The relationship between expected engagement and talking therapy outcome.

The aim of the study was to investigate whether client-reported expected engagement with therapy predicted therapy outcome. It was hypothesized that higher expected engagement with cognitive behavioural therapy (CBT) or person-centred experiential therapy (PCET) would predict more symptomatic improvement following therapy and higher likelihood of therapy completion. The Sheffield Expected Engagement with Therapy Scale was administered to 96 clients at pre-therapy assessment with all meeting a diagnosis of moderate or severe depression with 53 receiving CBT and 43 receiving PCET. Higher expected engagement predicted more symptomatic improvement in CBT but not PCET. Expected engagement only predicted improvement in CBT when clients rated the credibility of CBT as low or moderate. Expected engagement did not predict therapy completion in either therapy. Assessment of expected engagement could be a useful tool in prediction of symptomatic improvement in CBT.

RCT Entities:   Population Interventions Outcomes

The present study assesses clients' expectations of their engagement with therapy.

Expectations of engagement can predict improvement in CBT.

Personalized medicine may benefit from assessing expected engagement and its association with therapy outcome.

INTRODUCTION

Client expectations of therapy have been shown to have a significant influence on therapy outcome across a range of therapeutic modalities (Arnkoff, Glass, & Shapiro, 2002; Baekeland & Lundwall, 1975; Clarkin & Levy, 2004; Gaston, Marmar, Gallagher, & Thompson, 1989; Hardy et al., 1995; Noble, Douglas, & Newman, 2001). This association may be due to changes brought about by a patient's expectations, such as feeling optimistic about recovery, or higher adherence to treatment, which could contribute to a higher likelihood of improvement (Higginbotham, 1977). A positive influence of expectations on therapy outcome is more likely when the therapy received is in line with such expectations because discordance between expectations and experience can have a negative effect on outcome (Elkin et al., 1999; Horenstein & Houston, 1976).

However, typical assessments of expectations do not take into account a client's expectations of their experience of the therapy process, despite evidence that engagement with the therapy process predicts therapy outcome (Glenn et al., 2013; Gomes‐Schwartz, 1978; Persons, Burns, & Perloff, 1988). Rather, existing measures of expectations commonly only assess credibility and/or outcome expectancy (Devilly & Borkovec, 2000). These types of expectation have been found to be successful predictors of therapy outcome but do not provide insight into a client's expectations of themselves as an agent of change in the therapeutic context (Arnkoff et al., 2002; Baekeland & Lundwall, 1975). A client's insight into their own recovery is an important source of information in understanding which aspects of treatment will benefit them, although it has often been overlooked in measures of expectations (Bohart & Tallman, 2010). It is proposed that this insight could be present before therapy begins and could influence therapy engagement and outcome.

Client insight and anticipation of engagement with therapy is influenced by a multitude of factors, both practical and psychological. Additionally, as discussed by Duncan and Miller (2000), several of these factors, such as readiness to change, coping style, theory of change, and social support, are predictive of therapy success (Assay & Lambert, 1999; Carver, Scheier, & Weintraub, 1989; DiClemente & Prochaska, 1982; Miller, Duncan, & Hubble, 1997; Roehrle & Strouse, 2008). Hence, expectations influenced by such factors would also be hypothesized to predict therapy outcome. For example, a client may not expect to engage with a therapy that uses problem‐solving techniques because they have learned to adopt a denial coping style, whereas another client may not expect to engage with the same therapy because they have young children and do not have the time to conduct behavioural homework.

The influence of expected engagement on therapy outcome is based on the hypothesis that, as with credibility and expectancy, expected engagement can predict client engagement with therapy, which ultimately influences their symptomatic improvement and/or therapy completion (Abouguendia, Joyce, Piper, & Ogrodniczuk, 2004; Joyce, Ogrodniczuk, Piper, & McCallum, 2003; Meyer et al., 2002). It has been theorized that expectancy can predict therapy outcome due to increased adherence to the therapeutic model, better client–therapist alliance, and more persistent efforts towards change (Apfelbaum, 1958; Higginbotham, 1977; Meyer et al., 2002). Predictive factors such as client theory of change have also been found to influence engagement and, ultimately, therapy outcome (Duncan & Moynihan, 1994). Therefore, expected engagement could capture therapeutically relevant information about factors that may affect a client's engagement, as a process variable that influences therapy outcome.

