Clinical Consensus Strategies to Repair Ruptures in the Therapeutic Alliance

Abstract

Objective

The alliance been recognized as an essential common factor and robust predictor of outcome. The present study sought to further our knowledge of the alliance and to promote the integration of research and practice by assessing consensus among peer-nominated expert therapists of varying theoretical orientations on the effectiveness of clinical strategies to repair alliance ruptures.

Method

This study drew on the behavioral-analytic model (Goldfried & D’Zurilla, 1969) and the methodology of the Expert Consensus Guideline Series (Frances, Docherty, & Kahn 1997). In Phase I, 69 therapists submitted clinical situations describing alliance ruptures. In Phase II, 177 therapists generated responses to the situations, and clinical strategies underlying the responses were identified. In Phase III, 134 peer-nominated experts (a mean of 22.3 therapists per situation) rated the effectiveness of these clinical strategies.

Results

The experts reached consensus on the use of strategies that validated the client’s experience and explored the rupture during the rupture session. Change-oriented interventions (e.g., changing interpersonal interactions; highlighting patterns of behavior, thought, or emotions) were generally rated as less effective to use during the rupture, but effective for use in future sessions.

Conclusions

The findings are consistent with the growing literature on the value of using certain alliance-focused interventions during a rupture. The findings point to the importance of therapists’ awareness of the state of the alliance so that they can identify when ruptures are occurring.

Research has consistently identified the therapeutic alliance as a robust predictor of outcome across theoretical orientations (e.g., Horvath, Del Re, Flückiger, & Symonds, 2011) Indeed, as Wolfe and Goldfried observed, the alliance “is probably the quintessential integrative variable because its importance does not lie within the specifications of one school of thought” (Wolfe & Goldfried, 1988, p. 449). Research on the alliance as a common factor has also played a central role in the history of psychotherapy integration (see Castonguay et al., 2015). While there is debate as to whether the alliance has a causal effect on outcome or is simply correlated with outcome through prior symptom improvement (e.g., Feeley, DeRubeis, & Gelfand, 1999), many believe that both the relationship and therapy techniques are important and transact with each other (e.g., Goldfried & Davila, 2005).

There is evidence that ruptures, or strains in the alliance, occur often in treatment across theoretical orientations (see Eubanks-Carter, Muran, & Safran, 2010, for a review). Alliance researchers such as Safran and Muran argue that ruptures are in fact inevitable as clients and therapists negotiate the tasks and goals of therapy (Safran & Muran, 2000). Although poorer alliances, marked by ruptures, predict premature termination, there is evidence that successful resolution of ruptures is correlated with good outcome and with greater retention in treatment across theoretical orientations (Safran, Muran, & Eubanks-Carter, 2011). Repairing a rupture in session allows clients to work through interpersonal issues in vivo, which in turn may enhance outcome.

Rupture resolution may require therapists to deviate from the treatment they are employing in order to focus directly on the alliance rupture. Several investigators have found evidence that continuing with “techniques as usual” in the context of a rupture may further erode the alliance (e.g., Castonguay et al., 1996; Piper, et al., 1999). These findings point to the need to identify and test therapeutic techniques targeted to address alliance ruptures.

Much of the existing research on alliance ruptures and repairs draws on the work of Safran and Muran. Safran, Muran, and colleagues (e.g., Safran & Muran, 2000) have identified strategies for repairing two general classes of ruptures: withdrawal and confrontation ruptures. Withdrawal ruptures are characterized by client discontent expressed through avoidance and withdrawal behaviors (e.g., giving minimal responses or changing the topic). In contrast, confrontation ruptures occur when the client directly expresses negative feelings about the therapist and/or therapy. Comparisons of a brief relational treatment informed by Safran and Muran’s work to cognitive-behavioral and psychodynamic treatments found greater retention in the relational treatment (Muran et al., 2009). This suggests that use of rupture repair strategies can positively affect factors related to the alliance, such as dropout.

Building on Safran and Muran’s work, others are examining whether rupture repair strategies can enhance existing treatments such as cognitive therapy for depression (Constantino et al., 2008) and interpersonal-dynamic informed treatment for depression (Crits-Christoph et al., 2006). In a meta-analysis of these studies, Safran et al. (2011) found that rupture-resolution training or supervision resulted in significantly better outcomes for patients whose therapists had received this training compared to those who had not. However, some studies of alliance-focused training have not yielded positive results. Newman et al. (2011) combined CBT for Generalized Anxiety Disorder with an interpersonal and emotional processing module that included rupture repair strategies, and found that this integrated treatment did no better with respect to outcome than the control condition of CBT combined with supportive listening. Smith-Hansen et al. (2011) found that a brief alliance training workshop did not improve therapist-rated alliance or client engagement compared to a wait-list control group. The fact that efforts to train therapists to address alliance ruptures do not consistently lead to better outcomes could indicate a need to identify additional alliance rupture resolution strategies that are more efficacious and/or easier for therapists to learn and implement successfully.

Research on the alliance has also found that therapists vary in their ability to build and maintain strong alliances. Several studies (e.g., Baldwin, Wampold, & Imel, 2007; Dinger et al. 2008, Zuroff et al. 2010) and one meta-analysis (Del Re et al., 2012) have found that therapist variability in the alliance is a statistically significant predictor of outcome, while patient variability is not. These findings suggest that therapists who are particularly successful at building strong alliances, regardless of theoretical orientation, may be a valuable resource. As Kazdin (2008) has observed, “Clinical experience, wisdom, novel hypotheses, and knowledge are often lost because they are not in a form that we codify and accumulate. We are letting knowledge from practice drip through the holes of a colander” (p. 155). Studying expert clinicians’ evaluations of alliance resolution strategies has the potential to yield insights and to be a step toward rectifying this “colander effect.”

This study applied both the behavioral-analytic model (Goldfried & D’Zurilla, 1969) and the Expert Consensus Guideline Series (e.g., Frances et al., 1997) method to identify rupture repair strategies over three study phases. The behavioral-analytic model was developed to gather and organize information about effective functioning within the context of specific situations. To date, the behavioral-analytic model has been used in over 35 studies and applied to examine effective responses to a variety of situations (e.g., Cavell & Kelley, 1992). Data for the present study were collected simultaneously with another study that assessed clinical consensus on the treatment of interpersonal problems in young adults (citation removed for blind review).

