Accepted Manuscript Cumulative exposure to childhood stressors and subsequent psychological distress. An analysis of US Panel Data Emma Björkenstam, Bo Burström, Lars Brännström, Bo Vinnerljung, Charlotte Björkenstam, Anne R. Pebley PII:
S0277-9536(15)30058-7
DOI:
10.1016/j.socscimed.2015.08.006
Reference:
SSM 10189
To appear in:
Social Science & Medicine
Received Date: 2 May 2015 Revised Date:
29 July 2015
Accepted Date: 3 August 2015
Please cite this article as: Björkenstam, E., Burström, B., Brännström, L., Vinnerljung, B., Björkenstam, C., Pebley, A.R., Cumulative exposure to childhood stressors and subsequent psychological distress. An analysis of US Panel Data, Social Science & Medicine (2015), doi: 10.1016/j.socscimed.2015.08.006. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT Cumulative exposure to childhood stressors and subsequent psychological distress. An analysis of US Panel Data Emma Björkenstam1, 2, Bo Burström2, Lars Brännström3, Bo Vinnerljung3, Charlotte Björkenstam4, Anne R. Pebley1 1
RI PT
Department of Community Health Sciences, Fielding School of Public Health and California Center for Population Research, University of California Los Angeles, Los Angeles, California, United States 2 Department of Public Health Sciences, Division of Social Medicine, Karolinska Institutet, Stockholm, Sweden 3 Department of Social Work, Stockholm University, Stockholm, Sweden 4 Department of Epidemiology, Fielding School of Public Health, University of California Los Angeles, Los Angeles, California, United States
AC C
EP
TE D
M AN U
SC
Corresponding author: Emma Björkenstam Department of Community Health Sciences, Fielding School of Public Health and California Center for Population Research, University of California Los Angeles, Los Angeles, California, United States Telephone: +1 (323)229-6239 E-mail:
[email protected]
ACCEPTED MANUSCRIPT 1
Cumulative exposure to childhood stressors and subsequent psychological
2
distress. An analysis of US Panel Data
3 4 5
Key words: adverse childhood experience, childhood stressors, psychological distress, depression,
6
socioeconomic, latent class analysis,
RI PT
7 8
AC C
EP
TE D
M AN U
SC
9
1
ACCEPTED MANUSCRIPT Abstract
11
Research has shown that childhood stress increases the risk of poor mental health later in life. We
12
examined the effect of childhood stressors on psychological distress and self-reported depression
13
in young adulthood. Data were obtained from the Child Development Supplement (CDS) to the
14
national Panel Study of Income Dynamics (PSID), a survey of US families that incorporates data
15
from parents and their children. In 2005 and 2007, the Panel Study of Income Dynamics was
16
supplemented with two waves of Transition into Adulthood (TA) data drawn from a national
17
sample of young adults, 18–23 years old. This study included data from participants in the CDS
18
and the TA (n=2,128), children aged 4-13 at baseline. Data on current psychological distress was
19
used as an outcome variable in logistic regressions, calculated as odds ratios (OR) with 95%
20
confidence intervals (CI). Latent Class Analyses were used to identify clusters based on the
21
different childhood stressors. Associations were observed between cumulative exposure to
22
childhood stressors and both psychological distress and self-reported depression. Individuals
23
being exposed to three or more stressors had the highest risk (crude OR for psychological
24
distress: 2.49 (95% CI: 1.16-5.33), crude OR for self-reported depression: 2.07 (95% CI: 1.15-
25
3.71). However, a large part was explained by adolescent depressive symptoms. Findings support
26
the long-term negative impact of cumulative exposure to childhood stress on psychological
27
distress. The important role of adolescent depression in this association also needs to be taken
28
into consideration in future studies.
SC
M AN U
TE D
EP
AC C
29
RI PT
10
2
ACCEPTED MANUSCRIPT Introduction
31
Experiences of stressful or traumatic childhood experiences, often referred to as adverse
32
childhood experiences (ACEs), can negatively affect childhood development, and a series of
33
studies, primarily in the field of psychiatry, have shown that ACEs have negative long-term
34
health and social consequences throughout the life course (Anda, 2008; Bellis et al., 2014;
35
Chapman et al., 2004; Green et al., 2010; McLaughlin et al., 2010; Slopen et al., 2014;
36
Wadsworth & Butterworth, 2006). These studies have shown that children exposed to ACEs have
37
increased risk for depression (Anda, 2008; Chapman et al., 2004; Dube et al., 2003), alcohol
38
abuse (Anda, 2008; Dube et al., 2003), and psychological distress in general (Bellis et al., 2014).
SC
RI PT
30
Research has pointed out several explanations for the association between childhood stress
40
and negative health. Evidence from neurobiology suggests that early life stress causes enduring
41
brain dysfunction which, in turn, affects health and quality of life throughout the lifespan (Anda,
42
2008). This is congruent with the allostatic load theory, suggesting that the neurobiological stress
43
management systems can be permanently altered by cumulative or chronic stress in childhood
44
(Beckie, 2012; McEwen, 2004). Psychological and psychosocial explanations on the other hand
45
suggest that childhood adversity may damage emotional regulation and concept of self-worth,
46
reducing the child’s self- esteem (Wadsworth & Butterworth, 2006), leading to an increased
47
vulnerability for psychological distress. Another discussed explanation for the relationship is that
48
the physical or mental illness in childhood may precede the childhood stressors (e.g. marital
49
distress or financial problem) that in turn lead to health problems in young adulthood (Corman &
50
Kaestner, 1992).
AC C
EP
TE D
M AN U
39
51
Prior research has shown that people from disadvantaged family backgrounds are more likely
52
to accumulate risk factors associated with subsequent health problems, compared to peers born to
53
more privileged families (Anda, 2008; Halldorsson et al., 2000; Kuh et al., 2004; McMunn et al.,
54
2001).
3
ACCEPTED MANUSCRIPT 55
A life course approach offers a framework for studying long-term health effects of physical or social exposures during gestation, childhood, adolescence, young adulthood and later adult life
57
(Kuh et al., 2003; Lynch & Smith, 2005). It emphasizes the importance of time and timing in
58
understanding the causal links between exposure and outcome (Lynch & Smith, 2005). There are
59
several hypotheses on how the effect of exposure can be linked to health-related outcomes.
60
Accumulation of risks over the life span has been suggested as one etiologic pathway to
61
persistent health problems (Kuh et al., 2003). Risk factors in childhood tend to occur in clusters,
62
rather than as single events or experiences (Anda, 2008; Bjorkenstam et al., 2013; Dong et al.,
63
2004), and have a strong positive association with psychiatric and psychological distress (Anda,
64
2008; Bjorkenstam et al., 2013; Chapman et al., 2004; Dube et al., 2003; Slopen et al., 2014).
65
Although much is already known about childhood stressors and the risk for future adverse health,
66
less is known on what clusters of stressors are most closely associated with subsequent distress.
67
The current study uses data from the Panel Study of Income Dynamics (PSID) to examine the
M AN U
SC
RI PT
56
association between cumulative exposure to childhood stressors, and risk of psychological
69
distress in early adulthood. The studied indicators of childhood stress were parental death, single
70
parent household, fair or poor self-rated health in childhood, multiple school changes during the
71
school year, teenage parenthood, household public assistance recipiency, and long-term parental
72
unemployment. The indicators chosen were based on prior research that has shown them to have
73
significant adverse health or social implications (Berg et al., 2014; Conger et al., 1993; Duncan &
74
Brooks-Gunn, 1997; Halldorsson et al., 2000; Hodgkinson et al., 2014; Kuh et al., 2004; Sleskova
75
et al., 2006; Wadsworth & Butterworth, 2006; Weitoft et al., 2003; Wood et al., 1993).
