Adverse childhood experiences (ACEs) and later-life depression

CheongEV, etal. BMJ Open 2017;7:e013228. doi:10.1136/bmjopen-2016-013228

Open Access

ABSTRACT

Objective To investigate associations between adverse

childhood experiences (ACEs) and later-life depressive

symptoms; and to explore whether perceived social

support (PSS) moderates these.

Method We analysed baseline data from the

Mitchelstown (Ireland) 2010–2011 cohort of 2047

men and women aged 50–69 years. Self-reported

measures included ACEs (Centre for Disease Control

ACE questionnaire), PSS (Oslo Social Support Scale) and

depressive symptoms (CES-D). The primary exposure

was self-report of at least one ACE. We also investigated

the effects of ACE exposure by ACE scores and ACE

subtypes abuse, neglect and household dysfunction.

Associations between each of these exposures and

depressive symptoms were estimated using logistic

regression, adjusted for socio-demographic factors. We

tested whether the estimated associations varied across

levels of PSS (poor, moderate and strong).

Results 23.7% of participants reported at least

one ACE (95% CI 21.9% to 25.6%). ACE exposures

(overall, subtype or ACE scores) were associated with a

higher odds of depressive symptoms, but only among

individuals with poor PSS. Exposure to any ACE (vs

none) was associated with almost three times the odds

of depressive symptoms (adjusted OR 2.85; 95% CI

1.64 to 4.95) among individuals reporting poor PSS,

while among those reporting moderate and strong PSS,

the adjusted ORs were 2.21 (95% CI 1.52 to 3.22) and

1.39 (95% CI 0.85 to 2.29), respectively. This pattern

of results was similar when exposures were based on

ACE subtype and ACE scores, though the interaction was

clearly strongest among those reporting abuse.

Conclusions ACEs are common among older adults in

Ireland and are associated with higher odds of later-life

depressive symptoms, particularly among those with

poor PSS. Interventions that enhance social support, or

possibly perceptions of social support, may help reduce

the burden of depression in older populations with ACE

exposure, particularly in those reporting abuse.

INTRODUCTION

A life-course approach to mental health views

mental illness as a product of biological and

social factors that operate across the lifespan.

The stress sensitisation theory

suggests that

childhood adversity reduces an individual’s

threshold for developing depressive reactions

towards stressful events, causing one to have

depressive reactions towards current mild

stressors or greater reactivity towards severe

stressful events. For example, young women

who were exposed to childhood adversities

such as domestic violence, parent psychopa-

thology and alcoholism are at a higher risk for

depression following exposure to mild stress

than women without a history of adversity.

Women with a history of childhood abuse

have higher adrenocorticotropic hormone

(ACTH), cortisol and heart rate responses

to psychosocial stress such as public speaking

compared with those without a history of

childhood abuse.

The relationship persists

into older adulthood; data from the Health

and Retirement Study, a US population-based

study of adults age 50+, showed that in accor-

dance with the stress sensitisation theory,

childhood trauma (especially physical abuse)

amplifies the effect of stresses in adulthood

on depressive symptoms.

These psychosocial and neurobiological

findings converge on the idea that early-life

adversities have an enduring effect on how

Adverse childhood experiences (ACEs)

and later-life depression: perceived

social support as a potential protective

factor

E Von Cheong,

Carol Sinnott,

Darren Dahly,

Patricia M Kearney

To cite: CheongEV, SinnottC,

DahlyD, etal. Adverse

childhood experiences (ACEs)

and later-life depression:

perceived social support as a

potential protective

factor. BMJ Open

2017;7:e013228. doi:10.1136/

bmjopen-2016-013228

► Prepublication history and

additional material for this

paper are available online. To

view these les please visit the

journal online (http:// dx. doi.

org/ 10. 1136/ bmjopen- 2016-

013228).

Received 28 June 2016

Revised 26 March 2017

Accepted 29 March 2017

School of Medicine, University

College Cork, Cork, Ireland

Department of General Practice,

University College Cork, Cork,

Ireland

Department of Epidemiology

and Public Health, University

College Cork, Cork, Ireland

Correspondence to

DrE VonCheong;

111123401@ umail. ucc. ie

Research

Strengths and limitations of this study

► Awide range of demographic and health information

collected using validated, standardised instruments

and questionnaires.

► Large sample size (n=2047), study population is

representative of the source population reported in

national census data.

► Assessment of 10 types of adverse childhood

experiences(ACEs) under three ACE subtypes, that

is, abuse, neglect household dysfunction.

► Informs future interventions seeking to prevent or

manage mental ill-health among those with ACE

exposure.

► Risk of recall bias from retrospective self-report.

on August 3, 2024 by guest. Protected by copyright.http://bmjopen.bmj.com/BMJ Open: first published as 10.1136/bmjopen-2016-013228 on 1 September 2017. Downloaded from

CheongEV, etal. BMJ Open 2017;7:e013228. doi:10.1136/bmjopen-2016-013228

Open Access

one responds to stressful life events, hence, setting the

life-course trajectory for one’s mental health.

Adverse childhood experiences (ACEs) encompass

any acts of commission or omission by a parent or other

caregiver that result in harm, potential for harm or

threat of harm to a child in the first 18 years of life,

even if harm is not the intended result.

While the

association between ACEs and poor mental health has

been reported previously,

7–10

there remains a lack of

research on factors that may modify this relationship.

Identifying factors that alter the processing of stressful

events following exposure to ACEs may be a valuable

tool in developing interventions aimed at preventing or

mitigating the long-term mental health consequences

of ACEs.

Social support may have a protective or buffering

effect against the consequences of a stressful event

by enhancing cognitive and emotional processing of

the experience.

12 13

This facilitates reappraisal of the

stressful event in a manner that is psychologically adap-

tive.

12 13

Findings from meta-analyses have found a lack

of social support to be the single strongest predictor

of post-traumatic stress symptoms in both military

and civilian populations with a history of psycholog-

ical trauma.

