurrence, and those with different patterns of recurrence over time, may enable a better understanding of predictors of chronic or recurring mental health problems (Monroe & Harkness, 2012; Segal, Pearson, & Thase, 2003) and thus provide a basis for the development and targeting of early intervention and prevention.

In this analysis, we use trajectory-based modeling of longitudinal data over 16 years, from a population-based cohort of Australian women moving through early adulthood (Lee et al., 2005). This article examines (a) whether the women’s trajectories of mental health over time fall into distinguishable and meaningful groups, and, if so, (b) whether initial demographic, lifestyle, health-related and psychosocial variables predict membership of mental health trajectory groups, and (c) whether trajectory group membership predicts subsequent measures on these and related variables.

The Australian Longitudinal Study on Women’s Health (ALSWH) is a longitudinal cohort study that examines relationships among social and lifestyle factors and women’s physical health, emotional well-being, and use of and satisfaction with health services. The entire project began with three age-group cohorts of women, who were aged 18–23 (younger cohort, born 1973–1978), 45–50 (midage cohort, born 1946–1951), and 70–75 (older cohort, born 1921–1926) when first surveyed in 1996. This article focuses on mental health trajectories in the 1973–1978 cohort, and on variables at Survey 1 (1996, aged 18–23) and at Survey 6 (2012, aged 34–39) that may be associated with these trajectories.

Women were initially selected from the Australian national health insurance database (Medicare), which includes all citizens and permanent residents. The sampling strategy was stratified random, with systematic oversampling of women from rural and remote areas. Details of the samples and recruitment methods have been described elsewhere (Brown, Dobson, Bryson, & Byles, 1999; Lee et al., 2005). Comparison with national census data in 1996 indicated that the 1973–1978 cohort was demographically representative of Australian women in that age group, with a slight bias toward married, educated, and Australian-born women (Brown et al., 1999). Further comparisons of the retained cohorts with five-yearly census data in 2001, 2006, and 2011 have shown that ALSWH participants are more likely to be married and Australian born, have higher education, be employed, and work longer hours than Australian population norms for women in the same age range (Australian Longitudinal Study on Women’s Health, 2013). Women gave written consent at each survey. The study has ethical clearance from the Universities of Queensland and Newcastle, Australia.

Although 14,247 women were recruited to the 1973–1978 cohort, this analysis focuses on the 5,171 women in this cohort (36% of the initial sample) who responded to all of the first six surveys: Survey 1 (1996, aged 18–23); Survey 2 (2000, aged 22–27); Survey 3 (2003, aged 25–30); Survey 4 (2006, aged 28–33); Survey 5 (2009, aged 31–36); and Survey 6 (2012, aged 34–39).

Mental health was measured at each survey with the five-item Mental Health Index (MHI-5) from the SF-36 (Ware, Snow, Kosinski, & Grandek, 1993). This measure is designed for use in population research. It includes five of the 38 items of the validated long-form Mental Health Index and has been shown to correlate well (0.95) with the 38-item version (Ware et al., 1993). The MHI-5 has also been validated against depressive and anxiety disorders in a general population sample using the Munich Composite International Diagnostic Interview with a cutpoint of 60 (Rumpf, Meyer, Hapke, & John, 2001). Response options range on a 6-point scale from 1 (all of the time) to 6 (none of the time). After reverse scoring some items, the total is transformed to a score ranging from 0 to 100, with higher scores indicating better mental health. We adopted the widely used conservative cutpoint of ≤52 to identify individuals with low mental health. This cutpoint has been validated independently (Holmes, 1998; Silveira et al., 2005).

The following variables were measured at Survey 1 and at Survey 6 (unless indicated otherwise).

Area of residence, derived from respondents’ home address, was categorized using standard Australian classification as major cities, inner regional, outer regional, and remote/very remote (Australian Bureau of Statistics, 2006). Language spoken at home was categorized as English or other.

Parent education, asked at Survey 2 only but assumed to be relatively constant, was used as a proxy for social class of origin. We categorized the reported education level of the higher-educated parent as university, trade or technical college, and school only. This variable was included in the analysis of predictors of trajectories but not as an outcome. The woman’s own education level was categorized in the same way. This variable was used as an outcome, but not as a predictor, because many of the women were still completing their education at Survey 1 (at that time, slightly under 50% were studying, including 12% who were still at secondary school).

Ability to manage on available income was used as a proxy for financial circumstances. Five response options were grouped into three categories for analysis: no difficulty, difficult some of the time, and impossible/difficult all the time. Relationship status was categorized as married; cohabiting; separated, divorced, or widowed; and never married. A work/study variable was derived from a series of questions about time use and categorized as working only, both working and studying, studying only, and neither working nor studying. Motherhood was categorized as yes or no on the basis of a question about live births.