The first phase of the behavioral-analytic model is a situational analysis: a survey of meaningful, challenging clinical events. In Phase I, we surveyed a sample of practitioners to identify a set of alliance rupture situations representative of the types of ruptures therapists commonly encounter in their clinical work. Phase II of the behavioral-analytic model consists of response enumeration: in this phase, a second sample of therapists provided responses to address the rupture during the session, as well as responses to use in future sessions. Based on the Coding System of Therapeutic Focus (CSTF; Goldfried, Newman, & Hayes, 1989), we identified the clinical strategy reflected in each specific response. In the third phase of the behavioral-analytic model, response evaluation, peer-nominated expert clinicians rated the effectiveness of the strategies. For this final phase, we utilized the methods of The Expert Consensus Guideline Series (e.g., Frances et al., 1997), which was originally developed to establish consensus-based psychiatric treatment guidelines for a number of disorders.

The aim of the present study was to access the clinical knowledge of experienced clinicians to evaluate the resolution strategies that rank-and-file therapists use to address ruptures that they commonly encounter in their practices. We hoped that by sampling a broad sample of therapists—both psychologists and social workers from various theoretical orientations—we might identify novel resolution strategies that could inspire future research efforts. We also wanted to demonstrate that this methodology could serve as one way to integrate the experiences of practitioners into the research literature and foster greater collaboration between practitioners and researchers. Anticipating that many therapists might endorse supportive strategies such as empathic reflection in the immediate context of a rupture, we also assessed whether there was consensus about the use of rupture repair strategies in future sessions, after the rupture session. As a goal of this study was more focused on generating, rather than testing hypotheses, we did not have specific predictions about which strategies would be proposed, but we did explore whether therapist theoretical orientation was associated with endorsement of particular strategies.

Phase I: Situational Analysis

Method

Participants

We randomly selected 330 individuals from the membership directories of the National Association Social Workers (NASW) and of the American Psychological Association’s Divisions 29 (Psychotherapy) and 42 (Independent Practice). We mailed these individuals questionnaires that asked them to describe situations in which they had encountered alliance ruptures. Following an initial review of the responses, we found that situations describing withdrawal ruptures were under-represented relative to confrontation ruptures. To increase the likelihood that our sample was representative of the domain of withdrawal ruptures, another sample of 330 social workers and psychologists was selected using the same procedure described above, except that the questionnaires asked specifically for withdrawal ruptures. After the initial mailing and two reminder mailings to the 660 individuals, 133 responded (20% response rate), submitting 110 situations. Participants who failed to provide legible rupture situations, or who were no longer practicing therapy, were excluded, resulting in a final sample of 69 situations. Of these situations, psychologists submitted 51% (35 responses; 14 from Division 29, and 21 from Division 42), and social workers 49% (34 responses).

Phase I demographic information is detailed in Table 1, including demographics for the general memberships of Divisions 29 and 42 and NASW to allow for comparison. The demographics of the psychologist participants are comparable to the age and the experience of the Division 29 (American Psychological Association, 2008a) and 42 memberships (American Psychological Association, 2008b). Our social work participants were slightly older and more experienced than the average NASW member (Weismiller, Whitaker, & Smith, 2005). Participants also indicated the degree (%) to which a specific theoretical orientation informed their clinical work. The largest number of participants endorsed cognitive-behavioral (78%) and psychodynamic (70%) orientations. Most therapists identified as being strongly influenced by multiple orientations: Only 36% of participants indicated that over half their clinical work was informed by just one orientation. We did not ascertain to what extent participants were integrating different orientations at the theoretical level, or using different strategies following an eclectic approach.

Table 1.

Demographic Characteristics of Study Participants

	Study Participants
Variable	Phase I (n = 69)	Phase II (n = 177)	Phase III (n = 134)
Mean age in years (SD)	57.8 (8.8)	59.2 (9.0)	55.4 (8.9)
Division 29 Membership: 63.1 (11.)	60.6 (7.7)	61.4 (10.3)	61.6 (11.3)
Division 42 Membership: 61.8 (10.7)	58.8 (9.3)	60.1 (11.0)	59.8 (10.8)
NASW Membership: Median = 51 years	Median = 56 years	Median = 60 years	Median = 57.5 years
Mean years of experience as a therapist (SD)	28.8 (9.4)	30.2 (9.6)	26.3 (9.2)
Division 29 Membership: 29.7 (11.2)	32.78 (8.6)	35.6 (9.5)	26.8 (10.9)
Division 42 Membership: 28.5 (10.9)	29.4 (8.4)	30.6 (9.6)	30.6 (9.6)
NASW Membership: Median = 16 years	Median = 27.5 years	Median = 30 years	Median = 24.5 years
Gender (%)
Women	47.8	50.6	61.2
Men	52.2	49.4	34.3
Race (%)
White/European American	95.7	92.7	91.0
Black/African American	2.9	1.1	1.5
Latino/Latina/Hispanic American	0.0	1.1	2.2
Asian American	0.0	1.1	0.7
Native American	0.0	0.0	0.0
Multiracial/multi-ethnic	1.4	1.1	0.0
Other	0.0	1.1	0.0
Number of states representeda	28	41	31
Profession (%)
Psychologist	48.5	52.0	55.2
Social worker	51.5	48.0	44.8
Highest academic degree (%)
Master’s in social work	43.5	39.5	30.6
Doctorate in social work	2.9	2.8	0.7
PhD in social work	2.9	3.4	6.7
Doctorate in education	1.4	4.0	2.2
PhD in clinical psychology	30.4	29.9	41.8
PhD in counseling psychology	11.6	1.2	9.0
PsyD	4.3	5.6	4.5
Other	1.4	4.0	4.5
	Study Participants
Variable	Phase I (n = 69)	Phase II (n = 177)	Phase III (n = 134)
Hours spent providing therapy each week (%)
Fewer than 10 hr	13.0	19.8	25.4
10–15 hr	15.9	15.3	15.7
15–20 hr	18.8	15.8	13.4
20–25 hr	8.7	14.7	20.1
25–30 h	15.9	16.4	9.7
30–35 hr	17.4	9.0	6.0
More than 35 hr	8.7	7.3	6.7
Extent of guidance by theoretical orientationb
Psychodynamic	39.8	25.3	25.0
Cognitive-behavioral	35.1	40.4	37.0
Experiential/humanistic/client-centered	25.8	12.5	13.3
Family/systems	21.0	13.6	16.0
Other	21.3	22.0	8.9

^aParticipants’ state of residence.