12 14

The term social support encompasses

perceived and received support. It is suggested that

perceived support is best understood as an individual

difference variable, with evidence that those who report

that others will provide them with aid when they are

in need (perceived social support (PSS)) are protected

from the pathogenic effects of life stress.

Studies on PSS

have consistently shown it to be associated with reduced

stress and improved physical and mental health,

and

that perception of available social support was found to

be a better buffer of psychological distress than actual

availability of social support in some studies.

12 17

This

suggests that enhancing perception of available support

may be just as important, if not more, than increasing

actual social support in interventions aimed at moder-

ating psychological effects of stress. Although a number

of studies have generated findings supportive of the role

of PSS, most of these studies were focused on female

victims of childhood sexual abuse.

18 19

Therefore, PSS

is a potentially modifiable risk factor, with evidence to

show that social support interventions are associated

with improvements in measures of quality of life and

burden of illness.

This present study aims to build on prior research

by examining whether three ACE subtypes (abuse,

neglect and household dysfunction) are related to

later-life depressive symptoms, and if so, whether

these associations vary across levels of PSS. In line with

recent work that suggests that multiple ACEs have

an increasingly greater effect on mental health, this

study also aims to examine the association between

ACE scores and depressive symptoms and if PSS differ-

entially impacts depressive symptoms across an accu-

mulation of ACEs.

METHOD

Study design and population

Our analysis uses baseline data from the Mitchelstown

cohort,

a study of 50–69-year-old adults randomly

selected from patients attending the Livinghealth Clinic

in Mitchelstown, Ireland, in 2010–2011. The study popu-

lation is representative of the profile of the source popu-

lation reported in national census data.

A complete

description of the study was sent out to all selected partic-

ipants with a reply slip indicating acceptance or refusal.

After written, informed consent was obtained, the partic-

ipants completed a detailed health and lifestyle question-

naire and attended a physical examination conducted

by research nurses using standardised and validated

instruments. Participants were offered separate sealed

envelopes to submit their responses to the ACE ques-

tionnaire during data collection. Ethical approval for the

original study was granted by the Clinical Research Ethics

Committee of the Cork Teaching Hospitals.

Predictors

Adverse childhood experiences

Exposure to ACE was assessed using the ACE question-

naire

22 23

which addresses 10 individual ACEs under three

categories:

► abuse: emotional, physical and sexual abuse

► neglect: emotional and physical neglect

► household dysfunction: parental separation/divorce,

violence against mother, household substance abuse,

household mental illness and incarceration of house-

hold member.

The ACE questionnaire is a reliable and valid measure

of childhood adversity that has been used extensively in

large-scale ACE studies.

22 23

All questions about ACEs pertained to the respondents’

first 18 years of life and were binary (yes vs no). From

these, a dichotomous variable was created to reflect expo-

sure to any ACE, and similar variables were created to

reflect any exposure to each ACE subtype (abuse, neglect,

household dysfunction). We also calculated a total ACE

score for each participant (+1 for each of the 10 types of

ACE reported). ACE scores of ≥3 were combined into one

category due to small sample sizes in some strata, and the

score was treated as a categorical variable (0, 1, 2 or ≥3) to

capture any potential non-linearities in the relationship

with depressive symptoms. This method of ACE score

categorisation has been used previously.

24 25

Perceived social support

PSS was assessed using the self-administered Oslo Social

Support Scale with three questions

► Oslo 1: How many people are you so close to that you

can count on them if you have great personal prob-

lems? (none (1), 1–2 (2), 3–5 (3), 5+ (4))

► Oslo 2: How much interest and concern do people

show in what you do? (a lot (5), some (4), uncer-

tain (3), little (2), none (1))

on August 3, 2024 by guest. Protected by copyright.http://bmjopen.bmj.com/BMJ Open: first published as 10.1136/bmjopen-2016-013228 on 1 September 2017. Downloaded from

CheongEV, etal. BMJ Open 2017;7:e013228. doi:10.1136/bmjopen-2016-013228

Open Access

► Oslo 3: How easy is it to get practical help from neigh-

bours if you should need it? (very easy (5), easy (4),

possible (3), difficult (2), very difficult (1))

The response categories were assessed independently

for each of the three questions, and a sum score was

created by summarising the three scores. The Oslo

Social Support Scale has been used in several studies,

thus supporting its feasibility and predictive validity with

respect to psychological distress.

27 28

A sum score ranging

between 3 and 8 was categorised as poor, a score between 9

and 11 as moderate and a score of 12–14 as strong PSS.

This

categorisation was used previously in a study assessing PSS

in a population of older adults.

Outcomes

Depressive symptoms

The CES-D questionnaire

was used to assess for depres-

sive symptoms. The items of the scale are symptoms

associated with depression which have been used in

longer, previously validated instruments, and have been

tested in both household and clinical settings.

It has

very high internal consistency and adequate test–retest

repeatability.

The score is a sum of 20 questions. The possible range

of scores is 0–60, with higher scores indicating the pres-

ence of greater symptomatology. A score between 16 and

21 on the CES-D scale indicates the presence of mild to

moderate depressive symptoms while a score of ≥22 indi-

cates the possibility of major depression.

Participants

with a score of ≥16 were defined as having depressive

symptoms.

This cut-off point has been used extensively

in other studies in identifying individuals at risk of clinical

depression.

31 32

Covariates

Educational attainment was ascertained by the ques-

tion ‘What is the highest level of education you have

completed?’ and responses were categorised into primary,

secondary or tertiary level.

Current marital status was ascertained by the question

‘What is your current marital status?’ and the options were

single (never married), separated, cohabiting, divorced,

married or widowed.

Participants were asked whether they were covered

by the General Medical Services (GMS) scheme, which

entitles those covered to free medical care at the point of

access. Responses were categorised as GMS patient (yes/

no). GMS eligibility is based on low-income thresholds.

