An Examination of Perceived Therapeutic Bond as a Mediator of Psychological Outcomes

The therapeutic relationship, defined as “the feelings and attitudes that counseling participants have toward one another, and the manner in which these are expressed” (Gelso & Carter, 1985, p. 159), is an essential component of the therapeutic process. Specifically, the therapeutic bond, alongside faith in the therapeutic process and a mutual agreement between the therapist and client on therapeutic goals, constitutes an integral component of the working alliance (Bordin, 1975, 1989, 1994; Hatcher, Barends, Hansell, & Gutfreund, 1995; Luborsky, 1976). Past research has shown that having a working therapeutic alliance can decrease presenting stressors such as depression and anxiety (Horvath, Del Re, Fluckiger, & Symonds, 2011). Gelso and Carter (1985, 1994) argued that meaningful change occurs once the therapist and client have established a genuine relationship that includes an authentic working alliance. By doing so, the client will be able to experience transference and endure difficult emotions that naturally occur in the therapeutic process, thereby producing more desirable outcomes (e.g., decreased depression or anxiety).

Because greater percentages of students are pursuing higher education degrees than in previous decades (U.S. Bureau of Labor Statistics, 2017), college counseling centers have become one of the most in‐demand providers of mental health services to young adults (Center for Collegiate Mental Health, 2017). College counseling centers are frequently tasked with the competing goals of producing desirable mental health outcomes and simultaneously managing an extremely high student demand (Reetz, Bershad, LeViness, & Whitlock, 2016; Watkins, Hunt, & Eisenberg, 2011). Because of an overall presence of high demand across university counseling centers, assessing mental health outcomes throughout the therapeutic process has become increasingly common (e.g., Locke et al., 2011, 2012). Not only is ongoing assessment of mental health outcomes a critical component of accountability but it is also a mechanism that allows college counseling center administrators and staff to identify patterns of individual and group differences in response to treatment.

Perhaps the most common outcome measure used in university counseling center settings is the Counseling Center Assessment of Psychological Symptoms (CCAPS), developed by the Center for Collegiate Mental Health. As of present, the CCAPS is available as a 34‐item and a 62‐item measure (Locke et al., 2011, 2012), with the primary difference between the two consisting of a family distress measure found exclusively in the 62‐item version. Both versions of the CCAPS contain subscales measuring Depression, Generalized Anxiety, Social Anxiety, Academic Distress, Eating Concerns, Hostility, Alcohol Use, and an overall Distress Index measure (Nordberg et al., 2018). Past studies of the CCAPS's validity (e.g., McAleavey et al., 2012) have demonstrated validity for using the subscales with college populations, and the more recently developed Distress Index (Nordberg et al., 2018) has demonstrated promise for providing a parsimonious data point of the students’ symptomatology across domains.

Although it has become commonplace to assess progress toward goals and outcomes in university counseling settings, far less research exists that quantitatively examines factors that promote or impede progress. In their meta‐analysis of therapy dropout rates among adults seeking psychotherapy, Swift and Greenberg (2012) observed that the highest dropout rates occurred among young adults and students seeking services in university counseling settings versus clients seen in outpatient clinics or hospital settings. Examining factors that predict dropout in university counseling settings further reveals that students who experienced lower levels of validation from their therapist and expressed lower levels of connection with their therapist were less likely to return for follow‐up sessions (Luedke, Peluso, Diaz, Freund, & Baker, 2017). This finding aligns with an extensive body of research highlighting the importance of the therapeutic bond for remaining in therapy and making progress toward individual goals for clients regardless of setting (Gelso & Carter, 1994; Lambert & Barley, 2001; Norcross & Lambert, 2011; Norcross & Wampold, 2011). We present a summary of research that has examined the role of therapeutic bond in college counseling client outcomes to contextualize the present study within the existing literature.