Furthermore, existing measures of credibility and expectancy rarely specify the techniques and skills that distinguish one therapy from another, such as cognitive restructuring in cognitive behavioural therapy (CBT; Beck, Rush, Shaw, & Emery, 1979). It is not always the case that “one size fits all,” as some clients are likely to do better with a specific type of therapy (Norcross & Wampold, 2011, 2018). The inclusion of specific aspects of different therapies allows a client to compare their perceptions and ideas with what is being offered by different therapies.

It may be that some clients do not have strong insight into what would and would not engage them prior to therapy commencing or may even have expectations that are not a true reflection of their engagement when they do begin therapy (Norcross & Beutler, 1997). However, even no or false expectations of engagement would be hypothesized to capture information about how clients will engage with the therapy provided. For instance, holding no expectations of how well they might engage with two therapies would produce comparable scores that would predict that the client has an equal likelihood of success with any therapy. In comparison, a client who wrongly expects not to engage well with a therapy but then does engage would be expected to find it more difficult to engage and have a successful therapy experience than a client who does expect to engage. Predilection for a particular treatment has been associated with engagement and remaining in therapy when congruent with the treatment received (Elkin et al., 1999). In much the same way, expected engagement is hypothesized to affect engagement and outcome, even if initial expectations are not an accurate reflection of what occurs in therapy.

Furthermore, the limited research that has investigated a moderator effect of therapy type on the relationship between expectations of the process and outcome of therapy has found differences in association due to therapy type. For example, Gaston et al. (1989) found expectations of change using cognitive techniques to only be associated with improvement when cognitive therapy was provided, in contrast to behavioural or brief dynamic therapy. Such research has been sparse but is becoming increasingly desirable as healthcare focuses more on personalized medicine (Cohen & DeRubeis, 2018). Research into both the relationship between clients' expected engagement and therapy outcome and also how therapy type may moderate this relationship is important to advance personalized medicine as these themes recognize that there may be differential interactions between client and therapy that may produce variable outcomes.

In light of this literature, the aim of the present study was to investigate expected engagement as a predictor of therapy outcome, using a validated measure of expectations of engagement with aspects of CBT and person‐centred experiential therapy (PCET). Therapy outcome was assessed as symptomatic improvement and therapy completion. It was hypothesized that a higher level of expected engagement would predict greater symptomatic improvement and a higher likelihood of completing therapy. It was further hypothesized that those with higher levels of expected engagement with a cognitive approach would have better therapy outcomes with CBT and vice versa for PCET. Hence, clients who agreed that they would engage well with aspects of CBT or PCET were predicted to obtain more improvement and higher rates of completion, as long as they received the therapy that they rated as highly engaging.

METHOD

Design

The present study used a consecutive participant design by recruiting a subsample from the Pragmatic, Randomised Controlled Trial assessing the non‐Inferiority of Counselling and its Effectiveness for Depression (PRaCTICED; Trials Registry ID ISRCTN06461651; Saxon et al., 2017). The trial received a favourable ethical opinion from Yorkshire and the Humber—South Yorkshire Research Ethics Committee (REC reference: 14/YH/0001). The PRaCTICED trial recruited moderate to severely depressed clients who had been referred to the Improving Access to Psychological Therapies (IAPT) service in the U.K. National Health Service (NHS) for therapy for depression. The aim of the trial was to assess the non‐inferiority of PCET (known as counselling for depression in U.K. primary care services; Murphy, 2019; Sanders & Hill, 2014) compared with CBT (Beck et al., 1979) for the treatment of moderate to severe depression.