76 77
AC C
EP
TE D
68
We seek to answer the following research questions: -
psychological distress in young adulthood in a large nationally-representative US sample?
78 79 80
Is there an association between cumulative exposure to childhood stressors and
-
What clusters of childhood stressors are most closely associated with future psychological distress? 4
ACCEPTED MANUSCRIPT Methods
82
Study population
83
We used data from the three waves of the Child Development Supplement (CDS-I through CDS-
84
III) and the four waves of the Transition into Adulthood (TA) surveys from the Panel Study of
85
Income Dynamics (PSID). The PSID is a longitudinal study that began in 1968 with a nationally
86
representative sample of about 5,000 families in the United States, with an oversampling of
87
African American and low-income families (McGonagle et al., 2012). The household heads
88
(defined by PSID as the person, at least 16 years old, with the most financial responsibility in the
89
household) were reinterviewed annually until 1997 and every other year thereafter.
SC
In 1997, the PSID began collecting data on a random sample of the PSID families that had
M AN U
90
RI PT
81
children under the age of 13 in the Child Development Study (CDS)-I. The CDS was designed as
92
a nationally representative sample of children in the United States (McGonagle & Sastry, 2014;
93
McGonagle et al., 2012). All PSID families with a child aged 0-12 in the calendar year 1997 were
94
invited to participate, with up to two chosen children per family. Subsequent waves of interviews
95
were carried out in 2002-2003 and 2007-2008, in each case including only children who still were
96
under the age of 18 at the time of the study wave. Most information in the CDS is collected from
97
the participant’s primary caregiver, who must be living with the child. In over 90% this is the
98
child’s biological mother. The children are also interviewed. The entire CDS sample size in 1997
99
is approximately 3,500 children residing in 2,400 households. A follow-up study with these
AC C
EP
TE D
91
100
children and families was conducted in 2002–03 (CDS-II), and another one in 2007-2008 (CDS-
101
III). No new children were included in CDS-II and CDS-III.
102
In 2005, another supplementary study to the PSID was introduced, the Transition to
103
Adulthood (TA) study (Institute for Social Research). The TA component comprises young
104
adults who were children in the CDS and subsequently turned 18. These former children
105
themselves answer all questions in the TA. The TA data have been collected every other year
106
since 2005, with a final wave planned for 2015 (McGonagle & Sastry, 2014). 5
ACCEPTED MANUSCRIPT 107
The sample used in this study combined data from all three waves of the CDS, and all four available waves of the TA. Our sample included 2,128 individuals, born between 1984 and 1993,
109
who participated in the first CDS and at least in one of the TA. Of these individuals, 88% were
110
reinterviewed in CDS-II, and 36% in CDS-III (due to the fact that when the child turned 18,
111
she/he were no longer eligible to answer the CDS) (for additional information on data
112
missingness, see supplementary table 1).
RI PT
108
113
Exposure: childhood stressors
115
The studied indicators are principally based on questions answered by the primary caregivers
116
(PCGs) or other care givers (OCG). We used questions asked when the children were between the
117
ages of 4 and 14. Thus, we used CDS 1997 for all participants (during this year children were
118
between the ages of 4 and 13), CDS 2002 for participants born between 1988 and 1993 (children
119
were between the ages of 9 and 14) and CDS 2007 for those born in 1993 (children were 14 years
120
old). Supplementary table 2 illustrates which waves were used for the different birth cohorts.
121
Additionally, the original PSID studies were used to obtain information household public
122
assistance recipiency and long-term parental unemployment (see below).
125
M AN U
TE D
EP
124
To assess cumulative exposure to the studied indicators, the total number was summed. For each indicator, ever reporting an indicator during any interview was considered one exposure.
AC C
123
SC
114
126
Parental death: Death of a parent is a traumatic life event that is likely to increase stress levels in
127
children (Berg et al., 2014). Captured from the three waves of the CDS, the PCGs were asked if
128
the child’s biological mother and father were still alive. This indicator was coded as 1 if one or
129
both parents died and 0 otherwise.
130
Single parent household: Growing up in a single parent family has become increasingly
131
common, and it may entail disadvantages for the child in terms of socioeconomic circumstances
132
and health (Weitoft et al., 2003). In the CDS, the PCGs were asked whether the child was living 6
ACCEPTED MANUSCRIPT with both parents and, if not, which parent he/she was living with. If the child was living with
134
only one parent at the time of the interview this indicator was coded as 1.
135
Fair or poor self-rated health in childhood: Poor child health is often associated with childhood
136
adversity in terms of poverty and family economic adversity (Duncan & Brooks-Gunn, 1997).
137
The PCGs were asked about the child’s health in general, based on the survey questions: “In
138
general, would you say [child’s name]’s health is excellent, very good, good, fair, or poor?”.
139
Two or more school changes during the school year: Frequent changes of residence during
140
childhood implying that the child changed schools, are associated with an increased risk of
141
psychological distress (Wood et al., 1993). PCGs were asked: “Since the beginning of the school
142
year, how many times has your child changed schools?” This indicator was defined as two or
143
more school changes.
144
Child’s own teenage parenthood: Teenage parenthood is associated with excess stress, and risk
145
of adverse development in many areas (Hodgkinson et al., 2014). For this indicator, we used
146
questions from both the CDS and the TA for both female and male children. Teenage parenthood
147
was recorded as yes if one or more of the following was true: a) Have you ever been pregnant or
148
fathered a child? (from CDS); b) Have you ever been pregnant or fathered a child, and if so, how
149
old were you when the first child was born? (from TA). Pregnancies from age 12 up until age 18
150
were considered.
151
Household public assistance recipiency: Many studies indicate that financial problems during
152
childhood may have long-term influences on physical and mental health (Halldorsson et al.,
153
2000; Kuh et al., 2004; Wadsworth & Butterworth, 2006) . Receiving public assistance may be
154
considered an indicator of poverty (Ringback Weitoft et al., 2008) and children whose families
155
have received public assistance tend to have less satisfactory long-term health (Duncan &
156
Brooks-Gunn, 1997; Ringback Weitoft et al., 2008). An individual was classified as a recipient if
157
her/his family had received public assistance, including food stamps, ADF/AFDC/TANF,
158
Supplemental Security Income (SSI), Social Security Income or other welfare within the
AC C
EP
TE D
M AN U
SC
RI PT
133
7
ACCEPTED MANUSCRIPT preceding year. From the original PSID study, this indicator was obtained when the child was
160
between the ages of 0 and 14.
161
Long-term parental unemployment: Through continued stress, parental unemployment has been
162
associated with increased risk of psychological distress in their children (Conger et al., 1993;
163
Sleskova et al., 2006). In the PSID, the parents were annually asked if they had been unemployed
164
and, if so, for how many weeks. This indicator was defined as an unemployment spell of more
165
than six months during a year, i.e. if an individual had been unemployed for less than six month,
166
the child was not classified as exposed to this indicator. This indicator was obtained when the
167
child was between the ages of 0 and 14.
