Depressive Symptoms in Children and Adolescents with Chronic Physical Illness

Depressive Symptoms in Children and Adolescents with Chronic Physical Illness

Depressive Symptoms in Children and Adolescents with Chronic Physical Illness

Permalink: ‎

In the United States, the number of children and adoles-

cents with chronic health conditions has increased dramat-

ically in the past four decades (Perrin, Bloom, &

Gortmaker, 2007). Although results from epidemiological

studies differ considerably, an overview of articles found

that, on average, 15% of children and adolescents have a

chronic health condition (van der Lee, Mokkink,

Grootenhuis, Heymans, & Offringa, 2007).

Chronic illness is a risk factor for psychological prob-

lems, such as depressive symptoms (e.g., Bennett, 1994).

For example, the presence of physical symptoms, such as

pain and fatigue, combined with the need for disease man-

agement regimes, are likely to interfere with many aspects

of daily life, such as regular school attendance and main-

taining peer relations, and may cause frustration (e.g.,

Suris, Michaud, & Viner, 2004). Children with chronic

illness may feel different from his peers and experience

peer rejection, which may have detrimental effects on

their self-concept (e.g., Sandstrom & Schanberg, 2004).

In addition, chronic illnesses may foster inappropriate

parental attitudes and behaviors, ranging from overprotec-

tion to rejection, which may impair psychological

well-being (e.g., Holmbeck et al., 2002). In some cases,

poor prognosis may cause feelings of helplessness and

hopelessness. Finally, side effects of treatments may

cause psychological distress (e.g., Miller et al., 2008).

A meta-analysis by Bennett (1994) on 60 statistical

effects from 46 studies found that children and adolescents

with chronic medical problems have elevated levels of de-

pressive symptoms, but differences with test norms or

healthy control groups were small (mean d¼ .27 SD units). Because (a) the number of studies has increased

considerably since this meta-analysis, (b) the effects of

chronic illness on depressive symptoms may have changed

over time, and (c) the previous meta-analysis could not test

for moderating effects of many study characteristics, the

Journal of Pediatric Psychology 36(4) pp. 375–384, 2011 doi:10.1093/jpepsy/jsq104

Advance Access publication November 18, 2010 Journal of Pediatric Psychology vol. 36 no. 4 � The Author 2010. Published by Oxford University Press on behalf of the Society of Pediatric Psychology.

All rights reserved. For permissions, please e-mail: journals.permissions@oup.com

goal of the present study was to provide an updated

meta-analysis on the association between chronic physical

illness and depressive symptoms in children and

adolescents.

Depressive symptoms have to be distinguished from a

depressive disorder, such as major depression. Rating

scales assess depressive symptoms as a continuous vari-

able, and scores above defined cutoffs on valid scales

would imply a depressive disorder. However, depression

diagnosis is based on a clinical interview, and a number

of symptoms have to be present during a specified time

period (Emslie & Mayers, 1999). Because about 90% of the

available studies with chronically ill children used depres-

sion rating scales rather than clinical diagnoses, the present

meta-analysis focuses on depressive symptoms.

Research Questions

In the first research question we ask whether children and

adolescents with chronic physical illnesses have elevated

levels of depressive symptoms, and whether this would

differ between illnesses. Some authors have argued that

the nature of the child’s disorder is not important in de-

termining its psychological consequences, because chil-

dren with chronic physical disorders face common life

experiences and problems based on generic dimensions

of their conditions, rather than on idiosyncratic character-

istics of any specific disease entity (e.g., Stein & Jessop,

1982). Other authors have suggested that certain illness

characteristics or parameters may be more related to de-

pressive symptoms, such as neurologically related illnesses

(e.g., epilepsy; Plioplys, 2003), characteristics of illnesses

that have social implications (e.g., cosmetic effects, e.g.,

cleft lip; De Sousa, Devare, & Ghanshani, 2009), and

chronic pain (Eccleston, Crombez, Scotford, Clinch, &

Connell, 2004). Bennett (1994) reported moderate effect

sizes for asthma (d¼ .54) and sickle cell disease (d¼ .48), a small effect size for diabetes (d¼ .22), and no elevated levels of depressive symptoms in young people with cancer

(d¼ .00) and cystic fibrosis (d¼ –.04). However, due to the small number of studies per type of illness

(N¼ 4–13), effect sizes were not tested to see if they dif- fered significantly from zero or if illnesses differed from one

another.

In the second research question we analyze whether

the effect sizes would differ by other study characteristics.

Age

On the one hand, depressive symptoms are more common

in adolescence than in childhood, and adolescents may be

confronted with more illness-related stressors than chil-

dren (e.g., when chronic illnesses hinder the development

of peer groups and intimate relationships; Suris et al.,

2004). On the other hand, adolescents might also have

better coping abilities (e.g., because of higher cognitive

abilities; Skinner & Zimmer-Gembeck, 2007). Thus, aver-

age age differences in the association between chronic ill-

ness and depressive symptoms are probably small. Bennett

(1994) reported that age was generally unrelated to depres-

sive symptoms, but he did not provide results from a

statistical test of age differences.

