1
Hayes, L., Boyd, C. P., & Sewell, J. (2011). Acceptance and Commitment Therapy for the Treatment of Adolescent Depression: A Pilot Study in a Psychiatric Outpatient Setting. Mindfulness.doi:10.1007/s12671-011-0046-5
The final publication is available at
ORIGINAL PAPER
Acceptance and Commitment Therapy for the Treatment of Adolescent Depression: A Pilot Study in a Psychiatric Outpatient Setting
Louise Hayes Candice P. Boyd Jessica Sewell
______
L. Hayes J. Sewell
Psychology Department, University of Ballarat, P.O. Box 663, Mount Helen, Victoria, Australia, 3353
e-mail:
C. P. Boyd
Centre for Youth Mental Health, Orygen Youth Health Research Centre, University of Melbourne, Australia
e-mail:
Abstract
Based on promising results with adults, Acceptance and Commitment Therapy (ACT) presents as a treatment opportunity fordepressed adolescents. We present a pilot study that compares ACT with treatment as usual (TAU), using random allocation of participants who were clinically referred to a psychiatric outpatient service. Participants were 30 adolescents, aged M =14.9 (SD =2.55), with 73.6% in the clinical range for depression. At posttreatment on measures of depression participants in the ACT condition showed significantly greater improvement statistically (d=0.38), and 58% showed clinically reliable change with a response ratio of 1.59 in favor of ACT. Outcomes from 3-month follow-up data are tentative due to small numbers but suggest that improvement increased in magnitude. Measures of global functioning showed statistically significant improvement for both conditions, although clinical change measures favored only the ACT condition. The results support conductinga larger trial of ACT for the treatment of adolescent depression.
Keywords: Acceptance and Commitment Therapy, Adolescent psychopathology, Adolescent Depression,
Introduction
Acceptance and mindfulness treatments have revealed positive outcomes for the treatment of psychopathology in adults. Adult treatment studies demonstrate that Acceptance and Commitment Therapy (ACT) can achieve positive long-term outcomes for depression, anxiety, psychosis, chronic pain, work stress, stigma and burnout (S. C. Hayes, Luoma, Bond, Masuda, & Lillis, 2006; S. C. Hayes, Masuda, Bissett, Luoma, & Guerrero, 2004). In a review of 21 treatment studies the weighted mean effect size for ACT was d=0.66 at posttreatment (N=704) and this was maintained at follow-up(S. C. Hayes, et al., 2006). For depression, the ACT literature has shown positive outcomes in treating depressed adults individually and in group settings (Lappalainen et al., 2007; Zettle & Hayes, 1986; Zettle & Rains, 1989). ACT is now listed with modest research support as an empirically supported treatment for depression and chronic pain (Division 12 of the American Psychological Association,
ACT with adolescents is a new treatment frontier with empirical literature beginning to emerge. Wicksell and colleagues tested ACT for adolescents with chronic pain and found significant improvements in functional ability, pain intensity, and pain related discomfort (Wicksell, Melin, Lekander, & Olsson, 2009; Wicksell, Melin, & Olsson, 2007). They also found adolescents treated with ACT reported less catastrophizing and lowered perceived pain (Wicksell, et al., 2009). Other adolescent studies have been undertaken in school settings, with onerandomized controlled trial comparing ACT to a no treatment control group finding improved outcomes up to two years later on measures of stress and psychological flexibility (Livheim, 2004). Thus, previous studies suggest ACT for adolescents should be evaluated further.
A recent review of treatments for adolescent depression treatment argued that the ACT model could be suitably adapted (Hayes, L. L., Bach, & Boyd, 2010). We present a pilot study of ACT with adolescent depression. The aim was to examine ACT when compared to treatment as usual (TAU) provided in a psychiatric outpatient setting. The primary outcome variable was depressive symptoms and the secondary outcome variable global functioning (a measure that includes emotional symptoms, prosocial behavior, and peer interactions). Given that empirical literature on adolescent depression has shown that treatment as usual can perform similarly to an experimental treatment (Watanabe, Hunot, Omori, Churchill, & Furukawa, 2007) and that adolescents attending a psychiatric service are rarely discharged without improvement (they drop out instead), it was hypothesized that adolescents in both ACT and TAU groups would show significant improvements. However, given the empirical findings on ACT studies with adults, it was expected that adolescents receiving ACT would show larger gains than those receiving TAU. It was also hypothesized that at follow-up, participants who received ACT would continue to show significantly more improvement than those receiving TAU.