Sample size

The required sample size was determined by the power analysis programme G*Power (Faul, Erdfelder, Buchner, & Lang, 2009). The study required a minimum sample size of 69 participants to detect a medium effect (Cohen's f 2 = .15) at a power of 0.8 and a probability level of .05. This sample size provides sufficient power for 11 potential predictors to be included in the regression model as follows: tested predictors: (a) expected engagement and (b) Expected Engagement × Therapy Type; covariates: (c) therapy type (one dummy‐coded level), (d) age, (e) gender (one dummy‐coded level), (f) severity, (g) preference (two dummy‐coded levels), (h) credibility, (i) expectancy, (j) number of sessions, and (k) completion status (one dummy‐coded level)/symptomatic improvement.

Participants

Ninety‐six PRaCTICED trial clients were recruited for the present study. The sample was 55% (N = 53) female with a mean age of 38.64 years (SD = 12.58). White British participants comprised 69% (N = 66) of the sample. The sample comprised 40% (N = 38) full‐time workers, 25% (N = 24) unemployed, 7% (N = 7) students, and 4% (N = 4) retired. A large proportion of the sample (84%; N = 81) had attended further education, and 45% (N = 43) had attended higher education. Participants who had previously attended therapy comprised 57% (N = 55) of the sample, with 18 having received a form of cognitive therapy, 32 a form of counselling, and the remaining 5 other types of therapy.

Measures

Sheffield Expected Engagement with Therapy Scale (Harrison, Hardy, & Barkham, 2017)

The Sheffield Expected Engagement with Therapy Scale (ShEETS; Appendix A) is a measure of expected engagement with two broad types of therapy: a cognitive or person‐centred experiential therapeutic approach. The measure comprises 12 items that represent cognitive and PCET components as well as common therapy conditions found in all therapies. The items are taken from the Cognitive Therapy Scale—Revised (Blackburn et al., 2001), the Person Centred and Experiential Psychotherapy Scale (Freire, Elliott, & Westwell, 2013), and the Facilitative Conditions Scale from the Sheffield Psychotherapy Rating Scale (Shapiro & Startup, 1990). An example of a cognitive item is “Encouraging homework, such as trying out new ideas and experiences outside therapy,” a PCET example is “Being supportive when you experience negative or overwhelming experiences,” and a common condition example is “Showing warmth.” Participants are asked to rate each item on a 5‐point Likert scale for how likely an aspect of therapy is to engage them in the therapy process (0 = not at all likely and 4 = extremely likely). Moderate concurrent validity with the Client Involvement Scale, framed to be expected engagement rather than engagement, was found for each of the expected engagement scales, p < .001 (Harrison et al., 2017). The ShEETS also showed moderate concurrent validity with the Credibility Expectancy Questionnaire, p < .001. Internal consistency of the scales with three separate samples ranged from α = .73–.85 for the cognitive scale, α = .72–.83 for the PCET scale, and α = .71–.86 for common items. The overall internal consistency of the 12 items was α = .86–.91.

Patient Health Questionnaire‐9 (Kroenke & Spitzer, 2002)

The Patient Health Questionnaire‐9 (PHQ‐9) comprises nine items rated on a 4‐point Likert rating scale from 0 to 3 on the prevalence of certain depressive symptoms over the last 2 weeks. A 10th question assesses the extent to which any problems have impacted on everyday functioning. The clinical cut‐off scores are 0–4 (none), 5–9 (mild symptoms), 10–14 (moderate symptoms), 15–19 (moderately severe symptoms), and 20–27 (severe symptoms). The PHQ‐9 has good reliability (Cronbach's α = .89; PHQ Primary Care Study; Spitzer, Kroenke, Williams,, & Patient Health Questionnaire Primary Care Study Group, 1999). Sensitivity and specificity for testing for major depression are 95% and 84%, respectively (Löwe, Kroenke, Herzog, & Gräfe, 2004). The measure has also been found to have good test–retest reliability (r = .84) and criterion validity, with an incremental increase in positive likelihood ratios of PHQ‐9 scores for major depression. Finally, the measure was found to have good discriminant validity evidenced by a strong negative correlation between increasing PHQ‐9 scores and the six decreasing 20‐Item Short Form Health Survey scores (Medical Outcomes Study Short‐Form General Health Survey; Stewart, Hays, & Ware, 1988).

