Course,Predictors,and Outcomes of Depressed Mood in a Cohort of Norwegian Adolescents Followed From Age 13 Into Adulthood Age 30

Depressed Mood Development, Average Levels, and Individual Trajectories

The theoretical perspective we draw on in the present study focuses on development as a dynamic interaction between the individual and their context throughout the life course (Graber & Brooks-Gunn, 1996; Lerner, Almerigi, Theokas, & Lerner, 2005). Although development has been postulated by some scholars as mainly continuous within an individual, development may not always follow continuous smooth paths; there might for example be low points in an otherwise healthy trajectory (Howard et al., 2010).

Depressed mood development during the period of adolescence is well documented (e.g., Cole et al., 2002; Edwards et al., 2014; Castelao & Kröner-Herwig, 2013; Mezulis, Salk, Hyde, Priess-Groben, & Simonson, 2014). Although less is known about the emerging adulthood period and the extension into adulthood, research on the respective periods the last decades indicate that on average, depressed mood and depressive symptoms increase from ages 13 to 14, peak in mid-to-late adolescence (Adkins, Wang, Dupre, van den Oord, & Elder, 2009; Cole et al., 2002; Ferro, Gorter, & Boyle, 2015; Hankin et al., 1998; Natsuaki, Biehl, & Ge, 2009), and decline into emerging adulthood (Elovainio et al., 2012; Galambos, Barker, & Krahn, 2006; Ge, Natsuaki, & Conger, 2006; Gutman & Eccles, 2007; Rawana & Morgan, 2014; Vujeva & Furman, 2011). Females generally report higher levels of depressive symptoms at all ages commencing around ages 12–13 (Adkins et al., 2009; Hankin, et al., 1998; Rawana & Morgan, 2014; Twenge & Nolen-Hoeksema, 2002). In addition, females show in most studies a steeper increase in adolescence and a steeper decrease into young adulthood compared to males (Cole et al., 2002; Dekker et al., 2007).

Most of the recent studies complement findings from previous studies by also examining within- and between-individual differences in depressive symptom trajectories, thus providing a more nuanced picture of development. For example, Nocentini, Calamai, and Menesini (2012) found significant variances in individual intercepts and slopes in depressed mood during the period of adolescence. A robust body of evidence finds that most adolescents with initial low depressed mood levels follow a low trajectory throughout adolescence (Brendgen, Lamarche, Wanner, & Vitaro, 2010; Castelao & Kröner, 2013; Costello, Swendsen, Rose, & Dierker, 2008; Ferro et al., 2015; Mezulis et al., 2014). Other youth display a variety of individual trajectories; low and increasing, high and decreasing, or stable high depressive symptoms scores through adolescence (Mezulis et al., 2014). Studies reporting on heterogeneity during the emerging adulthood period find that most individuals report decreasing levels of depressive symptoms (Ferro et al., 2015; Yaroslavsky, Pettit, Lewinsohn, Seeley, & Roberts, 2013). Furthermore, findings have also demonstrated that a small group of individuals report relatively high and increasing depressed mood scores (Costello et al., 2008; Salmela-Aro, Aunola, & Nurmi, 2008).

Thus, whereas most adolescents and emerging adults experience stable low levels of depressed mood in line with healthy growth and adaptation (Lerner et al., 2005), others experience changes in development of depressed mood. Few studies have so far examined intraindividual changes in depressed mood in a cohort of adolescents followed through the period of emerging adulthood. The present study will thus extend research in this area.

Links Between Depressed Mood Trajectory, Adolescents’ Relationship With Parents and Peers, and Socioeducational Status

The levels and rates of change in depressed mood may differ according to individual and sociocultural factors. Previous studies have identified several factors that may influence levels of depressed mood and depressive symptoms. In the present study, we will focus on three important predictors, namely, adolescents’ relationship with parents and peers as well as parental education level.

