Professional Documents
Culture Documents
Ffects of Music Therapy For Children and Adolescents With Psychopathology - A Meta-Analysis - Adolescents With Psychopathology - A Meta-Analysis
Ffects of Music Therapy For Children and Adolescents With Psychopathology - A Meta-Analysis - Adolescents With Psychopathology - A Meta-Analysis
Sogn og Fjordane University College, Sandane, Norway; 2University of Vienna, Austria; 3Aalborg University,
Denmark
Background: The objectives of this review were to examine the overall efficacy of music therapy for
children and adolescents with psychopathology, and to examine how the size of the effect of music
therapy is influenced by the type of pathology, clients age, music therapy approach, and type of outcome. Method: Eleven studies were included for analysis, which resulted in a total of 188 subjects for
the meta-analysis. Effect sizes from these studies were combined, with weighting for sample size, and
their distribution was examined. Results: After exclusion of an extreme positive outlying value, the
analysis revealed that music therapy has a medium to large positive effect (ES .61) on clinically
relevant outcomes that was statistically highly significant (p < .001) and statistically homogeneous. No
evidence of a publication bias was identified. Effects tended to be greater for behavioural and developmental disorders than for emotional disorders; greater for eclectic, psychodynamic, and humanistic
approaches than for behavioural models; and greater for behavioural and developmental outcomes than
for social skills and self-concept. Conclusions: Implications for clinical practice and research are
discussed. Keywords: Behaviour problems, developmental delay, meta-analysis, music, psychotherapy, music therapy. Abbreviations: ES: effect size; AMT: analytical/analytically oriented music
therapy; GIM: guided imagery and music; CMT: creative music therapy.
ller &
oriented towards children and adolescents (Mu
Kehl, 1997). Music therapy is provided to children
and adolescents with psychopathology on a regular
basis in many countries (Aldridge, di Franco, Ruud, &
Wigram, 2001; Maranto, 1993).
While there are meta-analytic reviews on music
therapy in the treatment of dementia (Koger, Chapin,
& Brotons, 1999) and on music in general medical
treatment (Standley, 1986), a meta-analysis on the
effects of music therapy for children and adolescents
with psychopathology has not been performed.
Therefore this review was conducted to summarise
systematically previous findings on the efficacy of
music therapy in this field in the hopes of enabling
informed decisions on treatment provision. The goal
of this meta-analytic review was to examine the
overall efficacy of music therapy for children and
adolescents with psychopathology. Additionally, we
aimed to identify how the size of the effect of music
therapy is influenced by the type of pathology, clients
age, music therapy approach, and type of outcome.
Method
Literature search
All studies that focused on children and adolescents with
psychopathology, and compared music therapy with no
treatment or with a different treatment, or that compared
before and after music therapy treatment, were considered as possibly relevant. Eligible designs included
treatment versus control group pre-test post-test designs
and treatment group only pre-test post-test designs.
A systematic literature search, both computerised
and manual, was undertaken. Databases were
searched for the term music therap*, crossed with
child*, effect* or synonyms of these. Issues of relevant
journals were browsed by hand. Documents in any
language were considered, as were unpublished and
published articles. Searched databases included
Medline (19662000), Psyndex (19772000), PsycInfo
(18872000), the Cochrane Library (2001, issue 3),
Music Therapy Info CD-ROM 1 (1996), 2 (1999), and 3
(2001), a printed register of music therapy studies
(Jellison, 2000), various databases for ongoing and
unpublished studies, and programmes of music therapy conferences. The hand-searched journals included
Music Therapy (19811996), Journal of Music Therapy
(19642000), Music Therapy Perspectives (19822000),
British Journal of Music Therapy (19682000), Nordic
Journal of Music Therapy (19922000), Musiktherapeutische Umschau (19802000), and Revista
Internacional
Latinoamericana
de
Musicoterapia
(19952000). Reference lists of the included studies
were also checked to identify any additional studies.
Twenty-nine potentially relevant studies were identified
via these search strategies.
1055
1995
1991
1994
1989
1989
1989
1981
1992
1975
1970
(1973)
1998
Aldridge et al.
Clendenon-Wallen
Edgerton
Eidson
Gregoire et al.
