Behaviour Research and Therapy Essay

Abstract Cognitive-behavior therapy (CBT) for Social Phobia is effective in both group and individual formats. However, the impact of group processes on treatment efï ¬ cacy remains relatively unexplored. In this study we examined group cohesion ratings made by individuals at the midpoint and endpoint of CBT groups for social phobia. Symptom measures were also completed at the beginning and end of treatment. We found that cohesion ratings signiï ¬ cantly increased over the course of the group and were associated with improvement over time in social anxiety symptoms, as well as improvement on measures of general anxiety, depression, and functional impairment. In conclusion, ï ¬ ndings are consistent with the idea that changes in group cohesion are related to social anxiety symptom reduction and, therefore, speak to the importance of nonspeciï ¬ c therapeutic factors in treatment outcome. r 2006 Elsevier Ltd. All rights reserved. Keywords: Social phobia; Social anxiety disorder; Group c ohesion; Cognitive-behavior therapy Introduction Social phobia is characterized by an excessive fear of social or performance situations, during which a person may be scrutinized, judged, embarrassed, or humiliated by others. Evidence-based psychosocial treatments for social phobia have primarily come from a cognitive-behavioral orientation and include various combinations of four main components: (1) exposure-based strategies, (2) cognitive therapy, (3) social skills training, and (4) applied relaxation (for reviews, see   administered in either individual and group formats (e.g., Heimberg, Salzman, Holt, & Blendell, 1993; Turner, Beidel, Cooley, Woody, & Messer, 1994). However, the mechanisms of change, and effective ingredients of these treatments remain relatively understudied. Researchers have compared group and individual treatments for this condition, although evidence regarding the relative effectiveness of each approach has been inconsistent (see Scholing & Emmelkamp, 1993; Stangier, Heidenreich, Peitz, Laut erbach, & Clark, 2003; Wlazlo, Schroeder-Hartwig, Hand, Kaiser, & Munchau, 1990 for direct comparisons of individual and group cognitive-behavioral treatment for social  ¨ phobia). However, for some patients, group treatment may offer a number of advantages over individual treatment. For example, group treatment provides an opportunity to marshal group processes (e.g., encouragement, support, and modeling from other group members) that may aid in teaching cognitive strategies and facilitating exposure exercises. Further, there may be nonspeciï ¬ c effects that arise as a result of the relationships that form amongst group members that may contribute to therapeutic outcome. We decided to investigate how these group processes, particularly group cohesion, may be related to treatment outcome in cognitive-behavior therapy (CBT) groups for social phobia. Within the group therapy literature, one putative mechanism of change is that of group cohesion (Yalom, 1995). However, the construct of group cohesion has deï ¬ ed ready operational deï ¬ nition, especially with more technique-driven interventions like CBT. For example, a broad deï ¬ nition proposed to explain group cohesion is ‘‘the resultant of all forces acting on all the members to remain in t he group’’ (Cartwright & Zander, 1962, p. 74) or, in simpler terms, how attractive a group is for the members who are in it (Frank, 1957). Yalom (1995) conceptualizes group cohesion as the ‘‘we-ness’’ that is felt amongst the group members. Groups with higher levels of cohesion are presumed to have a higher rate of attendance, participation, and mutual support, and to be likely to defend group standards much more. Further, Yalom (1995) believes that group cohesion is necessary for other group therapeutic factors to operate. Researchers studying this construct have also included concepts such as a sense of bonding, a sense of working towards mutual goals, mutual acceptance, support, identiï ¬ cation, and afï ¬ liation with the group (e.g., Marziali, Munroe-Blum, & McCleary, 1997). Clearly then, cohesion is purported to be a critical ingredient for change and therefore would be expected to predict symptomatic outcomes. Some researchers investigating the relationship between group cohesion and treatment outcome have found positive results. Although some of these studies have investigated other nonspeciï ¬ c therapeutic factors as well (i.e., the therapeutic alliance), the present discussion will focus on ï ¬ ndings related to group cohesion processes. Studies have found that group cohesion is related to pre-treatment levels of symptomatic distress, improved self-esteem and reduced symptomatoloty (e.g., Budman et al., 1989). A recent study by Tschuschke and Dies (1994) found that the level of group cohesion in the second half of a long-term psychoanalytic treatment for inpatients was signiï ¬ cantly correlated with treatment outcome and patients who made therapeutic gains reported a high level of group cohesion that began shortly after the ï ¬ rst few sessions. In contrast, unsuccessful patients did not experience a high level of group cohesion at any time. Overall, these studies suggest that group cohesion may play a role in facilitating therapeutic change, though negative ï ¬ ndings