M AN U
168
SC
RI PT
159
Outcomes
170
Psychological distress was assessed using two different measures. First we used the K6 scale, a
171
self-rated 6-item scale that screens for mood and anxiety disorders (Kessler et al., 2002). K6 was,
172
together with the K10 scale, developed for use in the US National Health Interview Survey. This
173
measure is specifically designed for population-based surveys (Kessler et al., 2002). It has been
174
validated in the United States and around the world as a measure of psychological distress, and it
175
is used in many population-based surveys to estimate the prevalence of mental illness (Kessler et
176
al., 2002; Kessler et al., 2010a). Respondents are asked how often during the past 30 days they
177
felt: so sad that nothing could cheer them up (item A); nervous (item B); restless or fidgety (item
178
C); hopeless (item D); worthless (item E); that everything was an effort (item F). Each item is
179
scaled from 0 (none of the time) to 4 (all of the time). These six items were selected from a pool
180
of 135 items derived from the symptoms used in the diagnosis of major depression and
181
generalized anxiety disorder. The total psychological distress score is computed by summing up
182
the six items scores. Thus, the final score ranges from 0 to 24. Previous research has shown that
183
dichotomous scoring of responses in the range 13+ versus 0–12 discriminates between
184
respondents with and without serious psychological distress with good accuracy (Kessler et al.,
AC C
EP
TE D
169
8
ACCEPTED MANUSCRIPT 2003). Another approach has been to collapse K6 scores into different strata (Furukawa et al.,
186
2003; Kessler et al., 2010a), with scores 1-7 indicating a low score, 8-12 indicating a likelihood
187
of having mild to moderate mental illness, and a score of 13 and more indicating serious mental
188
illness. In this study, we dichotomized K6 and used the 13 as the cut-point indicating
189
psychological distress.
190
RI PT
185
As a second measure we used a question on indications of depression included in each TA study. In young adulthood, individuals were asked: “In the past 12 months, have you had two
192
weeks or longer when nearly every day you felt sad, empty, or depressed for most of the day?”.
193
The responses were dichotomized as “yes’ and “no’.
M AN U
194
SC
191
195
Potential confounders
196
In the analyses, we controlled for several potentially confounding factors, including age
197
(calculated based on birth year), sex, and parental socioeconomic position (SEP). Low socioeconomic position (SEP) in childhood is associated with increased risks of mental
TE D
198
disorder in adulthood (Marmot et al., 2001; McMunn et al., 2001) and is also associated with
200
exposure to a range of risk factors for both somatic ill-health and mental disorders (Bjorkenstam
201
et al., 2013). Hence, we adjusted for socioeconomic background, including parental education,
202
income and occupation during the respondent’s childhood (when the child was 15 years old).
203
Highest attained parental educational level was grouped into the following categories: 0-6th grade,
204
7th-11th grade, high school graduate, post-high school education, and missing. Parental occupation
205
was based on the PCG’s or the OCG’s self-reported occupation at the time of the interview. To
206
create occupational groups across multiple PSID waves, we used the Cross-National Equivalent
207
File’s (CNEF) 2-digit occupational codes prepared at the Ohio State University (Lillard et al.,
208
2015). From a total of approximately 100 codes, we collapsed occupations into seven larger
209
groups: professional, office, administrative, service/manual, skilled/semi-skilled manual, not
210
applicable (non-working) and missing/item non-response. Finally, family income was obtained
AC C
EP
199
9
ACCEPTED MANUSCRIPT 211
from the PSID main family data. As this variable was obtained over a 10-year period, we used
212
consumer-price index to calculate income in real dollars, i.e., to remove the effects of inflation.
213
Prior research has shown that most mental disorders begin early in life, usually during adolescence, even if they might not be discovered until later in life (Kessler et al., 2005; Suvisaari
215
et al., 2009). In the analyses, we adjusted for adolescent depressive symptoms based on the short
216
form of the Children’s Depression Inventory (CDI), which was included in the second and third
217
wave of the CDS. The CDI is a psychological assessment that rates the severity of symptoms
218
related to depression and/or dysthymic disorder in children and adolescents. This widely used
219
assessment with good reliability and validity was developed by Kovacs and was first published in
220
1979 (Allgaier et al., 2012). Adolescents were given 10 sets of three statements and were asked to
221
select one statement from each set to indicate how they had felt over the last two weeks.
222
Measured symptoms included being bothered by things; having no friends; hating oneself;
223
disliking how one looks; and feeling sad, alone, like crying, unloved, that things never work out,
224
and that they do things wrong. Responses were summarized, resulting in a scale ranging from 0 to
225
20. When dichotomizing this variable, a cutoff≥3 was chosen (Allgaier et al., 2012).
226
TE D
M AN U
SC
RI PT
214
Statistical analysis
228
Logistic regression analyses were used to statistically evaluate the association between childhood
229
stressors and psychological distress and self-reported depression, presented as odds ratios (OR)
230
with 95% confidence intervals (CI). Each indicator was analyzed separately. We also analyzed
231
cumulative effects of the studied indicators, in which the total number of indicators was summed.
232
As the PSID is oversampled for African American and low-income families, we used weights
233
provided by the PSID to allow the sample to approximate a representative sample of the US
234
population, i.e. to adjust analyses for attrition and oversampling of low-income families.
235 236
AC C
EP
227
Binary logistic regression models based on the KHB method proposed by Karlson, Holm and Breen (Karlson et al., 2012) were used to estimate OR. The KHB method ensures that the crude 10
ACCEPTED MANUSCRIPT 237
and adjusted coefficients presented are measured on the same scale. As the PSID data are
238
clustered, i.e. there are up to two siblings per household, our analysis must take within family
239
correlation into account. Therefore all models used robust standard errors obtained by using the
240
cluster-option in STATA. There were some missing data for the different childhood stressors. We used both STATA’s
242
multiple imputation command with iteratively chained equations, and the MI and MIANALYZE
243
procedure in SAS to impute missing information (Sterne et al., 2009). However, regression
244
analyses including the imputed values generated similar results as the original data. Thus, except
245
for data on adolescent depressive symptoms, we only present analyses based on the original data.
SC
In order to investigate whether classes or clusters of childhood stressors could be identified,
M AN U
246
RI PT
241
we conducted Latent Class Analysis (LCA), using Latent Gold 4.5 (Statistical Innovations Inc.,
248
Belmont, MA). The LCA included the 2,066 individuals, for which we had no missing data. The
249
goal of LCA is to identify the smallest number of latent classes that adequately describes the
250
association among the observed indicators (Magidson & Vermunt, 2004). We started with the
251
most parsimonious 1-class model and fitted successive models with increasing numbers of
252
classes. There are several strategies available to determine the number of classes (Magidson &
253
Vermunt, 2001; Tein et al., 2013). Goodness-of-fit statistics were used to select the optimal
254
model. We examined the Bayesian information criterion (BIC), the p-value-based likelihood ratio
255
tests, classification error, conditional bootstrap, reduction in L2, and the bivariate residuals
256
(BVRs) in order to determine the best-fitting model (Magidson & Vermunt, 2001, 2004). We first
257
fitted a one-class solution followed by a 2-class solution, and so on, until we reached the best
258
solution. We finally ended up with a 4-class model with two covariates (number of siblings and
259
sex), based on the values of the bootstrap p-value, the entropy R2, conditional bootstrap and the
260
BVRs (Tein et al., 2013) (table 1).
AC C
EP
TE D
247
11
ACCEPTED MANUSCRIPT 261
The next step was to determine whether there were significant differences in ORs across the
262
LCA classes. The modal assignment rule was used to assign cases to classes (Magidson &
263
Vermunt, 2004).
