Heterogeneous Trajectories of Depressive Symptoms in Late Middle Age: Critical Period,Accumulation,and Social Mobility Life Course Perspectives

Abstract

Objective: This study investigated patterns of depressive symptoms and whether socioeconomic status (SES) across the life course affects these trajectories using the critical period, accumulation, and social mobility models. Method: This study uses data from 8,532 adults, age 51 to 64, collected over 12 years from the Health and Retirement Study (observations = 25,887). A latent class analysis was performed to examine distinct depressive symptom trajectories; life course models were studied with multinomial logistic regression. Results: Four heterogeneous latent classes were identified for depression: Declining, Low, Increasing, and High and Increasing. The High and Increasing group was associated with a disadvantaged childhood SES, accumulated exposure to socioeconomic risks, and persistent SES disadvantage supporting the three life course models. Discussion: There was evidence of distinct profiles of depressive symptoms in late middle age and of interrelated life course mechanisms underlying the influences of childhood SES on later life depression.

Keywords

life course heterogeneous trajectories of depressive symptoms socioeconomic status late middle age

Introduction

Late middle age is an important period as social roles pressure adults toward retirement and require “socialization” into age-specific roles (Dannefer, 1987). In this life stage, people tend to experience major life events such as the death of loved ones, a decline in health status, and career setbacks, resulting in psychological stress or depression, often called “midlife crisis” (Choi & Lee, 2010). Across the adult age range, the average number of depressive symptoms was found to follow a U-shaped distribution: an intermediate level during early adulthood, a steep decline during early middle age, a rise back to an intermediate level during late middle age, and a consistent increase during old age (Miech & Shanahan, 2000; Sutin et al., 2013). As depressive symptoms appear to be less prevalent among the middle-aged compared with the younger or older age groups, there has been limited research on depressive symptoms specific to late middle age. However, a recent general population survey found that the rate of moderate or severe depressive symptoms increased by age, from 5.7% among young adults to 9.8% among middle-aged adults, but decreased in older adults (5.4%; Pratt & Brody, 2014).

The present study aims to contribute to current research on depressive symptoms among the middle-aged in two respects. First, a few studies have examined longitudinal changes in depressive symptoms in adulthood to gain a better understanding of their long-term course (Chen et al., 2011; Yang & George, 2005). Importantly, changes in depressive symptoms can be heterogeneous, and it is important to identify the underlying mechanisms of differential developmental changes. Recent studies on these heterogeneous trajectories were based on specific populations, most often adults with depressive symptoms (Cui, Lyness, Tang, Tu, & Conwell, 2008) or elders (ages 65+; Andreescu, Chang, Mulsant, & Ganguli, 2008; Byers et al., 2012). Even though Yang (2007) emphasized substantial age cohort heterogeneity in trajectories of depressive symptoms, most longitudinal studies on depressive symptoms used middle age or old age samples as if they were homogeneous, ignoring variations across the age cohorts (Liang, Xu, Quiñones, Bennett, & Ye, 2011; Sutin et al., 2013).

Second, from a life course perspective, the long-term relationship between early life socioeconomic deprivation and health in later life have most often been examined using the critical period, accumulation, or social mobility models (Kuh, Ben-Shlomo, Lynch, Hallqvist, & Power, 2003; Turrell et al., 2002). Prior research has suggested that these distinct life course models provide a more complete picture of the role of socioeconomic position on later life health (Lyu & Burr, 2016; Pudrovska & Anikputa, 2013). To date, no empirical research has examined to what extent these theoretical perspectives explain depressive symptoms in middle age. The life course perspective suggests that individual lives are shaped by sociohistoric environmental factors related to birth cohort (Elder & Johnson, 2003). However, most longitudinal research has not differentiated between the health trajectories in middle age and old age; their main focus has been on older adults (Byers et al., 2012). By using longitudinal data from the Health and Retirement Study (HRS) cohort of middle-aged adults, aged 51 to 64, and employing the three different models often used in life course studies on health, this study aims to provide a better understanding of the association between life course socioeconomic conditions and late middle-age depression.

Heterogeneity in Depressive Symptom Trajectory

Existing longitudinal studies of the course of depressive symptoms have followed individuals for several years (Chen et al., 2011; Liang et al., 2011). Evidence suggests that the long-term trajectories of depressive symptoms are heterogeneous. For most, depressive symptoms were transient; for others, symptoms were stable at some level or changed in varying degrees (Chin, Choi, & Wan, 2016; Lincoln & Takeuchi, 2010; Yang & George, 2005). A persistent increase in depressive symptoms has been found to be closely associated with other major health outcomes (Kuo, Lin, Chen, Chuang, & Chen, 2011), and some studies have yielded insights on the change patterns of depressive symptoms in terms of improvement and deterioration of health (Andreescu et al., 2008; Chen et al., 2011; Liang et al., 2011). There is limited literature on the heterogeneous trajectories of depressive symptoms in late middle age; the majority of existing studies have focused on children or older adults (Musliner, Munk-Olsen, Eaton, & Zandi, 2016). Prior studies including a late middle age sample explored depressive symptoms as if age groups were homogeneous (Hsu, 2012; Liang et al., 2011; Lincoln & Takeuchi, 2010) or followed a community-dwelling clinical sample diagnosed with depression (Cronkite et al., 2013). Middle age and old age cohorts experience differential exposure to social risk factors for depression, especially middle-aged adults who tend to face more life course changes, such as economic and employment status (Yang, 2007). One study followed a late middle age, nonclinical sample for a 13-year period and identified four distinct depression trajectory groups (persistently high, persistently low, increasing, decreasing; Melchior et al., 2013), but the sample was limited to employees of France’s national company who participated in the French GAZEL study. Using a national longitudinal sample, this study aims to explore heterogeneous trajectories of depressive symptoms over time during late middle age.

Critical Period, Accumulation, and Social Mobility Life Course Perspectives

The Critical Period, Accumulation, and Social Mobility models are conceptually distinguishable but closely interrelated, making it difficult to empirically distinguish differential effects, but together can provide a better understanding of how life course socioeconomic conditions can be related to later health (Hallqvist, Lynch, Bartley, Lang, & Blane, 2004). The critical period model examines the effects of early life adversity. The accumulation model introduces exposure to socioeconomic disadvantages throughout the life course, focusing on the impact of gradually accumulated risk on later health problems (O’Rand, 2003). The social mobility model adds changes in socioeconomic status (SES) like downward or upward mobility (Hallqvist et al., 2004). This study will explore how life course SES is associated with later depression at the same time generating an in-depth interpretation of the three models.

Research on the critical period model has explored the impact of early life on depressive symptoms in later life (Miech & Shanahan, 2000; Nicholson et al., 2008); findings suggest a long-lasting impact (Thompson, Syddall, Rodin, Osmond, & Barker, 2001). If early life SES had been found to yield a greater impact on later life depression than adult SES, the critical period model would have been strongly supported; however, evidence has been inconsistent, with some studies finding a strong impact on later depression independent of adulthood SES while others did not (Gale et al., 2011; Harper et al., 2002; Poulton et al., 2002). In most studies, the father’s occupation and/or parents’ educational attainment were used as early life SES indicators (Gilman, Kawachi, Fitzmaurice, & Buka, 2002), and most were cross-sectional or short follow-up (Luo & Waite 2005; Nicholson et al., 2008). This study adds a childhood poverty measure because it captures the overall SES of the family in early life (Haas, 2008), and we have little information on whether poverty exposure early in life shapes subsequent mental health irrespective of adult financial circumstances (Evans & Cassells, 2014).