Haines
Johnson
Laserer-Tschann
McQueen
USA
USA
UK
Austria
USA
USA
USA
USA
USA
USA
Germany
Country
no-treatment control;
non-randomised,
matched age, ability
and pre-test score
no-treatment control;
non-randomised,
matched age and
ability level
no control group
crossover design,
both groups received
music therapya
no control group
no control group
Design
learning and
behaviour
problemsb
emotional
disturbances,
learning
disabilities,
and/or attention
deficit disorder
1012 yrse
1114 yrse
26
16
disorders of
psychological
development
juvenile
delinquentsg
affective
problems in
gifted children
emotional
disturbance
mentally
handicapped
610 yrs
juvenile
1116 yrse
4th grade
emotionally
handicappedf
1116 yrse
eclectic
behavioural
psycho-dynamic
and humanistic
problem
behaviour
self-esteemc
cognitive abilityh
I. HamburgWechsler
Intelligence
Test
II. Marburg
Behaviour List
informal test
(identification
of pictures)
I. Coopersmith
Self-Esteem
Inventory
II. Coopersmith
Behaviour
Rating Form
Achenbach
Teacher
Report Form
I. intelligenceh
II. problem
behaviour
Orff
Interpersonal
Checklist
self-concept
behavioural
Piers-Harris
Childrens
Self-Concept Scale
Coopersmith
Self-Esteem
Inventory
Griffiths
Developmental
Scale
Adjective Checklist
Checklist of
Communicative
Responses/Acts
Score Sheet
behaviour
observation
Instrument
self-esteemc
self-concept
self-confidencec
communicative
responses and
acts in music
therapyd
social
behaviour in
classroomd
development
Type
Outcome
eclectic
GIM, Orff
behavioural
eclectic
CMT
sexual abuseb
autism
1419 yrs
69 yrs
children
24
14
26
16
11
25
11
11
Approach
CMT
46.5 yrs
Diagnosis
Therapy
developmental
delay
Age
valid n
Subjects
teacher;
blinding
unknown
independent;
blinding
unknown
I. self-report
II. teacher;
blinding
unknown
I. independent;
blinding
unknown
II. parent;
not blind
self-report
self-report
self-report
independent;
not blind
self-report
independent;
not blind
independent;
blind
Rater
Note: aClassified as a one-group pre-test post-test design. bClassified as mixed disorders. cClassified as self-concept. dClassified as social skills. eClassified as adolescents. fClassified as
emotional disorder. gClassified as behavioural disorder. hClassified as development.
Year
Author
Study
1056
Christian Gold, Martin Voracek, and Tony Wigram
Study
Eidson (1989)
25
-0.17
0.50
Laserer-Tschann (1992)
14
0.25
0.54
Haines (1989)
16
0.30
0.51
11*
0.30
0.43
26
0.32
0.40
Clendenon-Wallen (1991)
11*
0.53
0.43
8*
0.77
0.52
SE
Johnson (1981)
26
0.78
0.41
McQueen (1975)
24
0.86
0.49
16*
1.43
0.40
Edgerton (1994)
11*
4.56
0.81
188
0.99
0.13
177
0.61
0.14
1057
Results
Average effect of music therapy
The overall mean effect size of all included studies at
post-test was d .99 (SE .13). Following Cohens
(1988) benchmarks for interpretation and evaluation, this is a large effect, and the effect was significantly different from zero (t 7.47; df 243;
p < .001). However, it was not statistically homogeneous, as indicated by a significant Q test for
homogeneity (v2 88.02; df 10; p < .001). Inspection of the results (Figure 1) clearly showed that
the heterogeneity was caused by one extreme outlying value. The study was therefore excluded from the
analysis. The remaining studies provide a consistent
indicator of the effects of music therapy, as can be
seen from Figure 1.