AC C
EP
TE D
M AN U
SC
RI PT
264
12
ACCEPTED MANUSCRIPT
Results
266
Table 2 provides both unweighted and weighted descriptive characteristics for the 2,128 sample
267
members. The sample was equally distributed between the sexes. Forty-three percent of the
268
respondents reported symptoms related to depression in adolescence. Approximately half of the
269
PCGs attended school after high school, and one in five belonged to the high SEP group
270
(measured from occupation) during the participant’s childhood. Around 40 percent (52 percent
271
unweighted) had experienced at least one of the indicators of childhood stress. Living in a single
272
parent household during childhood was common (28 percent), as was growing up in a household
273
receiving public assistance (26 percent). Among the six percent who had experienced three or
274
more indicators in childhood, the combination of single parent household, household public
275
assistance and parental unemployment was most common. When comparing the unweighted and
276
weighted numbers in table 2, it appeared that all indicators were more common in our sample
277
than in the total population as a whole. This was probably due to the oversampling of low income
278
and African American families.
280
SC
M AN U
TE D
279
RI PT
265
The studied childhood stressors were highly inter-correlated (supplementary table 3). Specifically single-parent household and public assistance were highly correlated. Table 3 shows the distribution of psychological distress (based on the K6 scale and self-
282
reported depression), by childhood stressors. In total, seven percent reported a K6 scale of 13 or
283
more. Those who had experienced one or more indicators reported psychological distress more
284
frequently than did those who had not. As the number of indicators increased, the psychological
285
distress rate increased in a graded fashion, e.g. 14 percent of those with three or more indicators
286
reported psychological distress.
AC C
EP
281
287
Seventeen percent in the sample reported clear symptoms of depression in the last 12
288
months. As in the case of K6 scores, people who had experienced childhood stress reported
289
indications of depression to a larger extent. Among young adults with three or more indicators,
290
24 percent reported depression. The corresponding proportion of those with no indicators was 14 13
ACCEPTED MANUSCRIPT 291
percent. Individuals who had depressive symptoms in adolescence more often also reported
292
psychological distress in young adulthood. Results in table 4 suggest an association between almost all of the studied childhood
294
stressors and psychological distress and self-reported depression. Long-term parental
295
unemployment was related to the highest odds of psychological distress crude OR: 2.70; (95%
296
CI: 1.14-6.37) compared with those not experiencing parental unemployment, followed by
297
teenage parenthood crude OR: 2.67 (95% CI: 1.37-5.19). After adjustments were made for social
298
and demographic variables including parental SEP, and for symptoms of depression in
299
adolescence (Model II), OR decreased for a majority of the studied indicators and fewer were
300
statistically significant. The OR for parental unemployment remained significant (OR for
301
psychological distress: 2.67; 1.11-6.43). The ORs indicated an increased risk of psychological
302
distress with cumulative number of indicators. The crude ORs revealed that, compared to
303
adolescents with no indicators, those with one or two indicators had slightly increased risk of
304
psychological distress based on the K6 score (not statistically significant). Those who
305
experienced more indicators had 2.5 times higher risk of psychological distress (95% CI: 1.16-
306
5.33) (Model I).
TE D
M AN U
SC
RI PT
293
As for the outcome of self-reported depression (table 4), experiencing teenage parenthood
308
was associated with the highest OR, followed by long-term parental unemployment and
309
household receipt of public assistance. Further, the ORs increased in a graded fashion as the
310
number of indicators increased.
AC C
311
EP
307
312
Latent class analysis
313
Figure 1 depicts the identified classes along with the conditional probabilities for each of the
314
exposure variables. The classes were labeled according to the levels of the conditional
315
probabilities, and cases were assigned to classes using the modal assignment rule (Magidson &
316
Vermunt, 2001, 2004). Around 68 % (n =1,413) of the individuals were assigned class 1. This 14
ACCEPTED MANUSCRIPT class is mainly characterized by individuals with no childhood stressors and, therefore, the
318
conditional probabilities for the exposure variables are more or less zero for all indicators. As
319
shown in Figure 1, class 2 represents individuals who were mainly exposed to household public
320
assistance and single parent families (approximately 21% of the sample (n=434)). Class 3 is
321
characterized by people who were mainly exposed to single parent households (around 7% (n =
322
151)). Finally, class 4 represents the 3 % (n=69) experiencing teenage parenthood, single parent
323
households and household public assistance. In this class, girls constituted 85 %, whereas for
324
other classes the distribution was similar between the sexes.
SC
325
RI PT
317
Particularly individuals in class 4 reported psychological distress (table 3). Over 40 percent of the individuals in class 4 reported to have been depressed in the past 12 months, and 13
327
percent scored above 12 on the K6 scale. Individuals in class 4 had a fourfold risk of self-reported
328
depression compared to class 1 (crude OR: 4.42; 95% CI: 2.15-9.09). A substantial risk increase
329
remained even after controlling for demographic variables (OR: 3.61; 95% CI: 1.78-7.31).
M AN U
326
333 334 335 336 337
EP
332
AC C
331
TE D
330
15
ACCEPTED MANUSCRIPT
Discussion
339
Our study of 2,128 young adults shows that our indicators of childhood stress were associated
340
with an increased risk of psychological distress. The risk of psychological distress increased with
341
higher number of indicators. Application of LCA yielded four classes, where one class,
342
characterized by exposure to teenage parenthood, household public assistance and single parent
343
household, presented particularly increased risks for psychological distress in terms of self-
344
reported depression. Findings further suggest that the effects of childhood stress on psychological
345
distress in young adults to a large extent are explained by adolescent depression.
SC
RI PT
338
When the indicators were studied separately, experience of teenage parenthood was
347
particularly associated with future psychological distress. In our sample two thirds of the ones
348
reporting to have experienced teenage parenthood were girls, and when stratifying the analyses
349
by sex (data not shown), females with teenage parenthood had slightly higher ORs than males
350
who experienced teenage parenthood. Studies have shown that teenage parenthood may be
351
associated with adverse outcomes later in life, including single parenthood, and enduring long-
352
term poverty (Felice et al., 1999; Hodgkinson et al., 2014), both which are associated with
353
psychological distress (Felice et al., 1999; Hodgkinson et al., 2014). However, important
354
selection factors may contribute to both teenage pregnancy and later ill health. Public health
355
policies in the US and UK have included teenage parenthood as a national public health problem
356
requiring targeted interventions (Felice et al., 1999).
AC C
EP
TE D
M AN U
346
357
Consistent with prior research in various settings, both from the US and Europe (Anda, 2008;
358
Bellis et al., 2014; Bjorkenstam et al., 2013; Chapman et al., 2004; Dube et al., 2003; Slopen et
359
al., 2014), we found indications of a graded relationship between cumulative exposure to
360
childhood stressors and psychological distress. These findings provide evidence that childhood
361
stressors often are interrelated rather than occurring separately. This information has important
362
implications for intervention because it means that prevention or amelioration of only a single
16
ACCEPTED MANUSCRIPT 363
stressor in youths exposed to many stressors is unlikely to have important preventive effects
364
(Green et al., 2010; Kessler et al., 2010b). In an attempt to disentangle the complex associations between social factors and childhood
366
stress, we adjusted for important parental variables including parental education, occupation and
367
income. After these adjustments were made, risk estimates decreased to a large extent. Thus, our
368
findings support previous research (Gilman et al., 2002; Marmot et al., 2001) that has shown an
369
association between parental SEP and offspring’s mental health.
RI PT
365
The ORs were attenuated when taking adolescent depressive symptoms into account.