Preference question (Leykin et al., 2007)

The preference question asked participants how strong their preference was to receive CBT or PCET on a scale of 0 (no preference) to 5 (strong preference).

Credibility/Expectancy Questionnaire (Devilly & Borkovec, 2000)

The Credibility/Expectancy Questionnaire (CEQ) assesses credibility and expectancy of therapy. The CEQ comprises six questions, three regarding credibility and three for expectancy, each to be rated on a 1‐ to 9‐point Likert scale or from 0% to 100%. The measure was adapted for use with the current clinical sample with a simplified rating structure. The credibility and expectancy of a cognitive and a PCET approach were assessed prior to therapy allocation. Participants were provided with a brief description of both CBT and PCET and asked how logical and successful they believed the named therapies to be and how confident they would be in recommending them (credibility) and how much improvement they thought and felt would occur with that therapy (expectancy). Principal component analysis on the items revealed the six CEQ questions to create two factors (labelled “credibility” and “expectancy”), which accounted for 82.5% of the total variance. A test of reliability of the credibility factor gave a Cronbach's α internal consistency score of .86 and the expectancy factor a Cronbach's α of .90.

Socio‐demographics

Information including gender, age, ethnicity, employment status, education level (continuation of education after compulsory education and continuation to higher education), and previous therapy attendance and type was taken at the baseline assessment and from the service online system.

Number of therapy sessions

The research team collected data on the number of sessions that clients attended from therapists at the end of treatment.

PRaCTICED Therapist End of Therapy Form (Saxon et al., 2017)

The form asked the therapist for details of how therapy ended. The therapists' judgement of whether the client completed or dropped out of therapy early was used as the clients' completion status.

Therapy

PRaCTICED trial participants were randomized to two forms of manualized therapy routinely offered in the U.K. NHS IAPT service: 53 participants were randomized to CBT and 43 to PCET. Both treatments were provided by trained IAPT therapists and delivered according to therapy‐specific manuals written for the trial (see Saxon et al., 2017). All therapists received standard supervision at defined time intervals, and their adherence and competency were monitored by supervisors. The form of CBT delivered followed a Beckian CBT approach, and regular top‐up workshops were provided throughout the duration of the trial (Beck et al., 1979). PCET is a person‐centred experiential approach and focuses on emotions in the treatment of depression (Sanders & Hill, 2014). PCET therapists were previously trained in humanistic therapies but received an additional 5‐day intensive training plus 80 hr supervised practice in which audiotapes were externally rated. PCET therapists were required to pass this additional training programme prior to seeing clients in the trial. PRaCTICED participants could receive up to 20 weekly therapy sessions as per IAPT protocol.

Procedure

Participants provided informed consent for the study at their eligibility baseline assessment for the PRaCTICED trial. The inclusion of the ShEETS at baseline assessment followed a favourable ethical opinion on a substantial amendment to the original trial protocol. Participants first completed the preference measure to confirm that they had no treatment preference, followed by the Clinical Interview Schedule—Revised to assess their eligibility for the PRaCTICED trial (Lewis, 1994). Only those who were eligible proceeded to complete the questions on demographic details, the ShEETS, PHQ‐9, and expectations questions. Eligible participants were randomized to CBT or PCET and added to the IAPT therapy waiting list. Once therapy had begun, patients completed the PHQ‐9 at each therapy session, and the final session PHQ‐9 score was used to calculate symptomatic improvement from the baseline score. At treatment end, the therapist completed the Therapist End of Therapy form for each participant to provide information about whether the participant completed therapy or dropped out.

Analyses

IBM SPSS 23 statistical analysis programme was used for all analyses (IBM Corp, 2015). Preliminary data assessments were made prior to the main analysis. Kruskal–Wallis and Mann–Whitney U tests were conducted to identify differences in expected engagement between cognitive or PCET approach and common items or between expected engagement scores for those who received CBT and PCET.