College Counseling Outcomes and Therapeutic Bond

Researchers have studied the role of the therapeutic bond in predicting outcomes for clients in college counseling centers in the context of specific interventions. For example, two studies looked at the effectiveness of career counseling interventions with college students and whether the therapeutic bond explained any of the variances in outcomes. Alchin, Mcllveen, and Perera (2018) found that career interventions were most effective when therapeutic task agreement was high, but they also noted therapeutic bond did not play a role in predicting clients’ negative career thoughts. Milot‐Lapointe, Savard, and Le Corff (2018) found that working alliance (including both task agreement and therapeutic bond) significantly moderated the effectiveness of two career intervention strategies (i.e., individual feedback and written exercises) in reducing career decision‐making difficulties in college student clients. The interventions were only significant when the working alliance was either average or high.

Researchers have studied the therapeutic bond as it relates to successful outcomes for group therapy with college students. Uliaszek, Hamdullahpur, Chugani, and Rashid (2018) looked at working alliance factors as mechanisms of change in positive psychotherapy groups versus dialectical behavior therapy skills groups. Working alliance scores only significantly predicted outcomes for the positive psychotherapy groups. However, in a similar study (Uliaszek, Rashid, Williams, & Gulamani, 2016), it was dialectical behavior therapy skills group participants who outscored their counterparts in positive psychotherapy groups on retention, attendance, and working alliance scores.

Several other studies have examined the relation of the therapeutic bond to outcomes in individual counseling efforts with college students as it relates to alcohol use and depression. In a study of the effectiveness of a one‐session motivational interviewing intervention with college students who were heavy alcohol users, Feldstein and Forcehimes (2007) found that although the intervention yielded significant decreases in binge drinking (vs. the control group), working alliance played no role in predicting or moderating outcomes. In another study, researchers compared the effectiveness of computer‐based cognitive behavioral therapy interventions in decreasing depressive symptoms of college students. Richards, Timulak, and Hevey (2013) found that therapist‐assisted email cognitive behavioral therapy interventions were not superior to online cognitive behavioral interventions that had no therapist‐assisted component for improving depressive symptoms. Furthermore, although participants in the therapist‐assisted condition reported a higher therapeutic bond with their therapist than participants in the online condition did with the online interventions, the bond was not related to depression outcomes in either condition. Thus, some inconsistent findings have emerged concerning whether working alliance, in general, and therapeutic bond, in particular, are consistent predictors of therapeutic outcomes in studies examining outcomes for college student clients.

Additional studies have examined the inconsistent role of the therapeutic bond as it influences the outcomes of college student counseling clients. Tryon and Kane (1993) found that counselors’ ratings of working alliance subscores positively predicted mutual termination decisions, but clients’ ratings of working alliance did not predict termination decisions. Kim, Ng, and Ann (2009) found that with Asian American college student clients, the working alliance was one of several significant factors that predicted the likelihood of recommending the counselor to another client. Wade, Post, Cornish, Vogel, and Tucker (2011) found that self‐stigma was lower and the intention to seek help was higher for students when session depth and working alliance scores were higher. Bieschke et al. (1995) found that working alliance scores were predictive of client satisfaction with counseling regardless of the length of pretreatment waiting periods. Finally, Tokar, Hardin, Adams, and Brandel (1996) found that the therapeutic bond was unrelated to clients’ expectations for personal commitment to counseling. Thus, it appears that the therapeutic bond has a significant impact on the effectiveness of treatments or clients’ positive views of their therapeutic experience depending on the outcome being examined. However, in other cases, it has failed to emerge as a significant factor. One other observation with respect to the existing literature is that studies have infrequently examined bond as a distinct aspect of the working alliance in general, rendering an inability to conclude how the bond, as opposed to task or goal agreement, may be related to therapeutic outcomes.