Smoking status was categorised as never smoked,

current smoker or former smoker in response to the

questions ‘Have you smoked at least 100 cigarettes in your

entire life?’ and ‘Are you a current smoker?’

Alcohol consumption was derived from the question

‘During the past 7 days how many standard drinks of any

alcoholic beverage did you have each day?’ and was cate-

gorised as non-drinker (<1 unit/week), moderate drinker

(1–14 units/week) and heavy drinker (>14 units/week).

Physical activity was measured as metabolic equivalents

(METs) minutes per week using the short-form Interna-

tional Physical Activity Questionnaire

and was catego-

rised into three groups (low, moderate or high) based on

MET minutes per week in all activity types.

Height and weight were measured using standardised

methods by study personnel and used to calculate body

mass index (BMI, kg/m

). Participants were classified as

underweight if their BMI was <18.5 kg/m

, normal if 18.5

to <25 kg/m

, overweight if 25 to <30 kg/m

and obese

if ≥30 kg/m

Statistical methods

Descriptive data and crude tests of association

Continuous variables were described by means and

SD. Categorical variables were described by counts and

percentages. Student’s t-test, one-way analysis of variance

or Pearson’s χ

test were used as appropriate to test for

differences in the distributions of demographic, health

and lifestyle measures between the groups with and

without ACE.

Logistic regression

Associations between each ACE exposure (any ACE, each

ACE subtype and ACE score; and in a set of supplemental

models, each of the 10 individual ACEs) and depressive

symptoms were estimated using logistic regression in two

models, A and B. Model A included the main effects of

the ACE exposure, PSS and a product interaction term

between the two. Model B included additional effects to

adjust for potential confounding and selection biases.

These were age, gender and educational attainment,

current marital status and GMS cover. Results were

reported as ORs with 95% CIs. Interaction terms were

tested using type II sums of squares likelihood ratio χ

test (LRT), and exact p values were reported.

Quantile regression

Our main analyses use categorisations of the CES-D and

the Oslo Social Support Scale to define depressive symp-

toms (yes/no) and PSS (low/medium/strong). While

these categorisations are not uncommon, there will be

some loss of power associated with their use. We thus used

quantile regression

in a supplemental analysis to model

the association between CES-D scores and ACE (any ACE,

any abuse, any neglect or any household dysfunction),

conditional on the selected covariates. Quantile regres-

sion is analogous to multiple linear regression, except

that it models a given centile of the outcome’s distribu-

tion, the median in our case, rather than the outcome’s

mean. It is thus robust to departures from the normally

distributed errors assumption of linear regression, which

is relevant given the skewed distribution of CES-D scores

in our sample.

Missing data

Missing data were handled using multiple imputation, so

that all participants who completed the baseline question-

naire were included in the analytical sample, even if they

on August 3, 2024 by guest. Protected by copyright.http://bmjopen.bmj.com/BMJ Open: first published as 10.1136/bmjopen-2016-013228 on 1 September 2017. Downloaded from

CheongEV, etal. BMJ Open 2017;7:e013228. doi:10.1136/bmjopen-2016-013228

Open Access

were missing values for one or more variables. For each

estimated statistical model, 30 imputed data sets were

created, after a burn-in of 30 replications using predicted

mean matching.

Each imputation model included all

variables used in a given statistical model, allowed for

non-linear relationships using restricted cubic splines

with three knots, and included the key interaction of

interest (ACE × PSS). The statistical model of interest

was then estimated in each imputed data set, and param-

eter estimates were combined using Rubin’s rules.

complete case sensitivity analysis was also performed for

comparison.

Inference

All parameter estimates are reported with 95% CIs and/

or exact p values. While we have estimated a fairly large

number of parameters, we have not selectively reported

any of these, nor made any other decisions based on statis-

tical significance testing. This is consistent with current

practice in major epidemiological journals, particularly

with observational study designs, and recent guidance

from the American Statistical Association.

All analyses were conducted using the R Project for

Statistical Computing (V.3.3.1).

Research reporting

The Strengthening the Reporting of Observational

Studies in Epidemiology guidelines were used to inform

the study report.

RESULTS

Baseline characteristics

Of the 3051 people invited to participate in this study,

2047 (67%) completed the baseline assessment. The

mean age at baseline was 55.8 years and 51% of the partic-

ipants were female.

Of these, 1926 (94%) completed the ACE question-

naire. 23.7% (n=457; 95% CI 21.9 to 25.6) of the respon-

dents reported at least one form of ACE. 16.1% (n=302,

95% CI 14.4 to 17.7) of the participants reported depres-

sive symptoms.

The characteristics of respondents with and without

self-reported ACEs are summarised in table 1. Prevalence

of participants with a CES-D score indicative of major

depression was significantly higher among participants

who reported ACE compared with participants who did

not (14.1% vs 6.0%, p≤0.001). Prevalence of poor PSS

was also higher among participants who reported ACEs

(19.6% vs 10.7%, p≤0.001). Participants who reported

ACEs tended to be younger, separated/divorced, have

GMS cover, reported long-term illness/disability and had

attained tertiary education.

Logistic regression

Table 2 gives ORs and 95% CIs for models A and B where

any ACE was the exposure variable. Exposure to any

ACE was associated with almost three times the odds for

depressive symptoms among participants reporting low

PSS (adjusted OR 2.85, 95% CI 1.64 to 4.95). The asso-

ciation between any ACE and depressive symptoms was

substantially attenuated with moderate PSS (OR 2.21,

95% CI 1.52 to 3.22) and strong PSS (OR 1.39, 95% CI

0.85 to 2.29) (figure 1). The LRT p value for the inter-

action term was 0.19 in the full adjusted model. The esti-

mates were similar in the unadjusted and adjusted models

(table 2), as were results from the complete case analysis

(online supplementary table 1).

