Client and therapist views of contextual factors related to termination from psychotherapy: A comparison between unilateral and mutual terminators

Abstract

Contextual variables potentially influencing premature termination were examined. Clients (n = 83) and therapists (n = 35) provided parallel data on early working alliance, psychotherapy termination decision (unilateral vs. mutual), clients’ reasons for termination, and barriers to treatment participation. When clients unilaterally ended therapy, therapists were only partially aware of either the extent of clients’ perceived improvements or their dissatisfaction. When termination was mutually determined, there were no differences between client and therapist ratings of termination reasons. Although working alliance and barriers to treatment participation were rated as lower in the context of unilateral termination by clients and therapists, all clients rated the early alliance and barriers to treatment more highly than did therapists. Results have implications for understanding premature termination and suggest future research examining the utility of therapist feedback regarding contextual variables in terms of retaining clients in therapy.

Premature termination of treatment has been a perennial problem in psychotherapy. Up to 50% of clients discontinue psychological services prematurely (Barrett, Chua, Crits-Christoph, Gibbons, & Thompson, 2008; Swift, Callahan, & Levine, 2009; Wierzbicki & Pekarik, 1993), which undermines the potential benefits of treatment and reduces the cost-effectiveness of these services (Garfield, 1994; Ogrodniczuk, Joyce, & Piper, 2005; Pekarik, 1985a). Compared with clients who complete treatment, those who leave treatment prematurely tend to be less satisfied with services (Lebow, 1982), are less likely to have improved (Pekarik, 1986; Prinz & Miller, 1994; Saatsi, Hardy, & Cahill, 2007), and are more likely to be impaired and, therefore, more in need of services (Kazdin, Mazurick, & Siegel, 1994).

In order to intervene to prevent premature termination, we need to better understand why clients leave before their treatments are completed. Most research in this area has examined who leaves, focusing primarily on static client or therapist factors. Although few replicable results have been found, there is consistent evidence that premature termination is associated with socioeconomic disadvantage and non-White ethnicity (Wierzbicki & Pekarik, 1993; Williams, Ketring, & Salts, 2005). Closer examination of findings such as these reveals the possibility that the association with ethnicity can be largely accounted for by socioeconomic disadvantage (Garfield, 1994), which, in turn, may be at least partially explained by differences in client expectations for the duration of treatment (Pekarik, 1991; Pekarik & Stephenson, 1988; Pekarik & Wierzbicki, 1986). Thus, this line of evidence suggests that there may be considerable value in examining contextual factors potentially related to premature termination. In the present study, we examine three such factors: the reasons why clients terminate services, early treatment alliance, and possible barriers to clients’ involvement in therapy.

Premature termination has been defined a number of ways, including failing to attend a scheduled session, failing to complete a prescribed number of sessions, and making a unilateral decision to end treatment without agreement of the therapist (Wierzbicki & Pekarik, 1993). Pekarik (1985b) suggested that a unilateral decision on the part of the client to terminate best captures the construct of premature termination. This approach differentiates clients who unilaterally terminate from those who make a mutual decision with their therapist to end treatment. It also avoids the problem of defining premature termination as the failure to complete a prescribed number of sessions, because some clients achieve the necessary gains in functioning prior to the end of a set number of sessions. Defining premature termination according to the type of decision addresses the problem of appropriately classifying clients who meet their treatment goals with few therapy sessions as well as clients who may remain in therapy for a longer period of time but leave before their goals have been reached. Since Pekarik's suggestion, most researchers have used this operationalization (e.g., Callahan, Aubuchon-Endsley, Borja, & Swift, 2009; Chisolm, Crowther, & Ben-Porath, 1997; Keijsers, Kampman, & Hoogduin, 2001; Richmond, 1992; Smith, Subich, & Kalodner, 1995; Tryon & Kane, 1993).

A wealth of evidence indicates that obtaining data from both clients and therapists is necessary to understand the process of psychotherapy. Some perspective divergence between clients and therapists is expected, and a growing body of research documents that both similarities and differences in perspective can provide insight into the nature of client and therapist experiences in therapy (e.g., Reis & Brown, 1999; Tryon, Blackwell, & Hammel, 2007; Weiss, Rabinowitz, & Spiro, 1996). Accordingly, it is important to consider both client and therapist views in order to understand clients’ unilateral termination. For example, clients generally anticipate that they will require fewer sessions to address their problems than do their therapists (Garfield, 1994; Swift & Callahan, 2008), and therapists tend to overestimate treatment length and underestimate the number of clients who will terminate prematurely (Lowry & Ross, 1997; Mueller & Pekarik, 2000; Pekarik, 1992; Pekarik & Finney-Owen, 1987; Pulford, Adams, & Sheridan, 2008). Research has shown that any major discrepancy between client expectations and actual treatment content can lead to an increased risk of premature termination (Horenstein & Houston, 1976). Client-therapist divergences in estimations of problem severity also decrease the likelihood of mutual termination decisions (Corning, Malofeeva, & Bucchianeri, 2007). On the flipside, there is evidence that addressing clients’ role expectations prior to treatment can decrease the rate of dropout (e.g., Reis & Brown, 2006; Scamardo, Bobele, & Biever, 2004; Walitzer, Dermen, & Conners, 1999; Zwick & Attkisson, 1985). It is thought that this education may decrease unilateral termination by developing client expectations that are more congruent with what actually happens in therapy and more similar to the expectations therapists hold for clients (Reis & Brown, 2006; Swift & Callahan, 2008).