^bParticipants were asked to indicate the degree to which each theoretical orientation guided their work with clients as a percentage (e.g., 100% one orientation’ or 40% one orientation, 40% a second orientation, and 20% a third orientation). Mean percentages are presented here.

Measures

Participants were instructed to think about a rupture that occurred with an adult client in individual outpatient therapy and to describe the rupture. We asked participants to provide the client’s demographics and presenting problems, the context in which the rupture occurred (e.g., stage of therapy), and their own demographic information.

Coding

After each of the situations was reviewed and edited for clarity by an advanced graduate student (Author A), two advanced clinical psychology doctoral students (Author A and Author B) independently used Safran and Muran’s (2000) definitions of withdrawal and confrontation ruptures to code each situation to reflect the degree to which the client’s behavior depicted a withdrawal, a confrontation, or a mixture of the two. The coders also assigned each rupture to a subcategory. The three withdrawal subcategories included 1) shutdown (e.g., silence), 2) change of focus, and 3) failure to complete between-session assignments or to attend sessions. The three confrontation subcategories included 1) problems with the quality of the bond, 2) problems with collaborating on tasks, and 3) problems with agreeing on or achieving goals. This secondary coding was conducted to increase the likelihood that the domains of withdrawal and confrontation ruptures would be represented. Reliability was calculated for the two types of ruptures and reflected adequate agreement overall (kappa = 0.68). There was complete agreement on 56 (81%) of the 69 situations.

Results

For the next study phase, six situations representing the three withdrawal subtypes and the three confrontation subtypes were selected from the pool of 56 situations on which there was complete interrater agreement. Selection of the six situations was made by consensus of the three authors about which situations would represent the rupture subtypes well, were sufficiently clear to be easily understood by participants in the next phase of the study, and, in our judgement, were types of situations therapists commonly encounter in their clinical work. The situations were edited to ensure that characteristics such as gender, race, and stage of therapy were equally balanced among the sample of situations. The six representative situations that we selected appear in Table 2.

Table 2.

Clinical Situations Involving Ruptures in the Therapeutic Alliance

Rupture Type	Background	Situation
1. Confrontation – Bond	This client is a 35-year-old, white woman who was referred to therapy after her husband left her. This situation occurs during the beginning stage of therapy and I have a moderately good relationship with the client.	During session, the client starts to argue with the therapist about the length of the therapy session. She tells the therapist that if he “really cared” that he would allow the extra time she deems appropriate, considering everything she has been through.
2. Confrontation – Tasks	This client is a divorced, 55-year-old, white man, who is in therapy to address his depressed mood related to his estrangement from his children. He also describes that he felt rejected by his parents and siblings during his childhood. This situation occurs during the middle stage of therapy, and I have a strong relationship with the client.	In session, the patient reports that he had not heard from any of his adult children, despite strong and appropriate efforts on his part to establish good relationships. The situation has led to increased depressiveness and an angry, paranoid attitude towards his ex-wife, whom he believes has turned his children against him. When the therapist attempts to link his present reaction to his feelings of rejection by his parents and siblings as a child, the client angrily turns on the therapist, saying, “I don’t want to hear that. You always say that.”
3. Withdrawal - Shutdown	This client is a 54-year-old, white woman. She is in treatment due to her extreme frustration with her marriage. Her irritation centers on her husband’s failure to respond to her with “appropriate” sensitivity. This situation occurs during the beginning stage of therapy, and I am in the process of establishing a relationship with the client.	In session, the client repeats the same complaints and criticisms of her husband that she has detailed in each of our three prior sessions. After listening to her for a while, I ask her if she has ever experienced other situations where she found that people were unable to meet the expectations she had for them, or whether anyone has ever commented that she has high expectations. She replies sharply, “No, it’s not me who has the problem. It has to be him. It can’t be me.” Then she begins to cry, and refuses to discuss the situation.
4. Confrontation – Goals	The client is a 38-year-old, Latino man who has been in therapy for one year to address issues concerning poor self-confidence, periodic job loss, and fears about starting a new job. He feels emasculated because he cannot provide for his family and his wife is now the primary wage earner. This situation occurs during the middle stage of therapy and I have a moderately good relationship with the client.	During session, the client mentions that his fear and avoidance of taking a new job has grown to the point that he has resigned himself to staying home as a “house-husband.” He tells me that he is frustrated about the lack of progress on this issue. He then questions whether I am genuinely interested in him and his welfare.
5. Withdrawal – Change of Focus	This client is a 62-year-old, white man. He was pressured by his wife to come for therapy due to his irritability toward her, manifested in critical, confrontational communications directed toward her. This situation occurs during the beginning stage of therapy and I am in the process of establishing a relationship with the client. I am a female therapist.	In the first session, I ask the client about his therapy goals. He responds by effusively complimenting my intelligence, education, and expertise. Each time I return the focus of our discussion to the therapy goals, he resumes complimenting my abilities.
6. Withdrawal – Attendance and Completion of Assignments	This client is a 50-year-old, African American woman. She is in therapy to treat chronic depression. She has been in therapy a number of times prior to seeing me. This situation occurs during the middle stages of therapy and I have a strong relationship with the client.	Although the client has been attending sessions regularly, her attendance starts to become sporadic. In the session following one of the missed sessions, I point out that she has been missing sessions every other week. I further say that I wonder whether this pattern might be a sign that weekly sessions aren’t appropriate for her. In response, she becomes defensive and somewhat angry at me. She tells me that she does need the weekly support.

Phase II – Response Enumeration

Method

Participants

A second independent sample of social workers (100 NASW members) and psychologists (38 from Division 29 and 62 from Division 42) was randomly selected from the respective membership directories and mailed Phase II questionnaires.

