Background: Comorbidity is common among child clinical samples. Reviews on effective intervention for comorbid problems are lacking.
Method: Based on a literature search of three databases (PsycINFO, MEDLINE and ERIC), initial data analysis was carried out on 865 studies; of these,10 randomised trials fully met study inclusion criteria and were subject to final analysis, with quality assessments and effect sizes calculated.
Results: Overall, effect sizes for externalising (M = 1.12) and internalising (M = 1.09) outcomes were large. Effect sizes were large for family-based (M = 1.80) compared to individual (M = 0.78) and group-based (M = 0.54) interventions. Studies with homotypic comorbidity (M= 1.18) displayed larger treatment effect sizes than ones with heterotypic comorbidity (M = 0.54).
Conclusions: While the overall quality ratings of the reviewed studies varied from mediocre to good, with a variety of measures used across studies to assess the same outcomes, findings suggest that current interventions are effective for reducing internalising and externalising problems in children with comorbidity. More substantive evidence for the beneficial effects of psychosocial interventions for children with comorbid problems may arise as more robust studies, which more explicitly address and describe comorbidity, become available.
While there exists a consensus that comorbidity is common among clinical samples, a strict definition of the term has not yet been established. Originating in the field of psychiatry, the term comorbidity was coined by Feinstein (1970), who defined comorbidity as the impact of a co-existing disease (or clinical entity) on the treatment outcomes of patients with an already existing condition. In comparison, epidemiologists define comorbidity as the numerical risk of an individual with a condition/disorder to develop or acquire another one. Not only do definitions of comorbidity differ across fields of study, there also exist definitions of comorbidity based on the temporal onset of two or more conditions. These impairments are often classified as homotypic or heterotypic comorbidity (Boylan et al., 2007). Homotypic comorbidity occurs when disorders belong to the same diagnostic grouping (i.e. internalising or externalising) and co-occurring disorders not belonging to the same diagnostic grouping represent heterotypic comorbidity (e.g. internalising and externalising). These additional clinical impairments may impact on psychosocial functioning and responsiveness to treatment. In this study, comorbidity was defined as having a) one mental health disorder recognised by the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV; American Psychiatric Association, 1994) and b) at least one additional clinical impairment that may or may not meet criteria for a formal diagnosis (e.g. sub-clinical depression). Of particular interest for the first criterion was DSM-IV diagnoses most commonly encountered in childhood, namely mood, anxiety, conduct, and attention-deficit/hyperactivity disorder (ADHD), and likely to have high comorbidity rates (Armstrong & Costello, 2002; Wagner, 2003). The second criterion encompasses a broader spectrum of clinical impairment to include sub-clinical levels of psychopathology. This definition was utilised in this study as it reflects the experiences of many children (e.g. Briggs-Gowan et al., 2001).
At present, there exists empirical support for interventions targeting a variety of clinical problems of childhood and adolescence across individual (see Ollendick et al., 2008), group (Chu et al., 2009), family (Kazdin & Wassell, 2000; Sprang, 2009), school (Walter, 2010), and community-based interventions (see Jordans et al., 2009; Painter, 2009). As there is a substantial body of evidence for the utility of these modalities in treating single behavioural and emotional problems, an important next step is to understand the influence of treatment modes on clients’ co-existing difficulties. Since comorbidity is common in clinical practice, an appraisal of the potentially unique influences of treatment on comorbid conditions is critical.
Ollendick and colleagues (2008) reviewed comorbidity as a predictor and moderator of treatment outcome in which children were being treated for anxiety disorders, affective disorders, ADHD, or oppositional defiant disorder (ODD)/conduct disorder (CD). Overall, the results were unexpected in that the effects of comorbidity on treatment outcomes were not as negative as originally hypothesized. Rather than identifying consistently negative associations between comorbidity and treatment outcome, both neutral and positive associations were reported. However, differences in the ways in which researchers operationalised comorbidity and treatment outcomes may have influenced the findings (Ollendick et al., 2008). This qualitative review addressed the extent to which comorbidity predicted and moderated treatment outcome, yet it is unclear how individuals with comorbid problems are influenced by intervention efforts.