Reasons for Termination

Studies of client reasons for termination have shed much light (e.g., Bados, Balaguer, & Saldana, 2007; Hunsley, Aubry, Vestervelt, & Vito, 1999; Pekarik, 1983, 1992; Renk & Dinger, 2002; Roe, Dekel, Harel, & Fennig, 2006; Todd, Deane, & Bragdon, 2003). Although the proportion of clients reporting a given reason for leaving varies greatly across studies, common reasons are that they were satisfied with progress in treatment, they encountered circumstantial barriers (including any external obstacles such as difficulties with scheduling, child care complications, or financial barriers), or they were dissatisfied with the therapy or the therapist. In line with our emphasis on the importance of obtaining information from both client and therapist, research has shown that client and therapist perspectives on reasons for termination tend to diverge (e.g., Gager, 2004; Hunsley et al., 1999; Pekarik & Finney-Owen, 1987; Todd et al., 2003). Even when there is some general agreement on the reasons for termination, there are likely to be important differences in accounting for some termination factors. For example, Pekarik and Finney-Owen (1987) surveyed therapists and clients from community mental health clinics to compare the ratings of the primary reasons why clients left therapy. They asked therapists, in general, to list top reasons why clients leave and compared these with actual reasons given by a sample of clients. They found that therapists and clients tended to agree about positive reasons for termination (that the problem was solved or improved was endorsed by 39% of clients and 31% of therapists) and obstacles to treatment (environmental constraints was endorsed by 35% of clients and 37% of therapists). However, when the focus was on termination as a result of failed therapy, there was very little agreement between clients and therapists (resistance was endorsed by no clients and by 22% of therapists; dislike of therapy/therapist was endorsed by 26% of clients and 11% of therapists). Pulford et al. (2008) replicated these results in another adult outpatient sample.

Hunsley et al. (1999) also found that therapists and clients made different attributions about failed therapy. These researchers compared training clinic therapists’ reasons for client termination written in their final reports with reasons reported directly from interviews with former clients. Their results suggest that therapists were not aware of, or did not report, clients’ dissatisfaction with therapy as the primary reason for termination; no client was described by therapists as terminating because of dissatisfaction with therapy. However, 12% of clients reported that the fact that therapy made things worse for them was very important in their decision to end therapy. Almost half of these clients were described by therapists as terminating because they no longer had the time or interest to continue therapy. Fifteen percent of clients reported that the feeling that therapy was going nowhere was very important in their decision to end treatment. Based on therapists’ reports, 33% of these clients ended therapy because they had achieved many or all of their goals, and another 33% terminated because they no longer had the time or interest in continuing therapy. These results indicate that therapists were not accurate at detecting treatment failure, and the reasons for the failure, from the clients’ perspective. With respect to attributions for treatment success, among clients who were identified by therapists as leaving because they achieved their goals, 75% reported that this reason was important to their decision to leave. On the other hand, of clients who reported ending therapy because of having achieved their goals, only half were identified by therapists as having achieved their goals.

Todd et al. (2003) found a similar lack of concordance using a qualitative coding methodology to compare training clinic therapists’ reasons for client termination provided on routine clinic forms with reasons reported on similar forms given to clients at termination. Their results suggested only moderate overall agreement between therapist and client reasons (Cohen's κ = .43). More specifically, clients and therapists showed good agreement on client environmental and therapist environmental reasons, fair agreement on improvement reasons, and poor agreement on client negative and other reasons. Therapists were significantly more likely than clients to endorse improvement as a reason for termination, and clients were more likely to endorse client environmental and other reasons.

Both Hunsley et al. (1999) and Todd et al. (2003) utilized a file review methodology, using either client termination reports or standard clinic forms, to obtain therapist reasons for termination. Because of the possibility of the graduate student therapists trying to please supervisors, as well as other constraints on report writing and record keeping, actual therapist perceptions regarding reasons for termination might have been absent from the final report or clinic data. These authors’ results highlight the importance of examining both client and therapist perspectives on whether termination was unilateral or mutual. The methodologies used in this research to date have been either file review, general surveys about reasons for termination given to therapists or clients, or routine administrative forms used in clinic settings (Hunsley et al., 1999; Pekarik & Finney-Owen, 1987; Renk & Dinger, 2002; Todd et al., 2003). No study, to our knowledge, has used data from a research protocol that obtained parallel information from both members of the client-therapist dyad to examine specific reasons why the client terminated services and how perspective divergences may be related to unilateral termination.

Therapeutic Alliance

It is well established that therapeutic alliance, particularly agreement on therapeutic tasks, is strongly associated with psychotherapy outcome (e.g., Weerasekera, Linder, Greenberg, & Watson, 2001). In terms of predicting premature termination, although there have been inconsistencies in the research, working alliance (generally measured after the third treatment session) has been found to predict premature termination (Saatsi et al., 2007; Saltzman, Luetgert, Roth, Creaser, & Howard, 1976). In particular, problems with client-therapist agreement on therapeutic tasks have been found to be associated with ending treatment early (Tracey, 1986). Meta-analytic research on client and therapist ratings of working alliance suggest that client ratings, although higher than therapist ratings (d = .63), tend to be moderately positively correlated (r = .36) regardless of client disturbance, therapist experience, therapy length, alliance measure, or type of treatment (Tryon et al., 2007). To date, however, no research has examined how client-therapist congruence in ratings of the working alliance may differ as a function of mutual versus unilateral termination.

Barriers to Treatment Participation

Using a barriers-to-treatment model, Kazdin and colleagues have focused on the importance of therapy-specific factors in the search for causes of premature termination (Kazdin, Holland, Crowley, & Breton, 1997; Kazdin & Wassell, 2000). In this model, barriers include practical obstacles to participation in treatment (e.g., transportation difficulties, scheduling conflicts), perceptions of treatment as demanding, unhelpful, or irrelevant to the problems experienced by the client, and a poor therapeutic relationship with the therapist. Kazdin et al. (1997) found that consideration of these barriers added to the prediction of premature termination beyond the contribution of client characteristics (income, ethnicity, level of education), and that these findings were generally consistent across both parent and therapist perspectives for the reasons why families terminated therapy early. Large effect sizes were found for the contribution of the perceived relevance of treatment and stressors, and small and moderate effect sizes were found for the contribution of therapeutic relationship and treatment demands in discriminating between clients who completed treatment and those who left treatment prematurely. Interestingly, critical events that had occurred in a client's life while in treatment (e.g., moving, job loss, illness, change in marital status) were not found to contribute significantly to premature termination (Kazdin et al., 1997). Therefore, the present study examined whether these contextual factors found to be significant contributors to dropout in child and family therapy could also serve as useful ways to distinguish those adult clients who unilaterally terminated from those who mutually decided with their therapist to end treatment. We also examined barriers from both client and therapist perspectives; Kazdin et al. (1997) reported that the shared variance between parent-rated and therapist-rated barriers was only about 15%.