The accumulation model posits that socioeconomic disadvantages accumulate over the life course. It has been suggested there is a relationship between the intensity and duration of exposure to SES disadvantages and health outcomes in adulthood (Cohen, Janicki-Deverts, Chen, & Matthews, 2010). Accumulated socioeconomic disadvantages are likely to lead to adulthood SES disadvantages, such as low educational attainment, low occupational status, low income, high unemployment, and even unstable marital status (Kwag, 2009). To better understand, life course researchers developed a cumulative SES disadvantage index that combines childhood and adulthood SES into a single measure to capture the relationship between accumulated SES disadvantages and later health (Luo & Waite, 2005). Most of these studies have found that inequalities in health increase as socioeconomic disadvantages accumulate; however, the importance of cumulative SES for adulthood depression has been less extensively studied (Gale et al., 2011).

Very few studies have examined the relationship between socioeconomic mobility and later life depression. The social mobility model suggests that social disparities in childhood may be partially or wholly attenuated by upward social mobility (Luo & Waite, 2005). For example, stable low or downward social mobility was associated with depressive symptoms among adults (Ward et al., 2016) and reports of a major depressive episode (Nicklett & Burgard, 2009). Upwardly mobile adults had fewer depressive symptoms compared with those with low SES throughout their life (Luo & Waite, 2005). Upward mobility may partially compensate for early life disadvantage, while downward mobility increases risk (Turrell et al., 2002). This study examines distinct trajectories of depressive symptoms in relation to SES mobility as depressive symptoms in later life may change over time (Liang et al., 2011), with those changes being heterogeneous (Chin et al., 2016).

Present Study

This study will contribute to an understanding of heterogeneous trajectories of depressive symptoms in late middle age and how these trajectories are associated with different life course models. To our knowledge, no prior study has investigated these issues by examining closely interrelated life course mechanisms. This study addresses the following research hypotheses:

Hypothesis 1: Childhood SES disadvantages have a direct association with high or increasing depressive symptom trajectories (critical period model).

Hypothesis 2: Individuals with socioeconomic disadvantages across the life course are more likely to have a high or increasing depressive symptom trajectory compared with those with less exposure to socioeconomic disadvantages (accumulation model).

Hypothesis 3: Individuals with constantly high SES or upward SES mobility are more likely to experience a lower or decreasing depressive symptom trajectory (social mobility model).

Method

Data and sample

Data used in this study were obtained from seven waves of the HRS between 1998 and 2010. The HRS is a longitudinal survey with a nationally representative sample of older adults aged 51 years and older in the United States. (Juster & Suzman, 1995). The HRS includes individuals from four age cohorts: Asset and Health Dynamics Among the Oldest Old (AHEAD; born prior to 1924), Children of the Depression Age (CODA; born 1924-1930), the HRS cohort (born 1931-1941), and War Babies (WB; born 1942-1947; Soldo, Hurd, Rodgers, & Wallace, 1997). Prior to 1998, data collection on the original HRS cohort occurred on even years and for AHEAD in odd years; since 1998, the HRS has consolidated data collections. Our sample included adults aged between 51 and 64 years during our study period (observations = 53,607). For exploring trajectories of depressive symptoms, we restricted our sample to those who responded to the surveys in at least three waves (12,912 observations dropped) and reported their childhood circumstances at least once. Some childhood information was missing: poverty experience (0.18% of the sample), parental education (6.50%), and self-rated health (0.02%). Some adulthood information was also missing: work status (0.02% of the sample) and educational attainment (0.23%). Listwise deletion for missing information on either the dependent or the explanatory variables led to reductions (3,082 observations dropped), resulting in a total sample size of 25,887 observations among 8,532 adults. Because the range of age among our sample is between 51 and 64, only a few died or dropped out (N = 164; 0.63%) during the study period (1998 to 2010). Including attrition status as a control variable during sensitivity tests introduced little change in effect sizes for the other variables, so attrition was not included in the models.

Measures

Dependent variable: Depressive symptoms

The HRS depressive symptoms measure is a subset (eight items) of the Center for Epidemiologic Studies–Depression Scale (CES-D; 20 items) The negative indicators measure whether the respondent experienced symptoms all or most of the time: depression, everything is an effort, sleep is restless, felt alone, felt sad, and could not get going. The positive indicators measure whether the respondent felt happy and enjoyed life all or most of the time. A summary score (range = 0-8) was constructed with the positive items reverse coded such that a higher score indicated more negative symptoms. These scales were reliable (Cronbach’s α = .81).

Key independent variables: Measurement of life course socioeconomic position

Critical period model

Two measures of childhood SES were used in the critical period model: Parental education and financial status. Parental education was measured using the highest level of education completed by either mother or father; those who did not finish high school were coded 1, while those with a high school education or above made up the reference group (coded 0). Financial status in childhood has been used by many life course studies for the overall SES of the family during the respondent’s childhood (Haas, 2008). Financial status was assessed by whether the respondent indicated his or her family was poor (yes/no). The quality of this retrospective measure of childhood financial status has been confirmed in a robust sociological literature, and the variable from the HRS has been widely used (Luo & Waite, 2005). The educational attainment of the respondent was measured using the same binary categories as parental education. Household income was self-reported total income from all sources received in the previous year and was log-transformed due to high skewness.

Accumulation model

Cumulative SES disadvantage is based on the SES of respondents and their parents, measured using parental SES index, respondent’s education, and respondent’s income, similar to measures used in a study by Turrell and colleagues (2002). The measure reflects the respondent’s socioeconomic trajectory from childhood to adulthood. To create Cumulative SES disadvantage, the education and income levels of respondents and their parents were categorized into two groups: less than high school (1) versus high school or higher (0) and poor (1) versus nonpoor (0). To create the parental socioeconomic index, we first summed low parental education and childhood low financial status, ranging from 0 to 2, and created two categories based on a median split: the group at or above the median was coded 1 (the disadvantaged) and the group below the median was coded 0 (the nondisadvantaged). The cumulative disadvantage variable was created by summing the three socioeconomic indicators and grouping respondents into four categories: low socioeconomic position at all three points over the life course (labeled as greatest exposure; coded 3), low position at two points (moderate exposure; 2), low at one point (low exposure; 1), and never low at any point (no exposure; 0). We also considered using respondents’ occupation as another indicator of SES, but the addition of occupation did not yield additional information, consistent with previous studies (Turrell et al., 2002), so it was not included.

Socioeconomic mobility model

To create socioeconomic mobility patterns, respondents’ income and education were summed and divided into two categories based on a median split, ranging from 0 (no disadvantage) to 1 (disadvantaged). Based on the binary SES measures of parents and respondents, four SES mobility patterns were created: (0) high-high (below median childhood SES disadvantage and below median adulthood SES disadvantage), (1) low-low (above median childhood SES disadvantage and above median adulthood SES disadvantage), (2) low-high (upward; above median childhood SES disadvantage and below median adulthood SES disadvantage), and (3) high-low (downward; below median childhood SES disadvantage and above median adulthood SES disadvantage).

Covariates and Adulthood Factors

Respondents’ age at baseline was used in analyses. A dummy variable for gender (female = 1) was created. Race/ethnicity was coded into three categories: Caucasians (reference; 0), non-Hispanic African Americans (1), and others (2). There were several time-varying variables including marital status, work status, and adulthood health conditions. We created variables that reflected changes over time. Specifically, respondents who were married during the whole study period formed the reference group; those who were unmarried the whole time were coded 1; and those whose marital status changed were coded 2. Work status was measured in the same manner: constantly worked (0), constantly not worked (1), and work status changed (2). Finally, we created time-varying dummies for adulthood health outcomes. Self-rated health was measured at baseline and a variable was created to indicate no change in health (reference; 0), health became better (1); or health became worse (2). As there were no respondents whose number of chronic diseases decreased, just two time-varying dummies were created: no change in number of chronic diseases (reference; 0) and increased number of chronic diseases (1).