Symptomatic improvement was baseline PHQ‐9 score subtracted from final session PHQ‐9 score. Residualized change scores were not used for the current analysis as adjustments are made on the basis of scores from the whole sample without taking treatment group into account, which may result in misleading findings (Maxwell, Delaney, & Manheimer, 1985). Tests of association were conducted between symptomatic improvement/therapy completion: predictors: expected engagement, Expected Engagement × Therapy Type; covariates: therapy type (one dummy‐coded level), age, gender (one dummy‐coded level), severity, preference (two dummy‐coded levels), credibility, expectancy, and number of sessions. Tests were conducted separately for CBT and PCET to identify differences by therapy type and inform any required further interaction terms with therapy to be included in the regression models. Bonferroni corrections were applied to the tests to decrease the likelihood of type I error. Tests of association were not conducted for common items and improvement/completion as the research question was driven by the differences between cognitive therapy and PCET.

Models 1 and 2 from the PROCESS (v2.16 and v3.0) macro for SPSS were used to conduct two multiple ordinary least squares regression analyses: cognitive expected engagement and PCET expected engagement as predictors of PHQ‐9 change (Hayes, 2013, 2018). Two multiple logistic regressions were planned for cognitive and PCET expected engagement as predictors of therapy completion. Therapy type was included as a moderator in all analyses. Bootstrapping was performed to counteract non‐parametric data.

RESULTS

Missing data

Pairwise deletion was conducted on three missing scores on the credibility question, four on the expectancy question, and six missing responses on the completion question.

Descriptive statistics

Expected engagement

The mean expected engagement scores were as follows: cognitive: M = 13.32 (SD = 2.36), PCET: M = 12.99 (SD = 2.33), and common: M = 12.52 (SD = 2.51). The scales strongly correlated with each other: cognitive and PCET, r(96) = .59, p < .001, cognitive and common, r(96) = .43, p < .001, and PCET and common, r(96) = .74, p < .001. A Kruskal–Wallis test revealed a trend, although not significant, for a difference between expected engagement scores on the three scales, χ 2(2) = 5.34, p = .069. The mean expected engagement scores for those receiving CBT and PCET are shown in Table 1. Expected engagement scores with each scale did not differ for those who received CBT, χ 2(2) = 2.66, p = .264, or PCET, χ 2(2) = 2.68, p = .262, nor did expected engagement scores differ between CBT and PCET, as shown in Table 1.

Table 1

Means (M) and standard deviations (SD) for cognitive, PCET, and common expected engagement scores split by allocated therapy

Expected engagement scale	CBT	PCET	Difference between therapies
	M (SD)	M (SD)	U	Z	p
Cognitive expected engagement	13.09 (2.48)	13.60 (2.20)	1,006.5	−0.99	.321
PCET expected engagement	12.70 (2.52)	13.35 (2.05)	986.5	−1.14	.255
Common expected engagement	12.26 (2.70)	12.84 (2.25)	1,015.0	−0.93	.354

Abbreviations: CBT, cognitive behavioural therapy; Common, common components across both therapies; PCET, person‐centred experiential therapy.

Therapy outcome

The means for baseline PHQ‐9 and end of therapy PHQ‐9 were 19.19 (SD = 4.24) and 10.13 (SD = 6.85), respectively, creating a mean PHQ‐9 change score of 9.06 (SD = 6.88). Therapists rated 56% (N = 54) of participants as having completed therapy and 38% (N = 36) as having dropped out. The remaining 6% (N = 6) did not provide data on completion status so could not be included in completion analyses.

Regression analyses

Spearman's correlations were conducted prior to regression analysis between expected engagement and symptomatic improvement (Bonferroni‐adjusted p < .01). Higher cognitive expected engagement was significantly correlated with more symptomatic improvement, r(96) = .27, p = .009. There was a moderate but non‐significant trend association between PCET expected engagement and symptomatic improvement, r(96) = .23, p = .026. Other factors shown to be significantly related to symptomatic improvement were included in the model as covariates, p < .005.