Alcohol consumption was measured with standard questions on frequency and quantity, and respondents were categorized as low risk, nondrinker, rarely drinks, and risky/high-risk drinker (National Health & Medical Research Council, 2001). Smoking status was categorized as never smoked, exsmoker, and current smoker. Self-reported height and weight were used to calculate body mass index (BMI), which was categorized as healthy weight (18.5–<25); underweight (BMI <18.5); overweight (25–<30); obese (30+) (National Health & Medical Research Council, 2013). The ALSWH has assessed the accuracy of height and weight measures for our study among participants in the midage cohort and found strong agreement between self-reported and measured height and weight (Burton, Brown, & Dobson, 2010). At Survey 1 women were asked “How many times in a normal week do you engage in vigorous exercise lasting 20 minutes or more?” and “How many times in a normal week do you engage in less vigorous exercise lasting 20 minutes or more?” Response options were 1 (never), 2 (once per week), 3 (two or three times per week), 4 (four, five, or six times per week), 5 (once every day), and 6 (more than once every day) and were recoded as 0, 1, 2.5, 5, 7, and 10, respectively. A weighted sum of the recoded values of both questions combined was then calculated and categorized as high, moderate, low, and nil. At Survey 6, physical activity was measured using items on frequency and intensity of exercise from Active Australia’s 1999 National Physical Activity Survey (Armstrong, Bauman, & Davies, 2000). A detailed description of calculations is provided elsewhere (Brown, Ford, Burton, Marshall, & Dobson, 2005). The categories are high, moderate, low, and nil.

General health was measured with the SF-36 General Health subscale (range = 0–100, higher scores indicate better self-rated health). Because this was highly skewed, we created three groups, separately for Survey 1 and at Survey 6, both based on the distribution of this variable at Survey 1 among those women included in the analysis: high (>82); moderate (>67 and ≤82); low (≤67).

Primary care physician visits in the previous 12 months were self-reported and categorized as 0, 1–2, 3–4, and 5 or more.

Psychosocial factors

Women who reported ever having been told by a doctor that they suffered from depression were categorized as yes, others as no. Women who reported ever having been in a violent relationship with a partner/spouse were categorized as yes, others as no. Weight/shape dissatisfaction was assessed using two items from the Eating Disorder Examination Questionnaire (Fairburn & Beglin, 1994). Women were asked how dissatisfied they have felt in the past month about (a) their weight and (b) their shape; responses were on a 7-point Likert scale ranging from 1 (markedly satisfied) to 7 (not satisfied at all). Women with a mean score of 4 or more were categorized as low, others as high.

All analyses were performed in STATA 13.1 (Stata Corp, College Station, TX). First, mental health scores at Survey 1 were compared between those with complete data across all six surveys, who were included in these analyses, and all others, using analysis of variance, to assess any selection bias.

Mental health trajectory groups were identified using group-based trajectory modeling, a form of latent class growth modeling that identifies clusters of individuals following the same or similar trajectories (Nagin & Odgers, 2010). A logistic group-based trajectory model was used because the data were dichotomous. The optimal number of trajectory groups and polynomial order were assessed using the Bayesian information criterion (BIC), combined with knowledge of the data and the constraint that no group should include less than 5% of the sample (Nagin, 2005; Nagin & Odgers, 2010). Once trajectory groups had been identified, we used separate multinomial logistic regressions to determine Survey 1 predictors and Survey 6 correlates of trajectory group membership. Each trajectory group was compared to the reference group (consistently high mental health) with relative risk ratios. These models were adjusted for all listed covariates.

Of the initial 14,247 women recruited to the study, 5,171 (36%) responded to all six surveys and were included in the analyses. At Survey 1, there were small but at times statistically significant differences between those included and those excluded because of subsequent missing data. Women who had missing surveys were more likely to have low mental health, be from a non-English-speaking background, experience financial difficulty, be neither working nor studying, and have children. They were also more likely to be current smokers, to report poor general health, and to have experienced intimate partner violence (see Supplementary Table S1 for a comparison of those included and those excluded on the basis of missing data). Table 1 shows the distributions of all predictors and correlates, for those women included in the analysis, at Survey 1 and Survey 6.

Group-based trajectory modeling indicated four groups as the best solution, with the best model fit achieved using a polynomial order of: intercept, linear, cubic and cubic (BIC = −11,847.85, N = 5,171). Figure 1 shows the four patterns of mental health trajectories across the study period. The largest group, 2,856 women (55%; consistently high mental health), had high MHI-5 scores at all six survey points. These women served as the reference group in subsequent analyses. The others fell into three groups: improving mental health (472 women, 9%), who generally had low MHI-5 scores at Survey 1 or 2 only; varying mental health (1,228 women, 24%), who generally had high MHI-5 scores, but with low MHI-5 scores at one or two surveys, usually later in the trajectory period; and frequently low mental health (615 women, 12%), who had low MHI-5 scores at three or more surveys.