Statistical Analyses

We performed latent class analysis using Mplus Version 7 software (Muthén & Muthén, 2015). This allowed us to (a) identify study members who experienced distinct patterns of depressive symptoms between 1998 and 2010, (b) examine the depressive symptom trajectory patterns of each subgroup, (c) determine the proportion of the total sample in each group, and (d) profile the characteristics of individuals within each subgroup.

To identify long-term multiple trajectories of depressive symptoms, we determined whether there was a linear or nonlinear relationship between age and changes in depressive symptoms. Specifically, the models were run without covariates to establish the best fitting model and thus the number and shape of the trajectories. The analyses started with an intercept-only model, to which linear and quadratic growth factors were added to determine the forms of the growth model. A linear shape trajectory provided a better fit with the data than a quadratic shape. The next step was to establish the optimal number of latent classes to describe long-term depressive symptom patterns. Models one through six classes were estimated. Based on the smallest Bayesian information criterion (BIC; Schwarz, 1978), the Entropy Index (Muthén, 2003), the Lo-Mendell-Rubin (LMR) likelihood ratio test of model fit (Lo, Mendell, & Rubin, 2001), and the LMR test p value (Muthén, 2003), the model with five classes was considered the best for the developmental trajectories of depressive symptoms. However, as we defined adequate class size as at least 5%, a value typically considered the threshold for a meaningful class (Hipp & Bauer, 2006) and considered high posterior probabilities (near 1.0), the four-class model was chosen (Table 1) because the five-class model had two groups under 5%. Each class of depressive symptom trajectory was checked with plots to assure classes represented within-individual trajectories rather than variations at each time point.

Table 1.

Fit Statistics for CES-D Trajectories in Middle-Aged Population (1998-2010).

Number of classes	Log likelihood	AIC	BIC	SA-BIC	Entropy	LRT	Class proportions
Number of classes	Log likelihood	AIC	BIC	SA-BIC	Entropy	p value	Class proportions
1	−74,613.179	149,250.359	149,335.994	14,9297.860	^a	^a
2	−72,998.748	146,027.497	146,134.541	146,086.873	0.900	0.0000	Class 1: 1,280 (15%)Class 2: 7,252 (85%)
3	−72,430.495	144,896.991	145 025.444	144 968.242	0.869	0.0002	Class 1: 939 (11%)Class 2: 767 (9%)Class 3: 6,826 (80%)
4	−71,865.220	143,772.439	143,922.301	143,855.566	0.836	0.0003	Class 1: 1,024 (12%)Class 2: 6,143 (72%)Class 3: 853 (10%)Class 4: 512 (6%)
5	−71,581.944	143,211.889	143,383.159	143,306.891	0.828	0.0087	Class 1: 5,887 (69%)Class 2: 939 (11%)Class 3: 341 (4%)Class 4: 1,023 (12%)Class 5: 342 (4%)
6	−71,581.944	143,217.889	143,410.568	143,324.766	0.586	0.5140	Class 1: 2,474 (29%)Class 2: 3,413 (40%)Class 3: 940 (11%)Class 4: 341 (4%)Class 5: 1,024 (12%)Class 6: 340 (4%)

Note. CES-D = Center for Epidemiologic Studies–Depression Scale; AIC = Akaike information criterion; SA-BIC = sample-size adjusted Bayesian information criterion; LRT = likelihood ratio test.

Not estimable for a one-class model.

Next, we explored the characteristics of individuals within each subgroup according to independent variables included in the final analysis models. The overall differences in percentages and standard errors were examined using χ² tests and the overall differences in means were tested using Wald F tests. Our final analytic task was to examine the effects of different life course factors on trajectories of depressive symptoms using a multinomial logistic regression approach. To demonstrate the effects of life course factors on depression trajectories in late middle age, we used three different models: critical period, accumulation, and social mobility. Each model included differently measured childhood risk factors. Specifically, in the critical period model, two childhood SES risk factors, low parental education and low financial status, were used. In the accumulation model, the effects of cumulative SES status, combining childhood SES and adulthood SES (i.e., level of exposure to life course disadvantages over time), on depression in late middle age were examined. The social mobility model included measures of SES mobility between parental and respondent SES as predictors of depressive symptom trajectory group.

Finally, we conducted supplementary analyses to check the robustness of our findings. First, tests were conducted with two alternative reference groups for Model 3. Second, we rotated the reference categories to the latent classes of depressive symptoms to identify differences with alternative reference to the remaining trajectory groups.

Results

Trajectories of Depressive Symptoms

Using latent class analysis, we estimated the developmental trajectories of depressive symptoms in late middle age as summarized in Table 1; Figure 1 illustrates the observed trajectories for the four classes: Declining (Class 1, 11.8%), Low (Class 2, 71.7%), Increasing (Class 3, 10.6%), and High and increasing (Class 4, 5.9%). Individuals in Declining had about 4.5 depressive symptoms at baseline, but experienced substantial reduction over time. Low, the largest group, had only one symptom throughout the observation period, with a slightly negative linear slope. Increasing was distinguished by a substantial increase in CES-D from about 1.5 to nearly 4.5 symptoms over the 12-year period. The High and increasing group had 5.5 CES-D symptoms at baseline, which increased to nearly 6.5 over the observation period.

Figure 1.

Trajectory classes of CES-D in late middle age (HRS 1998-2010).

Table 2 presents results comparing characteristics such as demographic attributes and adulthood and childhood SES across identified CES-D trajectory groups. Uncontrolled examination of the association of characteristics with trajectory classes showed significant differences for all variables across classes. Specifically, Low (Class 2), the lowest depressive symptom group, included a larger proportion of Caucasians, men, those who were married throughout the study period, and those who had a high school education or higher. The economic status of this group was the highest among the trajectory classes, with the highest mean income. This group also had relatively better health, with the smallest proportion of respondents who reported their adult health as poor and had chronic diseases. In terms of life course factors, the parental education level of this group was significantly greater than that of other groups, and respondents in this group had experienced childhood poverty significantly less. This group also had the largest proportion of respondents with no exposure to cumulative SES disadvantages and those whose SES status was constantly high (high-high) over the life course.

Table 2.

Baseline Characteristics of Each Trajectory Class, M (SD) or n (%).