Similarly, each ACE subtype was also associated with

increased odds of depressive symptoms among individ-

uals reporting low PSS, and this association was reduced

among individuals reporting moderate and high PSS

(figure 1 and online supplemental tables 2-4). The odds

of depressive symptoms among those reporting abuse

(vs not) were more than five times greater in individuals

reporting low PSS (OR 5.20, 95% CI 2.71 to 9.99), three

times greater in those reporting moderate PSS (OR

3.22, 95% CI 2.11 to 4.92), but only slightly increased

in those reporting high PSS (OR 1.29, 95% CI 0.68 to

2.45). Among those reporting neglect, those figures

were OR 3.31 (95% CI 1.76 to 6.20), OR 3.31 (95% CI

1.63 to 5.73) and 1.81 (95% CI 0.83 to 3.95), respec-

tively; and among those reporting household dysfunc-

tion, the estimates were OR 2.24 (95% CI 1.24 to 4.06),

OR 1.87 (95% CI 1.20 to 2.90) and OR 1.24 (95% CI

0.68 to 2.28).

The odds of depressive symptoms were progressively

higher among individuals who experienced a greater ACE

score (figure 2 and online supplementary table 5). These

associations were again strongest in those reporting poor

PSS, for whom an ACE score of 1 (vs 0) was associated

with 1.44 times the odds of depressive symptoms (95% CI

0.61 to 3.37), the OR for an ACE score of 2 was 3.08 (95%

CI 1.24 to 7.65) and the OR for an ACE score of 3+ was

5.33 (95% CI 2.56 to 11.10). Overall, evidence for effect

modification of ACE exposures by PSS was strongest for

abuse, with an LRT p value of 0.011 (online supplemen-

tary table 2), while it was 0.16 and 0.23 for neglect and

household dysfunction, respectively (online supplemen-

tary tables 3 and 4).

Regarding individual ACEs, those reflecting

abuse (online supplementary table 6) and neglect

(online supplementary table 7) tended to be more

strongly associated with depressive symptoms than those

reflecting the various forms of household dysfunction

(online supplementary table 8). Further, the effect modi-

fication by PSS was most clearly demonstrated for the esti-

mated effects of abuse.

The quantile regression model for continuously

measured CES-D scores gave qualitatively similar esti-

mated to the logistic regression models (online supple-

mentary table 9): any ACE, any abuse, any neglect and any

household dysfunction were each associated with higher

median CES-D scores, as was lower PSS score. Further, the

association between ACE and CES-D score was highest in

those with lower PSS scores.

on August 3, 2024 by guest. Protected by copyright.http://bmjopen.bmj.com/BMJ Open: first published as 10.1136/bmjopen-2016-013228 on 1 September 2017. Downloaded from

CheongEV, etal. BMJ Open 2017;7:e013228. doi:10.1136/bmjopen-2016-013228

Open Access

Table 1 Baseline characteristics of participants in the Mitchelstown cohort study, 2010–2011, stratied by self-reported

adverse childhood experience(ACE) (n=1926)

Characteristics

With ACE

(n=457; 23.7%)

Without ACE

(n=1469; 76.3%) Group difference

Missing

data

Estimate* Test statistic† df p Value N (%)

Age (years) 53.6 (16.9) 56.6 (15.2) 3.54 1924 ≤0.001 0

Sex

Male 234 (51.2) 707 (48.1) 1.32 1 0.25 0

Female 223 (48.8) 762 (51.9)

Marital status

Single (never married) 45 (10.0) 116 (8.0) 28.85 3 ≤0.001 18 (0.9)

Cohabiting/married 334 (74.2) 1175 (80.6)

Separated/divorced 52 (11.6) 73 (5.0)

Widowed 19 (4.2) 94 (6.4)

Perceived social support

Poor 86 (19.8) 148 (10.6) 27.10 2 ≤0.001 101 (5.2)

Moderate 193 (44.5) 632 (45.4)

Strong 155 (35.7) 611 (43.9)

Socioeconomic

Education

Primary 116 (26.9) 382 (27.7) 10.57 2 0.01 117 (6.1)

Secondary 192 (44.4) 701 (50.9)

Tertiary 124 (28.7) 294 (21.4)

General Medical Services

cover

Yes 158 (34.6) 432 (29.4) 4.31 1 0.04 0

No 299 (65.4) 1037 (70.6)

Personal health behaviours

Smoking

Never smoked 200 (45.0) 750 (52.9) 8.93 2 0.01 64 (3.3)

Former smoker 175 (39.4) 462 (32.6)

Current smoker 69 (15.5) 206 (14.5)

Alcohol

Non-drinker 56 (17.7) 202 (21.4) 3.07 2 0.22 667 (34.6)

Moderate drinker 208 (65.6) 610 (64.8)

Heavy drinker 53 (16.7) 130 (13.8)

Physical activity

Low 211 (49.8) 667 (47.7) 0.70 2 0.70 104 (5.4)

Moderate 120 (28.3) 423 (30.3)

High 93 (21.9) 308 (22.0)

Personal health history

Self-rated health status

Very good 123 (27.2) 435 (30.1) 12.06 4 0.02 30 (1.5)

Good 235 (52.0) 804 (55.7)

Fair 81 (17.9) 178 (12.3)

Poor 11 (2.4) 20 (1.4)

Very poor 2 (0.4) 7 (0.5)

Continued

on August 3, 2024 by guest. Protected by copyright.http://bmjopen.bmj.com/BMJ Open: first published as 10.1136/bmjopen-2016-013228 on 1 September 2017. Downloaded from

CheongEV, etal. BMJ Open 2017;7:e013228. doi:10.1136/bmjopen-2016-013228

Open Access

Characteristics

With ACE

(n=457; 23.7%)

Without ACE

(n=1469; 76.3%) Group difference

Missing

data

Estimate* Test statistic† df p Value N (%)

Long-term illness/

disability

Yes 74 (19.6) 144 (11.8) 14.66 1 ≤0.001 331 (17.2)

No 304 (80.4) 1073 (88.2)

Hypertension

Hypertensive 198 (43.3) 699 (52.4) 2.62 1 0.11 2 (0.1)

Non-hypertensive 259 (56.7) 768 (47.6)

Diabetes

Diabetic 39 (8.8) 121 (8.4) 0.05 1 0.90 44 (2.3)

Non-diabetic 406 (91.2) 1316 (19.6)

Body mass index

Normal 97 (21.2) 329 (22.5) 2.82 3 0.28 4 (0.2)

Overweight 196 (42.9) 669 (45.7)

Obese 164 (35.9) 467 (31.9)

Mental health

Depression

No symptoms 309 (71.2) 1153 (84.8)

44.34 2 ≤0.001 133 (6.9)

Mild to moderate 64 (14.7) 125 (9.2)

Possibility of major

depression

61 (14.1) 81 (6.0)

*Estimates are reported as mean (SD) for continuous variables; and N (%) for categorical variables.