Table 2 shows the results of a multinomial logistic regression, in which each trajectory group was compared to the reference group (consistently high mental health) using relative risk ratios with simultaneous adjustment for all other variables.

At Survey 1, when the participants were aged 18 to 23, there were several demographic differences among the trajectory groups. Those with improving mental health had initially found it very difficult to manage on available income, compared to the reference group with consistently high mental health, and at this time were also more likely to married and less likely to be studying. Those with varying mental health were also more likely than the reference group to experience some difficulty managing financially. Those with frequently low mental health were likely to speak a language other than English and report difficulty managing financially, and were likely to be neither working nor studying. There were no differences between the trajectory groups in initial area of residence, parental education, or motherhood status.

Concerning health behaviors at Survey 1, those with improving mental health were less likely to be overweight; those with varying mental health were more likely to be current smokers; and the frequently low mental health group were more likely to be underweight than the reference group. There were no differences in alcohol use or physical activity between trajectory groups.

There were several differences on physical and psychosocial variables. Those with improving mental health were initially characterized by low general health, a history of partner abuse, and high weight or shape dissatisfaction. Those with varying mental health were characterized by low general health, high rate of primary care physician visits, higher likelihood of a previous diagnosis of depression, and weight or shape dissatisfaction. Those with frequently low mental health were initially characterized by low general health, higher likelihood of a previous diagnosis of depression, a history of partner abuse, and weight or shape dissatisfaction.

Table 3 shows the results of a similar multinomial logistic regression examining correlates of trajectory group membership at Survey 6, again using the consistently high mental health trajectory group as the reference group. On the demographic characteristics, those with improving mental health did not differ at Survey 6 from those with consistently high mental health. By contrast, those with varying mental health were more likely to find it very difficult to manage financially, less likely to be mothers, and more likely to be divorced, separated or widowed; those in the frequently low mental health trajectory group were more likely to find it very difficult to manage financially, more likely to be neither working nor studying, and less likely to be mothers.

Concerning health behaviors, at Survey 6 the improving mental health group were more likely to be nondrinkers or rare drinkers and to be former smokers than the reference group; they were also more likely to be underweight and less likely to be obese. The varying mental health group were more likely to drink alcohol rarely and to be former smokers; they were also more likely to be underweight and less likely to be overweight. The frequently low mental health group were characterized by greater likelihood of being in any alcohol consumption category other than low-risk drinking, and of being either current or former smokers.

The physical and emotional health variables again showed several differences. Those with improving mental health were more likely to report low general health, previous diagnosis of depression, and a history of partner abuse, than the reference group. Those with varying mental health and those with frequently low mental health were more likely to report low or moderate general health, previous diagnosis of depression, a history of partner abuse, and weight or shape dissatisfaction, than the reference group.

Longitudinal data from 5,171 women from a representative population-based study of young Australian women, initially aged 18 to 23 and followed on six occasions over 16 years, were used to identify mental health trajectory groups. We used the SF-36 Mental Health Index (MHI-5) to categorize each woman as having low or high mental health at each survey, according to previously established cut-offs. Using group-based trajectory modeling, we identified four distinct groups. Just over half the women met our criterion for high mental health at every survey. Around 10% experienced low mental health at the first or second survey but not subsequently, around a quarter showed a variable pattern over the 16 years, and around 10% experienced low mental health at most or all of the surveys.

We then used multinomial logistic regressions, first to identify variables at Survey 1 that might predict mental health trajectory over the next 16 years; and second to identify differences among the groups at Survey 6. The analysis has limitations that are typical of large-scale population-based cohort studies: a reliance on self-report, with some inconsistent reporting over time (e.g., some women changed their self-described smoking history from exsmoker to never-smoker between Surveys 1 and 6), some small demographic biases in inclusion, and biases in retention. However, the large sample size and use of well-validated measures designed for population surveys provide a degree of confidence in the pattern of findings.

A life course perspective is valuable in contextualizing the discussion of demographic, lifestyle, health-related, and psychosocial predictors and subsequent correlates of mental health patterns in early adulthood. This perspective emphasizes the cumulative importance of prior experiences as contributors to current health outcomes and health determinants, with the understanding that these factors unfold relative to one another (e.g., Umberson, Crosnoe, & Reczek, 2010). It also takes into account life stage, and the ages at which life events such as marriage and parenthood are normative for specific generational cohorts (e.g., Fussell & Gauthier, 2005). Within this approach, we consider not only individual characteristics over time, but also economic and social factors across the life course that affect health (Braveman, 2014).