In total, 177 individuals generated responses (described further below under Measures) to the six situations. An average of 30 participants responded to each situation. Each participant generated an average of 7.6 responses (SD = 2.3) per situation, yielding an average of 225 individual responses (SD = 33.0) per situation. Demographic information appears in Table 1. Similar to Phase I, the Phase II psychologist participants were comparable in age to the membership of their respective divisions, while social worker participants were on average older and more experienced than the NASW membership. As in Phase I, while these therapists indicated that psychodynamic and CBT orientations guided their clinical work the most, only 34% of participants indicated that any one orientation contributed more than 50% to their clinical work.

Measures

Participants were asked to review a description of one of the six situations identified during Phase I, and to indicate how they would advise a colleague seeking consultation to respond to this rupture 1) during the session in which the rupture occurred, and 2) in a future therapy session. Each participant could provide up to 10 responses in total: five responses indicating how they would respond during the current session, and five responses indicating how they would respond in a future session. Participants also rated how frequently they encountered clinical situations similar to the one described in the questionnaire.

Classification

Given the large number of responses generated, the overlap between some responses that expressed the same ideas using slightly different wording, and the use of terms specific to particular theoretical orientations, steps were taken to distill this large number of responses into a manageable number of strategies that could be easily rated by a new sample of participants in Phase III of the study. Two advanced clinical psychology graduate students and one experienced psychologist (Authors A, B, and C) sorted the individual responses into classifications taken from the Coding System of Therapeutic Focus (CSTF; Goldfried et al., 1989). The CSTF, which has been used in a number of studies (e.g., Castonguay et al., 1996), classifies therapist interventions into one of 17 types of interventions (e.g., providing support, focusing on emotions, focusing on expected/imagined response of another person). The CSTF uses theoretically neutral language and emphasizes the function of interventions as opposed to the extent to which the techniques look similar on the surface, because techniques that appear different (e.g., two chair technique vs. a thought record) could serve the same function (e.g., increasing patient awareness). This step of sorting into CSTF categories helped to highlight the functions of the interventions that were proposed in the individual responses. For example, responses that described techniques to enhance the bond (e.g., communication that the therapist cares, validation of the client’s feelings, etc.) were classified using the CSTF as “bond.”

As a next step, we translated individual responses into strategies that corresponded to the CSTF classifications into which the responses had been grouped. The goal was to capture the functions of the individual responses in theoretically neutral language, while also reducing the large number of individual responses into a manageable number of non-redundant strategies. For example, responses that were sorted into the CSTF classification of “bond” were translated into strategies such as “Explore whether the client feels rejected by the therapist’s suggestion to meet less frequently.” Responses directing the therapist to explore the transference were sorted into the CSTF classification of “pattern,” and were then translated into theoretically neutral language, such as “Explore whether the patterns in the client’s relationship with the therapist are similar to patterns in her relationship with her ex-husband.” All three authors evaluated all classifications and translations and any disagreements were resolved through consensus.

Results

The classification and translation process described above yielded between 35 and 45 strategies for each of the six situations (M = 41, SD = 4). The majority of clinicians (76.7%) indicated that these six situations occurred somewhat often to very often.

Phase III: Response Evaluation

Method

Participants

For this final phase, we utilized the methods of The Expert Consensus Guideline Series (e.g., Frances et al., 1997), which was originally developed to establish consensus-based psychiatric treatment guidelines for a number of disorders. In this approach, experts review clinical situations and rate the relative effectiveness of treatment options to address these situations. The ratings are then tallied and presented as a synthesis of expert opinion. We recruited our sample of experts using a variation on an existing method (Goldfried, Raue, & Castonguay, 1998). Individuals who oversee clinical training in clinical psychology and social work programs were invited via email to nominate expert clinicians. For the purposes of this study, “experts” were defined as therapists who had been practicing for a minimum of five years post-degree to whom training directors would refer close family and friends. We consulted directories of training and field directors in psychology and social work and contacted every director whose name, role, and individual email address we could identify from the directory, and invited them to submit nominations. In total, we requested nominations from the following: the training directors of 142 clinical psychology Ph.D. and Psy.D. programs listed in the American Psychological Association’s guide to Graduate Studies in Clinical Psychology (APA, 2006); the training directors at 383 psychology internship sites appearing in the online directory of the Association of Psychology Postdoctoral and Internship Centers (2006); and the directors of field education for 159 master’s level social work programs listed in the Directory of College and Universities With Accredited Social Work Degree Programs (Counsel on Social Work Education, 2003). These individuals were also invited to forward the email to colleagues who could also submit nominations. If the training directors failed to respond after three attempts, up to two additional faculty members who taught and/or supervised clinical practice in that program were contacted to provide nominations.

In total, 186 people provided nominations (48 nominators were from psychology graduate programs, 52 from internship programs, and 86 from social work programs); these 186 individuals nominated a total of 482 expert clinicians. Twenty nominees were excluded due to ineligibility or lack of contact information, resulting in a sample of 462 peer-nominated clinicians. This sample was divided into six groups of 77 clinicians (43 psychologists and 34 social workers) per group. Each group was invited to rate strategies for one of the six situations.

A total of 134 peer-nominated expert clinicians submitted effectiveness ratings for the six situations for an overall response rate of 29%. The response rates varied among the situations: Situation 1 (N = 25; 32.5%); Situation 2 (N = 22; 28.6%); Situation 3 (N = 21; 27.3%); Situation 4 (N = 21; 27.3%); Situation 5 (N = 22; 28.6%); Situation 6 (N = 23; 29.9%).

Demographic information for the Phase III experts is presented in Table 1. Similar to Phases I and II, the Phase III Divisions 29 and 42 member participants were comparable in age and experience to the general membership of their respective divisions, while the NASW participants were on average older and more experienced than NASW’s general membership (Weismiller et al., 2005). No significant differences were found across the situations with respect to age, gender, profession (social worker vs. psychologist), years of experience, theoretical orientation, or clinical activity.

Measures

The study materials contained 1) a client description, 2) the rupture situation, and 3) the clinical strategies generated to address the rupture. The 9 point rating scale (ranging from “usually ineffective” to “extremely effective”) and instructions were adapted from the Expert Consensus Guideline Series (e.g., Foa et al., 1999). Specifically, we asked the peer-nominated experts to review the situations and responses and to evaluate the effectiveness of the response 1) if used during the rupture session, and 2) if used in a future session.

