Group-Schematherapy for Adolescents: Results from a Naturalistic Multiple Case Study

A total of four patients (three girls and one boy) were included in the current study. All patients were referred by the GP to the community mental health care centre (Virenze-Riagg Maastricht) for treatment. Each patient underwent an assessment process, which involved a standardised intake involving a semi-structured clinical interview (i.e., Kid-SCID; Hien et al. 1994; Dutch version: Roelofs et al. 2015a), and an evaluation by a psychiatrist. The outcome of the Kid-SCID interview, the psychiatrists’ evaluation information from teachers and clinical observations made during assessment were all used by the multidisciplinary team in making the final clinical diagnoses. Inclusion criteria for the GST program were suffering from personality problems (i.e., having a research classification of a personality disorder not otherwise specified), IQ >80, age >14 years, and having a structured daily life (work or school). Contraindications for participating in the GST were acute psychotic symptoms or acute suicidal behaviours. The age of the four patients ranged between 16 and 18 at time of inclusion. All patients received medication treatment at the start of GST, which was left unchanged during treatment.

The GST program consisted of weekly group sessions complemented by individual treatment sessions with a frequency of once per week or per 2 weeks, depending on the need of the adolescent. The individual sessions were supportive of the group sessions. That is, during individual therapy, the patient could discuss what he or she had learned or experienced in the group and there was time to talk about relevant personal issues as well. The GST largely followed the protocol by Farrell and Shaw (2012). In short, the main phases were bonding and emotional regulation, schema mode change, and finally autonomy and changing behaviour. The aim of GST was to reduce maladaptive modes and develop and strengthen functional modes. The strategies comprise specific cognitive, experiential, and behavioural techniques. Cognitive schema change work involves techniques to identify and change automatic thoughts, identify cognitive distortions, and to empirically test maladaptive rules that have been developed from schemas. Experiential interventions include work with visual imagery, mode dialogues, creative work to symbolize positive experiences, limited-reparenting and the healing experiences of a validating psychotherapist. Behavioral techniques involve pattern breaking work to ensure that changes generalize to behaviors outside of the therapy setting. During the first phase the therapeutic relationship was built. In both individual and group sessions there was a focus on building connections between the therapist and the patient, and among the patients in the group. During group sessions, there were always two therapists, one schema therapist and a creative therapist who alternately took the lead while the other followed closely interactions among the patients as well as individual responses. In this stage, limited reparenting and experiential exercises were helpful to let patients experience what it is like to focus on the experience of a feeling rather than the cognitive process of talking about feelings. Setting limits, an important aspect of limited reparenting, was utilized in this stage to make agreements on issues like self-injurious behaviour, being late, and missing sessions. During the second and third phases the focus of therapy was on the six basic needs: safety, connectedness, autonomy and individuality, realistic boundaries, expression of emotions, and spontaneity and play. Each session started with a relaxation exercise and ended with patients choosing one or several colours that fit their feelings at that moment. A number of mode focused interventions were employed in each session based upon the presenting modes and needs of the patients. If desired, patients could bring in current problem situations.

Parental involvement consisted of group meetings once every 2 weeks that were guided by two therapists who were not the GST therapists. These sessions focused on education about the schema model so that parents would be familiar with schema therapy language. The most important modes were explored in terms of ‘mode clashes’, situations where conflict occurred between parents and the adolescent. The parents were trained to be aware of their own (maladaptive) schema and schema mode activation and were given tools to use when their schema modes conflict with their child’s. Emotion coaching skills were taught, which involve being sensitive to the need of the adolescent and using emotional moments as opportunities to become more connected with the child and to teach the adolescent how to regulate emotions.

All therapists were trained in (group) ST for at least 50 h of training. The training levels of the individual therapists ranged from standard to advance in the Dutch ST Registry as well as for the International Society for Schema Therapy (ISST). All therapists were biweekly supervised by an ISST certified supervisor and there was weekly intervision.

In order to analyse the course of schema modes during therapy, we determined for each patient and mode separately whether a significant trend over time could be detected. The scores on the individual schema modes, as reflected on 10-point Likert scales, represent a time series for which we expect either negative linear trend (for the unhealthy modes) or a positive linear trend (for the healthy modes). To examine statistical evidence for the existence of these expected trends, we fitted an autoregressive time series model for each patient and for each type of mode (unhealthy vs. healthy) separately, using AR1 to model the serial correlation. Although this clearly involves multiple testing, we opted not to correct for this, as conventional methods for correcting multiple testing all have an adverse effect on statistical power. As our study is exploratory rather than confirmatory in nature, our main goal is to highlight the existence of possible trends, and therefore our primary concern here is to guard against inflation of the Type II error rate. The significant changes that were found were discussed in interviews to get a better understanding of the changes that occurred during therapy from the perspective of the adolescent. In the interviews, patients were asked how they thought that their modes have changed during therapy.