also exist (e.g., Gillasp y, Wright, Campbell, Stokes, & Adinoff, 2002; Lorentzen, Sexton, & Hà ¸glend, 2004; Marziali et al., 1997). In the CBT literature, researchers are increasingly paying attention to nonspeciï ¬ c therapeutic factors contributing to treatment outcome (e.g., Ilardi & Craighead, 1994; Kaufman, Rhode, Seeley, Clarke, & Stice, 2005). One of the ï ¬ rst studies in this area was conducted by Hand, Lamontagne, and Marks (1974) in treatment groups for individuals presenting with agoraphobia. They found that members of the group in which cohesion was speciï ¬ cally fostered demonstrated greater improvement up to 6 months after treatment as compared to members of a less cohesive group who demonstrated a greater likelihood of relapse (see also Teasdale, Walsh, Lancashire, & Matthews, 1977, for a replication of these effects, albeit with weaker results). Other ï ¬ ndings from the CBT treatment literature include greater group cohesion ratings predicting lower physical and psychological abuse at follow-up in abusive men (Taft, Murphy, King, Musser, & DeDeyn, 2003), higher levels of group cohesion being signiï ¬ cantly related to decreased post-treatment systolic and diastolic blood pressure as well as improved post-treatment quality of life in cardiac patients (Andel, Erdman, Karsdorp, Appels, & Trijsburg, 2003). In addition, group cohesion ratings have been found to be associated with improvements on depressive symptoms at treatment midpoint, after controlling for initial depression level (Bieling, Perras, & Siotis, 2003). Overall, these studies indicate that group cohesion may play some role in facilitating change or enhancing long-term beneï ¬ ts in CBT-based treatments. Although it is not yet clear what factors are relevant for fostering group cohesion, certain disorders may present more challenges than others. For example, given that social phobia involves an intense fear of scrutiny from other people, these individuals may present with barriers to forming a collaborative alliance, such as poor social skills, extreme sensitivity to evaluations, or social avoidance (Woody & Adessky, 2002). Only one study thus far has examined the development of group cohesion and its relationship to outcome during a group CBT treatment of social phobia. Woody and Adessky (2002) treated individuals for social phobia in a group format using Heimberg’s (1991) protocol for group CBT for social phobia and had clients rate group cohesion using the Group Attitude Scale (GAS; Evans & Jarvis, 1986). The GAS measures the clients’ degree of attraction to the group. Measurements were conducted at three points during treatment (sessions 2, 5, and 9) and indicated that group cohesion remained static over time. They also found that the level of group cohesion clients reported was in no way related to outcome. It was suggested that the constructs and measurement of group process in cognitive-behavioral approaches might need to be further reï ¬ ned in order to more fully understand the degree to which group format and group process variables may add an important element to therapeutic outcome. It is important to note that the measure of group cohesion used by Woody and Adessky (2002) deï ¬ nes the construct unidimensionally. The GAS was designed to measure only attraction to group, deï ¬ ned as ‘‘an individual’s desire to identify with and be an accepted member of the group’’ (Evans & Jarvis, 1986, p. 204). Examples of items include: ‘ ‘I want to remain a member of this group,’’ ‘‘I feel involved in what is happening in my group,’’ and ‘‘In spite of individual differences, a feeling of unity exists in my group.’’ However, as discussed by Burlingame, Fuhriman, and Johnson (2002), elements of group cohesion may include both intrapersonal elements (e.g., group member’s sense of belonging and acceptance) as well as intragroup elements (e.g., attractiveness and compatibility felt among the group members). Therefore, by solely focusing on attraction to the group it is possible that the GAS fails to operationalize aspects of cohesion that are important for making therapeutic gains. The present study, therefore, examined the role of cohesion in group CBT for social phobia, using a measure that includes items that ostensibly assess a number of different constructs thought to be related to group cohesion. The Group Cohesion Scale-Revised (GCS-R), developed by Treadwell, Laverture, Kumar, and Veeraraghavan (2001), taps into several different aspects of group cohesion including: interaction and communication (including domination and subordination), member retention, decision-making, vulnerability among group members and consistency between group and individual goals. This self-report questionnaire has been shown to be both reliable and valid for detecting changes in group cohesiveness during the process of group development (Treadwell et al., 2001). Clients with a principal diagnosis of Social Phobia were treated and, based on the preceding literature, we explored: (1) group cohesion development during the course of the group and (2) the relationship of group cohesion to treatment outcome, broadly deï ¬ ned to include not only social phobia symptoms, but the overall experience of negative affect (e.g., general anxiety and depression) and functional impairment. We hypothesized that