The Present Study and Hypotheses

In an effort to better understand unilateral termination, the goal of the present study was to examine the congruence in perspectives of client–therapist dyads regarding important contextual factors, including clients’ reasons for termination, working alliance, and barriers to treatment between two groups where (a) both client and therapist agreed that termination was a unilateral decision on the client's part or (b) both client and therapist agreed that termination was mutual. Based on previous research, several specific hypotheses were formulated:

1. In dyads where both client and therapist agreed that termination was a mutual decision, compared with dyads where both client and therapist agreed that termination was a unilateral decision on the client's part, both clients and therapists would rate (a) having accomplished therapy goals as more important to the termination decision and circumstantial and therapy-specific reasons as less important to the termination decision; (b) the early working alliance as stronger; and (c) barriers to treatment participation as fewer.

2. When termination decisions were mutual as opposed to unilateral, client-therapist perspectives would be more congruent regarding (a) reasons for termination, (b) quality of the early working alliance, and (c) barriers to treatment participation.

Method

Participants

One hundred fifty-five adult clients seeking individual psychological services from a university clinical psychology training clinic were initially recruited for a study on the process of engagement and termination from psychotherapy. The training clinic serves as a community clinic and operates on the basis of a sliding-fee scale. Of this sample, 39 completed initial measures for the study while they were waiting for services but never attended an initial treatment session and nine received services but did not complete the final set of measures at the end of treatment (either because they could not be reached by the researchers or were no longer interested in participating). Therefore, data were available for a total of 107 client participants who received psychotherapy and completed all study measures. On 12 different demographic measures, there was only one statistically significant difference between the 107 study participants and the 48 individuals who did not complete final measures. Study participants had a slightly higher level of education (“some university coursework”; M = 7.05, SD = 1.68) than noncompleters (“college graduation”; M = 6.41, SD = 2.09). This finding is consistent with literature suggesting that individuals with higher education are less likely to drop out of therapy (Garfield, 1994). A comparison of these two groups on level of psychological distress prior to therapy from both client (SCL-10) and therapist (GAF) perspectives revealed no significant differences.

Thirty-five therapist participants (28 women, seven men) provided therapy to between one and 11 client participants. Therapists were practicum students and interns in a doctoral program in clinical psychology and were supervised by registered psychologists.

To determine the type of termination decision, both clients and therapists were asked whether the decision to terminate therapy was the client's unilateral decision or whether it was based on a mutual agreement with the therapist that treatment goals had been met. Decisions to end therapy based on the failure of the client to attend sessions or to schedule subsequent appointments were considered to be unilateral decisions, and decisions to refer the client to other services for any reason (including when practicum students or interns were ending their training) were considered to be mutual decisions. Thirty-one client-therapist pairs agreed that termination was a unilateral decision on the client's part, and 52 client-therapist pairs agreed that termination was a mutual decision made by both client and therapist together. Twenty-four client-therapist dyads (22.4%) did not agree on the type of termination decision, thus indicating the importance of collecting data from both perspectives (12 clients reported unilateral termination whereas their therapists reported mutual agreement, and 12 clients reported mutual agreement whereas their therapists reported unilateral termination). Data from these dyads were not used for analyses reported in this study.

The mean age of the final sample of 83 client participants (19 males, 64 females) was 31.7 years (SD = 9.9; range, 17–60). In general, clients were highly educated (approximately 40% (34) had completed some university or college education and 46% (37) had attained at least a university undergraduate degree): 28.9% (24) were students, 37.3% (31) were employed full time, 20.5% (17) were unemployed, 12%(10) were employed part time, and 1.2% (1) were homemakers. Most participants reported lower to middle income (based on a median income of $29,000 for persons 15 years of age or older in the study region; Statistics Canada, 2001a), with 33.8% (28) of clients earning less than $10,000, 25.3% (21) between $10,000 and $20,000, 16.8% (14) between $20,000 and $30,000, 14.4% (12) between $30,000 and $40,000, and 9.6% (8) more than $40,000. Most participants reported their ethnic background as White (71, 85.5%); other ethnic groups represented in the sample included Black (3, 3.6%), Asian (5, 6%), Aboriginal (1, 1.2%), and other (3, 3.6%). This level of ethnic diversity is consistent with census data for the study region (Statistics Canada, 2001b). Client participants reported a range of presenting problems: symptoms of anxiety (30, 36%), depressive symptomatology (26, 31%), relationship problems (24, 29%), sexual abuse (9, 11%), and anger management problems (8, 10%). Other identified problems included attention-deficit disorder, loneliness, personality disorder, posttraumatic stress disorder (PTSD), problems with sexual functioning, and shyness.

The 83 participants were treated by 31 different therapists, who provided therapy to between one and 11 different participants. To determine whether there was a problem of dependence in the data, 56 comparisons of independent sample means were conducted on eight different therapist variables comparing seven groups of therapists who had seen 1, 2, 3, 4, 5, 7, 8, and 11 different clients, respectively. On only one variable were two groups significantly different at p <.05, suggesting no important differences across study variables among therapists who provided treatment to different numbers of client participants. Three main therapeutic approaches were used: cognitive-behavioral (58, 69.9%), experiential (13, 15.7%), and interpersonal (12, 14.5%). No statistically significant difference in type of therapeutic approach used was found between participants who unilaterally terminated therapy and participants who mutually terminated therapy, χ²(2, N = 83) = 3.25, ns.