†Differences in means were tested using Student’s t-test)

; dependence between categorical variables was tested using Pearson’s χ

test

Table 1 Continued

DISCUSSION

In this population-based study, exposure to ACEs and

reporting poor social support were both related to a

higher odds of depressive symptoms later in life, even after

controlling for demographic and socioeconomic factors.

While other studies have reported an association between

ACE and later-life depression,

40–42

few have explored the

role of PSS as a potential effect modifier. Importantly, we

found that the deleterious impact of ACEs was typically

limited to those individuals who also reported poor and

moderate PSS. However, the statistical evidence for this

interaction was only strong among those reporting abuse.

In this sample, 23.7% reported having experienced

at least one form of ACE. This is low compared with

international and national estimates. The prevalence of

self-reported ACE was ~66% in the ACE study,

a collab-

oration between the Centers for Disease Contol and

Prevention (CDC) and Kaiser Permanente, while in the

Irish Longitudinal Study of Ageing (TILDA), the prev-

alence was 33.6%.

The higher prevalence of ACE in

TILDA may be explained by the broader nature of the

ACE measure. In contrast to the 10-item ACE question-

naire used in the current study, a 4-item measure was

used in TILDA to capture socioeconomic disadvantage,

parental substance abuse, physical abuse and sexual

abuse.

Prevalence of depressive symptoms and poor PSS were

substantially higher among participants who reported ACE

compared with those who did not. Perception of social

support level did not significantly differ between men and

women in our study, contrary to literature suggesting that

women tend to perceive their social support as stronger

than men do.

Prevalence of separation/divorce, GMS

cover and long-term illness/disability was found to be

markedly higher among participants who reported ACEs.

These findings suggest that exposure to childhood adver-

sity could affect a multitude of factors across the lifespan.

Animal studies have shown that early stressors result in

long-term changes in oxytocin, a peptide that regulates

pair bonding and social attachment.

45 46

It is possible that

this may extend to individuals with ACEs, accounting for

their impaired ability in forming long-term social attach-

ments, but it is important to note that our study does not

speak directly to this speculation.

Exposure to all ACE subtypes among those who

perceived poor social support was significantly related

to higher odds for later-life depression. Though the esti-

mated impact on depressive symptoms was strongest for

abuse, the less severe forms of childhood adversity such

on August 3, 2024 by guest. Protected by copyright.http://bmjopen.bmj.com/BMJ Open: first published as 10.1136/bmjopen-2016-013228 on 1 September 2017. Downloaded from

CheongEV, etal. BMJ Open 2017;7:e013228. doi:10.1136/bmjopen-2016-013228

Open Access

Table 2 Logistic regression analyses for adverse childhood experience(ACE) and later-life depression (n=2047*)

Model A (unadjusted) Model B (fully adjusted)

ORs 95% CI ORs 95% CI

ACE (and poor perceived social support)

No 1 – 1 –

Yes 2.82 1.64 to 4.85 2.85 1.64 to 4.95

Perceived social support (and no ACE)

Poor 1 – 1 –

Moderate 0.51 0.34 to 0.78 0.52 0.34 to 0.80

Strong 0.34 0.223 to 0.53 0.35 0.22 to 0.54

ACE × perceived social support† p=0.31 p=0.19

No ACE × poor social support 1 – 1 –

ACE × moderate social support 0.76 0.40 to 1.48 0.77 0.40 to 1.51

ACE × strong social support 0.52 0.25 to 1.09 0.49 0.23 to 1.02

Covariates

Age (years) – 0.98 0.96 to 1.00

Gender –

Male 1 –

Female 1.36 1.05 to 1.75

Education –

Primary 1 –

Secondary 0.91 0.68 to 1.21

Tertiary 0.67 0.46 to 0.96

Current marital status –

Cohabiting/married 1 –

Separation/divorce 0.73 0. to 1.21

Single (never married) 1.24 0.71 to 2.20

Widowed 1.53 0.84 to 2.80

General Medical Services (status) –

Not covered 1 –

Covered 1.11 0.85 to 1.45

*Parameter estimates based on the complete sample with missing data accounted for with multiple imputation with predictive mean

matching.

†pvalues are from the type II sum of squares likelihood ratio χ

test for the interaction term.

as neglect and household dysfunction may also have long-

term effects on mental health. Further, the experience of

any number of childhood adversity, from having experi-

enced one type of ACE to three or more types of ACE, is

associated with increased odds for later-life depression, in

the presence of poor PSS.

Consistent with the stress buffering model,

the rela-

tionship between ACE (overall, subtype or ACE score)

and depression later in life was found to vary according

to level of PSS, though this interaction was clearly stron-

gest among those reporting abuse. These findings high-

light that the strengthening of social support among

childhood adversity survivors may benefit mental health.

Social support may enhance cognitive and emotional

processing of the experience, hence, facilitating

reappraisal of the stressful event in a manner that is

psychologically adaptive.

12 13

Findings from experience

sampling method studies highlight the role of posi-

tive emotional content in buffering negative reactivity

of stress

47 48

and in improving responsiveness towards

antidepressant medications.