Gender

On average, female adolescents are more likely than males

to react to stressors with depressive symptoms (Piccinelli

& Wilkinson, 2000), which could lead to stronger effects

of chronic illness on depressive symptoms. Bennett (1994)

reported that the results of individual studies were incon-

sistent but he did not formally test for gender differences.

Race/Ethnicity

Because it is less clear whether race/ethnicity would mod-

erate the size of between-group differences, we did not

state a hypothesis.

Country

Because young people from industrialized, developed

countries may have better access to health care than their

peers from developing countries, we expected finding

lower between-group differences in depressive symptoms

in developed countries.

Year of Publication

Progress in the treatment of many diseases (e.g., Bleyer,

2002) and the development of services for young people

with chronic illness may lead to lower between-group

differences in more recent studies.

Rater and Assessment Methods

Bennett (1994) found higher between-group differences in

parent-rated depressive symptoms (d¼ 0.58) than in self-rated depressive symptoms (d¼ 0.02), which may either indicate that young patients tend to underreport

their psychological symptoms or that parents underesti-

mate their children’s ability to adapt to their illness.

Differences between raters were also expected to lead to

higher levels of depressive symptoms in young people with

chronic illnesses in studies that used parent ratings as a

measure of depressive symptoms (e.g., the Affective

Problems scale of the Child Behavior Checklist (CBCL);

376 Pinquart and Shen

Achenbach, Dumenci, & Rescorla, 2003) than in studies

that used self-reports of the child.

Duration of Illness

A longer duration of the disease gives more time for

psychological adaptation, but may also lead to an accumu-

lation of negative illness-related consequences, such as the

effect of repeated school absence on grades. Thus, we did

not state a specific hypothesis.

Study Quality

Associations of chronic illness with depressive symptoms

may be stronger in clinical convenience samples than in

representative community-based samples, because clinical

samples may overrepresent highly distressed young people

seeking treatment for their chronic disease. Similarly, the

size of between-group differences in depressive symptoms

may vary between studies that used groups matched on

sociodemographic variables and studies that did not con-

trol for these between-group differences, because the lack

of control for demographic variables may cause unsyste-

matic bias rather than a general overestimation or under-

estimation of between-group differences in depressive

symptoms.

Target of Comparison

Finally, between-group differences may be larger in studies

that compared children with chronic illnesses to healthy

peers than in studies that compared depressive symptoms

of chronically ill children to test norms, because the norm

population probably includes some children with chronic

illnesses. In fact, Bennett (1994) found such a difference

(d¼ 0.67 vs. d¼ 0.02).

Methods Sample

Studies were identified from the literature through elec-

tronic databases [PSYCINFO, MEDLINE, Google Scholar,

PSNYDEX (an electronic data base of psychological litera-

ture from German-speaking countries)—search terms:

(chronic illness or disability or aids or arthritis or asthma

or cancer or cleft or chronic fatigue syndrome or cystic

fibrosis or diabetes or fibromyalgia or hemophilia or hear-

ing impairment or HIV or epilepsy or inflammatory bowel

disease or migraine or rheumatism or sickle cell or spina

bifida or visual impairment) and (children or adolescents

or adolescence) and (depression or depressive or mental

health or psychological health)], and cross-referencing.

Criteria for inclusion of studies in the present

meta-analysis were:

(a) the studies have been published before September,

2010,

(b) they compared the levels of depressive symptoms

or the frequency of depression diagnoses between

children and adolescents with chronic physical ill-

ness and their healthy peers or test norms, or they

provided sufficient information for a comparison

with established normative data (e.g., by reporting

standardized T-scores),

(c) mean age of participants �18 years, and (d) standardized between-group differences in depres-

sive symptoms were reported or could be

computed.

Documentation of physician diagnosis within each study

was not a requirement, because of the need to include

broad-based survey studies for which medical documenta-

tion might not be available. However, studies were excluded

if they focused on young people with chronic illnesses that

have been referred to psychological services due to depres-

sive symptoms, or if sufficient information for computing

effect sizes was not reported. In order to include studies

from different regions around the world, we also did not

limit the included studies to those written in English.

Available unpublished studies were also included.

Approximately 25% of the total number of studies

surveyed were eliminated, mainly because they did not

assess depressive symptoms or depression diagnosis

(14%), provided insufficient information about the effect

sizes (4%), did not exclusively focus on children and ado-

lescents with chronic physical illnesses (3%), had an aver-

age age of participants >18 years (1%), duplicated results

of previously published studies (1%), or were not available

via interlibrary loan (1%). After the exclusion of such stud-

ies, we were able to include 340 studies in the

meta-analysis that provided results for 450 subsamples.