Measures

Demographic data. Age, gender, education level, employment status, annual income, and cultural/ethnic background were requested before commencing treatment.

Symptom Checklist-10 (SCL-10; Nguyen, Attkisson, & Stegner, 1983). Derived from the Symptom Checklist-90 (SCL-90; Derogatis, Lipman, & Covi, 1973), this 10-item measure yields a single global score reflecting the extent of psychological distress. In the present study, the SCL-10 was used as a general measure of client self-rated symptomatology. Items were chosen from the three SCL-90 factors that were found to be most interpretable and accounted for a large proportion of the variance in an outpatient population: Depression (six items; e.g., How much were you distressed by feeling lonely?), somatization (two items; e.g., How much were you distressed by feeling weak in a part of your body?), and phobic anxiety (two items; e.g., How much were you distressed by feeling afraid in open spaces or on the street?). Items are rated on a 5-point scale of distress (0 = not at all, 4 = extremely). Nguyen et al. (1983) and Rosen et al. (2000) found a high level of internal consistency (Cronbach's a = .88), indicating that the instrument is an internally consistent measure. In the current study, the alpha values were .78 at pretherapy assessment and .85 at posttherapy assessment. Rosen et al. (2000) found the SCL-10 to show good convergent validity with the well-developed SCL-90 (r = .92) and discriminant validity with several other measures of symptom distress that aim to capture more specific aspects of distress, including the Beck Depression Inventory (BDI; r = .67), Beck Anxiety Inventory (BAI; r = .68), and Mississippi PTSD scale (r = .50). As well, pre-post change scores on the SCL-10 were examined in relation to those of other measures and were found to correlate highly, indicating good sensitivity to change.

Global Assessment of Functioning Scale (GAF; American Psychiatric Association, 1994). For the purpose of the present study, the GAF was used as an overall measure of psychological distress from the therapist's perspective. The GAF is a rating of overall psychological functioning based on a scale of 1 (most distressed) to 100 (least distressed) published in the Diagnostic and Statistical Manual of Mental

Disorders (fourth edition), designed to be completed by clinicians or researchers. The scale is divided into 10 equal 10-point intervals. For example, a score in the 51–60 range indicates moderate symptoms, and a score in the 61–70 range indicates mild symptoms. Endicott, Spitzer, Fleiss, and Cohen (1976) reported that five studies revealed intraclass correlation coefficients ranging from α = .61 to .91. Assessments of validity of the GAF have indicated moderate to high correlations with other independently rated measures of overall severity and sensitivity to treatment change.

Working Alliance Inventory (Short Form; WAI-S; Tracey & Kokotovic, 1989). This 12-item inventory was used to assess working alliance. Based on the original 36-item scale (Horvath & Greenberg, 1986, 1989), the WAI-S was developed using the four highest loading items from each of three subscales (Tasks, Goals, and Bond) and has equivalent factor structure and internal consistency (Tracey & Kokotovic, 1989). Busseri and Tyler (2003), in a sample of client-therapist pairs from 54 university counseling centers, found high correlations between WAI and WAI-S scores, comparable descriptive statistics, internal consistencies, and subscale intercorrelations within and across rater perspectives. Predictive validity estimates for WAI and WAI-S total scales were also very similar, supporting the interchange-ability of scores on the WAI and WAI-S. The measure is designed to be administered in the early stages of therapy, between the third and fifth sessions. Items are rated on a 7-point scale ranging from 1 (does not correspond at all) to 7 (corresponds exactly). Two negative items (4 and 10) were reverse-scored, and all scores were summed to provide a global rating of the working alliance. Both a client version and a therapist version of the WAI-S were used. In the current study, total scale score reliabilities (Cronbach's a) were .93 for the client version and .92 for the therapist version.

Barriers to Treatment Participation Scale (BTPS). This was developed by Kazdin et al. (1997) for use in the context of the outpatient treatment of children and families. Written in an interview format, it can be administered in person or by phone and is phrased so that both treatment dropouts and treatment completers can answer questions. It was modified for use in the context of adult treatment by changing 11 of 44 items and eliminating six, for a total of 38 items (Best, 2003). In the present study, two versions of the BTPS were completed: one by the client and the other by the therapist. Items are rated on a 5-point scale ranging from 1 (never a problem) to 5 (very often a problem) and cover four general areas: (a) stressors and obstacles that compete with treatment, (b) treatment demands and issues, (c) perceived relevance of treatment, and (d) relationship with the therapist. Kazdin et al. (1997) found that principal-components analysis revealed a single global scale factor. Therefore, in the current study, analyses were conducted with the global score. In the current study, global scale score reliability (Cronbach's α) was .87 for the client-completed version and .83 for the therapist-completed version. Kazdin et al. (1997) found that the measure showed convergent validity with other measures of participation in treatment. Evaluation of the scale revealed either no or low correlation and little shared variance between perceived barriers and critical events occurring while in therapy and family, parent, and child characteristics, thereby demonstrating discriminant validity.

Reasons for termination. Client and therapist perspectives on reasons for termination were assessed with a measure developed by Hunsley et al. (1999). The 10-item measure was developed based on possible reasons for termination found in the literature. Using a 4-point scale (not at all important to very important), clients and therapists were asked, after the final session, to rate the importance of each of 10 possible reasons in their decision to end therapy: (a) accomplished what you/he/she wanted to do in therapy, (b) could no longer fit time for therapy into schedule, (c) just lost interest in therapy, (d) no longer had money or insurance coverage to pay for therapy, (e) felt therapy was going nowhere so ended therapy, (f) felt therapy was making things worse so stopped, (g) weren't confident in therapist's ability to help, (h) uncomfortable talking about personal matters with therapist, (i) therapy didn't fit with ideas about what would be helpful, and (j) decided to go elsewhere for services. The 10 reasons were examined separately in analyses because the measure was not designed to yield a summary score.