Purpose

The purpose of the present study was to examine a potential key process that facilitates positive mental health outcomes in college student counseling: the clients’ perceived therapeutic bond with their therapist. Given past research examining the importance of the therapeutic bond, the following hypothesis was developed for the present study: A stronger perceived therapeutic bond will partially mediate positive therapy outcomes in a college counseling setting. It is important to specify that the perceived therapeutic bond will only partially mediate the relationship between therapy outcomes because full mediation would imply that the therapeutic bond is the only factor that contributes to more desirable outcomes. Because the range of interventions providers use is designed to reduce presenting stressors, it is unrealistic to assume that simply establishing a strong therapeutic bond will attenuate distress by itself.

Method

Participants

The final sample consisted of 423 college students from a midsize private university in the midwestern United States. Among students who sought services during a recent 2‐year period (N = 596), their average age was 20.9 years (SD = 3.6) with a range of 17 to 50 years. Students’ gender breakdown was 63.4% female, 34.2% male, and 2.4% transgender/gender nonconforming. Students were mostly White (76.7%). Other racial identities included Asian American (5.9%), Hispanic/Latina/o (5.2%), multiracial (3.9%), African American (4.5%), American Indian (0.8%), and self‐identified as other (1.3%). An additional 1.7% of participants did not report their racial/ethnic identity. Students reported their current year in school as follows: first year (24.9%), second year (20.2%), third year (19.3%), fourth year (20.7%), and graduate/professional degree (12.2%); an additional 2.7% did not report present year. Students reported their presenting concerns from the following list of symptoms on their intake paperwork and were able to select more than one option: relationship difficulties (53.2%), career uncertainty (33.9%), self‐esteem issues (47.5%), existential concerns (21.3%), depression (58.1%), anxiety (75.5%), eating disorder (9.6%), substance use (6.0%), unwanted sexual experience (8.6%), sexual identity (2.9%), stress and psychosomatic symptoms (36.7%), academic concerns (31.7%), anger issues (11.6%), and grief/loss issues (13.1%). The 173 students not included in the present analysis either did not consent for their outcome data to be analyzed or were missing at least one item on any of the measures used in the present study.

Procedure

The university counseling center administered the measures to students who chose to participate in individual therapy from the fall semester of 2016 through the spring semester of 2018. As part of the center's policy, students had the option to allow their anonymous, numeric responses to psychological outcome data to be analyzed for research purposes. Of the 596 students who sought services during this time, 121 participants (20.3%) did not consent to their data being analyzed. An additional 52 participants (8.7%) contained missing responses, and their data were removed from the analysis. Among the participants removed for missing data (N = 52), 39 (75.0%) did not complete the measure assessing therapeutic bond. Little's missing completely at random test revealed there was no pattern to the missing items for the remaining 13 participants, χ2(2028) = 2,068.44, p = .26.

The clinicians consisted of 19 staff and graduate student trainees at the center during the time the data were collected. Past research has demonstrated that unpaid graduate trainees provide services as effective as those of professional staff (Ilagan, Vinson, Sharp, Havice, & Ilagan, 2014). Seven staff members were licensed psychologists. Trainees included a total of six predoctoral interns and six advanced practicum students completing their doctoral degrees in psychology. Licensed psychologists supervised interns, and interns supervised advanced practicum students. Interns received additional supervision from licensed psychologists regarding their supervision of advanced practicum students.

Students completed measures designed to assess psychological outcomes and perceived therapeutic bond before each session. Because the present study examined the students’ perceived therapeutic bond with their therapist, and because the therapeutic bond takes time to develop, only students who attended at least three sessions were included in the current data set. This threshold of excluding students who did not attend at least three sessions is consistent with previous research assessing the perceived therapeutic bond in the third session (Saunders, 2000). To capture the entirety of the clients’ perceived therapeutic bond with their therapist, only the perceived therapeutic bond at the time of the students’ final session was used in the present analyses.