Demographically, there were few differences among the four trajectory groups at Survey 1, when the women were aged 18 to 23. Those who would go on to have frequently low mental health were notable in that they were most likely to speak a language other than English at home and to be neither working nor studying; in addition, all groups other than those who maintained high mental health reported difficulty managing on their income. The improving mental health group, who had low mental health initially but not later in the survey period, were also more likely to be married and less likely to be studying without also working.

These findings are generally consistent with other research, both from Australia (e.g., Kiely & Butterworth, 2013) and from elsewhere (e.g., Molarius et al., 2009), which shows poor mental health to be associated with financial difficulties, unemployment, and receipt of welfare benefits. They are also consistent with evidence that young people not in education, employment, or training are at increased risk of poor mental health (e.g., UCL Institute of Health Equality, 2014) and social exclusion (e.g., Bynner & Parsons, 2002). Our finding of an association between poor mental health and ethnic minority or migrant status is also consistent with other research (e.g., Sirin, Ryce, Gupta, & Rogers-Sirin, 2013).

The finding that those with improving mental health—who generally had low mental health at Survey 1—were married might appear to contradict extensive evidence linking marriage with positive mental health (e.g., Liu & Umberson, 2008). However, from a life course perspective, this finding can be interpreted in light of their age (18 to 23) and the normative age of marriage in this cohort of women. Census data show that, in 1995, the median age of women at first marriage was 25.3 (Australian Bureau of Statistics, 1997); thus, these women’s low mental health may result from the social and educational disadvantages that are associated with early marriage (e.g., Maggs, Jager, Patrick, & Schulenberg, 2012). The lack of effect of motherhood may similarly be explained by the age of the cohort: although there is strong evidence from elsewhere (e.g., Schytt & Hildingsson, 2011) that pregnancy and early motherhood can have a profound effect on mental health, only 5% of these women were mothers, consistent with the median age of first-time mothers in 1996 being 28.7 (Australian Bureau of Statistics, 1996).

There were no clear patterns of health-related behaviors at baseline predicting membership of mental health trajectory groups. Alcohol consumption, smoking, body weight, and physical activity all showed no strong or consistent relationships with later mental health. This may be seen as inconsistent with studies of adolescents (e.g., Adrian, Charlesworth-Attie, Vander Stoep, McCauley, & Becker, 2014) showing a relationship between mental health and number of health behaviors endorsed. However, we also adjusted for sociodemographic factors in our analyses, whereas Adrian et al. (2014) did not, and it is well established that social disadvantage is a very strong correlate of smoking (e.g., Hefler & Chapman, 2014) and of obesity (e.g., Gustafsson, Persson, & Hammarström, 2012).

The health and psychosocial variables at Survey 1 may be more useful in understanding future trajectories of mental health. Although all comparison groups reported poor physical health by comparison with the consistently high trajectory group, the differences between trajectory groups on the psychosocial variables were more nuanced. Those in the varying and frequently low mental health groups were more likely to report a previous diagnosis of depression, although those in the improving group were not, suggesting that we may have in fact identified a group of women whose experience with poor mental health is a single isolated incident. The improving and frequently low groups—who both had low mean mental health at Survey 1—were more likely to report partner abuse and, given the age of these women at Survey 1, that abuse is likely to be relatively recent.

Patterns of association between trajectory groups and many of the same variables at Survey 6, when the women were aged 34 to 39, were somewhat different. Overall, there were very few differences between the improving mental health group and the consistently high group, both of whom were experiencing positive mental health at Survey 6. Those with improving mental health were more likely to have smoked in the past, to have been diagnosed with depression at some stage, and to have experienced partner abuse at some stage. Currently, the only differences were that the improving group were more likely than the consistently high group to drink alcohol rarely or not at all; to be underweight and not obese; and to be in the lowest tertile of general health.

By contrast, there were differences from the reference group on most of the variables for the varying mental health and the frequently low groups at Survey 6. Demographically, both groups reported difficulty managing on their incomes, and the varying group were also more likely to be divorced, separated or widowed. The largest demographic difference was in rates of motherhood. Additional analyses (not shown) indicate that 79% of those with consistently high mental were mothers by Survey 6, when they were aged 34 to 39, and the improving mental health trajectory group were not significantly different, with 75% being mothers. Although the effect for the varying mental health group was significant, the rate of motherhood at 73% was not meaningfully different from the reference group. By contrast, only 64% of those with frequently low mental health were mothers by this survey, despite the fact that this group did not differ from the reference group in relationship status. This is concerning in the light of separate analyses from this cohort (Johnstone & Lee, 2009) which show that, at each of Surveys 1, 2, and 3, over 90% of all participants reported that they would like to have children by the age of 35.