Mann–Whitney U test (Bonferroni‐adjusted α of .006) and point biserial correlations (Bonferroni‐adjusted α of .01) showed no significant relationships between therapy completion and expected engagement with a cognitive, U = 776.0, Z = −1.64, p = .102, or PCET approach, r(90) = .18, p = .099. Due to a lack of correlations, further regression analyses were not conducted for completion.

The assumptions of multicollinearity, outliers, and leverage were tested to understand if the data were appropriate to be entered into the regression models. There was no multicollinerity between the independent variables, variance inflation factor >5 for any of the models. There were some outliers and leverage, although none were found to be influential on the regression output. Levene's test of homoscedasticity showed no significant difference in variance of symptomatic improvement by the two significant covariates of gender, F = 0.48, p = .491, therapy type, F = 2.01, p = .160, or completion status, F = 0.23, p = .632.

Only the model including cognitive expected engagement was run, as PCET expected engagement did not correlate significantly with symptomatic improvement. A Credibility × Expected Engagement interaction was entered into the regression model due to the potential influence of credibility on the relationship between expected engagement and therapy outcome. A second power analysis to account for the added interaction revealed that a sample size of 78 would be required to ensure adequate power for detection of a medium effect size, at a probability level of .05. Hence, the sample size of N = 87 for the current analysis was acceptable.

The regression model was significant, F(8, 78) = 6.24, p < .001, R 2 = .39. Higher expected engagement with a cognitive approach predicted more symptomatic improvement, b = 0.72, SE = 0.36, t(78) = 2.03, p = .046. However, there was no moderator effect of therapy type, b = −0.40, SE = 0.53, t(78) = −0.75, p = .457, or credibility, b = −0.17, SE = 0.12, t(78) = −1.38, p = .172. Further details of the model are in Table 2.

Table 2

Regression model of cognitive expected engagement to predict symptomatic improvement

Variables	b	SE b	t	df	p	95% CI
Cognitive expected engagement	0.72	0.36	2.03	78	.046*	[0.01, 1.44]
Therapy type	−0.50	1.26	−0.40	78	.992	[−3.01, 2.01]
Cognitive Expected Engagement × Therapy Type	−0.40	0.53	−0.75	78	.457	[−1.46, 0.66]
Credibility	0.12	0.28	0.42	78	.673	[−0.44, 0.68]
Cognitive Expected Engagement × Credibility	−0.17	0.12	−1.38	78	.172	[−0.40, 0.07]
Gender	−2.57	1.28	−2.00	78	.049*	[−5.12, −0.01]
No. of sessions	−0.02	0.13	−0.13	78	.898	[−0.27, 0.24]
Completion	6.09	1.57	3.89	78	<.001*	[2.97, 9.21]

Abbreviation: CI, confidence interval.

Significant at <.05 level.

Significant at <.01 level.

Significant at <.001 level.

Despite a lack of interactions, there were some significant effects dependent on level of credibility and therapy received. Higher expected engagement significantly predicted more improvement in those who had received CBT and rated credibility as low, b = 1.06, SE = 0.39, t(78) = 2.72, p = .008, 95% CI [0.28, 1.83], or moderate, b = 0.72, SE = 0.36, t(78) = 2.03, p = .046, 95% CI [0.01, 1.44]. Higher expected engagement did not predict more improvement in those who received CBT and rated it as highly credible, b = 0.39, SE = 0.47, t(78) = 0.84, p = .403, 95% CI [−0.54, –1.33]. However, expected engagement did not significantly predict symptomatic improvement for those who received PCET regardless of credibility rating, low credibility: b = 0.66, SE = 0.46, t(78) = 1.43, p = .157, 95% CI [−0.26, 1.57], moderate credibility: b = 0.33, SE = 0.42, t(78) = 0.78, p = .436, 95% CI [−0.50, 1.16], and high credibility: b = 0.00, SE = 0.50, t(78) = −0.01, p = .993, 95% CI [−1.00, 0.99].