Few systematic reviews have focused on interventions for children with co-occurring disorders. In one review, the natural histories and treatment outcomes of those with anxiety disorders, depressive disorders, and mixed anxiety-depressive diagnoses were compared (Emmanuel, Simmonds, & Tyrer, 1998). Results suggested that those with anxiety and depressive disorders alone had better clinical outcomes than those with comorbid anxiety and depression. As the majority of studies in this review examined the natural course of a diagnosis (i.e. no intervention) and focused on adults, the findings may not be applicable to children and youth in treatment.
Three additional systematic reviews were examined, two of which targeted children and youth with ADHD (Danckaerts et al., 2010; Deault, 2010) and one focused on depression (Watanabe et al., 2007). Danckaerts et al. (2010) concluded that children with ADHD and co-occurring diagnoses had a lower quality of life than those without comorbid diagnoses. Similarly, Deault (2010) concluded that children with ADHD and comorbidity were more likely to have negative family characteristics than those who had ADHD only. Watanabe et al. (2007) examined the effectiveness of psychotherapy compared to no treatment, waiting list controls, and treatment as usual for children with depression. While several studies with comorbid samples were included, the impact of treatment on children with comorbid conditions was not examined.
Taken together, these findings suggest that not only is it rare for comorbid samples to be explicitly included in reviews of psychosocial interventions, but when they are, the effectiveness of interventions for comorbidity may not be assessed. Other than systematic reviews for the treatment of co-occurring substance abuse/dependence and severe mental health problems (Bender, Springer, & Kim, 2006; Cleary et al., 2009; Hesse, 2009), at present there does not appear to be a systematic examination of treatment effectiveness for other forms of comorbidity.
The purpose of the current systematic review was to examine the effectiveness of psychosocial (non-pharmacological) interventions to treat comorbid problems of childhood and adolescence. Since difficulties associated with mood, anxiety, conduct, and attention/hyperactivity appear to be particularly salient in childhood and adolescence, interventions designed to target these problems were the focus of this review.
To capture all recent interventions including a comorbid sample, peer-reviewed articles published in English between 1994 and August 2009 were gathered through a search of three relevant databases: PsycINFO, MEDLINE, and ERIC. In 1994 the DSM-IV was published and included major revisions since DSM-III-R was published in 1987 (American Psychiatric Association, 2010), such as changes in the classification of disorders of childhood and adolescence included in this study (e.g. ADHD). Keyword searches were conducted with the terms: comorbid*, co-occur*, ‘dual diagnosis’, adoles*, youth, child, treat*, interven*, and therap*, and RCT, random*, and ‘clinical trial’. In addition, reference lists of the final articles were hand searched for relevant studies.
Inclusion criteria were: (a) studies including any type of psychosocial intervention (trials with pharmacological interventions were included if there was a psychosocial intervention group); (b) randomized clinical or control trials; (c) samples of individuals with one DSM-IV diagnosis and at least one additional clinical impairment (no restrictions on the secondary impairment); (d) target population included individuals between the ages of 3- and 18-years; (e) externalising and/or internalising behaviours and/or psychosocial functioning measures (self-, parent-, or clinician-reported); (f) pre- and post-assessment points, for at least one of the comorbid problems.
There were important reasons for the definition of comorbidity used in this study. First, locating studies with samples of children and youth with DSM-IV diagnoses was important for the researchers to identify psychosocial intervention studies targeting children and youth with at least one confirmed diagnosis. Second, flexibility was maintained with the additional clinical impairment (i.e. did not need to be formally diagnosed problems) since it reflects the experiences of many children. In addition, it was expected that comorbidity may not have been measured or described within the research studies. This was particularly important since many of the studies did not include a focus on outcomes of concurrent problems among their samples.
Internalising and externalising behaviours were chosen as outcome measures of symptom reduction. This clustering of psychological problems of childhood and adolescence across these two broadband dimensions has received support in the literature (Achenbach, 1998). The characterisation of child psychopathology across these dimensions has been validated across a variety of measures (Achenbach, 1998; Edelbrock, 1979). As a result of the anticipated variability in measures of symptom reduction across studies, internalising and externalising behavioural outcomes was an appropriate means of including and comparing as many relevant studies possible across several problems (i.e. anxiety, attention, conduct, and depressive disorders).
