Therapist Self-Disclosure and the Therapeutic Alliance in the Treatment of Eating Problems

RUNNING HEAD: Therapist self-disclosure

Therapist self-disclosure and the therapeutic alliance in the treatment of eating problems

Laura M. Simonds1*Naomi Spokes2

1School of Psychology, University of Surrey, Guildford GU2 7XH, UK

2Tavistock and Portman NHS Foundation Trust, City and Hackney Primary Care Psychotherapy Consultation Service (PCPCS), 2nd Floor, A Block, St Leonard's Hospital, Nuttall Street, London N1 5LZ, UK

*Corresponding author: Dr Laura M. Simonds, School of Psychology, University of Surrey, Guildford GU2 7XH, UK. Telephone: +44(0)1483 686936. Email:

Abstract

Evidence is mixed regarding the potential utility of therapist self-disclosure. The current studymodelled relationships between perceived helpfulness of therapist self-disclosures, therapeutic alliance, patient non-disclosure, and shamein participants (n=120; 95% women) with a history of eating problems.Serial multiple mediator analyses provided support for a putative model connecting the perceived helpfulness of therapist self-disclosures with current eating disorder symptom severity through therapeutic alliance, patient self-disclosure and shame. The analyses presented provide support for the contention that therapist self-disclosure, if perceived as helpful, might strengthen the therapeutic alliance. A strong therapeutic alliance, in turn, has the potential to promote patient disclosure and reduce shame and eating problems.

Keywords:disclosure, therapist, therapeutic alliance, shame, eating disorder

1

RUNNING HEAD: Therapist self-disclosure

Introduction

Research suggests the ease with which a patient discloses in the therapeutic context is a function of multiple influences.The current study placed particular emphasis on perceived therapist self-disclosure in the treatment of eating problems as it has receivedlittle empirical attention in this context. Patient self-disclosure is considered central to effective psychotherapy (Barrett & Berman, 2001) because it facilitatesgreater emotional intimacy, increased insight, and reduces the negative effects of self-concealment (Farber, 2006). Empirical studies support this view, finding that patient self-disclosure ispositively correlated with therapeutic alliance quality (Farber & Hall, 2002). Moreover, meta-analytic findings suggest quality of alliance is strongly associated with intervention success in anorexia nervosa and, to some extent, bulimia (Antoniou & Cooper, 2013). Taken together, these studies underline the importance of promoting self-disclosure.

The use oftherapist self-disclosurehas beena more contentious topic (Barrett & Berman, 2001). Therapist disclosures might be of a personal nature, such as values or sexuality, or those emerging during therapy, such as countertransference reactions and acknowledgement of mistakes (Goldfried, Burckell & Eubanks-Carter, 2003; Skytte-Glue & O’Neill, 2010). The latter have been labelled‘disclosures of immediacy’(Knox & Hill, 2003) given that they happen in the moment. Self-disclosure might involve a ‘reciprocal disclosure’. This refers to a therapist revealing comparable personal experience, and might be done to strengthen the therapeutic alliance, normalise patient experience,and reduce shame (Hill & Knox, 2001).

Mixed views as to the appropriateness of therapist self-disclosureare often influenced by theoretical orientation (Knox & Hill, 2003). For some, therapist self-disclosure might be seen to shift the focus of therapy away from the patient, makingthem too aware of the therapist’s limitations and vulnerabilities (Barrett & Berman, 2001). Conversely, therapist self-disclosure might demystify the therapy process, render the therapist more humane, equalise the relationship, enhance the therapeutic alliance, model disclosure for the patient, normalise their experience, and challenge negative beliefs the patient might hold about their impact on others (Barrett & Berman, 2001; Goldfried et al., 2003; Hill & Knox, 2001; Simi & Mahalik, 1997).

In an early attempt to examine the potential effects of personal therapist disclosures on patient disclosure, Simonson (1976) recruited students to a simulated therapy study. Therapist disclosure ofimpersonal demographic information was associated with more student disclosure than therapist disclosure of personal information. Whilst this study suggests the nature of the disclosed material may be important, evidence indicates the effect of personal therapist disclosures might differ in real therapy contexts. In an interview study with thirteen long-term adult psychotherapy patients, beneficial effects of therapist personal disclosures were reported, although some participants expressed discomfort aboutthe increased intimacy engendered bysuchdisclosures (Knox, Hess, Petersen & Hill, 1997).Barrett and Berman (2001) deliberately manipulated levels of reciprocal therapist self-disclosure. Patients of therapists who were instructed to increase their level of reciprocal self-disclosures reported a greater reduction of symptoms and increased liking for their therapist relative to comparable patients of therapists instructed to limit reciprocal self-disclosure. No differences were found in the frequency or intimacy of patient disclosures between the disclosure conditions,however.