371
Consistent with prior research (Kessler et al., 2005; Suvisaari et al., 2009), depressive symptoms
372
in adolescence were in themselves associated with an increased risk of subsequent psychological
373
distress, also when childhood adversity and parental SEP were held constant. This result is in line
374
with past research, showing that many persistent mental disorders have antecedents in
375
adolescence (Kessler et al., 2005; Suvisaari et al., 2009). One could consider adolescent
376
depression an outcome in this study. However, as adolescent depression was captured in the
377
CDS, i.e. at the same time as the exposure (and possibly even before), we chose to include it as a
378
confounder instead.
TE D
M AN U
SC
370
The graded relationship between cumulative exposure to the studied indicators of childhood
380
stress and psychological distress was less evident when adjusting for demographic variables and
381
parental SEP. An inevitable question in the interpretation of these results is why the pattern is not
382
as evident as other studies have shown, i.e. the typical dose-response manner in cumulative
383
exposure and psychological distress. There may be several explanations to this finding. A recent
384
US study argued that using cumulative risk by summing dichotomous scores has its shortcomings
385
(Evans et al., 2013), some of which we may have encountered in this study. For instance we do
386
not capture the level of intensity of a particular adverse experience – for example, a parent’s
387
death may have a much more profound effect on a child than changing schools -- a shortcoming
388
that was pointed out in a recent US study (Slopen et al., 2014). Moreover, we do not consider the
AC C
EP
379
17
ACCEPTED MANUSCRIPT 389
timing of exposure to adversity although research on divorce, for example, suggests that the
390
effects on children differ by children’s age (Amato & Keith, 1991). As an alternative approach to the cumulative risk measure, we used LCA to identify
392
important subgroups with various patterns of coexisting indicators. Even though several studies
393
have shown that indicators of childhood stress tend to occur in clusters rather than as single
394
events, less is known about what these clusters may look like. Our LCA revealed four classes.
395
We identified one risk group (class 4) with particularly pronounced ORs for psychological
396
distress in terms of depression. The class was foremost represented by girls who had grown up in
397
households receiving public assistance, and who had experienced teenage parenthood. Although a
398
rather small class (n=69), this group had three times higher risk of self-reported depression
399
compared to those with no indicators. A recent US study identified a similar group as a high-risk
400
group for psychological distress, showing that adolescent parenthood is associated with a range of
401
mental health problems, and that teen mothers are more likely to reside in families that are
402
socially and economically disadvantaged (Hodgkinson et al., 2014). Our results reinforce the
403
importance of paying attention to the psychosocial stressors faced by young mothers, when
404
interacting with these women in primary care settings.
TE D
M AN U
SC
RI PT
391
EP
405
Strengths and limitations
407
The strengths of the current study include the use of the PSID, one of few nationally
408
representative longitudinal data sets in the U.S. spanning the entire life course (and multiple
409
generations). Another strength includes the ability to estimate latent classes to identify important
410
subgroups. We were also able to take multiple measures of parental SEP into account.
AC C
406
411
The study also has methodological weaknesses, some of which have been addressed above.
412
The structure of the data means that alternative causal pathways cannot be fully discounted. Thus,
413
it is possible that psychological distress in early adolescence also leads to future distress. We tried
414
to handle this by adjusting for adolescent depressive symptoms. Regarding the exposure 18
ACCEPTED MANUSCRIPT information obtained from the CDS: for the older individuals (i.e. those born before 1990), we
416
were only able to obtain exposure information once (i.e. in CDS-I), whereas for the younger,
417
additional information was captured in the CDS-II and CDS-III. Hence, the older children may
418
have been exposed to childhood adversity before the first interview took place. This potential
419
misclassification of exposure may lead to a dilution. The findings in our study are also limited by
420
the difficulty in truly capturing the concepts being measured. While tested and used widely, the
421
measures of psychological distress are only survey instruments, and not diagnostic tools. Further,
422
the prevalence of some of the indicators may be underreported, which would bias our findings
423
towards the null. Finally, the assessment of cumulative adversity was not comprehensive, and
424
important childhood stressors (e.g. substance abuse and crime in the home) might have been
425
omitted which would bias our results towards the null.
M AN U
SC
RI PT
415
426
Conclusion
428
Despite the above limitations, the association we found between cumulative exposure to
429
childhood stress and psychological distress calls for further attention. The important role of
430
adolescent depression in this association also needs to be taken into consideration in future
431
studies. Sources of childhood stress, including the indicators studied here, can be seen as key risk
432
factors for poor outcomes that policy can address. Sufficient evidence is already available for
433
governments to prioritize and invest in preventative interventions.
435
EP
AC C
434
TE D
427
19
ACCEPTED MANUSCRIPT References
AC C
EP
TE D
M AN U
SC
RI PT
Allgaier, A.K., Fruhe, B., Pietsch, K., Saravo, B., Baethmann, M., & Schulte-Korne, G. (2012). Is the Children's Depression Inventory Short version a valid screening tool in pediatric care? A comparison to its full-length version. J Psychosom Res, 73, 369-374. Amato, P.R., & Keith, B. (1991). Parental divorce and the well-being of children: a meta-analysis. Psychol Bull, 110, 26-46. Anda, R. (2008). The health and social impact of growing up with adverse childhood experiences. http://acestudy.org/files/Review_of_ACE_Study_with_references_summary_table_2_.pdf. Beckie, T.M. (2012). A systematic review of allostatic load, health, and health disparities. Biol Res Nurs, 14, 311-346. Bellis, M.A., Lowey, H., Leckenby, N., Hughes, K., & Harrison, D. (2014). Adverse childhood experiences: retrospective study to determine their impact on adult health behaviours and health outcomes in a UK population. J Public Health (Oxf), 36, 81-91. Berg, L., Rostila, M., Saarela, J., & Hjern, A. (2014). Parental death during childhood and subsequent school performance. Pediatrics, 133, 682-689. Bjorkenstam, E., Hjern, A., Mittendorfer-Rutz, E., Vinnerljung, B., Hallqvist, J., & Ljung, R. (2013). Multi-exposure and clustering of adverse childhood experiences, socioeconomic differences and psychotropic medication in young adults. PLoS One, 8, e53551. Chapman, D.P., Whitfield, C.L., Felitti, V.J., Dube, S.R., Edwards, V.J., & Anda, R.F. (2004). Adverse childhood experiences and the risk of depressive disorders in adulthood. J Affect Disord, 82, 217-225. Conger, R., Conger, K., Elder, G., Lorenz, F., Simons, R., & Whitbeck, L. (1993). Family economic stress and adjustment of early adolescent girls. Dev Psychol, 29, 206-219. Corman, H., & Kaestner, R. (1992). The effects of child health on marital status and family structure. Demography, 29, 389-408. Dong, M., Anda, R.F., Felitti, V.J., Dube, S.R., Williamson, D.F., Thompson, T.J., et al. (2004). The interrelatedness of multiple forms of childhood abuse, neglect, and household dysfunction. Child Abuse Negl, 28, 771-784. Dube, S.R., Felitti, V.J., Dong, M., Giles, W.H., & Anda, R.F. (2003). The impact of adverse childhood experiences on health problems: evidence from four birth cohorts dating back to 1900. Prev Med, 37, 268-277. Duncan, G.J., & Brooks-Gunn, J. (1997). Consequences of Growing Up Poor. New York: Russell Sage Foundation. Evans, G.W., Li, D., & Whipple, S.S. (2013). Cumulative risk and child development. Psychol Bull, 139, 1342-1396. Felice, M.E., Feinstein, R.A., Fisher, M.M., Kaplan, D.W., Olmedo, L.F., Rome, E.S., et al. (1999). Adolescent pregnancy--current trends and issues: 1998 American Academy of Pediatrics Committee on Adolescence, 1998-1999. Pediatrics, 103, 516-520. Furukawa, T.A., Kessler, R.C., Slade, T., & Andrews, G. (2003). The performance of the K6 and K10 screening scales for psychological distress in the Australian National Survey of Mental Health and Well-Being. Psychol Med, 33, 357-362. Gilman, S.E., Kawachi, I., Fitzmaurice, G.M., & Buka, S.L. (2002). Socioeconomic status in childhood and the lifetime risk of major depression. Int J Epidemiol, 31, 359-367. Green, J.G., McLaughlin, K.A., Berglund, P.A., Gruber, M.J., Sampson, N.A., Zaslavsky, A.M., et al. (2010). Childhood adversities and adult psychiatric disorders in the national comorbidity survey replication I: associations with first onset of DSM-IV disorders. Arch Gen Psychiatry, 67, 113-123. Halldorsson, M., Kunst, A., Köhler, L., & Machenback, J. (2000). Socioeconomic inequalities in the health of children and adolescents. A comparative study of the five Nordic countries. Eur J Public Health, 10, 281-288. Hodgkinson, S., Beers, L., Southammakosane, C., & Lewin, A. (2014). Addressing the mental health needs of pregnant and parenting adolescents. Pediatrics, 133, 114-122. Institute for Social Research. The Panel Study of Income Dynamics [data files and codebooks]. Ann Arbor, MI: University of Michigan. Karlson, K., Holm, A., & Breen, R. (2012). Comparing Regression Coefficients Between Same-sample Nested Models Using Logit and Probit: A New Method. Sociological Methodology, 42, 286313. Kessler, R.C., Andrews, G., Colpe, L.J., Hiripi, E., Mroczek, D.K., Normand, S.L., et al. (2002). Short screening scales to monitor population prevalences and trends in non-specific psychological distress. Psychol Med, 32, 959-976.