Consequences of Depressed Mood During Adolescence for Adulthood Functioning

Depressive symptoms during adolescence may lead to difficulties with the major developmental tasks associated with this life period. For example, individuals with depressed mood may avoid challenges within the educational, occupational, and relational arenas of life, which may also have consequences for adulthood SES. Previous cross-sectional research has found that depressive symptoms impair functioning in various domains such as work and family life (Fried & Nesse, 2014), and are associated with occupational difficulties (Judd, Paulus, Wells, & Rapaport, 1996) and social dysfunction (Leader & Klein, 1996). The long-term impact, however, of depressed mood trajectories during the transition to adulthood has rarely been studied. The few studies that exist indicate that depressive symptoms during adolescence are linked to later outcomes such as increased risk of educational underachievement and unemployment (Fergusson & Woodward, 2002), and lower probability of having graduated from higher education by age 30 (Jonsson et al., 2010). Also, in a longitudinal study of trajectories of depressive symptoms from the ages of 21 to 30, individuals with trajectories of high and moderate levels of depressive symptoms reported less success expectations in both achievement and social situations and higher levels of social avoidance than individuals with low levels of depressive symptoms (Salmela-Aro et al., 2008). In the same study, young adults with low- or moderate-depression trajectories reported higher salary levels than those with high-depression trajectories. A similar study of adolescents followed from the age of approximately 16–30 found that those in the high and moderate symptom groups attained a lower level of education and reported lower annual household income at age 30 (Yaroslavsky et al., 2013). Thus, depressive symptoms during adolescence may be associated with later functioning within work and family domains.

Life satisfaction is a more global indicator of adulthood functioning and can be viewed as a cognitive evaluation of one’s quality of life (Pavot & Diener, 2008). Studies indicate that having experienced depressive symptoms during adolescence may lead to lower life satisfaction in adulthood. For example, individuals who had major depressive episodes as adolescents reported reduced life satisfaction at age 24 (Lewinsohn, Rohde, Seeley, Klein, & Gotlib, 2003). In another study, poor trajectories of depressive symptoms from age 18 to 25 predicted life satisfaction at age 32 (Howard et al., 2010).

In adulthood (age 30), life satisfaction may also be tied to having completed key developmental tasks associated with this life period. Among the key developmental tasks in the period of transition to adulthood are obtaining employment, finding a partner, getting married, and becoming a parent (Schulenberg, O’Malley, Bachman, & Johnston, 2005). As already described, the probabilities of completing these tasks may be influenced by depressive symptoms during adulthood. However, completing these tasks may also act as turning points in trajectories of depressive mood, serving to shift the direction of individual paths at the time of the transition (Singer & Willett, 2003). This means that the link between trajectories of depressed mood during adolescence and emerging adulthood, and life satisfaction in adulthood, may be accounted for by completion of adulthood key tasks and/or SES. Therefore, Howard and colleagues studied the net effect of trajectories of depressed symptoms from the ages of 18 to 25 on life satisfaction at age 32, and found that levels of depressive symptoms at age 18 were associated with life satisfaction at age 32 for both genders, but that changes in depressive symptoms during this period of life carried over into lower life satisfaction at age 32 for women but not for men (Howard et al., 2010). Furthermore, lower depressive symptoms at age 23 and steeper declines in depressive symptoms from age 23 to 30 predicted higher life satisfaction at age 37, after accounting for family and labor experiences (Howard, Galambos, & Krahn, 2014). Howard and colleagues’ studies have focused on development of depressive symptoms from the age of 18, and we wanted to explore whether these processes begin even earlier. In addition, we argue that controlling for the effects of experiences during adolescence and emerging adulthood on depressive symptoms may limit the understanding of the consequences of depressive symptoms. If indeed experiences in adolescence have a substantial role in later development, one should not partial it out, as it can lead to removing the effects of the very variables one wishes to study (Spector, Zapf, Chen, & Frese, 2000). Thus, this study contributes to the literature by examining whether the natural course of depressed mood from the age of 13 have consequences for adulthood functioning.

Given the above reasoning and empirical findings we aimed to

describe the course of depressed mood from adolescence age 13 to emerging adulthood age 23,

Sample and Procedure

Data stem from the Norwegian Longitudinal Health Behavior Study (NLHB), a cohort longitudinal study started among seventh graders (13-year-olds) in Hordaland County, Norway, in 1990. Data collections in the NLHB study were performed in 1990, 1991, 1992, 1993, 1995, 1996, 1998, 2000, and 2007. Twenty-two schools were drawn from all 130 urban and rural schools in the county. The initial sample was 1,195 adolescents. Written informed consent was obtained from 927 students and their parents. Refusals from parents (n = 222; 18.6%), refusals from students (n = 46; 3.8%), and students with obviously inconsistent answers (crossing everywhere and anywhere in the survey; n = 3; 0.3%) resulted in a sample in 1990 of 924 participants (participation rate = 77%). In 1992, the students who were new in the participating classes were invited to participate. The total number of participants in the NLHB study then became 1,242. No data on race and ethnicity exist, but because Hordaland County has a rather homogenous population, it can be assumed that most of the participants were of Norwegian or Caucasian origin.