Evidence from the United States (Craig et al., 2014) shows that a substantial majority of childless women, including those in their late 30s and 40s, still want to become mothers. However, recent expert opinion (American College of Obstetricians and Gynecologists Committee on Gynecologic Practice and Practice Committee, 2014) is that women’s fertility begins to fall in the early 30s and to drop sharply from around 37. Thus, it is likely that many of these childless women will continue to be involuntarily childless throughout their lives, placing them at elevated risk of long-term depression and anxiety (e.g., Lechner, Bolman, & van Dalen, 2006).

The health behavior patterns of these two groups were generally less positive than those of the consistently high mental health group, with generally more extreme relative risk ratios for the frequently low mental health group than the varying mental health group. The varying mental health group were more likely to drink alcohol rarely, to be exsmokers, and to be underweight. The frequently low mental health group were more likely to report any pattern of alcohol use except low-risk drinking; they were also more likely to be current or exsmokers. Of the psychosocial variables, both these groups were also more likely to report moderate or low general health, previous depression, a history of partner abuse, and current weight or shape concerns. These differences suggest a pervasive and generalized range of problems among the women categorized in these groups, including unhealthy lifestyles and poor general and emotional health.

This research is of particular interest for health policy and health promotion interventions because of the potential to intervene to change life-course trajectories during early adulthood (e.g., Schulenberg et al., 2004). Others (e.g., Braveman, 2014; Klawetter, 2014) have argued that addressing the social determinants of health inequalities, such as financial inequality, is essential to improving the mental health of populations. We found no association between parent education, a proxy for socioeconomic status during childhood, and poor mental health in early adulthood. But current financial difficulty was associated with poor mental health for all three comparison groups at Survey 1, particularly for those experiencing poor mental health around that time. At Survey 6 current financial difficulty was only reported by those with chronic and recurrent poor mental health, not those whose mental health had improved. This suggests that chronic or recurrent poor mental health may be predictive of, rather than simply a consequence of, socioeconomic disparities, at least in early adulthood. This is broadly consistent with cross-sectional analyses from the U.S. National Epidemiologic Survey on Alcohol and Related Conditions, showing that respondents who reported a lifetime course of chronic major depression were more economically and educationally disadvantaged that those reporting 12-month prevalence but no lifetime history (Rubio et al., 2011).

The differences in health behaviors should also be considered from a life course perspective, because choices and changes made in early adulthood can have lasting effects into adulthood (Umberson et al., 2010). We found few differences between the trajectory groups at Survey 1 in either smoking or alcohol use, but by Survey 6, women from all comparison groups were less likely to describe low-risk drinking and more likely to have smoked in the past (only the frequently low mental health group were more likely to be current smokers). Exploration and risk-taking are well recognized as characteristics of early adulthood, whether or not an individual is experiencing poor mental health (e.g., Brodbeck, Bachmann, Croudace, & Brown, 2013). As with the demographic disparities, these findings suggest that chronic poor mental health may be a predictor, rather than a consequence, of health behavior choices in early adulthood.

All three comparison groups rated their general health as poor and reported dissatisfaction with their weight or shape at Survey 1, by comparison with those with consistently high mental health. These effects were maintained at Survey 6 by the frequently low and variable mental health groups, but the improving mental health group was no longer different from the reference group. This is consistent with the well-established link between mental and physical health (e.g., Robson & Gray, 2007) even in early adulthood (McCloughen, Foster, Huws-Thomas, & Delgado, 2012).

Overall, our findings seem consistent with the work of others (e.g., Hammen et al., 2008), suggesting that some women may experience poor mental health as an isolated event while for others it becomes chronic or recurrent; these groups are shown to have poorer outcomes in terms of physical and mental health, health behaviors, and childlessness. Our use of longitudinal data spanning 16 years of early adulthood has also enabled us to demonstrate that poor mental health is likely to predict rather than be a consequence of both financial difficulties and unfavorable health behaviors.

In summary, distinct trajectories of mental health can be identified in a population of young Australian women over a 16-year period: Women with frequently low and varying mental health are particularly at risk of other indicators of psychological distress, such as financial difficulty, and are more likely than others to remain childless. This population-level study provides a perspective that complements research focusing on individuals with clinical depression, demonstrating the potentially significant long-term outcomes among women of recurrent mental health problems across early adulthood.

Adrian, M., Charlesworth-Attie, S., Vander Stoep, A., McCauley, E., & Becker, L. (2014). Health promotion behaviors in adolescents: Prevalence and association with mental health in a state-wide sample. Journal of Behavioral Health Services Research, 41, 140–152. 10.1007/s11414-013–9370-y

American College of Obstetricians and Gynecologists Committee on Gynecologic Practice and Practice Committee. (2014). Female age-related fertility decline. Committee Opinion No. 589. Fertility and Sterility, 101, 633–634. 10.1016/j.fertnstert.2013.12.032

Armstrong, T., Bauman, A., & Davies, J. (2000). Physical activity patterns of Australian adults: Results of the 1999 National Physical Activity Survey. Canberra, Australia: Australian Institute of Health and Welfare.