To assess the quality of included trials, two authors (BM and PBR) independently coded articles using the Cochrane Collaboration’s tool for assessing risk of bias (Higgins & Altman, 2008). Six specific domains were assessed: sequence generation, allocation concealment, blinding, incomplete outcome data, selective outcome reporting, and ‘other issues’ (e.g. differential group attrition). Each domain was rated with a score of 0 (high risk), 1 (unclear), or 2 (low risk). Total scores could range from 0 (high risk of bias) to 12 (low risk of bias). Intercoder reliability ranged from 70% to 100% (M = 83%). All disagreements were discussed and consensus was achieved on discrepant items.
Effect sizes (ES) were calculated using the d-index (Cohen, 1988). The d-index can be calculated from various statistical results, such as mean differences, and is therefore intuitively appealing. Cohen’s interpretation of the size of the effect is: small (d = 0.20), medium (d = 0.50), large (d = 0.80), no effect (d = 0.00), and detrimental (d < 0.00). Effect sizes are important to understand the magnitude and direction of differences between means. As it is rare to locate studies in childhood intervention research in which pure, no intervention control groups are included in the study design, between group effect sizes could not be calculated. As a result, all cases were examined from baseline to post-intervention (and follow-up, where applicable). For each study, ES were calculated for each of the relevant outcomes using the following pretest-posttest formula:
where Mpost, Tx represents the mean post-intervention scores, Mpre, Tx represents the mean baseline scores, and SDpre, Tx represents the standard deviation of the mean baseline scores (Becker, 1988). Where follow-up intervals existed, ES were calculated using baseline and follow-up data.
Excluding duplicates, 865 studies were initially identified as relevant. Through review of titles, abstracts, and full articles, 10 studies were included for the final analyses (Figure 1). Reference lists of the 10 articles also were searched; no additional studies were located. Two raters (BM and PBR) independently coded each article for inclusion/exclusion at all stages of the selection process. Coders achieved consensus on all discrepant items. Agreement ranged from 83% to 92% (M = 88%) across stages.
Study characteristics and outcomes, intervention details, quality ratings, and ES appear in Table 1. Across all studies the average ES was 1.12 for externalising behaviour and 1.09 for internalising behaviour outcomes. In other words, the treatment effects on externalising and internalising behaviours appeared to be large following intervention in general. Among studies reporting significant psychosocial functioning outcomes the average ES was 0.40. Changes in psychosocial functioning from pre- to post-intervention were small, yet positive. Given that ES were large for internalising and externalising behaviours, it is surprising that ES for psychosocial functioning was small. Perhaps the large ES of internalising and externalising behaviours were a function of these behaviours/symptoms being directly targeted by the interventions. This may compare to psychosocial functioning, which may be considered a broader, global measure of functioning across a wider range of contexts, but not directly targeted among these interventions.
Table 1. Interventions to treat children with comorbidity
1The first values reported represent effect size at post-intervention; subsequent values represent later time points
nr = Statistical significance not reported;.n.s = Non significant; * = Statistically significant
aComparison was a Waitlist group (Pre: n = 15; Post: n = 12)
bComparison group was administered medication (Pre: n = 38; Post: n = 31)
cKendall (2001) conducted subgroup analyses comparing participants with a primary anxiety disorder + a comorbid anxiety disorder (i.e., ANX/ANX) to participants with a primary anxiety disorder + a comorbid externalising disorder (i.e., ANX/EXT)
I –Interpersonal psychotherapy for depressed adolescents
Baseline (n = 22) Post-Tx (4 mo.; n = 22)
There were several studies with high quality ratings and two studies appeared to have concerns regarding the potential for bias. The average quality rating was 7.4. There were several weaknesses of these included studies. The most salient limitation was the small sample sizes. While Kendall, Brady and Verduin (2001) had 136 participants at baseline, the remaining studies had comparably fewer participants, with between 15 to 45 participants in active treatment conditions. Furthermore, at least two of the studies did not include formal measures of treatment fidelity, and only one study reported a sample size calculation. Six studies reported analyses accounting for attrition. Overall, many of these trials were not well-designed, with quality ratings ranging from 5 to 10. Most of the studies were clinical trials. While none of the studies employed a no-treatment control group, the results suggest that attempts were made by the authors of the primary studies to limit investigator bias.