20
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
Kessler, R.C., Barker, P.R., Colpe, L.J., Epstein, J.F., Gfroerer, J.C., Hiripi, E., et al. (2003). Screening for serious mental illness in the general population. Arch Gen Psychiatry, 60, 184-189. Kessler, R.C., Berglund, P., Demler, O., Jin, R., Merikangas, K.R., & Walters, E.E. (2005). Lifetime prevalence and age-of-onset distributions of DSM-IV disorders in the National Comorbidity Survey Replication. Arch Gen Psychiatry, 62, 593-602. Kessler, R.C., Green, J.G., Gruber, M.J., Sampson, N.A., Bromet, E., Cuitan, M., et al. (2010a). Screening for serious mental illness in the general population with the K6 screening scale: results from the WHO World Mental Health (WMH) survey initiative. Int J Methods Psychiatr Res, 19 Suppl 1, 4-22. Kessler, R.C., McLaughlin, K.A., Green, J.G., Gruber, M.J., Sampson, N.A., Zaslavsky, A.M., et al. (2010b). Childhood adversities and adult psychopathology in the WHO World Mental Health Surveys. Br J Psychiatry, 197, 378-385. Kuh, D., Ben-Shlomo, Y., Lynch, J., Hallqvist, J., & Power, C. (2003). Life course epidemiology. J Epidemiol Community Health, 57, 778-783. Kuh, D., Power, C., Blane, D., & Bartley, M. (2004). Socioeconomic pathways between childhood and adult health. In D. Kuh, & Y. Ben-Shlomo (Eds.), A Life Course Approach to Chronic Disease Epidemiology. Oxford: Oxford University Press. Lillard, R., Christopoulou, R., Goebel, J., Freidin, S., Jaber, A., Lipps, O., et al. (2015). The CrossNational Equivalent Files 1970-2009. Ohio State University: http://cnef.ehe.osu.edu/ Lynch, J., & Smith, G.D. (2005). A life course approach to chronic disease epidemiology. Annu Rev Public Health, 26, 1-35. Magidson, J., & Vermunt, J. (2001). Latent Class Factor and Cluster Models, Bi-plots, and related Graphical Displays. Sociological Methodology, 31, 223-264. Magidson, J., & Vermunt, J. (2004). Latent Class Models. In D. Kaplan (Ed.), The Sage Handbook of Quantitative Methodology for the Social Sciences. Thousand Oaks: Sage Publications. Marmot, M., Shipley, M., Brunner, E., & Hemingway, H. (2001). Relative contribution of early life and adult socioeconomic factors to adult morbidity in the Whitehall II study. J Epidemiol Community Health, 55, 301-307. McEwen, B.S. (2004). Protection and damage from acute and chronic stress: allostasis and allostatic overload and relevance to the pathophysiology of psychiatric disorders. Ann N Y Acad Sci, 1032, 1-7. McGonagle, K.A., & Sastry, N. (2014). Cohort Profile: The Panel Study of Income Dynamics' Child Development Supplement and Transition into Adulthood Study. Int J Epidemiol, 1-8. McGonagle, K.A., Schoeni, R.F., Sastry, N., & Freedman, V.A. (2012). The Panel Study of Income Dynamics: Overview, Recent Innovations, and Potential for Life Course Research. Longit Life Course Stud, 3, 268-284. McLaughlin, K.A., Green, J.G., Gruber, M.J., Sampson, N.A., Zaslavsky, A.M., & Kessler, R.C. (2010). Childhood adversities and adult psychiatric disorders in the national comorbidity survey replication II: associations with persistence of DSM-IV disorders. Arch Gen Psychiatry, 67, 124-132. McMunn, A.M., Nazroo, J.Y., Marmot, M.G., Boreham, R., & Goodman, R. (2001). Children's emotional and behavioural well-being and the family environment: findings from the Health Survey for England. Soc Sci Med, 53, 423-440. Ringback Weitoft, G., Hjern, A., Batljan, I., & Vinnerljung, B. (2008). Health and social outcomes among children in low-income families and families receiving social assistance--a Swedish national cohort study. Soc Sci Med, 66, 14-30. Sleskova, M., Salonna, F., Geckova, A.M., Nagyova, I., Stewart, R.E., van Dijk, J.P., et al. (2006). Does parental unemployment affect adolescents' health? J Adolesc Health, 38, 527-535. Slopen, N., Koenen, K.C., & Kubzansky, L.D. (2014). Cumulative adversity in childhood and emergent risk factors for long-term health. J Pediatr, 164, 631-638 e631-632. Sterne, J.A., White, I.R., Carlin, J.B., Spratt, M., Royston, P., Kenward, M.G., et al. (2009). Multiple imputation for missing data in epidemiological and clinical research: potential and pitfalls. BMJ, 338, b2393. Suvisaari, J., Aalto-Setala, T., Tuulio-Henriksson, A., Harkanen, T., Saarni, S.I., Perala, J., et al. (2009). Mental disorders in young adulthood. Psychol Med, 39, 287-299. Tein, J.Y., Coxe, S., & Cham, H. (2013). Statistical Power to Detect the Correct Number of Classes in Latent Profile Analysis. Struct Equ Modeling, 20, 640-657. Wadsworth, M., & Butterworth, S. (2006). Early life. In M. Marmot, & R. Wilkinson (Eds.), Social Determinants of Health. Great Britain: Oxford University Press. Weitoft, G.R., Hjern, A., Haglund, B., & Rosen, M. (2003). Mortality, severe morbidity, and injury in children living with single parents in Sweden: a population-based study. Lancet, 361, 289-295.