Australian Bureau of Statistics. (1996). Births, Australia, 1996. Retrieved from http://www.ausstats.abs.gov.au/ausstats/free.nsf/0/6BFCAD8D7BD4945FCA257225000494D6/$File/33010_1996.pdf

Australian Bureau of Statistics. (1997). Australian social trends, 1997. Retrieved from http://www.abs.gov.au/AUSSTATS/abs@.nsf/2f762f95845417aeca25706c00834efa/a8d1bea8a2ff1b33ca2570ec001b0dc3!OpenDocument

Australian Bureau of Statistics. (2006). Australian standard geographical classification. Retrieved from http://www.abs.gov.au/websitedbs/D3310114.nsf/home/Australian+Standard+Geographical+Classification+(ASGC)

Australian Bureau of Statistics. (2008). National survey of mental health and wellbeing: Summary of results, 2007(cat no. 4326.0). Retrieved from http://www.abs.gov.au/AUSSTATS/abs@.nsf/Latestproducts/4326.0Main%20Features32007?opendocument&tabname=Summary&prodno=4326.0&issue=2007&num=&view=

Australian Longitudinal Study on Women’s Health. (2013). Tech. Rep. No. 36. Retrieved from http://www.alswh.org.au/images/content/pdf/technical_reports/report_36_alswh.pdf

Bonanno, G. A. (2005). Resilience in the face of potential trauma. Current Directions in Psychological Science, 14, 135–138. 10.1111/j.0963-7214.2005.00347.x

Braveman, P. (2014). What is health equity: And how does a life-course approach take us further toward it?Maternal and Child Health Journal, 18, 366–372. 10.1007/s10995-013-1226-9

Brodbeck, J., Bachmann, M. S., Croudace, T. J., & Brown, A. (2013). Comparing growth trajectories of risk behaviors from late adolescence through young adulthood: An accelerated design. Developmental Psychology, 49, 1732–1738. 10.1037/a0030873

Brown, W. J., Dobson, A. J., Bryson, L., & Byles, J. E. (1999). Women’s Health Australia: On the progress of the main cohort studies. Journal of Women’s Health & Gender-Based Medicine, 8, 681–688. 10.1089/jwh.1.1999.8.681

Brown, W. J., Ford, J. H., Burton, N. W., Marshall, A. L., & Dobson, A. J. (2005). Prospective study of physical activity and depressive symptoms in middle-aged women. American Journal of Preventive Medicine, 29, 265–272. 10.1016/j.amepre.2005.06.009

Burton, N. W., Brown, W., & Dobson, A. (2010). Accuracy of body mass index estimated from self-reported height and weight in mid-aged Australian women. Australian and New Zealand Journal of Public Health, 34, 620–623. 10.1111/j.1753-6405.2010.00618.x

Bynner, J., & Parsons, S. (2002). Social exclusion and the transition from school to work: The case of young people not in education, employment, or training (NEET). Journal of Vocational Behavior, 60, 289–309. 10.1006/jvbe.2001.1868

Campbell, S. B., Matestic, P., von Stauffenberg, C., Mohan, R., & Kirchner, T. (2007). Trajectories of maternal depressive symptoms, maternal sensitivity, and children’s functioning at school entry. Developmental Psychology, 43, 1202–1215. 10.1037/0012-1649.43.5.1202

Colman, I., Ploubidis, G. B., Wadsworth, M. E. J., Jones, P. B., & Croudace, T. J. (2007). A longitudinal typology of symptoms of depression and anxiety over the life course. Biological Psychiatry, 62, 1265–1271. 10.1016/j.biopsych.2007.05.012

Craig, B. M., Donovan, K. A., Fraenkel, L., Watson, V., Hawley, S., & Quinn, G. P. (2014). A generation of childless women: Lessons from the United States. Women’s Health Issues, 24(1), e21–e27. 10.1016/j.whi.2013.09.005

Eaton, W. W., Shao, H., Nestadt, G., Lee, H. B., Bienvenu, O. J., & Zandi, P. (2008). Population-based study of first onset and chronicity in major depressive disorder. Archives of General Psychiatry, 65, 513–520. 10.1001/archpsyc.65.5.513

Fairburn, C. G., & Beglin, S. J. (1994). Assessment of eating disorders: Interview or self-report questionnaire?International Journal of Eating Disorders, 16, 363–370.