Less research has focussed on disclosures of immediacy. Skytte-Glue and O’Neill (2010) found psychologists consideredpersonal disclosures inappropriate and irrelevant to the therapeutic relationship. Appropriate immediacy disclosureswere deemed useful for displaying empathy and in enhancing the therapeutic relationship, with acknowledgement of mistakes madeby the therapist considered particularly important. Findings reflected the views of a small number of psychologists, however, and did not considerpatients’views. A qualitative study focussing on anorexia nervosa (Oyer, O’Halloran & Christoe-Frazier, 2016) indicated that both patients and therapists considered therapist self-disclosure to promote greater normalisation and a stronger therapeutic relationship. However, only immediacydisclosures were considered appropriate.

In a sample of sixty-eight women witheating disorder symptoms (Swan & Andrews, 2003), forty-two percent reported that they had not disclosed specific information to their therapists. Non-disclosure was correlated with higher levels of shame. Non-disclosure of symptoms has been shown to be significantly correlated with current depression levels in those no longer undergoing therapy (Hook & Andrews, 2005). Taken together, these findings suggest that facilitating disclosure might ameliorate shame, and that reduced shame is likely to promote symptomatic improvement.Currently, there is little empirical evidence as to whether therapist self-disclosures of several different types have a positive impact on non-disclosure and shame in those with eating problems.

Theprincipal aim of the current study was to model relationships between different types of therapist self-disclosure, therapeutic alliance,patient self-disclosure, shame and severity of eating problems. Therapist personal disclosures and immediacy (i.e. therapy-related) disclosures were investigated. Modellingconnections between multiple variables in a cross-sectional study does not permitcausal conclusions. Instead, this method allows assessment of the potential validity of specific ways of modelling connections between variables (i.e. generating theoretical propositions) with the recognition that the variables could be placed in different configurations. As a result, although causal language is necessarily used to construct and discuss the putative theoretical processes under statistical scrutiny, no claim to have demonstrated causality is made. Based on existing evidence and theoretical plausibility, we hypothesized that the perceived helpfulness of therapist self-disclosure would enhance the therapeutic relationship. This, in turn, would promote patient self-disclosure that, in turn would promote a reduction of shame. A reduction in shame would be associated with reduced eating disorder symptoms. The conceptual and statistical diagram for this modelis shown in Figure 1.

Some authors have argued that a clinical sample is uncharacteristic of the majority of people who live with a disorder in the ‘real world’ (Goodman et al., 1997). For this reason, participants in the current study were recruited via an eating disorders charity to ensure the sample was not comprised of those with more severe problems who might be unrepresentative of all those individuals experiencing eating problems. Furthermore, given the study focused on the sensitive issues of shame and disclosure, anonymous recruitment via an independent organisation was considered less intrusive than face-to-face or therapist-based recruitment methods, and posed less of a risk of response bias.

Method

Prior to recruiting participants, the study protocol was reviewed and approved by the authors’institutional research ethics committee.

Participants

An invitation was sent to all current members on a UK eating disorder charity database who had previouslyconsented to be contacted. Inclusion criteria were: at least 16 years old; received psychotherapy for eating problems (current or past); and had at least two intervention sessions following assessment. Participants were not required to have an eating disorder diagnosis although most had. Given that comorbidity iscommon (Wade, Bulik, Neale & Kendler, 2000), participants were not excluded on the basis of additional diagnoses.

Measures

Demographic data were collected along withinformation on specific eating problems, treatment history, and the statistics needed to calculate BMI. Scores from all measures described below were used in hypothesis-testing with the exception of ‘reasons for patient non-disclosure’whichwas used descriptively.Given that participants could have a varied treatment history, those in therapyat the time of the study were asked to think about their current therapeutic experience. Those not in therapy were asked to think about their most recent therapeutic experience.

Reasons for patient non-disclosure. Following Hook and Andrews (2005), eleven pre-determined reasons for patient non-disclosure in therapy were devised. Participants were asked to specify on a scale of 0 (‘not at all’) to 10 (‘completely’) the extent to which each was a reason for non-disclosure. Four items related to aspects of the therapeutic relationship.

Patient non-disclosure in specific areas.There was no existing measureof specific patient non-disclosures, so thirteen items were devised based on previous non-disclosure research (Farber & Hall, 2002). Participants were asked if they were aware of having concealed information that they considered relevant to their difficulties. Items includedfeelings towards the therapist, past adverse experiences, and the psychological difficulties experienced. Ratings were made on a four-pointscale (1 ‘not at all’, 2 ‘a little’, 3 ‘mostly’, 4 ‘all of the time’) with a ‘not applicable’ option available.Individual non-disclosure items were initially used descriptively. For hypothesis-testing, a mean self-disclosure score was calculated. This was done by reverse coding the initial scale such that a higher mean score indicated more self-disclosure (i.e. less non-disclosure). The mean self-disclosure score was calculated by dividing the total by the number of applicable items. The 13 items showed good internal reliability (Cronbach’s alpha.85).