The overall mean effect size after exclusion of
Edgertons study was d .61 (SE .14), which can
be interpreted as a medium to large effect. The biascorrected, conservative effect size index Hedges g
was .56 (SE .14). The results remained statistically
significant (t 4.37; df 221; p < .001), and they
were statistically homogeneous after the exclusion of
Edgertons study (v2 10.31; df 9; p .48). The
observed effect size is attenuated by imperfect
reliability of outcome measures (cf. Hunter &
Schmidt, 1990). No reliability correction was performed, thereby making our effect size estimate a
-5.0
-2.5
0.0
2.5
5.0
1058
Table 3 Effect sizes of music therapy with children and adolescents, grouped by study characteristics
Study
Variable
Music therapy
Haines
Eidson
Gregoire
Michel
Clendenon
Laserer
Johnson
McQueen
Aldridge
Montello
(Edgerton1)
SE
9
20
11
14
11
6
13
18
8
16
(11)
126
NA
NA
.93
NA
.81
NA
.42
.82
.93
NA
(.39)
.09
.12
.30
.36
.53
.58
.59
.71
.77
1.43
(4.56)
.54
.19
.13
.11
.15
.19
.23
.30
.14
.13
.14
(.33)
.05
Other treatment
Haines
Johnson
Laserer
No treatment
McQueen
Michel
Eidson
All control groups
7
13
8
28
NA
.6
NA
).21
).14
.33
).01
.29
.25
.27
.16
6
12
5
23
51
.8
NA
NA
).12
.04
.29
.03
.01
.26
.22
.34
.16
.11
Moderator variables
The absence of statistical heterogeneity suggested
that the average effect size mentioned above was
representative for all conditions. However, clinical
heterogeneity between the included studies could
not be denied, and therefore the influence of moderator variables was examined for exploratory purposes (Table 3). The interpretation of these
comparisons requires caution because test power
Diagnosis
emotional
developmental
behavioural
mixed
Age
children
adolescents
Approach
behavioural
psychodynamic/
humanistic
mixed
Outcome
social skills
self-concept
development
behaviour
n (studies)
n (subjects)
SE
3
3
1
3
52
46
26
53
.16
.65
.78
.82
.27
.29
.39
.22
4
6
57
120
.54
.64
.24
.17
3
4
77
57
.38
.54
.24
.24
43
.89
.24
1
5
3
2
25
90
46
30
).17
.46
.76
.96
.50
.19
.29
.30
Note: Q tests for homogeneity revealed no significant heterogeneity either between or within any of the above groups. One
study (Laserer-Tschann) had two outcome measures. The
associated effect sizes were used separately in the sub-group
analysis by outcome, while in all other analyses the average of
both measures was used.
Discussion
This study was the first comprehensive meta-analytic review on the effects of music therapy for children and adolescents with psychopathology. The
study showed that music therapy with these clients
has a highly significant, medium to large effect on
clinically relevant outcomes. This finding remained
when sources of positive bias were examined and
eliminated.2 Alternatively, sources of negative bias
were not eliminated, making the observed effect size
a conservative estimate of the true effect (cf. Hunter
& Schmidt, 1990).
Because of the comprehensive literature search, it
is likely that this meta-analysis contained all relevant studies that have been conducted in the field.
It is possible that there are unpublished studies of
which we were not aware, but it is doubtful that their
existence would significantly alter our findings as we
did not observe a publication bias. Therefore, our
sample of studies can be considered as a representative sample of all studies conducted.
A particularly large effect was found for children
who suffer from either developmental or behavioural
disorders. The reason for this might be that music
therapy, especially active music making, helps these
children to focus and sustain attention. The noninvasive, non-judgmental setting in music therapy
gives them opportunities to show capacities that may
be hidden in other situations. Music therapy brings
them into a situation where they are set up to succeed. Music making is a highly motivational factor
for many of these children, as is documented in
anecdotal evidence (e.g., Wigram & de Backer, 1999).
Our results suggest that eclectic approaches to
music therapy, where techniques from different
models or theories are mixed, are particularly effective. As an interpretation of this finding, it may be
important that therapists have a flexible attitude and
openness to what a child brings into the music
therapy situation. Individualised music therapy
treatment, where a therapist chooses from a variety
of music therapy techniques to match the individual
2
1059
Limitations
The strength of a research reviews findings is
necessarily limited by the quality of the included
primary studies. Some of the most rigorous inclusion
criteria, such as concealed random allocation to
experimental and control groups, blind ratings, and
standardised measures (Cooper & Hedges, 1994;
The Cochrane Collaboration, 2002) could not be
applied in this review because of the small number of
studies and the limited quality of some of them.
Several studies included in this review were not
randomised, and in others randomisation procedures were not adequately reported. In some studies,
diagnoses and music therapy procedures were described with less than the desired precision and detail. In other studies the data were inadequately
reported for the calculation of effect sizes. Most of
these problems are likely to cause random error rather than bias, i.e., they are unlikely to have a systematic influence on our findings. Lack of
randomisation, however, may have a tendency of
inflating effect sizes, although there is generally good
agreement between prospective non-randomised and
randomised studies (Ioannidis et al., 2001).