To analyse change in quality of life, psychopathology, schema modes, early maladaptive schemas, and schema coping from pre- to post-treatment, scores at both measurement points as reported by adolescents and their parents (if available) were considered in the light of norm scores or findings from previous research. To analyse quality of life, scores from the KIDSCREEN-10 were converted to T-scores (i.e., M = 50, SD = 10). For the SDQ, the available cut-off scores were used to determine whether observed differences between pre- and post-treatment were meaningful (see Goodman 1997). For the SMI-A, YSQ, and SCI no published normative data is available. In order to interpret change on these variables from pre- to post treatment, data from previous studies in our research group were used (i.e., Roelofs et al. 2010; Roelofs et al. 2015b; Wijk-Herberink et al. in preparation). Percentile scores were computed for the SMI-A, YSQ, and SCI, which can be obtained from the first authors. Actual scores of the four patients were compared to these percentile scores. Clinically relevant change was defined as a change of at least two decile steps (i.e., 20 % change, for example a change from percentile 50–30). In addition, for the SMI-A, we relied on normative data for various patient groups from previous research with the SMI in adults (i.e., Lobbestael et al. 2010). We first determined the change in schema modes from pre-treatment to post-treatment and compared these changes to the range of scores for the various patient groups. To analyse changes in YSQ scores from pre- to post-treatment, data obtained in non-clinical adolescents was used (Roelofs et al. 2010). For the YSQ as well, the change in scores from pre- to post-treatment was assessed and compared to the range of scores that correspond with healthy controls.

For each separate mode, a timeseries analysis was carried out to examine the change in the mode scores over time. Table 1 (upper part) summarizes the significant findings. A significant increase was found for the Detached Self-Soother mode and the Healthy Adolescent mode, whereas a significant decrease was found for the Vulnerable Child mode, the Punitive Parent mode, and the Demanding Parent mode. In addition to these findings, data revealed that sensitivity from parents increased (i.e., talking with parents about problems made Emma feel less ashamed and parents more often noticed when she was worried about something; t = 2.27, p < .05 and t = 2.26, p < .05 respectively).

Information obtained through the interview with Emma, revealed that she confirmed most of the significant associations described above. What might have contributed to the change in the Vulnerable Child mode was that she learned to connect more with her inner feelings and that it is important not to block thoughts, feelings, and emotions. According to Emma, this might also have contributed to the positive change in the Healthy Adolescent mode and the decrease in the punitive parent and the demanding parent modes. The increase in the detached self-soother was explained by Emma in terms of the employment of more activities which brought her distraction from her inner feelings. With regard to the sensitivity from parents, Emma had experienced at start of the therapy that it was difficult for her to share emotions and experiences with her parents. However, during therapy she gradually realised that the contact and connection with her parents could be positive.

With respect to change from pre- to post-treatment, the lower part of Table 1 presents the clinically relevant findings. For symptoms of psychopathology as indexed by the SDQ, assessments of both Emma and her parents revealed reduction in symptoms of psychopathology. For quality of life, Kidscreen scores of Emma increased from more than two standard deviations below the mean score (T = 26.6) to less than one standard deviation below mean score (T = 42.1). Neither parent reported clinically relevant change so improvement was mainly reported by Emma.

Clinically relevant change in schema mode scores (i.e., change of at least two decile steps when compared to a normative sample) was found on a number of modes (see Table 1). In addition, at post-treatment, most scores were in the normative range for non-patient controls (see Lobbestael et al. 2010). Except for the schema domain of boundaries, all domains of the YSQ (early maladaptive schemas) showed a relevant change (see Table 1). At post-treatment, these scores were lower than mean scores of the non-patient group (see Lobbestael et al. 2010). Thus, with regard to change in schemas from pre- to post-treatment, results demonstrated a clinically relevant improvement on four out of five schema domains. Finally, for the SCI (schema coping), a relevant reduction in schema surrender and schema avoidance was found, indicating relevant changes in schema coping for Emma.