Measures

Counseling Center Assessment of Psychological Symptoms–34 (CCAPS‐34). The CCAPS‐34 (Locke et al., 2012) is a 34‐item measure designed to assess a variety of symptoms present in college populations, most of which are observed in former Axis I diagnoses from the Diagnostic and Statistical Manual of Mental Disorders (4th ed., text rev.; American Psychiatric Association, 2000). Participants respond with their level of agreement to each item on a 5‐point response scale (0 = not at all like me, 4 = extremely like me), with lower scores representing greater disagreement and higher scores representing greater agreement. Sample items include “I am shy around others” and “I feel tense.” The CCAPS‐34 takes approximately 5 to 10 minutes to complete and consists of seven subscales. Locke et al. (2012) reported acceptable internal consistency among all subscales. We averaged responses to compute an overall score for each subscale and the overall Distress Index. In the present study, Cronbach's alphas at the time of intake were .88 for Depression, .84 for Generalized Anxiety, .83 for Social Anxiety, .85 Academic Distress, .92 for Eating Concerns, .81 for Hostility, .85 for Alcohol Use, and .91 for the overall measure. Cronbach's alphas for the last session were .91 for Depression, .85 for Generalized Anxiety, .85 for Social Anxiety, .85 for Academic Distress, .92 for Eating Concerns, .82 for Hostility, .88 for Alcohol Use, and .93 for the overall measure.

Therapeutic Bond Scale. The Therapeutic Bond Scale (Kopta & Lowry, 1997) is one of eight scales contained within the Psychotherapy Outcome and Assessment Monitoring System (Kopta & Lowry, 1997). Participants indicate their level of agreement on a 5‐point response scale (0 = not at all, 4 = strongly agree) with each of the six Therapeutic Bond Scale items assessing the degree of the clients’ perceived therapeutic bond with their therapist. Sample items include “My therapist is confident in treating me” and “My therapist is accepting of me.” Responses were averaged to compute an overall therapeutic bond score. Kopta et al. (2010) reported an internal consistency (Cronbach's alpha) of .91 after two sessions and .95 after six sessions. Cronbach's alpha for the present sample was .92 with an eight‐session average.

Data Analysis

The central research question of determining partial mediation was assessed in three steps (Baron & Kenny, 1986). First, correlations between the initial and final CCAPS‐34 scores on each outcome were examined to determine whether there was a potential direct effect that might be mediated. Second, to determine which outcome variables were eligible to test for statistical mediation, the perceived therapeutic bond was correlated with the initial scores on each CCAPS‐34 domain. Any nonsignificant correlation indicated the lack of a potential mediating effect, and corresponding CCAPS‐34 domains were removed from further analysis. Finally, statistical mediation analyses were run for each of the remaining significant CCAPS‐34 outcomes using PROCESS (Hayes, 2017) Model 4 to determine partial mediation.

Results

Table 1 displays correlations for the observed variables. Under the first step for determining mediation, each of the eight CCAPS‐34 domains was significantly related to each other between the initial and final points of data collection. At the second step, examining correlations between the initial data point with each of the eight CCAPS‐34 domains and the perceived therapeutic bond revealed only three significant relationships. Specifically, the perceived therapeutic bond was significantly and inversely associated with the following CCAPS‐34 domains in the first session: depression, social anxiety, and academic distress. The other CCAPS‐34 domains collected at the first session (i.e., generalized anxiety, eating concerns, hostility, alcohol use, and overall distress) were nonsignificantly related to students’ perceived therapeutic bond by the final session and were thus ineligible to be assessed for mediation in the present study.

Table 1

Intercorrelations, Descriptive Statistics, and Internal Consistencies of Observed Variables

Note. Items 13–17 were omitted due to space limitations. A copy of the complete table is available upon request from the first author.

p < .05.

p < .01.

p < .001.