The small number of studies did not permit an
analysis of interactions between diagnoses, types of
music therapy interventions, and outcomes. It is
quite possible that the effect of music therapy in
general may depend on the specific combination of
diagnosis, type of music therapy, and type of outcome. For example, scales of self-esteem were found
to produce small effects, and these scales were often
used in children with emotional disorders. Therefore,
the small effect found for music therapy with emotional disorders might actually be an artefact of
study design. The findings concerning specific types
of music therapy and clinical subgroups should be
interpreted with caution and considered very preliminary. Studies that include multiple diagnoses,
interventions and outcomes would be needed to
examine these influences.
1060
The application of the findings of this meta-analysis is restricted to settings that are similar to those
in the primary studies. Most of the included studies
were conducted in the United States, and the
majority of studies examined the effects of an
experimental treatment rather than clinical practice.
Music therapy models used in Europe tend to be
more improvisational and less structured than those
used in America (cf. Wigram, 2002). Also, clinical
practice in psychotherapeutic interventions differs
from experimental therapy in various aspects. For
example, patient populations in clinics tend to be
more diverse, more severely disturbed, and more
often comorbid, therapists have higher caseloads
and less intense supervision, and the goals of therapy and techniques used may cover a broader range
than in experimental therapy (Weisz & Weiss, 1993).
Author note
Conclusion
The clinical implication of this meta-analysis is that
music therapy is an effective intervention for children and adolescents with psychopathology. Music
therapy produces a clinically relevant effect of a
considerable size and is therefore recommended for
clinical use.
Specifically, clients with behavioural or developmental disorders, or with multiple psychopathologies, may benefit from music therapy. Music
therapy for children and adolescents with psychopathology appears to be especially helpful when
techniques from different music therapy approaches
are combined.
Our review suggests the need for studies on the
efficacy of models of music therapy that are currently
practised in Europe, and the effectiveness of music
therapy in clinical settings. If the findings can be
replicated for varying models and settings, this will
strengthen their clinical applicability. Our findings
provide an empirical basis for planning and conducting larger-scale studies on music therapy with
children. Studies with larger sample sizes will not
only have the advantage of increased statistical
power, they will also help to answer the question of
how the different variables involved influence each
other. Further, it is important to use multiple outcome measures to reflect the various dimensions
that are involved in the disorders and addressed in
music therapy. Scales that address overt behaviour
and were successfully used in previous studies (such
as the Child Behavior Checklist) are very valuable
tools that can facilitate comparisons of research results across studies in a growing culture of music
therapy research. However, irrespective of the
measure, research results can be used most efficiently if design and reporting of music therapy
studies are in compliance with the general standards
for clinical trials (cf. Moher, Schulz, & Altman,
2001).
Acknowledgements
Christian Gold was supported by Aalborg University,
Denmark, Andreas Tobias Kind Foundation, Hamburg, Germany, and Viktor Frankl Foundation,
Vienna, Austria; Martin Voracek was supported by
the Austrian Steering Committee for Psychotherapy
Research. We thank Maryanne L. Fisher and Elisabeth Jandl-Jager for valuable comments on an earlier version of this paper.
Correspondence to
Christian Gold, Faculty of Health Studies, Sogn og
Fjordane University College, 6823 Sandane, Norway;
E-mail: christian.gold@hisf.no
References
*Aldridge, D., Gustorff, D., & Neugebauer, L. (1995). A
preliminary study of creative music therapy in the
treatment of children with developmental delay. The
Arts in Psychotherapy, 22, 189205.
Aldridge, D., di Franco, G., Ruud, E., & Wigram, T. (Eds.).
(2001). Music therapy in Europe. Rome: ISMEZ/Onlus.
Alvin, J. (1975). Music therapy. New York: Basic Books.
Becker, B.J. (1988). Synthesizing standardized meanchange measures. British Journal of Mathematical
and Statistical Psychology, 41, 257278.
References marked with an asterisk indicate studies included
in the meta-analysis.
1061
1062
Appendix
The variance of an effect size estimate d is given by
vi
1
1
d2
n1 n2 2n1 n2
vr
21 r
n
1063
2 14d 2 21 0
1 1
d2
vr
n n 2n n
n
n
Therefore, vi can be applied for both design
types, and the relation between sample size and
weight is retained. As the most reasonable choice
from both the clinical and the statistical perspective, vi was used in all analyses where effect
sizes from both design types were combined (all
analyses except those in Table 2). In the analyses
in Table 2, vr was used because all effect sizes
were from related samples. The sample size
displayed is always the number of subjects (i.e.,
n1 + n2 in controlled studies, n in one-group
studies).
vi