These findings suggest

that in addition to preventing depression improving

social support may serve as an important adjunct in the

medical management of depression. The role of social

support may be especially important for older persons

as this is a phase of major social transitions such as

retirement and bereavement.

There is also evidence for the role of PSS among

those with chronic medical conditions. Higher levels

of PSS have been associated with longer survival

following heart attacks

and improved well-being,

that is, mental health, perceived burden of illness and

on August 3, 2024 by guest. Protected by copyright.http://bmjopen.bmj.com/BMJ Open: first published as 10.1136/bmjopen-2016-013228 on 1 September 2017. Downloaded from

CheongEV, etal. BMJ Open 2017;7:e013228. doi:10.1136/bmjopen-2016-013228

Open Access

Figure 1 ORs and 95% CIs from a logistic regression model of later-life depressive symptoms and any adverse childhood

experience(ACE) or ACE subtypes, illustrating the interaction with perceived social support (PSS)(n=2047).

quality of life among patients with end-stage renal

disease.

Strengths and limitations

Demographic information and personal histories in this

study were obtained using validated, standardised instru-

ments and questionnaires of health and well-being.

Although the sample is a relatively homogenous, Cauca-

sian population taken from a single, large primary care

centre, it is representative of the profile of the source

population reported in national census data.

ACE was measured by retrospective self-report of events

that happened ~30 years previously. There may be a risk

of recall bias due to the time lapse between the events

in question and the survey. Questions that concerned

less objective events such as whether the participant felt

unloved may also be subject to greater recall bias and

individual interpretation. Despite the risk of recall bias,

the ACE questionnaire has been previously shown to have

good test–retest reliability.

The sensitive nature of the

questions and the participant’s perception of the ‘social

taboos’ of responding to such questions may also be an

important limitation. This was acknowledged during data

collection by offering patients a separate sealed envelope

in which to submit their responses.

We have reported models for each of the 10 individual

ACEs. However, given the relatively small number of partic-

ipants experiencing any one specific ACE, the respective

parameter estimates will be volatile. While those results

qualitatively conformed with the models for any ACE and

ACE subtypes, a larger study would be needed to further

examine the impact of the individual ACEs.

There was a non-negligible amount of missing data,

which is not uncommon for such studies. We have used

multiple imputation, rather than case-wise deletion, to

both improve the efficiency of analyses (by retaining

more observations in the analysis) and to reduce chances

on August 3, 2024 by guest. Protected by copyright.http://bmjopen.bmj.com/BMJ Open: first published as 10.1136/bmjopen-2016-013228 on 1 September 2017. Downloaded from

CheongEV, etal. BMJ Open 2017;7:e013228. doi:10.1136/bmjopen-2016-013228

Open Access

Figure 2 ORs and 95% CIs from a logistic regression model of later-life depressive symptoms and adverse childhood

experience(ACE) score, illustrating the interaction with perceived social support(PSS) (n=2047).

of bias. Multiple imputation, assuming the model was

correctly specified, is unbiased given an assumption

that data were missing at random, conditional on other

variables accounted for in the model (ie, the Missing at

random (MAR) assumption). This is a more defensible

position than the assumption that data were missing

completely at random required for valid estimates using

case-wise deletion. However, we cannot rule out the possi-

bility that missing data, particularly for ACEs and the

CES-D, were missing not-at-random.

Lastly, and most importantly, this is an observational

study where both the exposure and outcomes will certainly

share causes. We have tried to adjust for this through the

careful selection and adjustment for confounders, but

these in turn will be measured with some error and will

certainly not represent an optimal set of covariates to

adjust for, so it is important that these results are viewed

as part of a larger and still developing body of research.

Implications of ndings and further research

There is evidence that shows the efficacy of social support

intervention in improving PSS and psychological distress

symptoms, specifically among women who had experi-

enced intimate partner violence.

This intervention was

led by trained nurses and founded on the four modal-

ities of social support, that is, belonging, evaluation,

self-esteem and tangible support.

Belonging was done

through listening and responding to others who had expe-

rienced intimate partner violence.

Evaluation involved

helping women see themselves as others see them.

Self-esteem was promoted by focusing on their strengths

and achievements in surviving domestic violence.

Tangible support involved discussions of resources in the

community for help they need such as financial assistance

and healthcare.

Results of our study have potential implications for

clinicians seeking to prevent mental illness among survi-

vors of childhood adversity. Interventions that aim to

protect mental health among survivors of childhood

adversity might benefit from strengthening social support

or perhaps even just perception of social support. There

is an increasing literature showing the positive effect of

interventions that increase perceptions of social support

in patients with terminal disease or end-stage kidney

disease.

By showing the buffering effect of PSS on the

ACE–mental health relationship, our findings highlight

the potential for interventions targeting PSS to reduce

the likelihood of depression in patients who experienced

childhood adversity. Individuals with a history of child-

hood adversity may experience deficits in support-seeking

behaviour and social attachments. Hence, interventions

may include social skills training where participants

are equipped with skills to identify, invite and main-

tain healthy social connections with others.

12 52

Further

research on the implementation and efficacy of such

interventions is indicated.

In addition, our findings show that the contribution

of ACEs to mental health impairment persists across the

life course. Despite such findings, ACE screening is not

routinely undertaken in clinical practice. Twenty-five

per cent of primary care physicians in the Massachusetts

Academy of Family Physicians reported that they never or

rarely screened for childhood trauma in adult patients.

Further work on how this can be implemented effectively

in primary care and medical training is urgently needed.

Previous presentations

on August 3, 2024 by guest. Protected by copyright.http://bmjopen.bmj.com/BMJ Open: first published as 10.1136/bmjopen-2016-013228 on 1 September 2017. Downloaded from

CheongEV, etal. BMJ Open 2017;7:e013228. doi:10.1136/bmjopen-2016-013228

Open Access

1. Medicine and the Humanities and Social Sciences

Conference, Sam Houston State University, Huntsville,

Texas, USA; 4–5 March 2015.