Nine trials were conducted in community clinics and one trial was conducted in a school setting. Five studies included treatments delivered in an individual format. Three of these studies were CBT-based interventions (Azrin et al., 2001; Bernstein et al., 2000; Kendall et al., 2001). Azrin et al. (2001) included an Individual Cognitive Problem-Solving Therapy aimed at promoting appropriate problem-solving skills and self-control in adolescents with conduct problems. The intervention consisted of 15 sessions, each approximately 60 minutes in length. Similarly, Bernstein et al.’s (2000) individual 8-session CBT for school-refusing children and youth included homework assignments, a psycho-educational component, strategies to identify and replace negative self-statements, and behavioural contracting. Kendall and colleagues’ (2001) 16 to 20 sessions of CBT were designed to teach children with anxiety to recognise anxiety-related thoughts and behaviours through a variety of therapeutic exercises. The remaining two individual-focused studies included interpersonal psychotherapy for youth with depression (Young, Mufson, & Davies, 2006) and a coping skills intervention called Safety Seeking (SS, Najavits, Gallop, & Weiss, 2006). Young et al.’s (2006) 12-session interpersonal therapy for adolescent females was designed to address how the clients’ interpersonal problems may be contributing to depression. Najavits et al.’s (2006) SS consisted of 25 sessions in which youth were taught coping skills in cognitive, behavioural, and interpersonal areas of their lives. Overall the average ES for individual-based treatments was 0.85, suggesting a large treatment effect.
A total of five interventions were family-focused, where caregivers were included in treatment along with the target clients. Variations of family interventions among the studies included family behaviour therapy, family CBT, and Parent-Child Interaction Therapy. While the behaviour and cognitive-behavioural therapies were conceptually similar to the individual CBT-based interventions with the addition of a family component, the focus of Parent-Child Interaction Therapy was the development of a secure parent-child relationship. Azrin et al.’s (2001) family behaviour therapy for adolescents with externalising problems was administered across 15 sessions and Bor and colleagues’ (Bor, Sanders, & Markie-Dadds, 2002) family behaviour therapies (Standard and Enhanced Behavioural Family Interventions) aimed to support children and youth with ODD/CD and their families over 10 sessions. The Standard Behavioural Family Intervention included a parent management component, whereas Enhanced Behavioural Family Intervention included parent management as well as coping skills and partner support training. In general, ES were calculated for depression, anxiety, and externalising behaviours. Values ranged from 0.73 to 3.28 (M = 1.94) suggesting medium to large treatment effects.
Four interventions within two of the included studies were presented in group format. Levy, Hunt and Heriot (2007) examined 9 sessions over 11 weeks of either group CBT for aggression and anxiety or group CBT for anxiety only for a sample of children with anxiety and comorbid aggression. Effect sizes were small to medium and ranged from 0.18 to 0.52. In the CBT for anxiety and aggression group, ES ranged from 0.39 to 0.42 at post-intervention, indicating small treatment effects. Rhode et al. (2004) included 16 sessions of group CBT or a life skills/tutoring intervention for adolescents with major depressive disorders and comorbid CD. In the life skills/tutoring intervention, adolescents were taught life skills (e.g. filling out job applications) and were provided with academic tutoring. Effect sizes ranged from 0.10 to 1.06 (M = 0.44) and 0.48 to 1.58 (M = 0.76) at post-intervention for the CBT and life skills groups, respectively. On average there were medium treatment effects for these interventions. See Table 2 for treatment ES.
The sustainability of treatment effects has been a longstanding issue in determining the effectiveness of psychosocial interventions (Achenbach, 1982). Among included trials, 6 out of 10 included post-intervention follow-ups at 3-, 6-, and 12-months post-intervention. Effect sizes ranged from −0.53 to 0.33 (3-month), −0.15 to 0.48 (6-month), and −0.21 to 1.08 (12-month) at the post-intervention follow-ups on measures of internalising and externalising behaviours. Rhode et al. (2004) had two follow-ups: the first at 6-months post intervention and the second was 12 months later. At this second follow-up, ES ranged from −0.14 to 0.78 on measures of depression, externalising behaviour, social adjustment, and psychosocial functioning.
Homotypic and heterotypic comorbidity
Three studies included samples with homotypic comorbidity, and two studies included samples with heterotypic comorbidity. Participants in Bor et al.’s (2002) study presented with comorbid externalising problems whereas those in Young et al. (2006) and Bernstein et al. (2000) displayed comorbid internalising problems. On average, ES at post-intervention were small to large (Range = 0.38–2.06; M = 1.18). Among studies in which samples presented with heterotypic comorbidity (Levy et al., 2007; Rhode et al., 2004), both referred to interventions targeting internalising problems. In general, ES were small to medium across both studies at post-intervention and follow-up. The average treatment ES was 0.57. Of the remaining five studies, two included participants with comorbid pervasive developmental disorders (PDD), two studies had samples with comorbid substance abuse, and the final study included a sample of participants with both homotypic and heterotypic comorbid presentations.