Further interactions explored in separate models to ensure sufficient power between significant variables showed no moderator relationship between cognitive expected engagement and gender, b = −0.05, SE = 0.49, t(85) = −0.10, p = .919, 95% CI [−1.03, 0.93], or completion, b = 0.10, SE = 0.51, t(85) = 0.20, p = .843, 95% CI [−0.91, 1.11].

DISCUSSION

The present study has demonstrated the first implementation of the ShEETS in a clinical context, in which expected engagement predicted symptomatic improvement as hypothesized, although this effect was limited to those clients who received CBT. The effect was present for expectations of engagement with a cognitive approach when clients received CBT. Despite a moderate correlation between expected engagement with PCET and improvement, the relationship was only at trend level, leaving an inconclusive finding for the predictive effect of expected engagement with PCET on improvement. Additionally, there was no moderator effect of therapy type on expected engagement with CBT, meaning that, despite an effect only for those who received CBT, the effect of expected engagement on symptomatic improvement did not significantly differ between the two therapies. Additionally, unexpectedly, there was no predictive effect of expected engagement on therapy completion.

The presence of an effect of expected engagement on therapy outcome only in CBT may be due to inherent differences in the therapy processes of CBT and PCET. CBT, as a highly structured approach, may have been better placed to provide the tools for clients to recognize when expectations had been met. For instance, the focus in CBT on structure ensures that an agenda and goals are proposed and reviewed every session. Hence, such a therapy is built around explicit recognition of when goals are achieved. In contrast, PCET is more client led, which may mean the therapy progresses in a less structured and predictable manner. Expected engagement may require circumstances, which place a strong emphasis on achievement recognition in order to translate into improvement in therapy. As previously discussed, high engagement with the therapy process may mediate the relationship between expected engagement and improvement. The focus on achievement recognition that CBT provides may have increased engagement, hence making therapy type a moderator of the mediated relationship between expected engagement and therapy outcome via engagement. However, the current study did not find therapy type to be a mediator of the relationship between expected engagement and improvement, so future research that investigates moderated mediation through engagement is required.

Higher expected engagement with a cognitive approach only predicted more improvement in CBT in the absence of credibility. This finding provides evidence for expected engagement as a distinct concept from credibility as another form of expectations. Therefore, it appears that expected engagement can go some way to explaining why clients who have a low perception of credibility in CBT may still improve after CBT, if this low credibility is counteracted by moderate or high expected engagement. If a client has high expectations about their engagement with the specific techniques of CBT, rather than the credibility of the techniques themselves, experiencing such techniques in therapy may still produce the positive emotions and behaviours necessary for symptomatic improvement (Higginbotham, 1977; Horenstein & Houston, 1976). It may be that higher expected engagement with low credibility indicates more of a readiness to engage with any therapy and change as an autonomous client rather than being dependent on the therapy (DiClemente & Prochaska, 1982; Robins & Hayes, 1993). This would account for the trend found for expected engagement to predict PCET outcome. As a new concept, further research is necessary to understand the relationship of expected engagement with other forms of expectation.

Expected engagement did not significantly predict therapy completion. Previous research has shown clients who drop out to improve less than those who complete therapy (Pekarik, 1986); understanding why patients drop out of therapy, therefore, is important. The test of differences prior to regression showed that patients who complete therapy do not have higher expectations than patients who drop out of therapy, which suggests that other factors influence the decision to remain in or leave therapy, such as the progression of therapy itself, or perhaps that expectations of engagement are not realized once therapy begins.