Fussell, E., & Gauthier, A. (2005). American women’s transition to adulthood in comparative perspective. In R. A.Settersten, F. F.Furstenberg, & R. G.Rumbaut (Eds.), On the frontier of adulthood: Theory, research, and public policy (pp. 76–109). Chicago, IL: University of Chicago Press. 10.7208/chicago/9780226748924.003.0003

Galambos, N. L., Barker, E. T., & Krahn, H. J. (2006). Depression, self-esteem, and anger in emerging adulthood: Seven-year trajectories. Developmental Psychology, 42, 350–365.

Gustafsson, P. E., Persson, M., & Hammarström, A. (2012). Socio-economic disadvantage and body mass over the life course in women and men: Results from the Northern Swedish Cohort. European Journal of Public Health, 22, 322–327. 10.1093/eurpub/ckr061

Hamdi, N. R., & Iacono, W. G. (2014). Lifetime prevalence and co-morbidity of externalizing disorders and depression in prospective assessment. Psychological Medicine, 44, 315–324. 10.1017/S0033291713000627

Hammen, C., Brennan, P. A., Keenan-Miller, D., & Herr, N. R. (2008). Early onset recurrent subtype of adolescent depression: Clinical and psychosocial correlates. Journal of Child Psychology and Psychiatry, and Allied Disciplines, 49, 433–440. 10.1111/j.1469-7610.2007.01850.x

Hefler, M., & Chapman, S. (2014). Disadvantaged youth and smoking in mature tobacco control contexts: A systematic review and synthesis of qualitative research. Tobacco Control, 24, 429–435.

Herrman, H., Saxena, S., & Moodie, R. (Eds.). (2005). Promoting mental health: Concepts, emerging evidence, practice: Report of the World Health Organization. Geneva, Switzerland: World Health Organization. Retrieved from http://www.who.int/mental_health/evidence/MH_Promotion_Book.pdf

Holmes, W. C. (1998). A short, psychiatric, case-finding measure for HIV seropositive outpatients: Performance characteristics of the 5-item Mental Health Subscale of the S-20 in a male, seropositive sample. Medical Care, 36, 237–243. Retrieved from http://www.jstor.org/stable/3767185

Howard, A. L., Galambos, N. L., & Krahn, H. J. (2010). Paths to success in young adulthood from mental health and life transitions in emerging adulthood. International Journal of Behavioral Development, 34, 538–546. 10.1177/0165025410365803

Johnstone, M., & Lee, C. (2009). Young Australian women’s aspirations for work and family. Family Matters, 81, 5–14.

Kiely, K. M., & Butterworth, P. (2013). Social disadvantage and individual vulnerability: A longitudinal investigation of welfare receipt and mental health in Australia. Australian and New Zealand Journal of Psychiatry, 47, 654–666. 10.1177/0004867413484094

Klawetter, S. (2014). Conceptualizing social determinants of maternal and infant health disparities. Journal of Women and Social Work, 29, 131–141. 10.1177/0886109913516451

Kuchibhatla, M. N., Fillenbaum, G. G., Hybels, C. F., & Blazer, D. G. (2012). Trajectory classes of depressive symptoms in a community sample of older adults. Acta Psychiatrica Scandinavica, 125, 492–501. 10.1111/j.1600-0447.2011.01801.x

Lechner, L., Bolman, C., & van Dalen, A. (2006). Definite involuntary childlessness: Associations between coping, social support and psychological distress. Human Reproduction, 22, 288–294. 10.1093/humrep/del327

Lee, C., Dobson, A. J., Brown, W. J., Bryson, L., Byles, J., Warner-Smith, P., & Young, A. F. (2005). Cohort profile: The Australian Longitudinal Study on Women’s Health. International Journal of Epidemiology, 34, 987–991. 10.1093/ije/dyi098

Liu, H., & Umberson, D. J. (2008). The times they are a changin’: Marital status and health differentials from 1972 to 2003. Journal of Health and Social Behavior, 49, 239–253. 10.1177/002214650804900301

Maggs, J. L., Jager, J., Patrick, M. E., & Schulenberg, J. (2012). Social role patterning in early adulthood in the USA: Adolescent predictors and concurrent wellbeing across four distinct configurations. Longitudinal and Life Course Studies, 3, 190–210.

McCloughen, A., Foster, K., Huws-Thomas, M., & Delgado, C. (2012). Physical health and wellbeing of emerging and young adults with mental illness: An integrative review of international literature. International Journal of Mental Health Nursing, 21, 274–288. 10.1111/j.1447-0349.2011.00796.x

Molarius, A., Berglund, K., Eriksson, C., Eriksson, H. G., Lindén-Boström, M., Nordström, E., . . .Ydreborg, B. (2009). Mental health symptoms in relation to socio-economic conditions and lifestyle factors—A population-based study in Sweden. BMC Public Health, 9, 302. 10.1186/1471-2458-9-302

Monroe, S. M., & Harkness, K. L. (2012). Is depression a chronic mental illness?Psychological Medicine, 42, 899–902. 10.1017/S0033291711002066