	Total sample (N = 8,532)	Class1	Class 2	Class 3	Class 4	Statistics
	Total sample (N = 8,532)	Declining (11.8%, N = 885)	Low (71.7%, N = 6,409)	Increasing (10.6%, N = 760)	High and increasing (5.9%, N = 478)	Statistics
Covariates
Age	54.88 (0.03)	54.72 (2.84)	54.96 (3.02)	54.38 (2.80)	54.78 (2.97)	F(8, 532) = 9.63***
Race (ref: Caucasian)
African American	1,250 (14.64%)	187 (21.15%)	833 (13.00%)	135 (17.76%)	95 (19.87%)	χ²(6) = 102.98***
Others	517 (6.06%)	72 (8.14%)	341 (5.32%)	52 (6.84%)	52 (10.88%)
Gender
Female	5,046 (59.14%)	586 (66.21%)	3,615 (56.41%)	492 (64.74%)	353 (73.85%)	χ²(3) = 90.82***
Life course factors
Childhood
Low parental education^a	4,848 (56.83%)	596 (67.50%)	3,422 (53.39%)	470 (61.84%)	360 (75.31%)	χ²(3) = 146.15***
Low financial status	2,274 (26.66%)	293 (33.18%)	1,522 (23.75%)	241 (31.71%)	218 (45.61%)
Cumulative disadvantage (ref: No exposure)						χ²(3) = 144.68***
Least exposure	4,088 (47.91%)	399 (45.19%)	3,100 (48.37%)	391 (47.50%)	198 (41.42%)	χ²(9) = 537.99***
Moderate exposure	1,361 (15.96%)	240 (27.18%)	811 (12.65%)	158 (20.79%)	152 (31.80%)
Greatest exposure	283 (3.32%)	47 (5.32%)	128 (2.00%)	44 (5.79%)	64 (13.39%)
Social mobility (ref: High-High)
Low-Low	1,617 (18.96%)	285 (32.28%)	923 (14.40%)	198 (26.05%)	211 (44.14%)	χ²(9) = 475.20***
Low-High	3,764 (44.13%)	365 (41.34%)	2,892 (45.12%)	329 (43.29%)	178 (37.24%)
High-Low	321 (3.76%)	36 (4.08%)	224 (3.50%)	36 (4.74%)	25 (5.23%)
Adulthood factors
Low education (Less than college)	1,451 (17.01%)	254 (28.77%)	819 (12.78%)	169 (22.24%)	209 (43.72%)	χ²(3) = 424.05***
Household Income (US$, mean value)	6408.63 (104.01)	4166.57 (6666.93)	7191.15 (10433.32)	4762.40 (6064.66)	2688.09 (3348.49)	F(8, 532) = 62.94***
Logged household income (mean value)	10.63 (0.01)	10.079 (1.81)	10.848 (1.38)	10.234 (1.84)	9.513 (2.14)	F(8, 532) = 182.45***
Marital status (ref: Constantly married)						χ²(6) = 351.19***
Constantly unmarried	2,082 (24.41%)	318 (35.93%)	1,299 (20.27%)	244 (32.11%)	221 (46.23%)
Status changed	896 (10.49%)	111 (12.54%)	605 (9.44%)	111 (14.61%)	69 (14.44%)
Work status (ref: Constantly worked)						χ²(6) = 630.29***
Constantly not worked	1,795 (21.04%)	306 (34.58%)	1,018 (15.88%)	224 (29.47%)	247 (51.67%)
Status changed	2,945 (34.53%)	317 (35.82%)	2,162 (33.73%)	314 (41.32%)	154 (32.22%)
Self-rated health^b (mean value)	2.63 (0.01)	3.309 (0.90)	2.423 (0.85)	3.285 (0.92)	4.009 (0.80)
Changes in self-rated health (ref: no Δ)						F(8, 532) = 845.49***
Became better	1,780 (20.86%)	255 (28.81%)	1,293 (20.17%)	143 (18.82%)	89 (18.62%)	χ²(6) = 65.33***
Became poor	2,528 (29.62%)	230 (25.99%)	1,868 (29.15%)	289 (38.03%)	141 (29.50%)
Chronic disease (yes)	7,028 (82.38%)	802 (90.62%)	5,067 (79.06%)	696 (91.58%)	463 (96.86%)
Number of chronic diseases	1.49 (0.01)	2.04 (1.41)	1.26 (1.07)	2.02 (1.30)	2.83 (1.49)	χ²(3) = 203.35***
Changes in number of diseases (ref: no Δ)^c						F(8, 532) = 412.96
Increased	4,180 (48.99%)	488 (55.14%)	2,941 (45.89%)	452 (59.47%)	299 (62.55%)	χ²(3) = 106.67***

Parental education: For the parent with the highest level of education.

Self-rated health: The higher value indicates poorer health.

Dummy variables for chronic disease changes were originally three: no change (0); number of diseases increased (1); and number of diseases decreased (2). There were no observations for (2). For the multinomial analyses, we used change dummy for number of chronic disease rather than the presence of chronic disease because more than 90% of the sample have chronic diseases.

Significance level of p value: *p < .05. **p < .01. **p < .001.

However, the characteristics of High and increasing (Class 4) tended to be the worst. This group included a smaller proportion of Caucasians, a larger proportion of females, more who were unmarried throughout the study period and had less than a high school education, and the lowest mean household income. Their health status was also the poorest: poor self-rated health, a higher number of chronic conditions, and increased chronic conditions over time. With regard to life course factors, a greater proportion of those in the high and increasing group had low parental education and low financial status during childhood. They also had the largest proportion of people with the greatest exposure to cumulative disadvantages and those with constantly low SES status (low-low) over the life course.

The declining (Class 1) and increasing (Class 3) groups showed similar characteristics in terms of childhood and adulthood risk factors for depression. For example, these two groups had similar proportions of people with low parental education (67.5%-Declining group and 61.8%-Increasing group), low financial status during childhood (33.2%-Declining group and 31.7%-Increasing group), and a similar household mean income (US$4,167-Declining group and US$4,762-Increasing group). The patterns of cumulative SES disadvantages and the social mobility were also found to be very similar (Table 2).

Role of Life Course Mechanisms in Predicting Depression Symptoms Trajectory

Table 3 presents results for the association between three life course mechanisms and CES-D trajectories. The estimates were derived from multinomial logistic regression analyses where coefficient estimates were reported as relative risk ratios (RRR = e^b, where b is the logistic regression coefficient). Model 1 focused on the relationships between childhood SES and different CES-D trajectories controlling for adulthood factors and covariates. Low financial status in childhood was significantly associated with only High and increasing (Class 4), suggesting late middle age adults with low childhood family financial experiences had more and increasing depressive symptoms than those with better childhood family financial status. Before adjustment for adulthood socioeconomic circumstances, the association between low childhood financial status and/or low parental education with Declining, Increasing, and High and Increasing appeared to be significant, and those with low childhood SES had the highest probability of being in High and Increasing (Appendix A); most of these associations disappeared after adjustment for adulthood SES, except for the association between low childhood financial status and High and Increasing.

Table 3.

Multinomial Logistic Regression Models for the Association Between Life Course SES and Latent Classes of Depressive Symptoms.