For the first three data collections, the participants responded to a questionnaire, which was administered during school hours by the researchers and trained staff. The data collections in 1998, 2000, and 2007 were completed by mail. In this study, we used data from the collections in 1990 (mean age 13.3), 1992 (mean age 15.3), 1995 (mean age 18.3), 1998 (mean age 21.3), 2000 (mean age 23.3), and 2007 (mean age 30.5). The latest data collection, 2007, took place in November, the others in September. In addition, we used information from the adolescents’ fathers from 1996. The number of respondents was 963 in 1992, 779 in 1995, 634 in 1998, 627 in 2000, and 542 in 2007. Of the 1,242 respondents, some did not provide data on the measures used in this study, and the final sample that provided data for this study included 1,084 participants (54.3% males and 45.7% females).

Depressed Mood

Depressed mood was measured by applying a 7-item scale developed by Alsaker and Olweus (Alsaker, 1992). The items were as follows: (a) I often feel depressed without really knowing why, (b) I do not think that I have anything to look forward to, (c) sometimes I think everything is so hopeless that I do not feel like doing anything, (d) sometimes I am just so depressed that I feel most like staying in bed for the whole day, (e) I am often sad for no reason, (f) I think my life is mostly miserable, and (g) sometimes I think my life is not worth living. The answering categories for the depressed mood scale were as follows: (1) does not apply at all, (2) does not apply well, (3) applies somewhat, (4) applies fairly well, (5) applies well, and (6) applies exactly. Cronbach’s α was .81, .87, .90, .92, .91, in 1990, 1992, 1995, 1998, and 2000, respectively. This scale has shown satisfactory psychometric properties and high correlation of .82 with the “Center for Epidemiological Studies Depression Scale” (Holsen, Kraft, & Vittersø, 2000; Radloff, 1977).

Father’s Educational Status

The fathers reported in 1996 (when their sons/daughters were 19 years old) on their own present educational status; “No education after 9-year primary school,” coded as 0; “1–2 years of upper secondary school,” coded as 1; “3 years of upper secondary school,” coded as 2; “College or university, less than 4 years,” coded as 3; “College or university, 4 years or more,” coded as 4; and “Other” coded as 0.

Relationships With Parents

Adolescents’ perceptions of having a close, positive, and supportive relationship with their parents were measured using 5 items (Alsaker, Dundas, & Olweus, 1991; Birkeland, Breivik, & Wold, 2014; Lien, Jacobs, & Klepp, 2002). Examples of the items include the following: “There is good cohesiveness in my family,” “I enjoy myself when I am with my parents,” and “My parents praise and encourage me.” The response categories for three of the items ranged from applies exactly (6) to does not apply at all (1), and the remaining two items ranged from “very often” (6) to “seldom or never” (1). Cronbach’s α was .84.

Relationships With Peers

Relationships with peers were measured by 7 items (Alsaker et al., 1991; Holsen, Carlson Jones, & Skogbrott Birkeland, 2012; Lien et al., 2002), and examples of the items include the following: “I am doing fine with others of my age”; “I think most of my classmates like me”; and “If I went to a new school, it would be easy for me to find new friends.” The six response categories ranged from applies exactly (6) to does not apply at all (1). Cronbach’s α was .84.

Life Satisfaction

Life satisfaction at age 30 was measured using Diener’s Satisfaction With Life Scale (Diener, Emmons, Larsen, & Griffin, 1985). Examples of items include the following: “In most ways, my life is close to my ideal”; “The conditions of my life are excellent”; and “I am satisfied with my life.” The seven response categories ranged from strongly disagree (1) to strongly agree (7). Cronbach’s α was .87.