Procedure

Data collection took place over 35 months. Client participants were assessed at three different times: (a) following a request for therapy and before the intake session (demographics, client self-rated symptomatology [SCL-10]), (b) after the third therapy session (working alliance [WAI-S], therapist-rated client functioning [GAF]), and (c) at the end of therapy (to assess retrospectively for contextual factors that may have influenced the decision to terminate, including reasons for termination and barriers to treatment participation [BTPS]); also assessed posttherapy were client self-rated symptomatology (SCL-10) and therapist-rated client functioning (GAF). All client data were obtained via structured telephone interview by a research assistant. Therapist data were obtained by structured self-report. For Time 3 assessments, clients were contacted for a structured phone interview within a week of their last therapy session if they completed treatment in a planned manner. In cases where termination was not planned, clients were contacted within a month of their last session. The collection of data on therapists’ perspectives at this time point occurred at the same time as the client data were collected. As indicated previously, these data were collected in the context of a larger study that examined several other factors related to psychotherapy engagement and termination. Research ethics board approval was obtained for all phases of the study, and informed consent was obtained from all participants following a full presentation of the nature of the study

Results

Preliminary Analyses

Preliminary analyses compared unilateral and mutual terminators on demographic, psychological functioning, and service variables to (a) ensure that groups were equivalent at pretherapy and (b) examine therapy outcome for both groups. Before therapy, there were no significant group differences on client-rated (SCL-10) and therapist-rated (GAF) psychological distress, gender, ethnic origin, referral source, or duration of presenting problem. However, unilateral terminators attended significantly fewer sessions; with an average of 9.7 sessions (SD = 8.1), whereas mutual terminators attended an average of 20.8 sessions (SD = 12.2), t (79.50) = − 6.25, p <.001.

On the SCL-10, unilateral terminators reported a significant decline in distress over the course of therapy, with a mean of 17.2 (SD = 6.8) before therapy and 10.3 (SD = 6 9) posttherapy, t(30) = 5.49, p <.001. Mutual terminators reported a similar pattern, with a mean of 14.9 (SD = 7.2) pretherapy and 6.5 (SD = 5.6) posttherapy, t(52) = 7.72, p <.001. At posttherapy, mutual terminators were significantly less distressed than unilateral terminators when symptom distress scores before therapy were controlled for, F(l,80) = 5.46, p <.05. Therapists reported unilateral terminators on the GAF as remaining the same over the course of treatment (i.e., no significant change), with a mean of 66.6 (SD = 11.0) before therapy and 66.8 (SD = 10.6) posttherapy, t(30) = − 2.3, ns. Therapists reported mutual terminators’ psychological functioning on the GAF as having significantly improved over the course of treatment, with a mean of 61.8 (SD = 13.1) pretherapy and 73.5 (SD = 14.2) posttherapy, t(50) = − 8.47, p <.001 Posttherapy, mutual terminators were rated by therapists as having significantly higher functioning than were unilateral terminators when therapists’ GAF assessments before therapy were controlled for, F(l,79) = 28.60, p <.001.

We examined intercorrelations among variables within each of clients’ and therapists’ perspectives on the BTPS and WAI-S and among client- and therapist-rated outcome measures (SCL-10, GAF) and these variables. For both clients and therapists, the WAI-S and BTPS were moderately negatively correlated (clients r = −.52, p <.001; therapists r = − .28, p <.Q5). Both pre- and posttherapy SCL-10 scores were negatively correlated with client-rated WAI-S (pre: r = −.22, p <.Q5; post: r = −.30, p < .01), but not with therapist-rated WAI-S (pre: r = − .14, ns; post: r = −.19, ns). Both pre- and posttherapy SCL-10 scores were also positively correlated with BTPS from client (pre: r = .31, p <.01; post; r = .32, p <.01) but not therapist (pre: r = .04, ns; post: r = .01, ns) perspectives. The GAF, pre- and posttherapy, was positively correlated with both client WAI-S (pre: r = .22, p <.05; post: r = .39, p <.001) and therapist WAI-S (pre: r = .29, p <.01; post: r = .38, p <.001). Pretherapy, there was no association between the GAF and BTPS for clients (r = − .11, ns) or therapists (r = .04, ns). Posttherapy, GAF scores were associated with client BTPS (r = − .29, p <.01), but not therapist BTPS (r = −.12, ns).

Mutual Versus Unilateral Terminators: Reasons for Termination: Hypothesis 1a

We hypothesized that clients’ and therapists’ mean ratings of the importance of termination reasons would differ between unilateral and mutual termination groups. Specifically, we expected that both clients and therapists in the mutual group, compared with the unilateral group, would rate having accomplished therapy goals as more important and circumstantial and therapy-specific reasons for termination as less important. A one-way multivariate analysis of variance (MANOVA) was used to compare means between the two groups (unilateral vs. mutual). The omnibus test for client-rated reasons was significant, Wilks's λ = 0.37, F(10, 72) = 12.39, p <.001, partial η² = .63. Keeping the familywise alpha at .05, tests of between-subjects effects indicated that clients who terminated therapy unilaterally assigned less importance than mutual terminators to “accomplished what you wanted to do in therapy” as a reason for leaving, F(1, 81) = 15.75, p <.001, partial η² = .16. In contrast, unilateral terminators rated every other reason, with the exception of “Decided to go elsewhere for services,” F(1, 81) = 0.49, ns, as significantly more important than did mutual terminators. These reasons included the following: “Could no longer fit time for therapy into schedule,” F(1, 81) = 20.43, p <.001, partial η² = .20; “Just lost interest in therapy,” F(1, 81) = 13.95, p <.001, partial η² = A5; “No longer had money or insurance coverage to pay for therapy,” F(1, 81) = 9.66, p <.003, partial η² = .11; “Felt therapy was going nowhere so ended therapy,” F(1, 81) = 66.17, p <.001, partial η² = A5; “Felt therapy was making things worse so stopped,” F(1, 81) = 20.64, partial p <.001, η² = .20; “Weren't confident in therapist's ability to help,” F(1, 81) = 37.68, partial p <.001, η² = .32; “Uncomfortable talking about personal matters with therapist,” F(1, 81) = 25.78, p <.001, partial η² = .24; and “Therapy didn't fit with ideas about what would be helpful,” F(1, 81) = 25.68, p <.001, partial η² = .24 (see Table I).