Implications

The ability of expected engagement to predict symptomatic improvement in CBT provides some confirmation that there is a niche and a utility in researching expected engagement. Recognition of expected engagement as a determinant of therapy outcome is an important stage in the acknowledgement of clients' active role in their own recovery. Clients' insight and self‐awareness is an invaluable source of information in treatment decision making as clients are experts on their own abilities and limitations (Bohart & Tallman, 2010). Furthermore, such personal insight needs to be applied to a specific set of therapeutic techniques; otherwise, clients' expectations cannot be accurate. Previous client factors that have been explored as predictors of therapy outcome, such as gender, have been criticized for their selective irrelevancy to the therapeutic context (Beutler, 1991). The ShEETS has contributed to the field by successfully using clients' insight about themselves that is applied to the therapeutic context to predict symptomatic improvement.

The ShEETS is a tool designed to facilitate personalized treatment that is client led, in response to acknowledgement of the client as an active contributor to their treatment (Bohart & Tallman, 2010). This has long been recognized as an important aspect of therapy, but few practical steps have been taken to involve the client more in their own recovery (Norcross & Wampold, 2011, 2018). For example, the U.K. IAPT service currently has no evidence‐based client involvement in treatment decision making. Therefore, with further research, it would be possible for the ShEETS to be used for client involvement in deciding which therapy would be most effective for them as an individual.

Limitations and future research

Future research into expected engagement should endeavour to confirm whether the predictive ability of expected engagement with a cognitive approach is limited to CBT. The present study was limited in its application to only those receiving CBT or PCET. Future research should aim to replicate the study's findings in a wider range of cognitive therapies to understand whether the behavioural component in CBT plays a key role in the relationship between expected engagement and therapy outcome.

The present study raises further questions, namely, the reason for a CBT‐specific effect but no moderator effect by therapy type. There are likely other factors not investigated in the current research, such as therapist effects, that contributed to a lack of moderator effect. For example, variability in therapist effects within therapy type may have moderated the relationship between expected engagement and therapy outcome, which could have masked a moderator effect by therapy type. Future research should ensure that therapist skill and experience is controlled for in order to further explore moderation by therapy type on the effect of expected engagement on symptomatic improvement.

CONFLICT OF INTEREST

The authors have no conflicts of interest to declare.

PERMISSIONS

The ShEETS is in the public domain and is copyright free. Please contact the corresponding author so that use can be recorded if you wish to use the ShEETS.

	Not at all	A little	Somewhat	Quite a lot	Extremely
The therapist
1. Being empathic towards you	0	1	2	3	4
2. Supporting you emotionally	0	1	2	3	4
3. Helping you understand behaviours that contribute to your depression	0	1	2	3	4
4. Being supportive when you experience negative or overwhelming experiences	0	1	2	3	4
5. Encouraging homework, such as trying out new ideas and experiences outside therapy	0	1	2	3	4
6. Showing an understanding of your problems	0	1	2	3	4
7. Reflecting back the meaning of your thoughts and feelings	0	1	2	3	4
8. Helping you understand thoughts that contribute to your depression	0	1	2	3	4
9. Showing warmth	0	1	2	3	4
10. Encouraging exploration of underlying feelings	0	1	2	3	4
11. Supporting you developing new ideas and perspectives to help create solutions	0	1	2	3	4
12. Being involved in your therapy	0	1	2	3	4

Table A1

Regression model of PCET expected engagement to predict symptomatic improvement

Variables	b	SE b	t	df	p	95% CI
PCET expected engagement	0.55	0.37	1.49	78	.140	[−0.18, 1.27]
Therapy type	−0.63	1.29	−0.49	78	.624	[−3.19, 1.93]
PCET Expected Engagement × Therapy Type	−0.29	0.56	−0.51	78	.609	[−1.41, 0.83]
Credibility	0.08	0.28	0.27	78	.786	[−0.48, 0.63]
PCET Expected Engagement × Credibility	−0.16	0.11	−1.44	78	.153	[−0.38, 0.06]
Gender	−2.06	1.28	−1.61	78	.112	[−4.62, 0.49]
No. of sessions	0.02	0.13	0.17	78	.867	[−0.24, 0.28]
Completion	6.28	1.59	3.95	78	<.001***	[3.12, 9.45]

Abbreviations: CI, confidence interval; PCET, person‐centred experiential therapy.

p < .001.