Method
Participants
Participants were 38 adolescents aged between 12 and 18 years (M = 14.9, SD = 2.55) referred to a public child and adolescent psychiatric service. Demographic characteristics of the ACT and TAU groups were equivalent with no statistical differences found using t-test for means or chi square tests of expected frequencies (shown in Table 1). The participants in both groups were born in Australia, with only one participant from an indigenous background. Approximately half the adolescents were living with both parents.
Participants were referred to the service by parents (n=12) or by medical or school professionals (n=26). Eligibility for the study was determined in a two-step approach, initial allocation was at telephone triage and this was confirmed following clinical assessment. Participants were eligible for the study if they were experiencing moderate to severe depressive symptoms, including: depressed mood, loss of interest in pleasurable activities, feelings of worthlessness or guilt, irritability, lethargy, feeling wound up, sleeping too much or too little, social withdrawal and suicidal ideation. Exclusion criteria included being actively suicidal, as evidenced by any recent suicidal attempt or current plan, current substance abuse, active psychosis or schizophrenia, intellectual disability, and chronic illness. Adolescents meeting these exclusion criteria were given alternative treatments rather than individual psychotherapy (e.g. inpatient care, drug and alcohol services). Depression was assessed via clinician ratings from clinical interviews using the Development and Wellbeing Assessment (Goodman, Ford, Richards, Gatward, & Meltzer, 2000). The Development and Wellbeing Assessment is a clinical interview tool used to assess generate DSM-IV diagnosis ( At baseline, group comparisons revealedno statistically significant differences in the number of participants reporting depressed mood, irritability, or loss of interest between ACT and TAU. There was also no significant difference between the groups in contact hours with the service.
Insert Table 1 about here
Group Allocation. Treatment was delivered by eight clinicians, under the leadership of the team psychiatrists. Clinician occupations included psychologists, social workers, psychiatric nurses, and occupational therapists. Due to clinical and organizational constraints, clients were allocated to a treatment condition prior to the assessment of study criteria.
Referrals to the service are triaged by telephone and then reviewed at multidisciplinary intake meetings. Accepted referrals are allocated to the next available treating clinician. To avoid clinicians’ preference for one therapeutic mode over another, adolescent participants were agreed as suitable for the study at the above-described team meetings using triage information only (N=51). Participants were then allocated to treatment condition using a concealed SPSS generated random number table and the principal researcher advised the clinician of the treatment condition for their participant. A full clinical interview wasthen conducted with the adolescent. Thirteen participants (ACT = 6, TAU = 7) were excluded when the clinical interview revealed that they did not meet criteria for the study or the adolescent did not want to participate. No data were able to be collected on the 13 participants who were excluded prior to pretreatment measures. The time frame allowed for the study, along with this method of allocation, resulted in an imbalance in the groups. Nevertheless, this design was preferable to clinicians deciding on their client’s suitability for the study after conducting a clinical assessment, which can be up to three sessions, and is likely to be contaminated with therapeutic content. There were 22 participants who received individual ACT treatment and 16 who received TAU.The recruitment period was 13 months. The flow of participants from recruitment is shown in Figure 1.
Insert Figure 1 about here
Measures
Pretreatment measures were administered during the clinical assessment. Post-treatment was administered at the final discharge session. Follow-up measures were sent to participants by mail three months after discharge. Three attempts were made to gather follow-up data.
Reynolds Adolescent Depression Scale-2 (RADS-2).The Reynolds Adolescent Depression Scale-2 (RADS-2) is a brief 30-item self-report measure designed to measure depressive symptoms in adolescents (Reynolds, 2002). RADS-2 measures current levels of depressive symptomatology along four dimensions: Dysphoric Mood, Anhedonia/Negative Affect, Negative Self-Evaluation and Somatic Complaints, and it also provides a total depression score. Participants are asked to indicate how often the item describes how they feel on a 4-point Likert scale ranging from 1=Almost never to 4=Most of the time. Items are suited to adolescents as the questions are short and have an average reading level of grade 1.8 (for example, ‘I feel like crying’ or ‘I feel like hiding from people’). Internal reliability on a school sample of participants (N=9052) is strong with Cronbach’s alpha of .93 for the total score, and coefficients ranging from .80 to .87 for the four subscales (Reynolds, 2002). Test-retest reliability has been demonstrated over a 6 week period at .80, and over a 3-month period at .79 (Reynolds). The RADS-2 is a valid measure of depression when compared to clinical interviews, with an interrater reliability of .91 with the Hamilton Depression Rating Scale(Reynolds). RADS-2 also correlates highly with established self-report measures of depression with a median r of .73 when compared with the Beck Depression Inventory, and Children’s Depression Inventory(Reynolds). The RADS-2 manual provides normative data and clinical cut-offs for clinical and school samples and also grades scores into severity categories (normal, mild, moderate, severe) with higher scores indicating greater depressive symptomatology (Reynolds).