Results

Following the Expert Consensus Guideline Series, consensus was evaluated using a chi-square test of the distribution of scores across the three ranges of effectiveness: ineffective (scores of 1–3), somewhat effective (4–6), and usually effective (7–9). A significant result (p ≤ .05) indicated that the distribution was unlikely to occur by chance, and was interpreted as evidence of consensus. Consensus was achieved for the majority of during-session (55.6%) and future-session strategies (77.7%) (see Table 3). Peer-nominated experts reached consensus for all strategies within the “extremely effective” and “effective” ranges (described below) except for one during-session and one future-session strategy.

Table 3.

Distribution of Effectiveness Labels and Degree of Consensus Across the Six Situations

		Strategies During Session					Strategies in Future Sessions
Situation	Total Strategies (n)	Usually Ineffective	Somewhat Effective	Usually Effective	Extremely Effective	Consensus Reached	Usually Ineffective	Somewhat Effective	Usually Effective	Extremely Effective	Consensus Reached
1	35	22 (62.9%)	12 (34.3%)	0 (0.00%)	1 (2.9%)	22 (62.9%)	3 (8.6%)	30 (85.7%)	2 (5.7%)	0 (0.0%)	25 (71.4%)
2	44	12 (27.3%)	30 (68.2%)	1 (2.3%)	1 (2.3%)	20 (45.5%)	1 (2.3%)	31 (70.5%)	12 (27.3%)	0 (0.0%)	40 (90.9%)
3	40	19 (47.5%)	18 (45.0%)	1 (2.5%)	2 (5.0%)	28 (70.0%)	1 (2.5%)	35 (87.5%)	4 (10%)	1 (2.5%)	34 (85.0%)
4	38	13 (34.2%)	24 (63.2%)	1 (2.6%)	0 (0.0%)	14 (36.8%)	4 (10.5%)	34 (89.5%)	0 (0.0%)	0 (0.0%)	22 (57.9%)
5	44	23 (52.3%)	21 (47.7%)	0 (0.0%)	0 (0.0%)	20 (45.5%)	4 (9.1%)	35 (79.5%)	5 (11.4%)	0 (0.0%)	30 (68.2%)
6	45	9 (20.0%)	20 (44.4%)	6 (13.3%)	10 (22.2%)	33 (73.3%)	1 (2.2%)	33 (73.3%)	11 (24.4%)	1 (2.2%)	40 (88.9%)
Mean %		39.8%	50.8%	3.5%	5.4%	55.7%	5.7%	80.5%	13.8%	0.8%	77.7%

Effectiveness Ratings

Following the Expert Consensus Guideline Series, we identified effectiveness ratings for each strategy. “Extremely effective” strategies are those rated “9” by at least half of the expert clinicians. The ratings of effective, somewhat effective, or ineffective were determined by the 95% confidence interval (CI) of each strategy’s mean score. Specifically, “usually effective” strategies are those in which the entire CI falls at or above a score of 6.5 or greater; “somewhat effective” strategies are those strategies in which the CI falls between 3.5 and 6.5; “usually ineffective” strategies are those in which part of the CI falls below 3.5.

We screened the effectiveness ratings for outliers using boxplots. Outliers were more common in these ratings of alliance rupture situations than in the ratings of strategies to address interpersonal issues (citation removed for blind review). Across the six situations, 45% of the expert participants made at least one extreme rating. No specific therapist characteristics could be identified to differentiate these participants from others. Ultimately, since there were no a priori assumptions regarding the rating distributions, the outliers were included in the final analyses to reflect the diversity of ratings among our peer-nominated clinicians.

The distribution of effectiveness ratings is presented in Table 3. Using paired t test analyses and a corrected p-level of .001, we found that 36.6% of strategies were rated as significantly more effective in future sessions than during the rupture session. A higher percentage of strategies received extreme ratings (usually ineffective or extremely effective) for use during the session versus in future sessions (during-session = 44.7%, future-session 6.5%) (X², 246, 1, p < .003).

Categorization of strategies

In an effort to summarize our results in a manner that might be useful for informing clinical practice, we developed broader categories that reflected the therapeutic function of the strategies. The 16 categories appear in Table 4. Once the groupings were defined, the first two co-authors independently assigned each strategy to a specific category. An interrater reliability of .79 (Cohen’s kappa) was achieved. The third author resolved any coding discrepancies.

Table 4.

Category Descriptions and Examples of Corresponding Strategies

Category	Description
Assessment	Diagnostic assessment of the client’s symptoms or capacities. Examples: Assess the severity of the client’s depressive symptoms. Explore whether the client would benefit from an evaluation for psychotropic medication.
Gather Information	Interventions that aim to gather information about the client and his or her circumstances. Examples: Explore the meaning of the client’s relationship with her husband and his rejection of the client. Explore the history of the client’s childhood experiences in her family of origin.
Emotion	Interventions focused on clarifying and exploring the client’s emotional experience. Examples: Explore the link between the client’s anger and his depression. Explore the client’s feelings of guilt.
Articulate Need	Interventions that encourage a client to identify and/or communicate his or her needs and/or goals. Examples: Explore what goals the client wants to target in therapy. Encourage the client to identify what he needs from the therapist and identify ways those needs could be met.
Validate	Interventions that validate and legitimize the client’s position on an issue. Examples: Explore and empathize with the client’s feelings of anger toward the therapist. Validate the client’s right to make her own choices about her treatment.
Self-disclosure of therapist reaction	Interventions that involve the therapist communicating his/her reaction to the client and/or to the therapist-client interaction. Examples: Communicate to the client that the therapist cares about her. Communicate that the therapist is not perfect and will make mistakes and potentially disappoint the client.
Perspective/ Insight	Interventions that aim to help the client to gain a new perspective on his or her situation. Examples: Help the client to understand the relationship between his thoughts and feelings. Help the client to understand his wife’s perspective.
Foster Hope	Interventions that aim to enhance the client’s belief that his or her life can improve. Examples: Communicate that there is hope for a better life in the future. Explore and validate the client’s strengths and achievements.
Coping Strategies	Interventions that aim to help clients manage and cope with painful feelings and/or unpleasant thoughts. Examples: Focus on ways the client can manage and cope with her feelings of anxiety. Encourage the client to increase her activity level to alleviate her depression.
Practical Problem Solving	Interventions that seek to motivate the client to take practical, concrete steps to improve his or her situation. Examples: Help the client with time management so that she can find enough time for herself and her family. Help the client develop a specific plan for finding work.
Change Interaction	Interventions that seek to change the way the client interacts with a loved one. Examples: Explore whether the client is willing to change his attitude toward his ex-wife. Refer the client and his wife to a couple’s therapist.
Patterns	Interventions that highlight patterns in the client’s relationships and/or patterns of behavior, thoughts, or emotions. Examples: Explore the link between the client’s current feelings and the client’s unresolved feelings from her childhood experiences. Explore the link between the client’s anger and her depression.
Patterns-Therapist	Interventions that highlight patterns in the client’s relationship with the therapist and/or explore whether these patterns are present in the client’s other relationships. Examples: Highlight that the client’s reaction to the therapist’s comment provides an opportunity to understand how to address her problems with her husband. Explore how the client’s feelings toward the therapist are similar to her feelings toward others.
Rupture – Bond	Interventions that include those used to identify and explore the events surrounding a rupture in the emotional bond between the client and therapist. Examples: Explore and empathize with the client’s feelings of anger toward the therapist. Explore what about the therapist’s comment upset the client.
Communicate Limits	The therapist articulates his or her personal or professional limits regarding treatment parameters. Examples: Review the cancellation and no-show policy with the client. Educate the client about therapy process.
Rupture – Goals and Tasks	Interventions that focus on reaching agreement on treatment goals or the use of specific therapeutic methods following disagreement or tension concerning tasks or goals. Examples: Explore what goals the client wants to target in therapy and how these goals can be addressed in each session. Explore the client’s expectations about therapy.