The findings from this review suggest that psychosocial interventions are effective for comorbid conditions; average ES were large for internalising and externalising behaviours. In addition, overall quality ratings were in the upper middle range of risk of bias (i.e. 7.4 on a scale of 0 to 12). Results appeared positive, although caution should be taken in interpreting these ES since there were several inconsistencies across the findings, making cross-study comparisons difficult. An important limitation to note were the small sample sizes of reported studies. Furthermore, researchers utilised a variety of outcome measurements making it difficult to compare treatment effects. Where possible, inconsistencies among the examined studies were identified.
An overall examination of the ES of individual-based interventions revealed positive results from baseline to post-intervention on outcomes of symptom reduction and psychosocial functioning, regardless of the theoretical approach used to guide the treatment. The majority of treatment effects were in the medium to large range for relevant internalising and externalising behaviours and psychosocial functioning measures of children and youth in individual-focused treatments.
Similarly positive results were evident for family-based interventions. All family-based intervention effects were large for target problems and medium for comorbid problems. This finding suggests that family interventions appear to have strong treatment effects on primary outcomes and, in addition, positively affect comorbid problems. Likewise, group-based interventions displayed medium treatment effects at post-intervention for children and youth with comorbidity.
In terms of treatment effect sustainability, there appears to be inconsistencies with regard to the impact of treatment on participants’ behaviours and functioning across studies. Some ES were negative, suggesting detrimental effects. Other ES were medium to large, suggesting positive treatment effects. Perhaps different problems of childhood and adolescence vary in their developmental trajectories and were thus reflected in the mixed long-term treatment effects. As Kazdin (2002) noted, for some children with CD, symptoms are likely to escalate without intervention and therefore it is possible that ‘no change’ in symptoms following treatment may in fact be evidence of stability and be a clinically significant result. Thus these inconsistent findings over time may have reflected differences in the course of internalising and externalising problems.
An examination of studies in which samples had homotypic comorbidity, post-intervention treatment effects were larger than those with heterotypic comorbidity. Overlapping symptoms are less common in heterotypic compared to homotypic comorbidity (see Kopp & Beauchaine, 2007). When intervention effects were examined across diagnoses, improvements in internalising and externalising outcomes following treatment were evident, including intervention effects for children and youth with PDD (Bagner & Eyberg, 2007; Chalfant, Rapee, & Carroll, 2007).
Overall the findings from this systematic review revealed that children and adolescents with comorbidity tend to improve on measures of internalising and externalising behaviours following psychosocial intervention. Often, positive treatment effects were evident not only for specified intervention targets; but also for comorbid problems. This is consistent with previous research suggesting that clients generalise learned coping skills to problems other than the target ones (Kendall et al., 1997).
Some important limitations of this review are worth highlighting. First, the results of this study were based on a relatively small number of studies and so caution should be taken in interpreting the findings. As researchers of future empirical studies continue to be explicit about addressing and describing comorbidity among their samples, more substantive claims may be made. In addition, this search was limited to studies that reported internalising and/or externalising behaviours or psychosocial functioning outcomes, and only studies in English were included. As a result studies with other relevant reported outcomes would have been missed. Not only therefore was there a small number of studies included in this review, but a limitation of the primary studies was that there were relatively small numbers of participants within these studies. Also, psychosocial functioning outcomes were seldom reported. As mentioned, it was common to see a variety of measures used across studies to assess the same outcomes, which may have had an impact on the comparability of studies. Furthermore, the results from the risk of bias assessment revealed variable quality ratings. Caution, therefore, should be taken in interpreting the findings presented in this review.
This review provides some evidence for the beneficial effects of psychosocial interventions for children with comorbid problems. In terms of practice implications, monitoring changes in comorbidity may provide clinicians with useful information on the utility of treatments for conditions that may not necessarily be an intervention target. A thorough understanding of comorbidity may provide insight into how psychosocial interventions may benefit children with comorbidity.