Strengths and Difficulties Questionnaire (SDQ-S).The Strengths and Difficulties Questionnaire – Student Version (Goodman, Meltzer, & Bailey, 2003) is a brief questionnaire that assesses the emotional and behavioral functioning of adolescents (the measure is available at are five subscales – Conduct Problems, Hyperactivity, Emotional Symptoms, Peer Problems and Prosocial Behavior – and these create a Total Difficulties score. Sample items include: ‘I worry a lot’, or ‘I have one good friend or more’, I have many fears, I am easily scared’. Adolescents self-rate 25 items on a 3-point scale, ranging from 0=Not true to 2=Certainly True. The full-scale score has the strongest reliability and will be the focus of analysis for this study. The full scale of the youth self-report version has sound reliability at = .80 (Goodman, 2001) and is considered a valid screen for psychopathology (Goodman, Ford, Simmons, Gatward, & Meltzer, 2003). Australian test-retest reliability for the measure is .84 for the self-report student version(Mellor, 2004). Reliability and validity have been subsequently established cross-culturally (Hawes & Dadds, 2004; L. L. Hayes, 2007; Klasen, Woerner, Rothenberger, & Goodman, 2003; Malmberg, Rydell, & Smedje, 2003; van Widenfelt, Goedhart, Treffers, & Goodman, 2003) and the measure is routinely used in many countries for epidemiological, developmental, and clinical research. In Australia the SDQ is a required measure in child and adolescent psychiatric services(L. L. Hayes, 2007).
Procedure
Participants were invited into the study during their clinical assessment period, which is typically up to three sessions. The assessment comprises family, adolescent and parent interviews that cover problem and history of treatment, developmental history, family and relationship history, educational history, mental status examination, risk assessment and psychiatrist review. All participants were required to provide written parental consent to be involved in the study.
The ACT treatment comprised individual sessions using published ACT treatment manuals (including: Greco, 2006; Harris, 2007; S. C. Hayes, Strosahl, & Wilson, 1999; Zettle, 2007). All clinicians were required to undertake a 2.5 day ACT training course provided by an accredited ACT Trainer and read the following texts:(a) ACT for teens group program (Greco, 2006), and (b) The Happiness Trap (Harris, 2007). Clinicians were able to use any components of these three treatment resources according to the needs of the client. Resource copies of ACT texts were also available for additional learning(S. C. Hayes, et al., 1999; Zettle, 2007). An external consultantprovided ongoing group supervision during the course of the research to ensure skill development and treatment adherence. This consultant was a clinical psychologist experienced in ACT and in psychiatric services.
TAU was the approved psychotherapy provided by the psychiatric service. This comprises manualized CBT that includes: psychoeducation, early warning signs planning, coping with unpleasant thoughts, increasing pleasant activities, problem analysis, problem solving, goal setting, and session by session problem solving and crisis management (Ballarat Psychiatric Services Treatment Manual, detailed information available from first author). OneTAU participant received the above standard treatment plus family therapy sessions. Ongoing group and individual supervision was also provided for the TAU condition by consultant psychiatrists and senior clinical staff. To eliminate overlap, therapists were instructed that TAU must be delivered in accordance with the service’s prescribed treatment manuals, which do not include mindfulness, acceptance, defusion or values work, as used in ACT. Treatment overlapped with regard to goal setting and behavioral activation.