	Model 1			Model 2			Model 3
	Critical period model			Accumulation model			Social mobility model
	Class1: Declining	Class 3: Increasing	Class 4: High and increasing	Class1: Declining	Class 3: Increasing	Class 4: High and increasing	Class1: Declining	Class 3: Increasing	Class 4: High and increasing
	RRR			RRR			RRR
Covariates
Age (baseline)	0.93 (.01)***	0.90 (.01)***	0.92 (.01)***	.93 (.01)***	.90 (.01)***	.93 (.01)***	.93 (.01)***	.90 (.01)***	.92 (.01)***
Female	1.30 (.10)**	1.26 (.10)**	1.82 (.22)***	1.30 (.10)**	1.24 (.10)**	1.78 (.21)***	1.29 (.10)**	1.25 (.10)**	1.79 (.21)***
Race (Caucasian)
African American	1.01 (.10)	0.85 (.09)	0.64 (.09)**	1.01 (.10)	.86 (.09)	.66 (.09)**	1.01 (.10)	.86 (.09)	.66 (.09)**
Others	1.04 (.15)	0.89 (.14)	1.02 (.19)	1.09 (.16)	.89 (.15)	1.08 (.20)	1.07 (.16)	.90 (.15)	1.12 (.21)
Life course factors
Model 1
Low parental education	1.17 (.10)	1.01 (.09)	1.24 (.16)
Low financial status	1.09 (.09)	1.15 (.10)	1.48 (.17)**
Model 2 (ref: None at any)
Least exposure				1.20 (.11)	1.12 (.11)	1.60 (.25)**
Moderate exposure				1.69 (.19)***	1.21 (.15)	2.04 (.36)***
Greatest exposure				1.38 (.28)	1.53 (.32)*	3.02 (.72)***
Model 3 (ref: High-High)
Low -Low							1.67 (.18)***	1.26 (.15)*	2.21 (.37)***
Low -High							1.20 (.11)	1.12 (.11)	1.63 (.26)**
High -Low							1.07 (.22)	1.10 (.22)	1.48 (.41)
Adulthood factors
Edu (ref: College or higher)
Low education	1.36 (.13)**	1.05 (.11)	1.58 (.20)***
Logged household income	0.93 (.02)**	0.93 (.02)*	.92 (.02)**
Marital status (ref: married)
Constantly unmarried	1.87 (.17)***	1.71 (.16)***	2.69 (.34)***	1.95 (.17)***	1.81 (.17)***	2.81 (.34)***	1.95 (.17)***	1.81 (.17)***	2.84 (.34)***
Status changed	1.57 (.19)***	1.82 (.22)***	2.36 (.39)***	1.58 (.19)***	1.84 (.22)***	2.34 (.39)***	1.584 (.19)***	1.84 (.22)***	2.37 (.39)***
Work status (ref: worked)
Constantly not work	1.61 (.17)***	1.60 (.19)***	1.90 (.31)***	1.69 (.18)***	1.66 (.19)***	2.01 (.32)***	1.69 (.18)***	1.67 (.19)***	2.05 (.33)***
Status changed	1.31 (.12)**	1.49 (.14)***	1.43 (.22)*	1.31 (.12)**	1.50 (.14)***	1.43 (.22)*	1.31 (.12)**	1.50 (.14)***	1.45 (.22)*
SRH at baseline	1.92 (.09)***	2.03 (.10)***	3.61 (.27)***	1.96 (.09)***	2.04 (.10)***	3.72 (.28)***	1.95 (.09)***	2.04 (.10)***	3.75 (.28)***
SRH change (ref: no Δ)
Became better	0.91 (.08)	0.61 (.06)***	.40 (.05)***	.91 (.08)	.61 (.07)***	.40 (.05)***	.91 (.08)	.61 (.07)***	.40 (.05)***
Became poor	1.61 (.16)***	2.46 (.24)***	3.43 (.49)***	1.64 (.16)***	2.48 (.24)***	3.55 (.50)***	1.64 (.16)***	2.48 (.24)***	3.54 (.50)***
Chronic at baseline	1.16 (.04)***	1.16 (.04)***	1.32 (.05)***	1.15 (.04)***	1.16 (.04)***	1.33 (.05)***	1.15 (.04)***	1.16 (.04)***	1.32 (.05)***
Chronic disease (ref: No Δ)
Number of diseases increased	1.27 (.10)**	1.34 (.11)***	1.55 (.17)***	1.27 (.10)**	1.35 (.11)***	1.56 (.17)***	1.27 (.10)**	1.34 (.11)***	1.55 (.17)***
Childhood SRH	1.11 (.04)**	1.10 (.04)*	1.24 (.05)***	1.12 (.04)**	1.10 (.04)**	1.27 (.06)***	1.12 (.04)**	1.11 (.04)**	1.28 (.06)***
Constant	0.31 (.26)	1.14 (1.02)	.00 (.00)***	.11 (.09)**	.53 (.44)	.00 (.00)***	.12 (.09)**	.53 (.44)	.00 (.00)***
Pseudo R²	.1742			.1724			.1719
LR chi² (df)	2456.91 (54)***			2430.72 (51)***			2424.46 (51)***

Note. Reference group = Low class. SRH = self-rated health; LR = likelihood ratio.

Significance level of p value: *p < .05. **p < .01. ***p < .001.

Model 2 included the results for the relationship between the cumulative SES index and depressive symptom trajectories. Respondents with higher SES disadvantages across the life course were more likely to be in the higher depression classes than those who were never exposed to SES disadvantages, those who experienced the greatest exposure to adverse socioeconomic circumstances across the life course tended to be in the High and Increasing group, and the probability of experiencing more depressive symptoms in late middle age appeared to become higher as the number of exposures increased.

Model 3 included the association between socioeconomic mobility and depressive symptoms. Those who had a disadvantaged childhood and continued to experience limited or no upward mobility tended to experience more depressive symptoms. Interestingly, those who had a socioeconomically disadvantaged childhood but experienced upward mobility over their life course (low-high) still tended to be High and Increasing. In supplemental analyses, we rotated the social mobility categories to compare those who were upwardly mobile with those in the constantly disadvantaged group (low-low) and found that those who had better SES conditions in adulthood after a disadvantaged childhood were less likely to be High and Increasing than those with persistent socioeconomic disadvantages across the life course (Appendix B). The comparison between downward mobility and upward mobility (Low-High is the reference) was not significantly different (Appendix B).

With regard to demographic and health factors, poor childhood self-rated health significantly increased the risk of being in a group other than Low (Class 2). Most demographic factors such as age, gender, marital status, and work status were associated with different CES-D trajectories. Those who were male and younger were more likely to be Low (Class 2; Table 3) as were respondents who were married and worked over the whole period. In the post hoc analyses with different reference groups, we found that respondents who were male and constantly married were least likely to be in High and increasing (Class 4; results not shown). Respondents who experienced changes in marital status were less likely to be in the Increasing class than High and increasing (results not shown). An increase in the number of chronic diseases was associated with the highest probability of being in High and increasing (Class 4), followed by Increasing (Class 3) and Declining (Class 1; Table 3). Improvement in self-rated health over time was associated with a lower probability of being in the Increasing (Class 3) or High and increasing (Class 4) groups compared with Low (Class 2; Table 3). Respondents whose self-rated health became worse over time were most likely to be in High and Increasing (Class 4).

In supplementary analyses (Appendix C), we rotated the reference groups to check the robustness of our findings, and these models were generally consistent with our interpretation. Setting Declining (Class 1) as the reference group to compare it with Increasing (Class 3), no significant effects were found in Model 1 (critical period model) with adverse childhood among Increasing compared with Declining; instead, those who had poorer financial status in childhood tended to be in High and Increasing rather than Declining. Those who experienced the greatest number of exposures to SES risks across the life course were also more likely to be High and Increasing rather than Declining. Other significant associations among demographic and health factors remained the same in different directions depending on the reference group (Table 3).

Discussion

This study identified trajectory patterns of depressive symptoms in late middle age using nationally representative data over a 12-year period of time, providing new information concerning the heterogeneity of late middle age depression. Our group-based trajectory approach yielded an evidence-based approximation that allowed us to gain a better understanding of how life course SES is associated with each depressive symptom trajectory. By performing separate analyses of three life course models—critical period, accumulation, and social mobility models—this study comprehensively examined the effects of the different aspects of life course SES on depression trajectory patterns in later life.

Depressive symptoms have been found to be less prevalent among late middle age adults (Melchior et al., 2013; Sutin et al., 2013), resulting in limited research focusing on this subpopulation. In this respect, our findings on the four distinct patterns of depression trajectories in late middle age expand the current knowledge on mental health among the middle-aged. Consistent with most previous studies, a majority (71.7%) of our middle-aged respondents belonged to the Low depressive symptoms group, with one or no symptoms, and the High and increasing group was the smallest group (5.9%). The Low depression and the High and increasing groups were found to be opposite: While the Low depressive symptoms group tended to have low childhood SES, the High and increasing group was most likely to have parents with low education and low economic status, and tended to be exposed to cumulative disadvantage over time. The characteristics of the Declining and the Increasing groups tended to be between the Low and the High and increasing groups, but we found that status changes in marriage, work, and self-rated health were significantly related to the Increasing group. These results from longitudinal analyses support previous findings that life events, often accompanied by stress, could cause depression (Hammen, 2005).