Nagin, D. S. (2005). Group-based modeling of development. Cambridge, MA: Harvard University Press. 10.4159/9780674041318

Nagin, D. S., & Odgers, C. L. (2010). Group-based trajectory modeling in clinical research. Annual Review of Clinical Psychology, 6, 109–138. 10.1146/annurev.clinpsy.121208.131413

National Health & Medical Research Council. (2001). Australian alcohol guidelines: Health risks and benefits. Endorsed October 2002. Retrieved from http://www.nhmrc.gov.au/_files_nhmrc/publications/attachments/ds9.pdf

National Health & Medical Research Council. (2013). Clinical practice guidelines for the management of overweight and obesity in adults. Retrieved from https://www.nhmrc.gov.au/_files_nhmrc/publications/attachments/n57_obesity_guidelines_131204_0.pdf

Rhebergen, D., Batelaan, N. M., de Graaf, R., Nolen, W. A., Spijker, J., Beekman, A. T., & Penninx, B. W. (2011). The 7-year course of depression and anxiety in the general population. Acta Psychiatrica Scandinavica, 123, 297–306. 10.1111/j.1600-0447.2011.01677.x

Robson, D., & Gray, R. (2007). Serious mental illness and physical health problems: A discussion paper. International Journal of Nursing Studies, 44, 457–466. 10.1016/j.ijnurstu.2006.07.013

Rose, G. (1985). Sick individuals and sick populations. International Journal of Epidemiology, 14, 32–38. 10.1093/ije/14.1.32

Rubio, J. M., Markowitz, J. C., Alegría, A., Pérez-Fuentes, G., Liu, S.-M., Lin, K.-H., & Blanco, C. (2011). Epidemiology of chronic and nonchronic major depressive disorder: Results from the national epidemiologic survey on alcohol and related conditions. Depression and Anxiety, 28, 622–631. 10.1002/da.20864

Rumpf, H. J., Meyer, C., Hapke, U., & John, U. (2001). Screening for mental health: Validity of the MHI-5 using DSM–IV Axis I psychiatric disorders as gold standard. Psychiatry Research, 105, 243–253. 10.1016/S0165-1781(01)00329-8

Schulenberg, J. E., Sameroff, A. J., & Cicchetti, D. (2004). The transition to adulthood as a critical juncture in the course of psychopathology and mental health. Development and Psychopathology, 16, 799–806. 10.1017/S0954579404040015

Schytt, E., & Hildingsson, I. (2011). Physical and emotional self-rated health among Swedish women and men during pregnancy and the first year of parenthood. Sexual & Reproductive Health, 2, 57–64. 10.1016/j.srhc.2010.12.003

Segal, Z. V., Pearson, J. L., & Thase, M. E. (2003). Challenges in preventing relapse in major depression. Report of a National Institute of Mental Health Workshop on state of the science of relapse prevention in major depression. Journal of Affective Disorders, 77, 97–108. 10.1016/S0165-0327(02)00112-X

Silveira, E., Taft, C., Sundh, V., Waern, M., Palsson, S., & Steen, B. (2005). Performance of the SF-36 Health Survey in screening for depressive and anxiety disorder in an elderly female Swedish sample. Quality of Life Research: An International Journal of Quality of Life Aspects of Treatment, Care & Rehabilitation, 14, 1263–1274. 10.1007/s11136-004-7753-5

Sirin, S. R., Ryce, P., Gupta, T., & Rogers-Sirin, L. (2013). The role of acculturative stress on mental health symptoms for immigrant adolescents: A longitudinal investigation. Developmental Psychology, 49, 736–748. 10.1037/a0028398

Tanner, J. L., Reinherz, H. Z., Beardslee, W. R., Fitzmaurice, G. M., Leis, J. A., & Berger, S. R. (2007). Change in prevalence of psychiatric disorders from ages 21 to 30 in a community sample. Journal of Nervous and Mental Disease, 195, 298–306. 10.1097/01.nmd.0000261952.13887.6e

UCL Institute of Health Equality. (2014). Local action on health inequalities: Reducing the number of young people not in employment, education or training (NEET); Health Equity Briefing 3: September 2014. Retrieved from https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/356062/Review3_NEETs_health_inequalities.pdf

Umberson, D. L., Crosnoe, R., & Reczek, C. (2010). Social relationships and health behavior across the life course. Annual Review of Sociology, 36, 139–157. 10.1146/annurev-soc-070308-120011

Ware, J., Snow, K., Kosinski, M., & Grandek, B. (1993). SF-36 Health Survey Manual and Interpretation Guide. Boston, MA: The New England Medical Centre.

Submitted: January 8, 2015 Revised: August 26, 2015 Accepted: September 9, 2015