Table 2 (upper part) presents the change in schema modes over time. A significant increase was found in the vulnerable child mode, the impulsive child mode, and the undisciplined child mode, whereas for the punitive parent mode both a quadratic association as well as a significant linear increase was observed. For the healthy adolescent mode a significant decrease as well as a quadratic association was observed. In addition, sensitivity from parents pertaining to feelings of shame when talking about problems increased (t = 6.73, p < .01). At the same time, a significant decrease was seen in Sensitivity from parents in terms of taking feelings into account (t = −2.82, p < .01).

The interview with Mary provided confirmation of all of the significant associations described above. The significant quadratic effect that was observed in the Punitive parent and Healthy adolescent modes was caused by a clash in the group during therapy when new patients had entered the group. The increase in the Punitive Parent mode was due to a strong manifestation of the Punitive Parent mode halfway therapy. She explained this as a kind of inner voice telling her that she was less than others that was triggered when new patients entered the group. A decrease in the Healthy Adolescent mode was also experienced at that time. A possible explanation for the unexpected increase in the Impulsive and Undisciplined Child mode might be that the patient learned during therapy to speak up for herself even though this could sometimes hurt other people. With respect to opposite effects of therapy on two variables that index sensitivity from parents, she started sharing more with her parents about her feelings, but at the same time she realized that her parents had difficulty dealing with her feelings.

With respect to the changes from pre- to post-treatment (see Table 2, lower part), findings for symptoms of psychopathology (SDQ) demonstrated clinically relevant change but the domains of clinically relevant change were different for Mary and her mother. For quality of life, the criteria for a clinically relevant change were not met and scores from both parents were absent at post-treatment. Relevant change in scores on the SMI-A from pre- to post-treatment are shown in Table 2. SMI scores at pre-treatment revealed scores higher than mean scores of axis II patients for the Vulnerable Child mode, the Angry Child, the Enraged Child, the Undisciplined Child, the Compliant Surrender, the Detached Protector, the Detached Self-Soother, the Self-Aggrandiser, the Punitive Parent and the Demanding Parent. At post-therapy scores remained higher than mean scores of axis II patients, except scores for the Self-Soother mode, which was reduced to scores below mean scores of axis I patients. At pre- and post-treatment, scores for the Happy Child mode and the Healthy Adolescent mode were lower than mean scores of axis II patients. Scores for the Bully and Attack mode at post-treatment remained lower than mean scores of control patients. Scores for the Impulsive Child at pre-treatment were comparable to mean scores of axis I patients, but increased to post-treatment scores lower than mean scores of axis II patients (see Lobbestael et al. 2010). Taken together, most modes at post-therapy remained comparable with mean scores of axis II patients. Scores on the YSQ did not support significant change from pre- to post-treatment. Finally, for the SCI (schema coping), an increase in schema overcompensation was found.

With respect to the change in mode scores over time, a significant decrease was seen for the Vulnerable Child and the Detached Protector, and a borderline significant increase was found in the Happy Child mode (see Table 3). A significant quadratic association was obtained for Sensitivity from parents related to parents noticing that Isabel is worried about something (t = 2.49, p < .05).

Information obtained during the post hoc interview with Isabel confirmed most of the significant associations described above. Although the patient did not know what might have contributed to the positive change in the Vulnerable Child mode, the Detached Protector and the Happy Child mode, the patient indicated that therapy made her stronger. With regard to the change in Sensitivity from parents, the patient had learned to share her feelings more with her parents. When new patients entered the group, she experienced difficulty in opening up to the new members and to discuss this difficulty with her parents. This might have contributed to the positive change in sensitivity from parents and explains the quadratic trend of this variable (Table 4).

With regard to change from pre- to post-treatment, symptoms of psychopathology (SDQ) decreased with more symptom reduction reported by Isabel than her parents. For quality of life a substantial improvement was reported by Isabel (i.e., an increase from three standard deviations below the mean score (T = 22.2) to approximately one standard deviation below mean score (T = 39.9). Unfortunately, quality of life scores were not available for parents. A change in scores on the SMI was found for the Angry Child mode (see Table 3). Scores on modes at pre-treatment were higher than mean scores of axis II patients for the Vulnerable Child mode, the Angry Child, the Undisciplined Child, the Compliant Surrender, the Detached Protector, the Detached Self-Soother, the Punitive Parent and the Demanding Parent. At post-treatment, these scores remained at this level except for the Angry Child mode which reduced to below the mean score of axis I patients. Scores for the Impulsive Child mode increased from a score lower than the mean score of axis II patients, to a score higher than the mean score of axis II patients. At pre- and post-treatment, scores for the Happy Child mode and Healthy Adolescent were lower than mean scores of axis II patients. Scores for the Bully and Attack mode and Self-Aggrandising mode remained at post-treatment lower than mean scores of control patients (see Lobbestael et al. 2010). Thus, these pre- to post-treatment comparisons revealed little positive change in manifestation of dysfunctional modes. In addition, at post-treatment, most modes remained comparable to the mean scores of axis II patients. Scores on the YSQ demonstrated relevant change on the domain Boundaries and for the SCI (schema coping), a relevant decrease was found for schema avoidance.