After grouping together strategies that belonged to the same category, we conducted the same analyses at the category level that we that we had employed in the analysis of the individual strategies. First, we determined consensus at the category level. In contrast to the individual strategy findings, consensus was reached for all of the during-session and future-session categories. Within each situation, we assigned effectiveness labels based on the means and CIs of the strategies included in that category. For Situations 1, 4, and 5, no categories received a “usually effective” rating. Consistent with the Expert Consensus Guideline Series, categories at the high end of the “somewhat effective” range, whose confidence intervals overlapped with the “usually effective” range, were designated as “high, somewhat effective.”

Similar to the analysis at the strategy level, over half (58.2%) of categories received significantly higher ratings when used in future sessions versus during the session. The reverse pattern was observed two times for the Validate category: for Situations 5 and 6, Validate strategies received significantly higher effectiveness ratings during session than in future sessions.

The categories that received effectiveness ratings in the “usually effective”, “high, somewhat effective”, or “ineffective” ranges appear in Table 5. Addressing the rupture in the bond (Rupture-Bond) and employing validation strategies (Validate) were rated “effective” or “high, somewhat effective” most often to address ruptures occurring during the session. For future sessions, other strategies emerged as effective, including Coping, Gathering Information, and use of Emotion. Changing the Interaction, Patterns, and Assessment were rated as “ineffective” to use during the session for multiple situations.

Table 5.

Categories Rated as Usually Ineffective or Usually Effective Based on Means and CIs

	Usually Ineffective		Usually Effective
Situation	During session	Future sessions	During Session	Future sessions
1. Confrontation - Bond	Fostering Hope Assessment, Rupture – Tasks and Goals	None	Rupture-Bondb, Validateb	Rupture-Bondb, Gather Information, Patterns-Therapist, Articulation of Need, Validate
2. Confrontation – Tasks	Patterns	None	Rupture-Bondb	Emotion, Coping, Perspective/Insight, Gathering Information
3. Withdrawal - Shutdown	Perspective/Insight, Patterns, Assessment, Fostering Hope, Coping, Interaction	None	Rupture-Bond	Rupture-Bond, Emotion
4. Confrontation – Goals	Interaction, Perspective/Insight, Problem Solving	Interaction	Emotionb, Rupture-Bondb	Emotionb, Copingb, Problem Solvingb, Rupture-Goals/Taskb
5. Withdrawal – Change of Focus	Communicate Limits, Patterns-Therapist, Assessment, Patterns, Rupture-Bond, Interaction, Articulation of Need, Self-Disclosure, Interaction	Communicate Limits, Rupture-Bond	Validateb	Coping, Patterns
6. Withdrawal – Attendance and Completion of Assignments	Articulation of Need	None	Validate c, Rupture-Bond, Rupture-Tasks/Goals	Problem Solving, Patterns-Therapist, Validate, Coping, Rupture-Goals/Tasks

^aWhere two or more categories received the same effectiveness designation, they appear in order of descending mean score.

^bHigh, somewhat effective

^cExtremely Effective

Effectiveness Ratings – Rupture Type

We also examined category effectiveness ratings for each rupture type (withdrawal vs. confrontation). For both withdrawal and confrontation ruptures, Rupture-Bond and Validate were the highest rated strategies for use during the session (highly effective for withdrawal ruptures; high, somewhat effective for confrontation ruptures). In future sessions, Coping and Patterns were rated as usually effective for both withdrawal and confrontation ruptures. In addition, Problem-solving and Rupture-Goals/Tasks were rated as usually effective in future sessions following a withdrawal, and Emotion was rated as usually effective in future sessions following a confrontation.

For both withdrawal and confrontation ruptures, Communicate Limits was rated as an ineffective strategy to use during the rupture session. In addition, Self-Disclosure and Interaction were rated as ineffective during a withdrawal rupture, and Hope was rated as ineffective during a confrontation rupture. For both withdrawal and confrontation ruptures, no strategies were rated as usually ineffective in future sessions. A comparison of the ratings for the two rupture types found that confrontation rupture strategies received significantly lower effectiveness ratings than withdrawal rupture strategies across the situations both during the session, t(5413) = −6.69, p < .001, and for future-sessions, t(5278) = −5.14, p < .02.