Table I.

Means and Standard Deviations of Client and Therapist Ratings of the Importance of Termination Reasons

	Client		Therapist
Reason for termination	Mutual	Unilateral	Mutual	Unilateral
Accomplished goals	3.2 (1.2)a	2.2 (1.1)	3.1 (1.1)1	1.6 (0.9)2
Could no longer fit time or therapy into schedule	1.1 (0.6)a	2.1 (1.2)b	1.3 (0.6)1	2.5 (1.3)2
Just lost interest in therapy	1.1 (0.4)a	1.8 (1.2)	1.2 (0.5)1	2.1 (1.0)2
No longer had money or insurance coverage	1.1 (0.6)a	1.8 (1.2)b	1.1 (0.3)	1.4 (1.0)
Felt therapy was going nowhere so ended therapy	1.0 (0.3)a	2.5 (1.2)	1.1 (0.4)1	1.6 (1.1)2
Felt therapy was making things worse	1.0 (0.2)a	1.8 (1.3)b	1.1 (0.3)	1.5 (1.1)
Weren't confident in therapist's ability to help	1.1 (0.3)a	2.2 (1.3)	1.1 (0.4)1	1.7 (1.0)2
Uncomfortable talking about personal matters	1.0 (0.2)a	1.9 (1.2)	1.1 (0.3)1	1.6 (1.0)2
Therapy did not fit with ideas of what would be helpful	1.2 (0.5)a	2.1 (1.2)	1.2 (0.6)	1.6 (1.0)
Decided to go elsewhere for services	1.3 (0.9)	1.5 (0.8)	1.3 (1.0)	1.5 (1.0)

Note. Within clients, columns with different superscript letters differed from each other at least at p <.005. Within therapists, columns with different superscript numbers differed from each other at least at p <.005

When comparing therapist ratings across groups, the omnibus test was also significant, Wilks's λ = 0.320, F(10, 70) = 14.89, p <.001, partial η² = .68. Keeping the familywise alpha at .05, tests of between-subjects effects indicated a pattern of findings similar to those obtained with the client ratings. Compared with therapists in the mutual group, those in the unilateral group assigned less importance to “Accomplished what you wanted to do in therapy” than the mutual group, F(1, 79) = 36.76, p <.001, partial η² = .32, and more importance to all other reasons except “Went elsewhere for services,” F(1, 79) = .92, ns; “No longer had money or insurance coverage to pay for therapy,” F(1, 79) = 5.07, ns; “Felt therapy was making things worse so stopped,” F(1, 79) = 6.98, ns; and “Therapy didn't fit with ideas about what would be helpful,” F(1, 79) = 4.02, ns. Reasons rated significantly more important by therapists of unilateral terminators included the following: “Could no longer fit time for therapy into schedule,” F(1, 79) = 32.25, p <.001, partial η² = .29; “Just lost interest in therapy,” F(1, 79) = 37.27, p<.001, partial η² = .32; “Felt therapy was going nowhere so ended therapy,” F(1, 79) = 10.84, p <.001, partial η² = .12; “Weren't confident in therapist's ability to help,” F(1, 79) = 15.83, p <.001, partial η² = .17; and “Uncomfortable talking about personal matters with therapist,” F(1, 79) = 13.82, p <.001, partial η² = .15 (see Table I).

Congruence between Client and Therapist Views on Reasons for Termination: Hypothesis 2a

It was expected that client − therapist perspectives regarding termination reasons would be more similar in dyads that made mutual decisions to terminate therapy compared with dyads in which both client and therapist agreed that termination was a unilateral decision on the client's part. To test this hypothesis, difference scores were calculated by subtracting therapist ratings from client ratings for each reason for termination separately (see Table II); positive values indicate that, on average, the client assigned higher importance to the reason than did the therapist, and negative values indicate that the therapist assigned higher importance to the reason than did the client. A series of one-sample t tests was conducted to determine whether difference scores were significantly different from zero. In light of the number of analyses, the alpha level for each comparison was set at .005. Difference scores that were significantly different from zero are shown in Table II.

Table II.

Difference Scores and Standard Deviations (Client Importance Ratings Minus Therapist Importance Ratings) of Reasons for Termination

Reason for termination	Unilateral (n = 29)	Mutual (n = 52)
Accomplished goals	0.6 (1.2)	0.1 (0.9)
Could no longer fit time or therapy into schedule	−0.3 (1.4)	− 0.1 (0.5)
Just lost interest in therapy	−0.3 (1.4)	− 0.0 (0.4)
No longer had money or insurance coverage	0.3 (1.0)	0.1 (0.6)
Felt therapy was going nowhere so ended therapy	0.9 (1.4)*a	− 0.0 (0.3)
Felt therapy was making things worse	0.4 (l.l)a	− 0.0 (0.3)
Weren't confident in therapist's ability to help	0.6 (1.4)a	− 0.0 (0.5)
Uncomfortable talking about personal matters	0.3 (1.4)	− 0.1 (0.3)
Therapy did not fit with ideas of what would be helpful	0.6 (1.4)a	− 0.1 (0.6)
Decided to go elsewhere for services	−0.1 (1.2)	0.0 (0.7)

Note. Columns with different superscripts differed from each other at least at p < .005.

*p < .001; indicates significant differences from zero.