We explored the important question of to what extent socioeconomic position across the life course influences heterogeneous trajectories of depressive symptoms. In the critical period model, low financial status in childhood appeared to be directly associated with membership in the High and Increasing group even after controlling for low adulthood socioeconomic conditions. Although some aspects of depressive symptoms in adulthood may have socioeconomic roots in childhood (Harper et al., 2002), the critical period model highlights that childhood disadvantages matter even after controlling for adulthood risk factors. A number of studies have examined the model with respect to depression with mixed results. Some studies showed that the impact of childhood SES disadvantage remained significant after adjustment for adult SES (Gale et al., 2011; Kuh et al., 2003), while others found that the impact of early life SES on depression was attenuated by adulthood socioeconomic conditions and health behaviors (Poulton et al., 2002). This inconsistency may be due to different measures of early life SES, years of follow-up, and birth cohorts. Our finding of the significant association between childhood financial status and the High and Increasing group provides powerful evidence of the role of long-lasting economic inequalities in the development of higher levels of depressive symptoms. As childhood poverty has been shown to be associated with harmful exposures, such as parental conflict and loss and violence (Harper et al., 2002), these early adverse experiences may lead to a cynical, hostile, and hopeless view of the world (Lipman & Offord, 1997) and may have an additive impact with life time stressors on depression (Cronkite et al., 2013). In sensitivity analyses (results not shown), we included subjects with missing information on childhood and found that the characteristics of those respondents appeared to be more disadvantaged, consistent with previous research (Brown, 2010). When we included these respondents in analyses, those with missing information were significantly more likely to be in the High and Increasing group. This may indicate that the exclusion of those with missing childhood information resulted in an underestimation of the effects of childhood adversity on depression.

SES in adulthood, respondents’ education and income, was also associated with the trajectory groups. Although these two measures are used widely, often interchangeably, in the health inequalities literature, they reflect distinct socioeconomic domains that may require different interventions. For example, education is achieved in early adulthood and measures skills requisite for acquiring positive social, psychological, and economic resources, which increase well-being in later life (Lynch & Kaplan, 2000). Higher educational levels were found to have a protective effect against depression throughout life (Bjelland et al., 2008). However, income reflects purchasing power and the level of financial strain and has been considered as a strong correlate of depression (Zimmerman & Kanton, 2005). The results presented here are stronger as income did not lose much of its relationship to the distinct trajectories when all the other adulthood variables were controlled.

The accumulation model focused on exposure to poor SES across life course. Increased adversity over the life course was associated with Increasing and High and increasing trajectories of depressive symptoms. These results make a unique contribution by demonstrating that cumulative disadvantage is associated with strong and consistent graded effects on changes in depressive symptoms, especially for those with High and increasing depressive symptoms. While the critical period model focused only on childhood risk factors, the accumulation model suggests the importance of additive effects of socioeconomic conditions measured at different stage of life. There is strong interdependence between the critical period and accumulation models because there must be exposure to socioeconomic disadvantages in the critical period to have accumulated disadvantages (Hallqvist et al., 2004). In our sample, about 35% had two or more cumulative SES disadvantages. It is likely that adverse SES circumstances in early life increase the risk of depressive symptoms through gradually accumulated exposures to disadvantages over the life course.

The social mobility model investigated if mobility from childhood to adulthood was an important determinant of distinct depressive symptom trajectories in late middle age. As expected, respondents who were disadvantaged both in their childhood and adulthood exhibited more depressive symptoms than those who had high SES positions in both life stages. Although it was expected there would be a protective effect of upward mobility on depressive symptoms (Poulton et al., 2002), being disadvantaged in childhood was significantly associated with High and increasing depression regardless of better socioeconomic conditions in adulthood. These results are consistent with the critical period model that stresses the importance of childhood risk factors. They also support the social-origins hypothesis that children from low-SES families have poorer health in adulthood, regardless of their adulthood SES (Lyu & Burr, 2016). However, the protective effects of upward SES mobility were still found in that those in the upward mobility group were significantly less likely to be in the High and Increasing group than those in the persistently low SES group.

There are several limitations. First, the childhood variables used in this study are based on respondents’ retrospective reporting, which is subject to recall bias. For example, depressed individuals may recall their childhood negatively and rate childhood SES conditions accordingly, which means our findings may be overestimated. Second, to better capture the association between life course SES and depression, future research should expand the SES measures beyond those used in our study to include occupational prestige, wealth, and debt (Harper et al., 2002; Wilson, Shuey, & Elder, 2007). Third, despite strong evidence establishing the association between childhood poverty and high and increasing depressive symptoms in late middle age, much remains unknown regarding specific pathways. Future studies need to include information about mediators assessed prospectively over the life course; for example, health behaviors or harmful exposures (i.e., life events) could be explored as pathways (Korkeila et al., 2010). Fourth, this study focused on the association between life course factors and depressive symptoms only among middle-aged adults; other age groups or other health outcomes may be affected differently. Fifth, various protective factors may have contributed to declines in depressive symptoms, such as timely clinical treatment, positive environmental changes (e.g., social support), or improved coping ability (Taylor & Lynch, 2004). Although this study could not examine factors related to such positive change patterns (Declining), the examination of possible protective factors is an important inquiry for future research.

Despite these limitations, this study uncovers heterogeneous depression trajectories among late middle age individuals and illustrates the utility of life course mechanisms as a framework for further investigation of depressive symptoms. The study’s innovative empirical approach should inform future research that compares and contrasts the conceptually interconnected life course models as related to multigroup health trajectories. Our findings from the three life course models emphasize the enduring influence of childhood disadvantage on depression. Policy interventions addressing socioeconomic inequality among children may be critical in alleviating mental health disparities in later life.

Footnotes

Appendix

Appendix C

Multinomial Logistic Regression Models for the Association Between Life Course SES and Latent Classes of Depressive Symptoms (Reference: Declining)