2. The Atlantic Medical Corridor Conference, University

College Cork, Western Road, Cork, Ireland;

10 November 2014.

3. The 18th International Conference on Public Health,

The World Academy of Science, Engineering and

Technology, London; 23–24 May 2016.

Acknowledgements The authors thank the participants, study nurses,

administrators and clinical staff at the Livinghealth Clinic, Mitchelstown, Cork,

Ireland.

Contributors PMK conceived of the study. PMK, DD and CS provided statistical

expertise in the study design. EVC conducted the secondary data analysis and

interpretation and drafted the manuscript. All authors contributed to critical revision

of the article and approved the nal manuscript.

Funding The Mitchelstown cohort study is supported by a research grant from the

Irish Health Research Board (reference HRC/2007/13).

Competing interests Health Research Board, Ireland (SSS'2014'781 to EVC). DD

was further supported by a HRB Interdisciplinary Capacity Award (ICE/2012/12).

CS was supported by the HRB National SpR Academic Fellowship Award

(NSAFP/2011/3).

Patient consent Obtained.

Ethics approval Clinical Research Ethics Committee of the Cork Teaching

Hospitals.

Provenance and peer review Not commissioned; externally peer reviewed.

Data sharing statement No additional data are available.

Open Access This is an Open Access article distributed in accordance with the

Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which

permits others to distribute, remix, adapt, build upon this work non-commercially,

and license their derivative works on different terms, provided the original work is

properly cited and the use is non-commercial. See: http:// creativecommons. org/

licenses/ by- nc/ 4. 0/

expressly granted.

REFERENCES

1. Brundtland GH. The implications for training of embracing; A life

course approach to health. Population 2050;2000:2025.

2. Monroe SM, Simons AD. Diathesis-stress theories in the context

of life stress research: implications for the depressive disorders.

Psychol Bull 1991;110:406–25.

3. Hammen C, Henry R, Daley SE. Depression and sensitization to

stressors among young women as a function of childhood adversity.

J Consult Clin Psychol 2000;68:782–7.

4. Heim C, Newport DJ, Heit S, et al. Pituitary-adrenal and autonomic

responses to stress in women after sexual and physical abuse in

childhood. JAMA 2000;284:592–7.

5. Thalida EM, Arpawong JL, Phillips DF, et al. Prescott. longitudinal

test of the stress sensitization hypothesis for depression in older age:

evaluating the combined impacts of Childhood Trauma and recent

stress. Society for Prevention Research 24th annual meeting, 2016.

Pacic D/L (Hyatt Regency San Francisco).

6. Gilbert R, Widom CS, Browne K, et al. Burden and consequences of

child maltreatment in high-income countries. Lancet 2009;373:68–81.

7. Lansford JE, Dodge KA, Pettit GS, et al. A 12-year prospective study

of the long-term effects of early child physical maltreatment on

psychological, behavioral, and academic problems in adolescence.

Arch Pediatr Adolesc Med 2002;156:824–30.

8. Herrenkohl TI, Herrenkohl RC. Examining the overlap and

prediction of multiple forms of Child Maltreatment, Stressors, and

socioeconomic status: a longitudinal analysis of Youth Outcomes. J

Fam Violence 2007;22:553–62.

9. Banyard VL, Williams LM, Siegel JA. The long-term mental health

consequences of child sexual abuse: an exploratory study of the

impact of multiple traumas in a sample of women. J Trauma Stress

2001;14:697–715.

10. Thornberry TP, Ireland TO, Smith CA. The importance of timing: the

varying impact of childhood and adolescent maltreatment on multiple

problem outcomes. Dev Psychopathol 2001;13:957–79.

11. Scott J, Varghese D, McGrath J. As the twig is bent, the tree inclines:

adult mental health consequences of childhood adversity. Arch Gen

Psychiatry 2010;67:111–2.

12. Evans SE, Steel AL, DiLillo D. Child maltreatment severity and adult

trauma symptoms: does perceived social support play a buffering

role? Child Abuse Negl 2013;37:934–43.

13. Thoits PA. Social support as coping assistance. J Consult Clin

Psychol 1986;54:416–23.

14. Brewin CR, Andrews B, Valentine JD. Meta-analysis of risk factors for

posttraumatic stress disorder in trauma-exposed adults. J Consult

Clin Psychol 2000;68:748–66.

15. Cohen S. Social relationships and health. Am Psychol

2004;59:676–84.

16. Haber MG, Cohen JL, Lucas T, et al. The relationship between self-

reported received and perceived social support: a meta-analytic

review. Am J Community Psychol 2007;39–133–44.

17. Cohen S, Wills TA. Stress, social support, and the buffering

hypothesis. Psychol Bull 1985;98:310–57.

18. Murthi M, Espelage DL, abuse C. Social support, and

psychological outcomes: a loss framework. Child abuse & neglect

2005;29:1215–31.

19. Hyman SM, Gold SN, Cott MA. Forms of social support that

moderate PTSD in childhood sexual abuse survivors. J Fam Violence

2003;18:295–300.

20. Cohen SD. editor Social support interventions will improve the quality

of life of ESRD patients. Seminars in dialysis: Wiley Online Library.,

2013.

21. Kearney PM, Harrington JM, Mc Carthy VJ, et al. Cohort prole: the

Cork and kerry diabetes and heart disease study. Int J Epidemiol

2013;42:1253–62.

22. Anda RF, Butchart A, Felitti VJ, et al. Building a framework for global

surveillance of the public health implications of adverse childhood

experiences. Am J Prev Med 2010;39:93–8.

23. Centers for Disease Control and Prevention NCfIPaC, Division

of Violence Prevention. Adverse Childhood experience study

Atlanta2006. 2014 http://www. cdc. gov/ violenceprevention/ acestudy/

(accessed 1 Jul 2015).