For mutual terminators, none of the difference scores differed significantly from zero, indicating that client and therapist ratings of the importance of each reason for termination were very similar. In client-therapist dyads who agreed that the client made a unilateral decision to end therapy, clients rated the importance of one termination reason, “Felt therapy was going nowhere so ended therapy,” t(28) = 3.55, p <.001, d = 0.64, significantly higher than therapists (Table II). The magnitude of differences between client and therapist importance ratings, although in the expected direction, was not large enough to be considered meaningful for the following reasons: “Accomplished what you wanted to do in therapy,” t(28) = 2.51, ns; “Felt therapy was making things worse so stopped,” t(28) = 2.05, ns; “Weren't confident in therapist's ability to help,” t(28) = 2.12, ns; and “Therapy didn't fit with ideas about what would be helpful,” t(28) = 2.16, ns.

To test whether client-therapist perspectives on reasons for termination differed to a greater extent in the unilateral compared with the mutual termination group, a one-way MANOVA was conducted to compare the magnitude of difference scores between groups. The multivariate test of between-subjects effects was significant, Wilks's λ = 0.58, F(10, 70) = 5.11, partial η² = .42, p <.001. Follow-up univariate analyses, keeping the familywise alpha at .05, indicated that client-therapist difference scores were significantly larger in the unilateral group for reasons of “Felt therapy was going nowhere so ended therapy,” F(1, 79) = 22.37, p <.001, partial η² = .22; “Felt therapy was making things worse so stopped,” F(1, 79) = 8.14, p<.005, partial η² = .09; “Weren't confident in therapist's ability to help,” F(1, 79) = 7.93, p <.005, partial η² = .09; and “Therapy didn't fit with ideas about what would be helpful,” F(1, 79) = 8.44, p<.005, partial η² = .10. Overall, it appears as though clients who unilaterally decided to end therapy rated reasons related to the unhelpfulness of therapy as more important to their termination decisions than did their therapists. It seems as though, even when therapists recognized that the client made a unilateral decision to leave, therapists may not have been aware of the full extent of the importance of clients’ negative perceptions of the therapy experience and of the therapist.

Congruence between Client and Therapist Views on the Quality of the Working Alliance: Hypotheses 1b and 2b

Repeated measures analysis of variance¹ with dyad member as the repeated factor was used to examine hypotheses that (lb) client-therapist dyads who mutually terminated therapy would report a stronger working alliance than that reported by the unilateral decision dyads and (2b) clients’ and therapists’ ratings of the working alliance would be more discrepant when termination was a unilateral decision on the client's part than when the decision was mutual. The test of between-subjects effects indicated that client-therapist dyads in the mutual termination group rated the working alliance slightly but significantly higher than dyads in the unilateral termination group, F(1, 78) = 5.39, p <.05, η² = .07. Dyads in the mutual group reported a mean of 69.71 (SE = 1.30), whereas those in the unilateral group reported a mean of 64.71 (SE = 1.72). The test of within-subjects effects indicated that, across termination groups, clients rated the working alliance significantly higher than did therapists, F(1, 78) = 5.08, p <.05, η² = .06. Clients reported a mean of 68.67 (SE = 1.38), whereas therapists reported a mean of 65.74 (SE = 1.12). The Dyad Member x Termination Status interaction was not significant, indicating that the magnitude of the difference between client-therapist ratings of the working alliance was similar in unilateral and mutual terminators, F(1, 78) = 3.44, ns.

Congruence between Client and Therapist Views on Barriers to Treatment Participation: Hypotheses 1c and2c

Repeated measures analysis of variance with dyad member as the repeated factor was used to examine hypotheses that (1b) client-therapist dyads who mutually terminated therapy would report fewer barriers to treatment participation than did the unilateral decision dyads and (2c) clients’ and therapists’ ratings of barriers to treatment would be more discrepant when termination was a unilateral decision on the client's part than when it was a mutual decision. The test of between-subjects effects indicated that client-therapist dyads in the unilateral termination group reported more barriers to treatment than those in the mutual termination group, F(1, 81) = 35.41, p<.001, η² = .30. Dyads in the unilateral group reported a mean of 61.50 (SE = 1.56), whereas dyads in the mutual group reported a mean of 49.78 (SE = 1.20). The test of within-subjects effects indicated that, across termination groups, clients reported significantly more barriers to treatment than did their therapists, F(1, 81) = 4.94, p <.05, η² = .06. Clients reported a mean of 57.48 (SE = 1.30), whereas therapists reported a mean of 53.80 (SE = 1.27). The Dyad Member x Termination Status interaction was not significant, indicating that the magnitude of the difference between client-therapist ratings of barriers to treatment was similar in unilateral and mutual terminators, F(1, 81) = 1.59, ns.

Discussion

In an effort to better understand unilateral termination, the present study examined the congruence in perspectives of client-therapist dyads regarding important therapeutic variables, including clients’ reasons for termination, working alliance, and barriers to treatment between two groups where (a) both client and therapist agreed that termination was a unilateral decision on the client's part or (b) both client and therapist agreed that termination was mutual. As hypothesized, results of our study indicated that unilaterally terminating clients, compared with mutual terminators, rated the importance of having accomplished their goals in therapy as less important to their decision to end therapy and reasons related to circumstantial barriers and dislike of therapist and therapy as more important to their decision. Therapists reported a similar pattern of results; therapists of unilateral terminators, compared with those of mutual terminators, reported that their clients’ accomplishing goals in therapy was less important to their decisions and that reasons related to circumstantial barriers and dislike of therapist and therapy were more important to their decisions.

When client-therapist assessments were examined within each dyad, small but systematic differences in attributions of clients and their therapists became evident. When termination decisions were mutual, there was no difference between client and therapist ratings of the importance of any termination reason. When clients terminated therapy unilaterally, compared with their therapists, they rated four of 10 reasons for termination as significantly more important to their decision to leave. They ascribed higher importance to all of the reasons related to dislike of therapy or therapist: “Felt therapy was going nowhere so ended therapy,” “Felt therapy was making things worse,” “Weren't confident in therapist's ability to help,” and “Therapy did not fit with ideas about what would be helpful.” Clients and therapists rated the importance of more benign and circumstantial barriers similarly.