	Model 1			Model 2			Model 3
	Critical period model			Accumulation model			Social mobility model
	Class 2:	Class 3:	Class 4:	Class 2:	Class 3:	Class 4:	Class 2:	Class 3:	Class 4:
	Low	Increasing	High and increasing	Low	Increasing	High and increasing	Low	Increasing	High and increasing
	RRR			RRR			RRR
Covariates
Age (baseline)	1.06 (.01)***	0.97 (.01)	0.98 (.02)	1.06 (.01)***	0.97 (.01)	0.99 (.02)	1.06 (.01)***	0.97 (.01)	0.99 (.02)
Female	0.76 (.06)**	0.96 (.10)	1.40 (.18)*	0.76 (.06)**	0.95 (.10)	1.36 (.18)*	0.77 (.06)**	0.96 (.10)	1.38 (.18)*
Race (Caucasian)
African American	0.98 (.10)	0.84 (.11)	0.64 (.09)**	0.98 (.10)	0.84 (.11)	0.65 (.10)**	0.98 (.10)	0.85 (.11)	0.65 (.10)**
Others	0.95 (.14)	0.85 (.16)	0.981 (.20)	0.91 (.13)	0.81 (.16)	0.99 (.20)	0.93 (.13)	0.84 (.16)	1.04 (.21)
Life course factors
Model 1
Low parental education	0.85 (.07)	0.86 (.09)	1.06 (.15)
Low financial status	0.90 (.07)	1.05 (.11)	1.35 (.17)*
Model 2 (ref: None at any)
Least exposure				0.83 (.08)	0.93 (.11)	1.34 (.23)
Moderate exposure				0.59 (.06)***	0.71 (.10)*	1.20 (.23)
Greatest exposure				0.72 (.14)	1.10 (.27)	2.18 (.57)**
Model 3 (ref: High-High)
Low-Low							0.59 (.06)***	0.75 (.11)	1.32 (.24)
Low-High							0.82 (.08)	0.93 (.11)	1.35 (.23)
High-Low							0.92 (.19)	1.02 (.26)	1.38 (.42)
Adulthood factors
Edu (ref: College or higher)
Low education	0.73 (.07)**	0.77 (.09)*	1.15 (.16)
Logged household income	1.06 (.02)**	0.98 (.02)	0.98 (.02)
Marital status (ref: Married)
Constantly unmarried	0.53 (.04)***	0.91 (.10)	1.43 (.19)*	0.51 (.04)***	0.92 (.01)	1.44 (.19)**	0.51 (.04)***	0.92 (.10)	1.45 (.19)**
Status changed	0.63 (.07)***	1.15 (.17)	1.50 (.27)*	0.63 (.07)***	1.16 (.17)	1.48 (.27)*	0.63 (.07)***	1.16 (.17)	1.49 (.27)*
Work status (ref: Worked)
Constantly not work	0.61 (.06)***	0.99 (.14)	1.17 (.21)	0.58 (.06)***	0.98 (.14)	1.18 (.21)	0.58 (.06)***	0.98 (.14)	1.21 (.21)
Status changed	0.76 (.07)**	1.14 (.14)	1.09 (.18)	0.75 (.07)**	1.14 (.14)	1.09 (.18)	0.75 (.07)**	1.14 (.14)	1.10 (.18)
SRH at baseline	0.52 (.02)***	1.05 (.06)	1.87 (.15)***	0.51 (.02)***	1.04 (.06)	1.89 (.15)***	0.51 (.02)***	1.05 (.06)	1.92 (.15)***
SRH change (ref: No Δ)
Became better	1.09 (.10)	0.66 (.08)**	0.44 (.06)***	1.09 (.10)	0.66 (.08)**	0.44 (.06)***	1.09 (.10)	0.67 (.08)**	0.44 (.06)***
Became poor	0.61 (.06)***	1.52 (.19)**	2.12 (.33)***	0.60 (.06)***	1.50 (.19)**	2.15 (.34)***	.60 (.06)***	1.51 (.19)**	2.15 (.34)***
Chronic at baseline	0.86 (.03)***	1.00 (.04)	1.14 (.05)**	0.86 (.03)***	1.00 (.04)	1.15 (.05)**	0.86 (.03)***	1.00 (.04)	1.14 (.05)**
Chronic disease (ref: No Δ)
Number of diseases increased	0.78 (.06)**	1.05 (.11)	1.21 (.15)	0.78 (.06)**	1.06 (.11)	1.22 (.15)	0.78 (.06)**	1.05 (.11)	1.21 (.15)
Childhood SRH	0.89 (.03)**	0.99 (.04)	1.11 (.05)*	0.88 (.03)**	0.98 (.04)	1.12 (.05)*	0.88 (.03)**	0.98 (.04)	1.14 (.05)*
Constant	3.17 (2.62)	3.65 (3.94)	0.02 (0.03)**	8.41 (6.49)**	4.47 (4.57)	0.01 (0.01)**	7.99 (6.18)**	4.26 (4.36)	0.01 (0.01)***
Pseudo R²	.1742			.1724			.1719
LR chi² (df)	2456.91 (54)***			2430.72 (51)***			2424.46 (51)***

Note. Reference group = Declining class. RRR = relative risk ratios. SRH = self-rated health; LR = likelihood ratio.

Significance level of p value: *p < .05. **p < .01. **p < .001.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

References

Andreescu

Chang

C. H.

Mulsant

B. H.

Ganguli

(2008). Twelve-year depressive symptom trajectories and their predictors in a community sample of older adults. International Psychogeriatrics, 20, 221-236.

Bjelland

Krokstad

Mykletun

Dahl

A. A.

Tell

G. S.

Tambs

(2008). Does a higher educational level protect against anxiety and depression? The HUNT study. Social Science & Medicine, 66, 1334-1345.

Brown

M. T.

(2010). Early-life characteristics, psychiatric history, and cognition trajectories in later life. Gerontologist, 50(5), 646-656.

Byers

A. L.

Vittinghoff

Lui

L. Y.

Hoang

Blazer

D. G.

Covinsky

K. E.

. . . Yaffe

(2012). Twenty-year depressive trajectories among older women. Archives of General Psychiatry, 69, 1073-1079.

Centers for Disease Control and Prevention. (2015). National vital statistics system-mortality.

Chen

C. M.

Mullan

Griffiths

Kreis

I. A.

Lan

T. Y.

Chiu

H. C.

(2011). Trajectories of depression and their relationship with health status and social service use. Archives of Gerontology and Geriatrics, 53, e118-e124.

Chin

W. Y.

Choi

E. P. H.

Wan

E. Y. F.

(2016). Trajectory pathways for depressive symptoms and their associated factors in a Chinese primary care cohort by growth mixture modeling. PLoS ONE, 11, e0147775.

Choi

Lee

Y. H.

(2010). Depression, powerlessness, social support, and socioeconomic status in middle-aged community residents. Journal of Korean Academy of Psychiatric Mental Health Nursing, 19, 196-204.

Cohen

Janicki-Deverts

Chen

Matthews

K. A.

(2010). Childhood socioeconomic status and adult health. Annals of the New York Academy of Sciences, 1186, 37-55.

10.

Cronkite

R. C.

Woodhead

E. L.

Finlay

Timko

K. U.

Moos

R. H.

(2013). Life stressors and resources and the 23-year course of depression. Journal of Affective Disorders, 150, 370-377.

11.

Cui

Lyness

J. M.

Tang

Conwell

(2008). Outcomes and predictors of late-life depression trajectories in older primary care patients. American Journal of Geriatric Psychiatry, 16, 406-415.

12.

Dannefer

(1987). Aging as intracohort differentiation: Accentuation, the Matthew effect, and the life course. Sociological Forum, 2, 211-236.

13.

Elder

G. H.

Johnson

M. K.

(2003). The life course and aging: Challenges, lessons, and new directions. In: Settersten

R. A.

(ed), Invitation to the life course: Toward new understandings of later life (pp. 49-81). Amityville, NY: Baywood.

14.

Evans

G. W.

Cassells

R. C.

(2014). Childhood poverty, cumulative risk exposure, and mental health in emerging adults. Clinical Psychological Science, 2, 287-296.

15.

Gale

C. R.

Sayer

A. A.

Cooper

Dennison

E. M.

Starr

J. M.

Whalley

L. J.

, . . . the HALCyon Study Team. (2011). Factors associated with symptoms of anxiety and depression in five cohorts of community-based older people: The HALCyon (Healthy Ageing across the Life Course) Programme. Psychological Medicine, 41, 2057-2073.

16.

Gilman

S. E.

Kawachi

Fitzmaurice

G. M.

Buka

S. L.

(2002). Socioeconomic status in childhood and the lifetime risk of major depression. International Journal of Epidemiology, 31, 359-367.

17.

Haas

(2008). Trajectories of functional health: The “long arm” of childhood health and socioeconomic factors. Social Science & Medicine, 66, 849-861.

18.

Hallqvist

Lynch

Bartley

Lang

Blane

(2004). Can we disentangle life course processes of accumulation, critical period and social mobility? An analysis of disadvantaged socio-economic positions and myocardial infarction in the Stockholm Heart Epidemiology Program. Social Science & Medicine, 58, 1555-1562.

19.

Hammen

(2005). Stress and depression. Annual Review of Clinical Psychology, 1, 293-319.

20.