24. Dube SR, Felitti VJ, Dong M, et al. The impact of adverse childhood

experiences on health problems: evidence from four birth cohorts

dating back to 1900. Prev Med 2003;37:268–77.

25. Anda RF, Whiteld CL, Felitti VJ, et al. Adverse childhood

experiences, alcoholic parents, and later risk of alcoholism and

depression. Psychiatr Serv 2002;53:1001–9.

26. Brevik J, Dalgard O. The health prole inventory. Oslo: University of

Oslo, 1996.

27. Dalgard OS, Dowrick C, Lehtinen V, et al. ODIN Group. Negative

life events, social support and gender difference in depression: a

multinational community survey with data from the ODIN study. Soc

Psychiatry Psychiatr Epidemiol 2006;41:444–51.

28. Korkeila J, Lehtinen V, Bijl R, et al. Establishing a set of mental health

indicators for Europe. Scand J Public Health 2003;31:451–9.

29. Bøen H, Dalgard OS, Bjertness E. The importance of social support

in the associations between psychological distress and somatic

health problems and socio-economic factors among older adults

living at home: a cross sectional study. BMC Geriatr 2012;12:1–12.

30. Radloff LS. The CES-D Scale: a Self-Report Depression Scale

for Research in the General Population. Appl Psychol Meas

1977;1:385–401.

31. Lewinsohn PM, Seeley JR, Roberts RE, et al. Center for

Epidemiologic Studies Depression Scale (CES-D) as a screening

instrument for depression among community-residing older adults.

Psychol Aging 1997;12:277–87.

32. Shinar D, Gross CR, Price TR, et al. Screening for depression

in stroke patients: the reliability and validity of the Center for

Epidemiologic Studies Depression Scale. Stroke 1986;17:241–5.

33. Khaw KT, Wareham N, Bingham S, et al. Combined impact of health

behaviours and mortality in men and women: the EPIC-Norfolk

prospective population study. PLoS Med 2008;5:376–7.

34. Craig CL, Marshall AL, Sjöström M, et al. International physical

activity questionnaire: 12-country reliability and validity. Med Sci

Sports Exerc 2003;35:1381–95.

35. Hallock K, Koenker R. Quantile regression. Journal of Economic

Perspectives 2001;15:143–56.

36. Van Buuren S. Flexible imputation of missing data: CRC press;, 2012.

on August 3, 2024 by guest. Protected by copyright.http://bmjopen.bmj.com/BMJ Open: first published as 10.1136/bmjopen-2016-013228 on 1 September 2017. Downloaded from

CheongEV, etal. BMJ Open 2017;7:e013228. doi:10.1136/bmjopen-2016-013228

Open Access

37. Rubin DB. Multiple imputation for Nonresponse in surveys: Harvard

Univ, 1987.

38. Wasserstein RL, Lazar NA. The ASA's Statement on p -Values:

context, process, and purpose. Am Stat 2016;70:129–33.

39. Team RDC. Alanguage and environment for statistical computing. R

foundation for statistical computing, Vienna, Austria 2011.

40. Chapman DP, Whiteld CL, Felitti VJ, et al. Adverse childhood

experiences and the risk of depressive disorders in adulthood. J

Affect Disord 2004;82:217–25.

41. Anda RF, Whiteld CL, Felitti VJ, et al. Adverse childhood

experiences, alcoholic parents, and later risk of alcoholism and

depression. Psychiatr Serv 2002;53:1001–9.

42. Gilbert R, Widom CS, Browne K, et al. Burden and consequences

of child maltreatment in high-income countries. Lancet

2009;373:68–81.

43. McCrory C, Dooley C, Layte R, et al. The lasting legacy of childhood

adversity for disease risk in later life. Health Psychol 2015;34:687–96.

44. Cutrona CE. Social support in couples: marriage as a resource in

times of stress: Sage Publications;, 1996.

45. Insel TR, Winslow JT. Serotonin and neuropeptides in afliative

behaviors. Biol Psychiatry 1998;44:207–19.

46. Francis DD, Young LJ, Meaney MJ, et al. Naturally occurring

differences in maternal care are associated with the expression of

oxytocin and vasopressin (V1a) receptors: gender differences. J

Neuroendocrinol 2002;14:349–53.

47. Wichers MC, Myin-Germeys I, Jacobs N, et al. Evidence that

moment-to-moment variation in positive emotions buffer genetic risk

for depression: a momentary assessment twin study. Acta Psychiatr

Scand 2007;115:451–7.

48. Höhn P, Menne-Lothmann C, Peeters F, et al. Moment-to-moment

transfer of positive emotions in daily life predicts future course of

depression in both general population and patient samples. PLoS

One 2013;8:e75655.

49. Glasgow N, Pillemer K, Moen P, et al; Social integration in the second

half of life. Baltimore: Johns Hopkins University Press, 2000.

50. Lett HS, Blumenthal JA, Babyak MA, et al. Social support and

coronary heart disease: epidemiologic evidence and implications for

treatment. Psychosom Med 2005;67:869–78.

51. Constantino R, Kim Y, Crane PA. Effects of a social support

intervention on health outcomes in residents of a domestic violence

shelter: a pilot study. Issues Ment Health Nurs 2005;26:575–90.

52. Cohen S, Janicki-Deverts D. Can We improve our Physical Health by

Altering Our Social Networks? Perspect Psychol Sci 2009;4:375–8.

53. Weinreb L, Savageau JA, Candib LM, et al. Screening for childhood

trauma in adult primary care patients: a cross-sectional survey. Prim

Care Companion J Clin Psychiatry 2010;12:PCC.10m00950..

54. Witt PL, McGrain P. Comparing two sample means t tests. Phys Ther

1985;65:1730–3.

55. McHugh ML. The chi-square test of independence. Biochem Med

2013;23:143–9.

on August 3, 2024 by guest. Protected by copyright.http://bmjopen.bmj.com/BMJ Open: first published as 10.1136/bmjopen-2016-013228 on 1 September 2017. Downloaded from