The studies included are listed in the Appendix S1 (see

the Supplementary Data).

We entered the number of patients and control group

members, mean age, percentage of girls and of members of

ethnic minorities, the country of data collection, year of

publication, type of illness, duration of illness, the sam-

pling procedure (1¼ probability samples, 0¼ convenience samples), the use of a control group (0¼ yes, 1¼ compar- ison with test norms), equivalence of patients and control

group (1¼ yes, 2¼ not tested, 3¼ no), the rater of depres- sive symptoms (1¼ child, 2¼ parent, 3¼ teacher, 4¼ cli- nician), the measurement of the variables, and the

standardized size of between-group differences in

Depression and Chronic Illness 377

depressive symptoms. If between-group differences were

provided for several subgroups within the same publication

(e.g., for different illnesses), we entered them separately in

our analysis instead of entering the global association.

If data from more than one rater were collected, we entered

the effect sizes separately because we were interested in

whether the effect size would vary by the source of infor-

mation. However, in order to avoid a disproportional

weight of these studies, we adjusted the weights of the

individual effect sizes so that the sum of the weights of

the effect sizes was equal to the weight of the study if

only one effect size had been reported (Lipsey & Wilson,

2001). Based on one third of the coded studies, a mean

inter-rater reliability of 93% (range 86–100%) was

established.

Measures

Depressive symptoms were most often assessed with the

Child Depression Inventory (CDI; Kovacs, 1992; 203 sam-

ples), structured clinical interviews (46 samples), the Beck

Depression Inventory/Beck Youth Inventory (Beck, Beck,

& Jolly, 2001; 41 samples), the Behavior Assessment

System for Children (Reynolds & Kamphaus, 2004;

39 samples), the Affective Problems scale of the CBCL

(13 samples), and the depression scale of the Minnesota

Multiphasic Personality Inventory (MMPI) (Tellegen et al.,

2003; 10 samples).

Information from the World Bank (2010) was used for

coding countries as developed or developing/threshold

countries.

Statistical Integration of the Findings

Calculations for the meta-analysis were performed in six

steps, using random-effects models and the method of mo-

ments (for computations, see Lipsey & Wilson, 2001).

1. We computed effect sizes d for each study as the

difference in depressive symptoms between the

sample with chronic illness and the control

sample divided by the pooled SD. If the authors

provided only test scores for children and adoles-

cents with chronic illness, we used the norms

from the test manual for comparison. However,

because Twenge and Nolen-Hoeksema (2002) pro-

vided norms for the CDI based on a much larger

sample than the original manual (Kovacs, 1992),

we used the norms by Twenge and

Nolen-Hoeksema (2002). Outliers that were more

than two SD from the mean of the effect sizes

were recoded to the value at two SD (Lipsey &

Wilson, 2001).

2. Effect size estimates were adjusted for bias due to

overestimation of the population effect size in

small samples.

3. Weighted mean effect sizes and 95% confidence

intervals (95% CIs) were computed. The signifi-

cance of the mean was tested by dividing the

weighted mean effect size by the SE of the mean.

To interpret the practical significance of the re-

sults, we used the Binomial Effect Size Display

(BESD; Rosenthal & Rubin, 1982) and Cohen’s

criteria (Cohen, 1988). According to Cohen, differ-

ences of d� .8 are interpreted as large, of d¼ .50–.79 as medium, and of d¼ .20–.49 as small.

4. For testing whether the results may be influenced

by publication bias (a trend for nonsignificant re-

sults not being published), we used the ‘‘trim and

fill’’ algorithm (Duval & Tweedie, 2000), which

estimates an adjusted effect size in the presence of

publication bias.

5. Homogeneity of effect sizes was computed by use

of the Q statistic.

6. In order to test the influence of moderator vari-

ables, we used an analogue of analysis of variance

and weighted ordinary least squares regression

analyses.

Results

Data from 33,047 children and adolescents with chronic

illnesses were included. The largest subgroups had asthma

(N¼ 9,274), diabetes (N¼ 4,058), cancer (N¼ 3,400), mi- graine or tension-type head ache (N¼ 2,300), and epilepsy (N¼ 2,096). The participants had a mean age of 12.6 years (SD¼ 2.6 years); 50.2% of them were girls and 32.5% were members of ethnic minorities.

On average, children and adolescents with chronic

physical illnesses had higher levels of depressive symptoms

than their healthy peers—a small to very small effect

(Table I). According to the BESD, 54.8% of children with

chronic illnesses and 45.2% of their healthy peers would

show depressive symptoms above the median. The

trim-and-fill algorithm did not find any evidence for a

file-drawer problem, and the original effect size remained

unchanged after applying this procedure.

We computed separate effects in cases where at least

five studies were available for a particular chronic disease.

Separate effect sizes could be computed for 16 illnesses.

Strongest between-group differences were found for

chronic fatigue syndrome, fibromyalgia, migraine/tension

378 Pinquart and Shen