Harper

Lynch

Hsu

Everson

S. A.

Hillemeier

M. M.

Raghunathan

T. E.

. . . Kaplan

G. A.

(2002). Life course socioeconomic conditions and adult psychosocial functioning. International Journal of Epidemiology, 31, 395-403.

21.

Hipp

J. R.

Bauer

D. J.

(2006). Local solutions in the estimation of growth mixture models. Psychological Methods, 11, 36-53.

22.

Hsu

H. C.

(2012). Group-based trajectories of depressive symptoms and the predictors in the older population. International Journal of Geriatric Psychiatry, 27, 854-862.

23.

Juster

F. T.

Suzman

(1995). An overview of the Health and Retirement Study. Journal of Human Resources, 30, S7-S56.

24.

Korkeila

Vahtera

Nabi

Kivimäki

Korkeila

Sumanen

. . . Koskenvuo

(2010). Childhood adversities, adulthood life events and depression. Journal of Affective Disorders, 127, 130-138.

25.

Kuh

Ben-Shlomo

Lynch

Hallqvist

Power

(2003). Life course epidemiology. Journal of Epidemiology & Community Health, 57, 778-783.

26.

Kuo

S. Y.

Lin

K. M.

Chen

C. Y.

Chuang

Y. L.

Chen

W. J.

(2011). Depression trajectories and obesity among the elderly in Taiwan. Psychological Medicine, 41, 1665-1676.

27.

Kwag

K. H.

(2009). The cumulative influences of socioeconomic status on health outcomes in late adulthood: A latent growth curve analysis (Doctoral dissertation). Human Development and Family Studies, Iowa State University, Ames.

28.

Liang

Quiñones

A. R.

Bennett

J. M.

(2011). Multiple trajectories of depressive symptoms in middle and late life: Racial/ethnic variations. Psychology and Aging, 26, 761-777.

29.

Lincoln

K. D.

Takeuchi

D. T.

(2010). Variation in the trajectories of depressive symptoms, results from the Americans’ Changing Lives Study. Biodemography and Social Biology, 56, 24-41.

30.

Lipman

E. L.

Offord

D. R.

(1997). Psychosocial morbidity among poor children in Ontario. In Duncan

G. J.

Brooks-Gunn

(Eds.), Consequences of growing up poor (pp. 239-287). New York, NY: Russell Sage Foundation.

31.

Mendell

N. R.

Rubin

D. B.

(2001). Testing the number of components in a normal mixture. Biometrika, 88, 767-778.

32.

Luo

Waite

L. J.

(2005). The impact of childhood and adult SES on physical, mental, and cognitive well-being in later life. Journals of Gerontology. Series B, Psychological Sciences and Social Sciences, 60B, 93-101.

33.

Lynch

Kaplan

(2000). Social epidemiology. New York, NY: Oxford University Press.

34.

Lyu

Burr

J. A.

(2016). Socioeconomic status across the life course and cognitive function among older adults: An examination of the latency, pathways, and accumulation hypotheses. Journal of Aging and Health, 28, 40-67.

35.

Melchior

Chastang

J. F.

Head

Goldberg

Zins

Nabi

Younes

(2013). Socioeconomic position predicts long-term depression trajectory: A 13-year follow-up of the GAZEL cohort study. Molecular Psychiatry, 18, 112-121.

36.

Miech

R. A.

Shanahan

M. J.

(2000). Socioeconomic status and depression over the life course. Journal of Health and Social Behavior, 41, 162-176.

37.

Musliner

K. L.

Munk-Olsen

Eaton

W. W.

Zandi

P. P.

(2016). Heterogeneity in long-term trajectories of depressive symptoms: Patterns, predictors and outcomes. Journal of Affective Disorders, 192, 199-211.

38.

Muthén

(2003). Statistical and substantive checking in growth mixture modeling: Comment on Bauer and Curran (2003). Psychological Methods, 8, 369-377.

39.

Muthén

L. K.

Muthén

B. O.

(2015). Mplus user’s guide (7th ed.). Los Angeles, CA: Author.

40.

Nicholson

Pikhart

Pajak

Malyutina

Kubinova

Peasey

. . . Bobak

(2008). Socio-economic status over the life-course and depressive symptoms in men and women in Eastern Europe. Journal of Affective Disorders, 105, 125-136.

41.

Nicklett

E. J.

Burgard

S. A.

(2009). Downward social mobility and major depressive episodes among Latino and Asian-American immigrants to the United States. American Journal of Epidemiology, 170, 793-801.

42.

O’Rand

A. M.

(2003). Cumulative advantage theory in life-course research. In Crystal

Shea

(Eds.), Annual review of gerontology and geriatrics: Vol. 22. Focus on economic outcomes in later life (pp. 14-30). New York, NY: Springer.

43.

Poulton

Caspi

Milne

B. J.

Thomson

W. M.

Taylor

Sears

M. R.

Moffitt

T. E.

(2002). Association between children’s experience of socioeconomic disadvantage and adult health: A life-course study. Lancet, 360, 1640-1645.

44.

Pratt

L. A.

Brody

D. J.

(2014). Depression in the U.S. household population, 2009-2012 (National Health and Nutrition Examination Survey Data Brief No. 172). Hyattsville, MD: National Center for Health Statistics.

45.

Pudrovska

Anikputa

(2013). Early-life socioeconomic status and mortality in later life: An integration of four life-course mechanisms. The Journals of Gerontology. Series B, Psychological Sciences and Social Sciences, 69B, 451-460.

46.

Schwarz

(1978). Estimating the dimension of a model. The Annals of Statistics, 6, 461-464.

47.

Soldo

B. J.

Hurd

M. D.

Rodgers

W. L.

Wallace

R. B.

(1997). Asset and health dynamics among the oldest old: An overview of the AHEAD Study. The Journals of Gerontology. Series B, Psychological Sciences and Social Sciences, 52, 1-20.

48.

Sutin

A. R.

Terracciano

Milaneschi

Ferrucci

Zonderman

A. B.

(2013). The trajectory of depressive symptoms across the adult lifespan. JAMA Psychiatry, 70, 803-811.

49.

Taylor

M. G.

Lynch

S. M.

(2004). Trajectories of impairment, social support, and depressive symptoms in later life. The Journals of Gerontology. Series B, Psychological Sciences and Social Sciences, 59B, S238-S246.

50.

Thompson

Syddall

Rodin

Osmond

Barker

D. J.

(2001). Birthweight and risk of depressive disorder in late life. British Journal of Psychiatry, 179, 450-455.

51.

Turrell

Lynch

J. W.

Kaplan

G. A.

Everson

S. A.

Helkala

E. L.

Kauhanen

Salonen

J. T.

(2002). Socioeconomic position across the lifecourse and cognitive function in late middle-age. The Journals of Gerontology. Series B, Psychological Sciences and Social Sciences, 57B, S43-S51.

52.

Ward

J. B.

Haan

M. N.

Garcia

M. E.

Lee

T. M.

Aiello

A. E.

(2016). Intergenerational education mobility and depressive symptoms in a population of Mexican origin. Annals of Epidemiology, 26, 461-466.

53.

Wilson

A. E.

Shuey

K. M.

Elder

G. H.

(2007). Cumulative advantage processes as mechanisms of inequality in life course health. American Journal of Sociology, 112, 1886-1924.

54.

Yang

(2007). Is old age depressing? Growth trajectories and cohort variations in late-life depression. Journal of Health and Social, 48, 16-32.

55.

Yang

George

L. K.

(2005). Functional disability, disability transitions, and depressive symptoms in late life. Journal of Aging and Health, 17, 263-292.

56.

Zimmerman

F. J.

Kanton

(2005). Socioeconomic status, depression disparities, and financial hardship: What lies behind the income-depression relationhip? Health Economics, 14, 1197-1215.