Outcome data collected in the study also reflect a perspective divergence between clients and therapists in the unilateral, but not the mutual, termination group; unilateral terminators rated their distress as significantly lower at posttherapy, whereas their therapists indicated no change in functioning. In contrast, clients in the mutual termination group reported a similar decline in distress from pretherapy to posttherapy, and their therapists agreed with them, reporting a significant increase in functioning.

These results build on previous research showing that therapists tend to perceive both treatment success and failure differently than clients (Hunsley et al., 1999; Pekarik & Finney-Owen, 1987). Directly comparing client and therapist ratings, results from the present study indicate that these differences in perception occur exclusively around unilateral termination. When termination was a unilateral decision on the clients’ part, therapists appeared not to be aware of the extent to which clients perceived either success or failure in therapy (i.e., symptom improvement). Given the small differences in client and therapist ratings, therapists were largely aware of clients’ dissatisfaction but tended to rate the importance of clients’ dissatisfaction reasons as less important than they actually were. This could reflect both self-serving biases (whereby therapists are not as likely to rate themselves too negatively) and differing expectations about what will be accomplished in therapy. It likely also reflects the limited communication inherent in unilateral decision making; clients may be unlikely to share the extent of their negative perceptions of therapy and the therapist.

Results from the present study regarding therapeutic alliance data were in line with previous research; the early alliance, rated after the third therapy session by both client and therapist, was related with type of termination decision. As we hypothesized, client-therapist dyads who made mutual decisions to end therapy reported a stronger working alliance early in treatment than did client-therapist dyads where the client terminated unilaterally. Contrary to our expectations that mutually terminating dyads would have more similar perceptions of the working alliance, regardless of how clients terminated therapy, all clients rated the early alliance significantly higher than did their therapists. It seems as though the tendency, well-documented in the literature (e.g., Bachelor & Salame, 2000; Fitzpatrick, Iwakabe, & Stalikas, 2005; Hersoug, Høglend, Monsen, & Havik, 2001; Hilsenroth, Peters, & Ackerman, 2004; Tryon et al., 2007), for clients to rate the working alliance as higher than their therapists holds true in spite of eventual unilateral decisions to leave and poorer therapeutic outcome. Fitzpatrick et al. (2005) assessed client-therapist alliance ratings in early, middle, and late phases of therapy and found that, once formed, divergence between client and therapist remained unchanged, and alliance ratings for both clients and therapists increased linearly. Although there has been no systematic investigation into why clients rate the alliance as higher, Tryon et al. (2007) suggested that therapists may rate clients relative to alliances formed with other clients, whereas clients may rate therapists in comparison to other health professionals who may take a less collaborative, more paternalistic role or to friends and family members experienced as (naturally) less collaborative.

A similar pattern of results was found for barriers to treatment participation. Both client and therapist dyads who made mutual decisions to end therapy indicated fewer barriers to treatment than did clients and therapist dyads where the client made a unilateral decision to leave. Contrary to hypotheses, there was no difference in client-therapist perspective congruence between unilateral and mutual decision groups. In general, clients rated barriers to treatment participation as higher than did therapists. This was the first study, to our knowledge, that examined Kazdin's BTPS in adult clients. More barriers to treatment reported by clients and therapists of adult clients are associated with unilateral termination decisions, just as more barriers to treatment reported by parents of children and adolescents with conduct problems are associated with premature termination in Kazdin and colleagues’ (1997) research on dropping out of child treatment.

Our pattern of results was different than that of Kazdin et al. (1997) in that our clients reported significantly more barriers than did their therapists, whereas parents of conduct-disordered children in Kazdin et al.'s study reported significantly fewer barriers than did their therapists. This may be due to differences in client demographics and presenting problems, or therapist experience (our study took place in a training clinic). Further research should be conducted to replicate our results; however, our study indicates that therapists can expect that as clients experience more barriers, they are more likely to make unilateral decisions to leave therapy.

As mentioned by Kazdin and Wassell (2000), the timing of assessment of perceived barriers (and of reasons for termination) raises issues, because it was conducted at the end of therapy, after termination decisions had already been made. Retrospective reporting always runs the risk of biased recall, however, given the relatively short time frame of treatment, and because the timing of the posttherapy assessment was within 1 month after termination, recall bias of retrospective reporting is less likely. As a result of logistical constraints, the time lag between end of therapy and completion of the BTPS and the reasons for termination measure was within 1 month for unilateral terminators and within 1 week for mutual terminators, potentially adding further measurement biases of an unknown nature. It is possible that treatment outcome influenced our results: Mutually terminating clients were less symptomatic and higher functioning posttherapy and, therefore, may have reported fewer barriers as a result of experiencing greater improvement. As Kazdin and Wassell (2000) discussed, assessing barriers at other therapy points (e.g., early in treatment or on multiple occasions throughout treatment) have their own methodological and practical liabilities (e.g., clients not having a complete idea of barriers early in treatment, confounding number of assessment administrations with duration in treatment, possibly sensitizing clients to the challenges of attending psychotherapy). Future research should examine other methods of assessing barriers to treatment throughout the therapy process.

In conclusion, this was the first study to obtain parallel information from both members of the client-therapist dyad about specific reasons why the client terminated services and to examine how these perspective divergences regarding reasons for termination, early working alliance, and barriers to treatment participation are related to unilateral termination. When clients made unilateral decisions to end therapy, therapists were only partially aware of either the extent of clients’ perceiving success in therapy or with their dissatisfaction. Although working alliance and barriers to treatment participation were rated as lower in the context of unilateral termination by both clients and therapists, all clients, in general, rated the early alliance and barriers to treatment as higher than their therapists. Future research should examine the utility of providing therapists with feedback regarding barriers to treatment and other process variables in terms of retaining clients in therapy. Preliminary research (Manfred-Gilham, Sales, & Koeske, 2002) suggests that therapists use more engagement strategies (particularly direct discussion of barriers) when they perceive clients to have more barriers; however, no research has examined how therapists’ use of these strategies impacts client perception of barriers or influences treatment retention.