21
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
Wood, D., Halfon, N., Scarlata, D., Newacheck, P., & Nessim, S. (1993). Impact of family relocation on children's growth, development, school function, and behavior. JAMA, 270, 1334-1338.
22
ACCEPTED MANUSCRIPT Figures and tables
M AN U
SC
RI PT
Figure 1. Conditional Probabilites for Latent Classes
Table1. Model Fit Statistics for Latent Class Analysis L² 848.71
2 3 4 5 6 7
7,271.74 7,259.81 7,306.22 7,356.22 7,412.33 7,467.33
326.11 237.84 207.92 181.58 161.36 140.02
Classification Error Reduction in L 0.00 0.0%
TE D
BIC(LL) 7,718.01
0.09 0.08 0.09 0.09 0.10 0.13
AC C
EP
Model 1
23
61.6% 72.0% 75.5% 78.6% 81.0% 83.5%
2
2
Bootstrap P-value 0.000
Entropy R 1.000
0.000 0.004 0.018 0.084 0.180 0.346
0.613 0.724 0.742 0.753 0.732 0.774
ACCEPTED MANUSCRIPT Table 2. Sample characteristics of the participants in the Child Development Supplement Study of the Panel Study of Income Dynamics, absolute numbers and percent. Unweighted
Weighted
1,076 (51) 1,052 (49)
1,084 1,044
Age at baseline (i.e. 1997) 4-6 7-9 10-13
568 (27) 658 (31) 902 (42)
542.7 693.1 892.2
Adolescent depressive symptoms 0 1-2 ≥3
412 (19) 795 (37) 921 (43)
411.3 (19) 756.1 (36) 960.6 (45)
SC
Parental educational level None - 6th grade 7th - 11th grade High school graduate Post high school education Missing
RI PT
Demographic characteristics Sex Women Men
133.9 (6) 166.7 (8) 591.3 (28) 1,186.0 (56) 50 (2)
415 (20) 478 (22) 489 (23) 359 (17) 290 (14) 89 (4) 8 (0)
471.0 (22) 503.3 (24) 464.6 (22) 321.9 (15) 280.9 (13) 84.5 (4) 1.9 (0)
528 (25) 530 (25) 529 (25) 528 (25) 13 (1)
405.3 (19) 481.4 (23) 554.8 (26) 677.2 (32) 9.3 (0)
Indicators of childhood stress Parental death Single parent household Poor/fair health in childhood Two or more school changes Teenage pregnancy Household receiving public assistance Long-term parental unemployment (i.e. ≥6 months of a year) Household receiving public assistance (excluding food stamps)
68 (3) 756 (36) 88 (4) 48 (2) 226 (11) 702 (33) 96 (5) 459 (22)
40.6 (2) 603.7 (28) 82.0 (4) 37.5 (2) 137.5 (6) 548.9 (26) 48.6 (2) 336.6 (16)
Cumulative number of childhood stressors 0 1 2 3+
993 (47) 507 (24) 384 (18) 218 (10)
1,179.0 (56) 472.9 (22) 323.9 (15) 118.1 (6)
M AN U
104 (5) 198 (9) 694 (33) 1,085 (51) 47 (2)
AC C
EP
Family income < 31,416 31,417 - 56,437 56,438 - 98,422 98,423> Missing
TE D
Parental occupation 1-9: Prof. 10-21: Office workers 30-49: Admin. 50-64: Service/Manual 70-101: Skilled/semi-skilled manual 0: Not applicable (non-working) Missing/Item non-response
24
ACCEPTED MANUSCRIPT
Table 3. Psychological distress (based on the K6 scale and self-reported depression), by indicators of childhood stress. Weighted numbers and row percent. K6-scores 8-12 550.3 (26)
13+ 159.0 (7)
74.8 (4) 25.4 (4) 11.7 (14) 1.4 (4) 3.9 (3) 27.1 (5) 1.0 (2)
1,318.0 (63) 364.1 (60) 39.8 (49) 24.4 (65) 58.9 (43) 302.4 (55) 25.8 (53)
540.4 (26) 162.8 (27) 25.1 (31) 6.4 (17) 52.1 (38) 164.0 (30) 13.3 (27)
154.0 (7) 51.2 (8) 5.4 (7) 5.3 (14) 22.6 (16) 55.5 (10) 8.5 (17)
Cumulative number of childhood stressors 0 1 2 3+
40.0 (3) 11.3 (2) 16.2 (5) 8.4 (7)
789.0 (67) 282.0 (60) 184.4 (57) 60.1 (51)
Adolescent depressive symptoms 0 1-2 ≥3
21.8 (5) 35.4 (5) 18.6 (2)
303.1 (74) 527.9 (70) 511.9 (53)
Latent classesa Class 1: “No indicators” Class 2: “Public assistance/single parent household” Class 3: “Single parent household” Class 4: “Teenage parenthood, public assistance, and single parent household”
1.5 (3)
2,087.2 (100) 603.5 (100) 82.0 (100) 37.5 (100) 137.5 (100) 549.0 (100) 48.6 (100)
33.9 (84) 483.8 (80) 66.6 (81) 27.7 (74) 86.7 (63) 417.7 (76) 35.8 (74)
6.6 (16) 119.9 (20) 15.4 (19) 9.8 (26) 50.8 (37) 131.2 (24) 12.8 (26)
40.5 (100) 603.7 (100) 82.0 (100) 37.5 (100) 137.5 (100) 548.9 (100) 48.6 (100)
M AN U 70.0 (6) 43.0 (9) 29.1 (9) 16.0 (14)
1,179.0 (100) 473.0 (100) 323.9 (100) 118.0 (100)
1,019.0 (86) 372.0 (79) 247.3 (76) 89.1 (76)
160.1 (14) 100.9 (21) 76.6 (24) 28.9 (24)
1,179.1 (100) 472.9 (100) 323.9 (100) 118.0 (100)
75.4 (18) 162.2 (21) 312.7 (33)
11.0 (3) 30.6 (4) 117.4 (12)
411.3 (100) 756.1 (100) 960.6 (100)
377.1 (92) 693.4 (92) 690.0 (72)
34.2 (8) 62.8 (8) 270.5 (28)
411.3 (100) 756.1 (100) 960.6 (100)
890.1 (66) 275.0 (58) 113.3 (65)
327.9 (24) 131.3 (28) 46.6 (27)
91.9 (7) 46.0 (10) 10.7 (6)
1,352.8 (100) 477.1 (100) 175.5 (100)
1,154.0 (85) 370.6 (78) 144.5 (82)
199.6 (15) 106.7 (22) 31.0 (18)
1,353.6 (100) 477.3 (100) 175.5 (100)
25.6 (43)
25.3 (42)
7.5 (13)
59.9 (100)
34.0 (57)
26.0 (43)
60.0 (100)
Based on the 2,066 individuals included in the Latent Class Analysis
AC C
a
42.9 (3) 24.8 (5) 4.9 (3)
Self-reported depression No depression Depression Total 1,760.0 (83) 367.5 (17) 2,128.0 (100)
280.3 (24) 136.7 (29) 94.2 (29) 33.5 (28)
TE D
Indicators of childhood stress Parental death Single parent household Poor/fair health in childhood 2 or more school changes Teenage pregnancy Household receiving public assistance* Long-term parental unemployment (i.e. ≥6 months of a year)
EP
N (%)
Total 2,128.0 (100)
RI PT
1-7 1,343.0 (63)