For the third and final step to determine partial mediation, three mediation models were examined to assess whether the therapeutic bond was significantly associated with the final CCAPS‐34 outcomes of depression, social anxiety, and academic distress in the presence of the initial CCAPS‐34 scores for each respective domain. Each of the mediation models (see Figure ) produced significant indirect effects between the initial and final CCAPS‐34 scores through the mediating perceived therapeutic bond variable. Bootstrap analyses (5,000 bootstrap samples, bias‐corrected intervals) revealed a significant indirect effect between pre‐ and post‐CCAPS‐34 depression scores (Figure ; B = .02, 95% CI [.003, .04]), pre‐ and post‐CCAPS‐34 social anxiety scores (Figure ; B = .02, 95% CI [.01, .03]), and pre‐ and post‐CCAPS‐34 academic distress scores (Figure ; B = .02, 95% CI [.003, .03]) via the mediating role of perceived therapeutic bond.

Figure 1

Significant Partial Mediation Models

Note. The figure examines the relationship between perceived therapeutic bond and psychological outcomes. Standardized path coefficients are shown.

Discussion

The present study examined how a client's perceived therapeutic bond with his or her therapist may potentially mediate the relationship between psychological outcomes in a college counseling setting. The perceived therapeutic bond was found to partially mediate the relationship between decreased CCAPS‐34 scores in Depression, Social Anxiety, and Academic Distress. In other words, as the perceived therapeutic bond increased, post‐CCAPS‐34 Depression, Social Anxiety, and Academic Distress scores decreased. The therapeutic bond was not found to mediate the relationship between decreased CCAPS‐34 scores in Generalized Anxiety, Eating Concerns, Hostility, Alcohol Use, or overall distress.

To digest the implications of these results, it is important first to consider the presenting concerns students endorsed at the time of intake. First, a relatively low number of students who sought counseling at this particular counseling center presented with concerns related to an eating disorder (9.6%), hostility (11.6%), or substance use (6.0%). It is quite possible the nonsignificant mediation results in the domains of eating concerns, hostility, and alcohol use would look much different if these presenting concerns were represented with greater frequency. Furthermore, because the overall Distress Index represents a cumulative score of students’ responses across all domains (Nordberg et al., 2018), it logically follows that the nonsignificant mediation result for overall distress would include variance from the nonsignificant results in the underrepresented domains of eating concerns, hostility, and alcohol use. Although these results support previous research on the nonsignificance of the therapeutic bond as it related to alcohol‐related outcomes in college students (Feldstein & Forcehimes, 2007), no existing studies have examined the relationship between bond and eating concerns and hostility as outcomes in college student counseling efforts.

Perhaps the most surprising result is the perceived therapeutic bond failing to partially mediate the relationship between generalized anxiety outcomes given the high prevalence of students in this sample reporting anxiety concerns at the time of intake (75.5%). However, past research suggests that diagnoses of generalized anxiety disorder have lower overall success rates in terms of effectiveness, treatment, and recidivism (Allgulander, Florea, & Trap Huusom, 2006; Evans et al., 2008; Francis, Moitra, Dyck, & Keller, 2012). For example, Ballenger (2001) reported that the remission rates for clients diagnosed with generalized anxiety disorder are between 33% and 50%. These statistics suggest that therapeutic progress for clients who have such diagnoses is below the typical rates seen for adolescent and young adult clients in general (Lee, Horvath, & Hunsley, 2013). The mechanisms of change for clients with these conditions may be qualitatively different from more “typical” types of presenting problems and, thus, less affected by the therapeutic bond.

Implications for College Counselors

Although the results of the present study may not provide new insights about the mechanisms of change in the treatment of the aforementioned conditions, they demonstrate the importance of the therapeutic bond in mediating outcomes of depression, social anxiety, and academic concerns in college students, which is an important contribution to the literature. Because these types of presenting problems are more commonplace and representative of the reasons college students seek psychological services (Center for Collegiate Mental Health, 2017), understanding the importance of the therapeutic bond has implications for therapists who work with college students with these concerns. These results support literature that has found that common factors, including the bond between therapist and client, explain more variance in outcomes than do specific types of interventions (Norcross & Wampold, 2011).