The Young Adulthood Measures

Milestones at age 30 were as follows:

“Married/cohabiting”: At age 30, the participants were asked “Whom do you live with?” and provided with a list of possible living arrangements. Answers indicating marriage or cohabitation with a partner were coded as 1; all other answers were coded as 0.

Socioeconomic Status

Three indicators of SES at age 30 were used: the participant’s self-reported income in the previous year, the number of years of education undertaken after upper secondary school, and vocational status. Vocational status was estimated by coding the participant’s vocation using the Norwegian Standard for Occupational Classification (NOS C 539), which is based on the International Standard Classification of Occupations (ISCO 88). This standard classifies vocations that involve responsibilities or higher education, such as company leaders and academics, as high (10), and vocations that demand little responsibility or education, such as shop assistants and transport workers, as low (1).

Statistical Analyses

All of the data modeling were performed with the Mplus program, Version 7.2, using maximum likelihood estimation (Muthén & Muthén, 1998–2014). To correct for the skewed distributions in depressed mood, maximum likelihood estimation with robust errors was applied.

In the first step, we assessed gender similarities and differences in level of depressed mood at each age from 13 to 23 years using Wald tests. In the second step, we used latent growth modeling (LGM) to describe and analyze general development and predictors of the development of depressed mood over time. In LGM, one hypothesizes that scores over time are captured by the initial status and slope of individual changes. Individuals are allowed to have different intercepts (initial status) and linear, quadratic, and cubic slopes (growth).

After the unconditional growth model was fitted, gender, relationships with parents, relationships with peers, and father’s level of education were added as predictors of both intercept and slopes. All predictors were brought into the models and estimated. To assess the appropriateness of the growth models, root mean square error of approximation (RMSEA) and comparative fit index (CFI) were assessed. Values of RMSEA below 0.05 and values of CFI above 0.95 were considered to denote a well-fitting model (Browne & Cudeck, 1992; Hu & Bentler, 1999). To test differences between models, Satorra–Bentler χ² difference tests were used (Satorra & Bentler, 2001).

Next, we estimated associations between depressed mood trajectories during adolescence and probabilities of having achieved adult milestones (marital status, parenthood, having full-time work) and SES at age 30. Due to computational difficulties with estimating relationships between quadratic and cubic slopes on the dichotomous variables difficulties (probably due to high model complexity relative to size of sample), we decided to estimate linear growth curves and to study whether the overall developmental course without quadratic and cubic slopes could predict adulthood functioning.

In the last step, we predicted levels of life satisfaction at age 30 with intercept and slopes of depressed mood, with and without controlling for influences of relevant predictors such as gender, marital status, parenthood, having full-time work, income, vocational status, and educational level.

Missing Data

The total sample consisted of 589 males and 495 females, making the total N = 1,084. The percentage of missing across variables ranged from 15% (items at age 15) to 58% (items at age 30). Most of the missing data were due to wave nonresponse.

To assess selective participation, participation in each of the waves was regressed on scores of depressed mood at previous ages. Participation at age 15 was not dependent on depressed mood at age 13. Participation at age 18 was not dependent on depressed mood at ages 13 or 15. Participation at age 21 was not dependent on depressed mood at ages 13, 15, or 18. Participation at age 23 was not dependent on depressed mood at ages 13, 18, or 21, but higher depression at age 15 decreased the probability of participating at age 23 (p = .05). Participation at age 30 was not dependent on depressed mood at ages 13, 15, or 18, but higher depression at ages 21 and 23 decreased the probability of participating at age 30 (p = .01 and p = .04). This indicates that participation in the later waves may be selective and that we have a situation in which missing is dependent of some of the measured variables—a situation that is commonly termed missing at random (MAR; Enders, 2010).

The Mplus 7.11 inbuilt full information maximum likelihood (FIML) estimation with robust standard errors was used to handle missing data. This approach assumes that data are MAR and that all of the observed information is used to produce the maximum likelihood estimation of parameters. This is one of the best approaches available to handle missing data (Graham, 2009).