Expert Participant Theoretical Orientation

We conducted exploratory analyses to evaluate the association between category effectiveness and expert participant theoretical orientation. For these analyses, we collapsed the data across the six situations because there were no significant differences in expert participant theoretical orientation across the situations. As therapists were able to endorse multiple orientations, we correlated the extent to which participants endorsed a particular orientation (e.g., being 50% psychodynamic) and ratings of particular strategies. To account for multiple correlations, we employed a Bonferroni-adjusted alpha level of .002. Here we present only correlations that were both statistically significant (p<.002) and at least small-to-moderate (r > .20) in order to focus on the strongest relationships. Behavioral: The more expert participants endorsed a behavioral orientation, the lower they rated the effectiveness of using Pattern strategies in future sessions, r(189) = −.25. CBT: The more expert participants endorsed CBT, the greater they rated the effectiveness of the use of strategies such as Hope, r(259) = .20 and Coping strategies, r(420) = .23 and the lower their ratings for Patterns-Therapist, r(342) = −.24 in future sessions. Experiential: The more expert participants endorsed an experiential influence, the higher their effectiveness ratings for the use of Validate during, r(253) = .21, and in future-sessions, r(245) = .24. Psychodynamic: The more an expert participant endorsed a psychodynamic orientation, the lower the effectiveness ratings for strategies focusing on Rupture-Goals/Task, r(685) = −.20 during session and a trend in future sessions, r(674) =.10, p = .013 as well. In future sessions, the more that dynamic theory influenced practice, the higher the effectiveness ratings for Patterns, r(189) =.28 and Patterns-Therapist, r(342) = .30, and the lower the ratings for use of Coping, r(424) = −.20, Problem solving, r(56) = −.46, and Communicate Limits, r(108) = −.35.

Discussion

These results suggest that a group of peer-nominated psychologists and social workers of various theoretical orientations can reach some degree of consensus on rupture-repair strategies. Overall, the strategies they considered effective to use during the rupture session emphasized exploring the client’s experience of the rupture (Rupture – Bond), and acknowledging the client’s perspective (Validate). Strategies rated less effective to use during the rupture included interpretation (Patterns), changing interpersonal interactions (Interaction), and cognitive restructuring (Perspective-Insight). In future sessions, participants rated coping strategies (Coping), gathering information (Gather Information), and focusing on emotional experience (Emotion) as effective. In general, these therapists rated more strategies effective to use in future sessions than during the rupture session. It is important to note that none of the strategies to address confrontation ruptures were rated in the “effective” range.

The one exception to the high endorsement of exploration of the client’s experience of the rupture (Rupture-Bond) was Situation 5: For this situation involving an avoidant client, the Rupture-Bond strategy was considered usually ineffective to use during the rupture and only somewhat effective to use during later sessions. This result may be due to the fact that only one response represented the Rupture-Bond category for this situation, and this specific response was not considered to be effective: “Ask the client whether he feels more comfortable focusing on the therapist than focusing on himself.” For this particular situation, validation strategies were rated highest for use during the session (e.g., “Explore topics the client is willing to discuss”) and both coping strategies (e.g., “Focus on ways the client can cope with and manage his feelings, particularly anger and frustration”) and strategies focused on patterns (e.g., “Explore other situations in which the client has been disappointed by others when they have not met his expectations”) were rated highest for use in future sessions.

Although the same sample was used to evaluate the effectiveness of strategies to address problems in young adults’ relationships with their parents (citation removed for blind review), generally therapists in this study provided more extreme ratings (both extremely effective and usually ineffective). This finding may reflect the nature of the situations: ruptures center on the therapist and therapy, and may tend to pull for more extreme responses.

The impact of ruptures on the therapist may also explain why participants submitted more confrontation rupture situations than withdrawals (61% confrontation vs. 19% withdrawal; 20% were a mix). Therapists may notice and remember confrontation ruptures more, and/or may find them more difficult to manage than withdrawal ruptures. This may also explain why participants failed to achieve consensus for any strategies within the “usually effective” range during a confrontation. Therapists may have difficulty with confrontation ruptures, and as a result they may agree more on what not to do (i.e., consensus on ineffective strategies) versus what to do (i.e., consensus on effective strategies). Consistent with this idea, Coutinho, Ribeiro, Hill and Safran (2011) found that during a confrontation rupture, therapists often feel less present and experience more guilt and feelings of incompetence, factors that could impair their effectiveness.

Conversely, the fact that fewer withdrawal ruptures were reported may indicate that therapists have greater difficulty identifying these ruptures. The small percentage of withdrawal situations submitted (19%) contrasts with results from studies employing observational coding, which have consistently found higher rates of withdrawal ruptures than confrontation ruptures (Eubanks-Carter, et al., 2010). It is also worth noting that of three withdrawal situations used in the present study, two of them (situations 3 and 6) also contain elements of confrontation (e.g., patient defensiveness and hostility). The difficulty we had obtaining examples of “pure” withdrawal from our participants may indicate that therapists need to learn how to identify the subtle signs of withdrawal ruptures. This is consistent with recommendations (e.g., Safran & Muran, 2000) that therapists engage in mindfulness practice to help them cultivate greater awareness in the therapeutic relationship. Strategies to resolve ruptures can only be useful if the clinician is aware that the rupture is happening.

However, another possibility that must be acknowledged is that our conceptualization of withdrawal may differ from how many therapists often think about alliance ruptures. Therapists may disagree about the extent to which clients’ avoidance behaviors in session are interpersonal (withdrawal ruptures) versus intrapsychic (e.g., resistance) phenomena. This possibility suggests that additional work at the theoretical level as well as empirical research would be beneficial to clarify what constitutes a withdrawal rupture.

These findings also suggest that use of change-oriented interventions during a rupture may be less effective than use of those strategies at other times, consistent with findings in cognitive (e.g., Castonguay et al., 1996) and psychodynamic therapies (e.g., Piper, et al., 1999). Specifically, strategies such as problem-solving, focusing on interpersonal consequences, generating hope, or making interpretations were rated less effective to use during a rupture than in future sessions. While these strategies may be effective in other contexts, using them during a rupture may lead the client to feel criticized or invalidated, in turn deepening the rupture.

The strategies participants rated as effective are consistent with parts of Safran and Muran’s rupture repair model (Safran & Muran, 2000). Three out of the four processes associated with their model were rated as effective for use during the rupture: 1) addressing the rupture (Rupture-Bond), 2) exploring the rupture experience (Rupture-Bond, Validation), and 3) exploring the avoidance (Rupture-Bond). The fourth process, articulation of need, was rated as usually effective only in future sessions for one of the confrontation situations, and as usually ineffective during the rupture for two withdrawal situations. It should also be noted that participants rated “self-disclosure of therapist reaction” as usually ineffective during a withdrawal session. This is in contrast to Safran and Muran’s approach, which encourages therapists to metacommunicate during a withdrawal rupture by disclosing their experience of the patient-therapist interaction, which in turn facilitates self-disclosure and self-assertion by the patient (see Safran et al., 2014).