Some scholarship has explored ways to increase the therapeutic bond in counseling relationships with college students. Perhaps unsurprisingly, although some general recommendations have been offered, these studies have, for the most part, found that student characteristics and intervention contexts are important influences. For example, Bedi, Cook, and Domene (2012) studied factors that hindered alliance development in college students. They found that several common factor–related issues were generated by college students, including nongenuine engagement, unwanted counselor directiveness, nonverbal counselor disengagement, and a lack of trust or comfort with one's therapist. Many of these specific findings would be predicted by the larger body of literature that underscores the critical importance of therapist‐client relationship factors in explaining counseling outcomes (Norcross & Wampold, 2011).

Tailoring counseling efforts to the specific needs of clients is an important element for enhancing the therapeutic bond. Previous studies have researched this issue with culturally diverse clients. For example, Kim, Li, and Liang (2002) and Li and Kim (2004) studied the working alliance as an outcome with Asian American college students. They found that directive intervention styles yielded higher working alliance scores than did insight attainment interventions. Zhang and McCoy (2008) found that processing racial differences between counselors and clients increased working alliance scores for clients compared with clients who did not process such differences with their counselors. Thus, attending to both universal and culturally specific ways to enhance working alliance and therapeutic bond may be important to maximize the effectiveness of therapy for depression, social anxiety, and academic distress with college students.

Limitations

Further research is needed to examine other factors, above and beyond the therapeutic bond, involved in successful outcomes for college students seeking counseling with a variety of presenting concerns. Some limitations of the current investigation include demographic restrictions that preclude generalization to the U.S. population. For example, the sample was overwhelmingly White, and future research should consider examining how the therapeutic bond may develop differently among marginalized racial and ethnic groups. Additionally, because this was an exploratory study examining the perceived therapeutic bond exclusively concerning therapy outcomes, other factors may contribute variance to the perceived therapeutic relationship. Much research has explored the impact of “matching” clients with similar therapists on the basis of demographic variables such as gender, race/ethnicity, or sexual minority status (Bhati, 2014; Johnson & Caldwell, 2011; Owen, Imel, Adelson, & Rodolfa, 2012; Presnell, Harris, & Scogin, 2012; Stracuzzi, Mohr, & Fuertes, 2011).

Further studies examining the impact of the therapeutic relationship on therapy outcomes in a college counseling setting should consider variables that represent perceived similarities or differences between the client and therapist. Our criteria for removing participants with at least one missing item may be considered an additional limitation. However, per Little's missing completely at random test, after removing participants who did not complete measures in their entirety, the remaining missing items were completely random.

Additional limitations include the existence of potential confounding variables that were not examined in the present study. For example, in a meta‐analysis examining underlying processes of cognitive behavioral therapies, Kazantzis et al. (2018) noted strategies such as goal collaboration, provision of feedback, assignment of homework, and modification of cognitive processes improved therapeutic outcomes across a range of presenting problems. Although their meta‐analysis was not restricted to college counseling settings, the abovementioned strategies are used by college counselors (e.g., Chugani, 2017; Iarussi & Shaw, 2016; Wrape, Jenkins, Callahan, & Nowlin, 2016) and may thus contribute to additional variance in the study of college counseling psychological outcomes.

Conclusion

Limitations notwithstanding, the present study has implications for researchers studying the importance of therapeutic relationship factors in college counseling settings as well as practitioners working in college counseling environments. Specifically, the present study demonstrated the salience of a perceived therapeutic bond with college students presenting to counseling with depression, social anxiety, or academic distress. Therefore, college counselors working with clients presenting with such concerns should assess clients’ perceptions of the therapeutic bond to demonstrate intentional concern for and desire to more holistically understand their clients. Examples of interventions include directly asking the client about his or her reactions with respect to the working relationship and collaboratively identifying strategies for feeling understood and accepted by the therapist. By doing so, clients may experience additional support as they work toward achieving their therapy goals.