Before addressing the results of Josh, it is important to note that he experienced a relapse in drug addiction half way therapy. He stayed in the group but this relapse clearly influenced the outcome assessments. No changes in schema modes occurred during treatment. However, a significant increase was seen in Sensitivity from parents related to taking feelings more into account (t = 2.32, p < 0.05).

Information obtained during the interview with Josh, confirmed this positive association. However, in his opinion sensitivity from parents was especially shown to the outside world and he did not feel that his parents’ expression of concern and interest was sincere.

Marginal change in symptoms of psychopathology and quality of life were found from pre to post treatment. For the SMI-A, the Vulnerable Child mode increased whereas the Compliant Surrender, the Punitive Parent, and the Healthy Adolescent modes showed a decrease. SMI-A scores were higher than the mean score of axis II patients for the Angry Child mode, the Impulsive Child, the Undisciplined Child, the Detached Protector, the Detached Self-Soother, the Self-Aggrandiser, Bully and Attack and the Demanding Parent. The post-treatment scores remained higher than mean scores of axis II patients. Scores for the Vulnerable Child mode, the Punitive Parent mode and Healthy Adolescent mode, increased at post-therapy from scores comparable to the mean score of a non-patient control group to scores comparable with mean scores of axis I patients. Scores for the Enraged child increased from higher than the mean score of axis I patients to a score higher than the mean score of axis II patients. At pre-treatment, scores for Compliant Surrender appeared to be comparable with the mean score of axis I patients and reduced to a score below the mean score of non-patient controls. Scores for the Happy Child mode remained below the mean score of non-patient controls (see Lobbestael et al. 2010). Thus, from pre- to post-treatment, some change in manifestation of modes was observed with more frequent manifestation of most modes. Scores on the domains of the YSQ did not change throughout treatment. Finally, schema avoidance was found to increase from pre- to post-treatment.

Admittedly, the current study suffers from a number of limitations. First, we only described four cases of adolescents who received GST and no control group or control condition was included. Therefore, it remains uncertain whether changes are due to treatment or to non-specific factors such as the therapeutic relationship, attention from group members, or time-effects. Given that the treatment lasted for 6 months to 1 year, spontaneous improvement or maturation might be of greater concern than for case studies where the treatment is of shorter duration. The use of a naturalistic treatment design enabled us to explore the effects of treatment in an ecologically valid way, but future research should evaluate the effects of GST in a more controlled manner. Second, we did not asses the long-term effects of the group intervention, thus we cannot draw any conclusions about whether the gains were maintained after termination of therapy. Third, we relied entirely on self-report measures. Future research should include structured interview assessments and ratings of treatment adherence. Finally, another limitation is the lack of formal norms or cut-off scores for the adolescent versions of the questionnaires utilized. Although we used the results available from previous research and conducted analyses to compare adolescent means to those of existing samples, it was not possible to indicate whether gains or declines in our patients were indeed clinically significant.

Despite the aforementioned limitations, the present study adds to the emerging body of evidence, showing that GST, a promising treatment for adults, might also benefit young people with personality disorders or personality disorder features. Our findings that GST in combination with individual treatment sessions can be effective in changing symptoms of psychopathology, schemas and modes, replicates previous findings (van Vreeswijk et al. 2014; Farrell et al. 2009; Renner et al. 2013). The cost-effectiveness of GST compared to the more intense and long-term individual treatment and the potential for making a specialized treatment more available, adds to its potential value for adolescents. Although further evaluation of this model with a larger sample size and under more controlled conditions is warranted, the present study should be considered as a preliminary exploration of the effects of ST for patients under the age of 18 years with personality problems. The findings suggest that ST warrants further examination under more controlled conditions and with a larger sample as a treatment approach for the adolescent population.