The fact that participants endorsed strategies generally consistent with the literature may indicate some consensus between practicing clinicians and researchers. However, it should also be noted that this study did not generate any novel interventions. This could be partly due to the study’s methodology: methods to identify consensus highlight commonalities, and we realize in retrospect that they are not the best way to identify innovative interventions. In addition, our efforts to make the findings accessible by distilling responses into theoretically neutral strategies, and then grouping strategies into categories, inevitably led to the loss of some of the richness and specificity of the original strategies. Furthermore, our methodology relied on asking experts to rate strategies submitted by other clinicians. We did not ask those experts to submit their own strategies, which may have been a better way to identify strategies that are novel and likely to be extremely effective. Along these lines, it may be that our experts did not endorse any of the strategies for resolving confrontation ruptures as effective because they preferred other strategies that we did not include on our questionnaire. Another possible explanation for the paucity of novel and extremely effective strategies is that our field needs to develop and disseminate more effective rupture resolution interventions. Did our participants agree on the importance of validating the client and addressing the rupture because they believed that these were the best strategies we currently have to use during a rupture? Or is noting that a rupture is happening and validating the patient a fallback position when one does not know what else to do?

While participants generally agreed on addressing the rupture and validating the client during the rupture, they endorsed a broader range of interventions for future sessions. Also, when thinking about future sessions, participants’ endorsements were often consistent with their theoretical orientations. For example, a psychodynamic orientation was correlated with a focus on patterns related to the therapist, which would encompass transference interpretations, a key psychodynamic technique. A CBT orientation was correlated with a focus on coping strategies, which is a common feature of CBT treatments. An experiential orientation was correlated with a focus on validation, which is a central element of humanistic therapy approaches (e.g., Rogers) that are part of the experiential tradition. These findings suggest that during a rupture, participants of all orientations focus on addressing the rupture and validating the client, but when they look ahead to future sessions, they may shift to the strategies of their theoretical orientations. An important question for future research would be to understand how therapists decide when to make this shift—how do they determine that the rupture has been at least partly resolved? This question points to the importance of therapists’ awareness of and responsiveness to the state of the alliance. The findings also suggest that an alliance rupture may be a prime opportunity for integration—a moment in therapy when skilled clinicians are particularly open and receptive to integrating alliance-focused strategies in order to address a rupture.

There are several limitations to note. Effectiveness ratings do not necessarily equal effective practice. Researchers will need to test these hypotheses using more rigorous methods to determine how use of these strategies and contextual factors, such as timing, affect outcomes. Another limitation is how consensus was determined. Specifically, we defined consensus as agreement through statistical methods as opposed to consensus through discussion, which might have yielded different results. Also, while we met the Expert Consensus Guideline Series criteria for achieving consensus, agreement on 55.7% of strategies does not indicate a very high degree of unanimity. Finally, while our peer-nominated therapists were considered to be experts by their peers, we did not directly measure their expertise through clinical outcomes.

Another limitation concerns the rupture situations. It is possible that including more contextual information might have altered both the responses generated and the effectiveness ratings. Also, these ruptures were from the therapists’ perspective and may miss critical aspects of the client’s perspective, in turn, limiting the comprehensiveness of the sample of situations. Finally, written descriptions fail to capture body language and vocal tone.

Although this sample was diverse in many respects, only five percent of participants endorsed being an ethnic or racial minority. While this statistic is consistent with other clinician-informed research (e.g., Norcross, Hedges, & Castle, 2002), it highlights the need for research to include the perspectives of individuals belonging to minority groups (APA Presidential Task Force on Evidence-Based Practice, 2006) when examining phenomena that rely heavily on interpersonal processes.

This study’s response rates were lower than other studies of practicing social workers (e.g., Arrington & Whitaker, 2008) and psychologists (e.g., Norcross et al., 2002). The response rate was low despite the explicit use of methods to increase clinician involvement, including obtaining endorsements from leaders in social work and psychology organizations, conducting extensive pilot testing of measures, and consulting with researchers involved in clinician-informed research. It may be that the demands of practice restrict the time practitioners have to devote to extra activities, an effect that is probably heightened among “expert” clinicians. In addition, the fact that we collected data simultaneously for two research projects increased participant burden and may have reduced recipients’ willingness to complete our questionnaires.

These findings suggest a number of avenues for future research. Consistent with on-going efforts (e.g., Safran et al., 2014) studies could examine whether coupling these rupture repair strategies with existing treatments improves outcome. Future studies could evaluate the impact of timing on effectiveness or apply this methodology to specific client populations. Specifically, the kind and frequency of ruptures may differ depending on the specific symptoms, acuity, and problems clients experience (Safran & Muran, 2000).

In addition, these findings have implications for therapist training. Specifically, they point to the importance of helping therapists to recognize ruptures as they are occurring. In particular, therapists may benefit from training to recognize more subtle ruptures such as withdrawals. The expert therapists in our sample responded differently to ruptures during the rupture session, when they endorsed a focus on empathy and validation, as compared to future sessions, when they tended to favor strategies congruent with their theoretical orientation. This finding suggests that in addition to recognizing ruptures, therapists also need to be able to track the therapeutic process and determine when a rupture has been resolved sufficiently to shift to different clinical strategies. Videotape analysis of rupture moments and awareness-oriented role plays in supervision as well as mindfulness practices have been suggested as ways to increase therapists’ abilities to detect and respond appropriately to ruptures (Eubanks-Carter, Muran, & Safran, 2015). There is a growing body of research demonstrating that alliance-focused training programs that incorporate these ideas lead to small but significant patient improvements (Safran et al., 2011). However, alliance-focused training approaches have not yet been widely disseminated. Greater dissemination of such training might increase therapists’ awareness of withdrawal ruptures, as well as their confidence in specific strategies for addressing confrontation ruptures.

Finally, this study builds on the findings of [citation removed for blind review] to further illustrate the use of a method to draw on expert clinician experience systematically. We hope that this study, and others like it, may help to promote clinician-informed research and the integration of research and practice.