Data Analysis
Data were analyzed in four stages using the SPSS statistical package. First using pretreatment data, the clinical characteristics of the groups were compared to each other, and then compared to published normative data. Second, the effect of the intervention was examined for each outcome variable obtained at the three times using linear mixed models. While an ANCOVA is the most commonly used method to analyze treatment data, linear mixed modeling(LMM, also known as hierarchical linear modeling or HLM) is increasingly being seen in treatment outcome publications because it has a number of strengths for outcome analysis (some examples of recent uses in high ranking journals include, Aldea, Rice, Gormley, & Rojas, 2010; Neacsiu, Rizvi, & Linehan, 2010; Webb, de Ybarra, Baker, Reis, & Carey, 2010). While ANCOVA requires replacing missing data using an intention-to-treat strategy (e.g. last observations carried forward) LMMs allow all data to be analyzed without the need to generate lost observations. Therefore, LMM has less distortion in the data than intention-to treat methods (for a comparison of methods using clinical data see Beunckensa, Molenberghsa, & Kenward, 2005). Unlike ANCOVA, LMM do not require homogeneity of slopes and therefore the model can tolerate variability in both the slope and intercepts. Finally, LMM is considered superior when the design is unbalanced (Field, 2009).
The first step in LMM is to test for the fit of the data to the model. Data models were tested in a stepped approach with the improvement in model fit assessed using 2 tests of deviance and the final model accepted when no significant improvement in 2 was evident. That is, the chi-square demonstrates that the data model is the best fit of the data. In this model, pretreatment scores were used as a covariate. The model then tested two factors (group and time), together with the group by time interaction. In the final model the necessary assumptions of normality and homogeneous variance of errors were examined by analysis of residuals. At the third stage of analysis, post hoc tests of the main effects and interactions were performed using Bonferroni corrected alpha levels.
The final step examined the clinical significance of the change at posttreatment and follow-up. This was based on the Jacobson and Truax(Jacobson & Truax, 1991) reliable change index, with RCI set at +1.96 and adopting their change criterion (c), where the level of functioning places the participant closer to the mean of the functional population. The RCI scores were calculated using software provided by Agostinis, Morley and Dowzer (2008) and using the published normative reliability data on each measure (as shown in Table 3). Response ratio was then calculated as a ratio of responders to non-responders, where 1 indicates no difference between treatment and control.
Results
Participant Flow Demographics and Clinical Characteristics
The two groups were equivalent in clinical severity on pretreatment measures (Table 1 and 3). Of the 38 participants included in the study, posttreatment data were available on 30 participants (n=8, or 21% missing). As shown in Figure 1, comparatively more TAU participants did not complete posttreatment measures. Participants dropped out of the study because they were no longer seeking services (n=4), or had moved house (n=4).Attrition analysis showed no difference between completers and dropouts (n=8) on pretreatment measures.There were no significant correlations between the outcome measures and the sample characteristics of age, gender, or hours of treatment provided.
On the RADS-2 and SDQ there was no statistical difference in the pretest mean scores for the ACT and TAU conditions (Table 2). These scores were then compared with published normative data (Table 3). On the RADS-2, clinical levels of depression are evident above the 85th percentile (Reynolds, 2002); there were 16 ACT participants (73%) in this clinical range and 12 TAU participants (75%). Established Australian norms are available for the Strengths and Difficulties Questionnaire (Mellor, 2005). Scores are in the borderline-clinical range at the 80th percentile; there were 18 ACT participants (82%) above this cut-off and 16 TAU participants (100%).
Intervention Effects on Depressive Symptomatology
The best fitting linear models for the RADS-2 total and each of the subscales were models with pretreatment scores used as a covariate, a random intercept allowed in the model, and time used as a nonlinear factor (this allowed the model to fit varying slopes for the ACT and TAU conditions). For the RADS-2 total scale score across the three time points, the final model was a significantly better fit than the independence model, 2 (3) = 17.09, p .001. As shown in Table 2, participants in the ACT condition showed greater improvement than those in the TAU condition, with a significant group x time interaction, F (47.15) = 5.59, p =.007. Main effects comparisons revealed significant group differences between the ACT and TAU conditions (p =.013). There was also a significant effect for time on the RADS-2 (p = .001). Post hoc comparisons using Bonferroni corrected significance revealed that the ACT group improved at posttreatment (pretreatment to posttreatment, Mdiff= 12.86, p=.007, 95%CI 2.32, 23.39), but the TAU group did not significantly improve (pretreatment to posttreatment, Mdiff= 9.92, p=.42, 95%CI -3.62, 23.46). The same pattern was evident at follow-up, the ACT group continued to improve (pretreatment to follow-up, Mdiff= 21.99, p.001, 95%CI 7.41, 36.56), whereas the TAU group deteriorated (pretreatment to follow-up, Mdiff= -4.62, p=1.00, 95%CI -24.79, 15.53).The effect size for ACT using Cohen’s d was 0.38 at posttreatment and 1.45 at follow-up.