group cohesion would increase from the midpoint of treatment to the endpoint of treatment and that group cohesion ratings would be signiï ¬ cantly related to positive treatment outcome (i.e., symptom reduction). Method Participants There were a total of 34 outpatient individuals in this study. The average age of participants was 36 years (range 19–64 years; 19 female, 15 male). All individuals reported symptoms meeting criteria for a principal diagnosis (i.e., the diagnosis causing the most distress or impairment) of Social Phobia, as determined by the Structured Clinical Interview for the Diagnostic and Statistical Manual of Mental Disorders—4th edition (SCID-IV; First, Spitzer, Gibbon, & Williams, 2001). One individual also had symptoms meeting criteria for a co-principal primary diagnosis of Dyssomnia Not Otherwise Speciï ¬ ed. For 32 of the participants, the social phobia was generalized (i.e., occurring in most social situations), whereas for the other two participants, it was nongeneralized, occurring in several, but not most social situations. Of the 34 participants, 57% reported symptoms meeting criteria for one or more additional mood disorder (Major Depressive Disorder, 47%; Bipolar Disorder, 6%; Dysthymic Disorder, 3%), 62% had one or more additional anxiety disorder (Speciï ¬ c Phobia, 47%; Generalized Anxiety Disorder, 26%; Obsessive Compulsive Disorder, 21%; Panic Disorder, 12%; Panic Disorder with Agoraphobia, 9%), and 27% had one or more additional other diagnoses (Hypochondriasis, 6%; Eating Disorder Not Otherwise Speciï ¬ ed, 6%; Cannabis Dependence, 6%; Paraphilia Not Otherwise Speciï ¬ ed, 3%; Intermittent Explosive Disorder, 3%; Impulse Control Disorder Not Otherwise Speciï ¬ ed, 3%). The values for the anxiety disorders sum to greater than 100% as several participants had multiple anxiety disorders. Measures Depression Anxiety Stress Scales, 21-item version (DASS-21; Lovibond & Lovibond, 1995). This short form of the original 42-item DASS is a 21-item self-report measure designed to assess depression, anxiety and stress that an individual has experienced over the past week. Each scale consists of seven items and respondents indicate how much each statement applied to them over the past week on a four-point Likert scale. The Depression scale (DASS-21-D) measures dysphoria, hopelessness, devaluation of life, self-deprecation, lack of interest/involvement, anhedonia, and inertia. The Anxiety scale (DASS-21-A) measures autonomic arousal, skeletal musculature effects, situational anxiety, and the subjective experience of anxiety affect. The Stress scale (DASS-21-S) measures difï ¬ culty relaxing, nervous arousal, and a tendency to become easily upset/ agitated, irritable/over-reactive, and impatient. Strong internal consistency with a clinical sample has been demonstrated with the DASS-21 (a’s ranging from .87 to .94), and the factor structure is well supported (Antony, Bieling, Cox, Enns, & Swinson, 1998). Construct validity of the three scales has also been demonstrated (see Brown, Chorpita, Korotitsch, & Barlow, 1997). In the current sample, reliability was acceptable at pre-treatment for the Depression (a  ¼ :91), Anxiety (a  ¼ :87), and Stress (a  ¼ :87) scales. Illness Intrusiveness Ratings Scale (IIRS; Devins, 1994). The IIRS is a 13-item questionnaire that measures the extent to which a disease, its treatment, or both interfere with activities in 13 important domains considered essential to a positive quality of life. These domains include health, diet, work, active recreation (e.g., sports), passive recreation (e.g., reading), ï ¬ nances, relationship with partner, sexual and family relations, other social relations, self-expression/self-improvement, religious expression, and community and civic involvement. For each item, an individual rates the intrusiveness on a scale for 1–7 with higher scores indicating more intrusiveness. The IIRS has been shown to have good psychometric properties in both medically ill populations (Devins et al., 2001) and anxiety disorders groups (Antony, Roth, Swinson, Huta, & Devins; 1998; Bieling, Rowa, Antony, Summerfeldt, & Swinson, 2001). In the current sample, reliability was acceptable at pre-treatment (a  ¼ :87). Social Phobia Inventory (SPIN; Connor et al., 2000). The SPIN is a 17-item questionnaire designed to assess symptoms of social phobia. Each item measures the severity of a particular symptom during the past week, using a ï ¬ ve-point scale ranging from 0 (not at all) to 4 (extremely). It consists of three subscales: fear, avoidance, and physiological arousal. The SPIN has been shown to have good empirical support (Antony, Coons, McCabe, Ashbaugh, & Swinson, 2006; Connor et al., 2000) and enables the assessment of a wide range of social anxiety symptoms, making it an ideal measure for generalized social phobia. The total score measure of the SPIN has recently been shown to have excellent internal consistency for the total score for individuals meeting criteria for Social Phobia (with a  ¼ :92) and a combined sample of individuals meeting criteria for Social Phobia (Generalized Type), Panic Disorder with Agoraphobia and Obsessive-Compulsive Disorder (with a  ¼ :95) (Antony et al., 2006; Connor et al., 2000). It has also been shown to have good test–retest reliability (r  ¼ :86, po:001), convergent and discriminant validity as well as being able to distinguish well between those with Social Phobia as opposed to Panic Disorder with Agoraphobia or Obsessive-Compulsive Disorder. The SPIN has also been shown to be sensitive to changes in the severity of social phobia following cognitive-behavior treatment (Antony et al., 2006). In the current sample, pre-treatment reliability was acceptable (a  ¼ :93). Group Cohesion Scale-Revised (GCS-R; Treadwell et al., 2001). The GCS-R is a 25-item questionnaire designed to assess group cohesion in terms of interaction and communication among group members (including domination and subordination), member retention, decision-making, vulnerability among group members, and consistency between group and individual goals. Each item is rated on a scale from 1 (strongly disagree) to 4 (strongly agree). Examples of items include: ‘‘Group members usually feel free to share information,’’ ‘‘There are usually feelings of unity and togetherness among the group members,’’ and ‘‘Many members engage in ‘back-biting’ in this group.’’ This scale was recently revised (Treadwell et al., 2001) in order to modify one item, discard another item, and change the wording of the anchor points. In a validation study, internal consistency (as measured by Cronbach’s alpha) ranged from .48 to .89 on pre-test assessment and .77–.90 on post-test assessment (Treadwell et al., 2001). In the current sample, reliability was acceptable at both treatment midpoint (a  ¼ :84) and at treatment endpoint (a  ¼ :79), and the reliability of the change score was .56 (Williams & Zimmerman, 1996). Procedure All individuals completed a 10-session CBT treatment group for social phobia.1 Treatment administered was based on protocols described by Heimberg and Becker (2002) and Antony and Swinson (2000). The key components of therapy included: psychoeducation, cognitive restructuring, in-session and between-session exposure exercises, as well as social skills training. Groups were run by two therapists and consisted of ï ¬ ve–eight patients per group. A total of 11 groups were included in the study. It should be noted that initially 76 individuals were enrolled in these 11 groups. However, of these 76 individuals, there were only 67 individuals from whom any measures were received at all (i.e., nine individuals did not return any data). In order to conduct the analyses that will be described below, it was possible to include only 34 of these 67 individuals. This was due to a need to have received both mid- and post-GCS measures as well as pretreatment outcome data. Therefore, the working sample that will be discussed in this study encompasses 34 individuals who completed the treatment as well as these various measures2. Participants completed the GCS questionnaire at the midtreatment session of each group (i.e., session 5) as well as during the last session of each group. Questionnaires assessing symptom severity (i.e., the DASS-21, IIRS a nd SPIN) were completed prior to the beginning of group treatment as well as during the last session of each group. Data analysis Multilevel regression analyses (i.e., generalized mixed modeling) using the software program HLM 6 (Raudenbush, Bryk, Cheong, & Congdon, 2004) were conducted. We used this approach because it allowed us to assess and control for nonindependence of data that might arise from being nested into treatment groups (Hedeker, Gibbons, & Flay, 1994; Herzog et al., 2002). Another advantage of HLM was that it can accommodate unequal group sizes and employs maximum likelihood estimation instead of least squares. Before examining change over time in the outcome variables and GCS, intraclass correlations (ICCs) were estimated to examine the interdependence of data due to nesting (see Herzog et al., 2002). The intraclass correlation depicts how much variance in the outcome variable is due to within-subjects, between-subjects, and between-groups variance. Results For each analysis, items were included from each scale for each individual, unless 20% or more of data were missing. Missing values for a particular scale item were replaced by calculating the mean value for that scale item and using this mean value in place of the missing value. Outcome measures A series of three-level regression models were evaluated to examine change over time in the outcome variables. Level 1 consisted of repeated measures (i.e., two assessment occasions) that were nested within 1 Two groups completed 12-session CBT treatment groups and one group completed a 9-session treatment group. The use of a multilevel regression approach (i.e., HLM) allowed us to examine whether or not number of sessions per group affected any of the relationships reported. Results indicated that number of sessions did not moderate any of the results reported in the paper. 2 These treatment groups were not conducted as part of a formal treatment outcome study, which accounts for the number of patients who failed to return their post-treatment questionnaires. Therefore, the individuals who completed both pre-treatment and post-treatment measures provide a naturalistic and ecologically valid cross section of moderate to severe social phobia patients typically seen in an acute outpatient clinic, presenting with anxiety disorders and related problems.