Average Changes in Depressed Mood and Gender Differences

Overall, the average level of depressed mood is low through the entire period of the present study. In line with most previous studies (e.g., Cole et al., 2002; Natsuaki et al., 2009), the findings revealed a small general increase in depressed mood during adolescence. This was especially evident among females who consistently reported higher levels of depressed mood compared to males during adolescence and emerging adulthood. This confirms previous studies revealing that gender differences in prevalence of depressed mood develop before the age of 13 (Dekker et al., 2007; Nolen-Hoeksema & Girgus, 1994). Although the period of adolescence provides opportunities for positive growth, the present finding supports previous findings revealing increased emotional difficulties encountered in adolescence, particularly among females (Elovainio et al., 2012; Nolen-Hoeksema & Girgus, 1994).

Consistent with previous studies, we found a decline in depressed mood in emerging adulthood between ages 18 and 21 (Elovainio et al., 2012; Meadows et al., 2006; Rawana & Morgan, 2014). As in the study by Rawana and Morgen (2014), our data supported a cubic growth model among females; hence, the present study revealed a tendency toward stabilization of depressed mood levels among females after the age of 21.

This suggests that emerging adulthood may be a period with increased positive mental health experiences (Galambos et al., 2006; Pettit et al., 2011). Among the possible mechanisms for this decline in depressed mood may be that with age and increased independence, the emerging adults experience increased capacity for emotional regulation, new opportunities in life in education, work, social relationships, and living arrangements.

Relationship With Parents and Peers and Father’s Education Level as Predictors of Depressed Mood

Closeness and supportive relationship with parents predicted baseline levels of depressed mood at age 13 for both males and females. This may be interpreted in light of the attachment perspective of development (Bowlby, 1988), which posits that a close and supportive relationship with parents may constitute a secure base from which adolescents can build stable, positive emotional well-being. Unlike other studies (Meadows et al., 2006; Needham, 2008), we did not find that relationship with parents predicted changes in depressed mood. This may indicate that quality of relationships with parents may be important for establishing a baseline stable trait-like component of depressive mood. We did not find evidence that relationship with parents lead to further changes in depressed mood during this age period. It might be that these associations reflect stable relationships between the same individuals but this might also reflect that the main function of relationships with parents may be to constitute a profound base for further development.

Among males, fathers’ education levels predicted baseline depressed mood level at age 13. The findings among females revealed a similar tendency, but the estimate did not reach significance. It may be that the family of origin’s socioeducational status has a launching effect that sets off a particular developmental course.

Despite being associated with initial level of depressed mood, neither relationships with parents or peers nor socioeducational status predicted rate of change in depressed mood development, indicating that other and perhaps dynamic processes occurring in adolescence and emerging adulthood may explain the further course of depressed mood. Among these may be late maturational timing (Alsaker, 1992; Natsuaki et al., 2009), level of self-esteem (Morin, Maïano, Marsh, Janosz, & Nagengast, 2011; Rawana & Morgan, 2014), genetically based factors (Rutter et al., 1990), substance use and abuse (Schulenberg & Zarrett, 2006), and level of school stress (Brendgen et al., 2005; Repetto, Caldwell, & Zimmerman, 2005).

Depressed Mood Trajectory Predicting Outcomes in Adulthood Age 30

Strengths and Limitations

Among the major strengths of this study are its 17-year longitudinal design capturing development in depressed mood from 13 to 30 years, its relatively large sample size, and its use of sophisticated statistical analyses. The study also has some limitations. A considerable amount of the data was missing because of dropouts, and we also found evidence of what is termed MAR (Enders, 2010). In such situations, it is regarded appropriate to use FIML estimation with robust standard errors to handle missing data. However, the possibility of data missing not at random (MNAR) cannot be rejected. There are ways to minimize possible biases resulting from this situation by specifying a model that accounts for the missing data. However, these MNAR models rely on other untestable assumptions that are much narrower than the assumptions for an MAR analysis (Enders, 2010). Therefore, some methodologists have asserted that a well-conducted MAR analysis may be preferable to an MNAR analysis, even when there is reason to believe that missingness is related to the variable of interest (Enders, 2010). Thus, we argue that utilizing FIML estimation seems to be the best options for handling missing data in our study.

Additionally, due to computational difficulties when estimating relationships between quadratic and cubic slopes on the dichotomous variables (probably due to high model complexity relative to size of sample), we estimated linear growth curves and studied whether the overall developmental course could predict adulthood functioning. This may result in a crude and rudimentary estimation process that might conceal important relationships between depressed mood during important phases in adolescence and emerging adulthood and adulthood functioning.