SC
0 75.8 (4)
25
ACCEPTED MANUSCRIPT
Table 4. Indicators of childhood stress and the risk of psychological distress, weighted odds ratios (OR) with 95% confidence intervals (CI). Psychological distress (K6 score >12) Model Ia Model IIb
Self-reported depression Model Ia Model IIb
Indicators of childhood stress Parental death Single parent household Poor/fair health in childhood 2 or more school changes Teenage parenthood Household receiving public assistance Long-term parental unemployment (i.e. ≥6 months of a year)
1.79 (0.81-4.54) 1.22 (0.77-1.93) 0.87 (0.36-2.08) 2.07 (0.73-5.87) 2.67 (1.37-5.19) 1.60 (1.01-2.53) 2.70 (1.14-6.37)
1.83 (0.73-4.59) 1.02 (0.64-1.62) 0.97 (0.38-2.49) 1.96 (0.67-5.74) 1.84 (0.98-3.45) 1.29 (0.75-2.23) 2.67 (1.11-6.43)
Cumulative numbers of childhood stressors 0 1 2 3+
1 (REF) 1.58 (0.96-2.62) 1.57 (0.84-2.90) 2.49 (1.16-5.33)
1 (REF) 1.26 (0.74-2.14) 1.42 (0.71-2.83) 2.05 (0.94-4.47)
1 (REF) 1.73 (1.21-2.46) 1.97 (1.27-3.06) 2.07 (1.15-3.71)
1 (REF) 1.41 (0.94-2.11) 1.95 (1.15-3.30) 1.84 (1.01-3.36)
Adolescent depressive symptoms No Yes
1 (REF) 3.77 (2.36-6.01)
1 (REF) 3.27 (2.01-5.30)
1 (REF) 4.33 (3.14-5.87)
1 (REF) 3.66 (2.62-5.12)
a
TE D
M AN U
SC
RI PT
0.83 (0.32-2.10) 1.10 (0.74-1.63) 1.26 (0.59-2.72) 1.54 (0.70-3.40) 2.28 (1.44-3.60) 1.63 (1.07-2.46) 1.49 (0.75-2.96)
Psychological distress (K6 score >12) Self-reported depression Model Ia Model IIb Model Ia Model IIb 1 (REF) 1 (REF) 1 (REF) 1 (REF) 1.46 (0.89-2.41) 1.29 (0.72-2.31) 1.66 (1.15-2.40) 1.71 (1.09-2.66) 0.89 (0.39-2.06) 0.93 (0.41-2.13) 1.24 (0.71-2.18) 1.37 (0.72-2.58)
EP
Latent classes Class 1: “No indicators” Class 2: “Public assistance/single parent household” Class 3: “Single parent household” Class 4: “Teenage parenthood, public assistance, and single parent household”
0.94 (0.42-2.08) 1.28 (0.90-1.81) 1.10 (0.54-2.25) 1.67 (0.74-3.80) 3.09 (1.96-4.88) 1.78 (1.26-2.53) 1.74 (0.83-3.64)
1.97 (0.56-6.91)
1.14 (0.37-3.52)
4.42 (2.15-9.09)
AC C
Crude b Adjusted for sex, birth year, parental SEP (income, education and occupation), and adolescent depressive symptoms
26
3.61 (1.78-7.31)
ACCEPTED MANUSCRIPT
TE D
M AN U
SC
RI PT
Supplementary figure 1. A conceptual framework for investigation of the effect of childhood stressors on psychological distress.
Supplementary table 1. Participants age at the different interviews and percentage missingness for each interview. CDS-II Age at interview % Missing
CDS-III Age at interview % Missing N/A
N/A
21
N/A
N/A
20
1984 (n=204)
13
0
N/A
32
1985 (n=232)
12
0
17
13
1986 (n=232)
11
0
16
11
1987 (n=234)
10
0
15
1988 (n=230)
9
0
14
1989 (n=215)
8
0
13
1990 (n=213)
7
0
12
1991 (n=228)
6
0
1992 (n=220)
5
1993 (n=120)
4
EP
Birth year
CDS-I Age at interview % Missing
TA 2005 Age at interview % Missing
TA 2007 Age at interview % Missing
TA 2009 Age at interview % Missing
TA 2011 Age at interview % Missing
13
23
12
25
12
27
16
12
22
16
24
16
26
20
N/A
19
12
21
13
23
14
25
16
N/A
N/A
18
38
20
15
22
14
24
15
16
N/A
N/A
N/A
N/A
19
11
21
11
23
12
12
N/A
N/A
N/A
N/A
18
46
20
8
22
13
9
17
14
N/A
N/A
N/A
N/A
19
8
21
10
11
9
16
11
N/A
N/A
N/A
N/A
18
30
20
7
0
10
8
15
9
N/A
N/A
N/A
N/A
N/A
N/A
19
0
0
9
6
14
9
N/A
N/A
N/A
N/A
N/A
N/A
18
0
AC C
N/A
9
Note. CDS = Child Development Supplement; TA = Transition into Adulthood
27
ACCEPTED MANUSCRIPT
Exposure/background info from CDS
Exposure/background info from PSID
Follow-up
1984 1985 1986 1987 1988 1989 1990 1991 1992 1993
CDS 1997 CDS 1997 CDS 1997 CDS 1997 CDS 1997 and 2002 CDS 1997 and 2002 CDS 1997 and 2002 CDS 1997 and 2002 CDS 1997 and 2002 All waves of CDS
PSID 1984-1998 PSID 1985-1999 PSID 1986-2000 PSID 1987-2001 PSID 1988-2002 PSID 1989-2003 PSID 1990-2004 PSID 1991-2005 PSID 1992-2006 PSID 1993-2007
All waves of TA All waves of TA All waves of TA All waves of TA TA2007 – TA2011 TA2007 – TA2011 TA2009 – TA2011 TA2009 – TA2011 TA2011 TA2011
M AN U
SC
Birth year
RI PT
Supplementary table 2. Study sample from Child Development Supplement (CDS) and Transition into Adulthood (TA), supplemental studies to the Panel Study of Income Dynamics (PSID), 1997-2011
Supplementary table 3. Pearson correlation matrix of sample characteristics and indicators of childhood stress
1.00 -0.01 0.04* -0.01 -0.03 0.02 0.00 0.02 0.04 0.00 0.00
3
4
5
6
7
8
9
10
11
12
1.00 -0.30** 0.30** -0.09** -0.19** -0.09** -0.04** -0.13** -0.29** -0.10**
1.00 0.21*** 0.05* 0.09*** 0.01 -0.02 0.06** 0.13*** 0.07**
1.00 -0.07** -0.25*** -0.07** -0.06** -0.13*** -0.29*** -0.10***
1.00 0.18*** 0.02 0.04* 0.06** 0.14*** 0.01
1.00 0.06** 0.07** 0.14*** 0.42*** 0.14***
1.00 0.06** 0.04 0.16*** 0.02
1.00 0.06** 0.08** -0.02
1.00 0.21*** 0.13***
1.00 0.25***
1.00
TE D
2
EP
1 1.00 0.13*** -0.01 0.00 0.02 -0.01 -0.02 0.00 -0.01 0.00 0.00 0.01
AC C
1. Sex 2. Adolescent Depression Symptoms 3. Parental education 4. Parental occupation 5. Parental income 6. Parental death 7. Single parent household 8. Fair/poor health in childhood 9. Two or more school changes 10. Teenage parenthood 11. Public assistance 12. Long-term parental unemployment Note: * p