Shame. The Experience of Shame Scale (ESS; Andrews, Qian & Valentine, 2002; Swan & Andrews, 2003) is a 28-item measureassessing characterological, behavioural, bodily shame, and shame around eating. Items are rated on a four-point scale (1 ‘Not at all’to 4 ‘Very much’). The ESS has shown high internal and test-retest reliability (Andrews et al., 2002). Cronbach’s alpha in the current study was:.95(full scale), .93 (characterological), .92 (behavioural), .83 (bodily), and .93 (eating).

Therapeutic alliance. The Working Alliance Inventory-Short version (WAI-S; Tracey & Kokotovic, 1989) is a 12-item measure of three alliance components: ‘agreement on the tasks of therapy’, ‘agreement on the goals of therapy’, and ‘development of an affective bond’. Each item is rated on a 7-point scale (1 ‘never’, 4 ‘sometimes’, 7 ‘always’). Higher scores represent a stronger therapeutic alliance. Cronbach’s alpha in the current sample was .96.

Therapist self-disclosure.Nine items were devised, assessing both personal and immediacy disclosures,based on previous research (Henretty & Levitt, 2010). Participants rated the extent to which they perceived their therapist to have disclosed on a 4-point scale (1 ‘not at all’, 2 ‘a little’, 3 ‘sometimes’, 4 ‘often’). Cronbach’s alpha for the total scale was .73.If therapist disclosure occurred, participants were also asked to rate on a 5-point scale how helpful they had found it (1 ‘very unhelpful’, 3 ‘neutral/no impact’, 5 ‘very helpful’).

Eating problems. The Eating Attitudes Test (EAT-26; Garner, Olmsted, Bohr & Garfinkel, 1982)comprises 26 items rated on a 6-point scale (‘always’ to ‘never’).Responses for each item are given a score from 0 to 3 with the higher scores assigned to responses closest to the symptomatic direction (e.g. ‘always’ for the item‘I am terrified about being overweight’and ‘never’ for the item‘I enjoy trying new rich foods’). A total higher score indicates greater eating problems. The EAT has been used to identify eating disturbances in non-clinical samples (Garner, Rosen & Barry,1998).An error during construction of the survey resulted in item 17 (‘I eat diet foods’) being omitted. Cronbach’s alpha for the 25 items completed was .93.

Procedure

The measures were formatted for online completion. Participant information and consent screens preceded administration of the study measures. Participants were advised that the survey was anonymous and they could withdraw at any time during the survey. In this instance, their data were not used as it was considered to be withdrawn. Contact details for a number of eating disorder and mental health support organisations were made available to participants throughout the survey.

Data analysis

Aside from the missing EAT-26 item, there were no missing values. Data distribution was assessed using histograms and normality tests. Hypotheses were tested using parametric correlationand Hayes (2013) PROCESS Model 6 for serial multiple mediator analysis. Two-tailed tests were used throughout.

Results

Sample characteristics

In total, 208 people accessed the survey with 120 (58%) completingit. Sample characteristics are displayed in Table 1.

INSERT TABLE 1 ABOUT HERE

Extent of patient non-disclosure in specific areas

Table 2 indicates the extent to which participants considered that each of eleven reasons for not disclosing in therapy applied to them. Participants could also rate ‘other’. Most frequently endorsed reasons for not disclosing related to self-consciousness (i.e. shame, guilt, fear of negative judgment). Therapist qualities and aspects of the intervention were rated as applicable, but less so.

INSERT TABLE 2 ABOUT HERE

Table 3indicates that participants reported not disclosing most issues ‘a little’to ‘mostly’with non-disclosure being at a similar level for most items.

INSERT TABLE 3 ABOUT HERE

For the correlation analysis, a patient self-disclosure score was calculated (by reverse coding the non-disclosure items and dividing each participants’total score on all applicable items by the number of applicable items).

Helpfulness of perceived therapist self-disclosures (Table 4)

Participants were asked to indicate if they perceived their therapist to have made a self-disclosure in a number of areas. A large majority (84%) perceived that their therapist had disclosed positive feelings towards them in therapy. Therapist disclosure of their sexuality was uncommon (14%). On average, most therapist disclosures were rated between neutral to helpful.

INSERT TABLE 4 ABOUT HERE

To examine whether perceived helpfulness of therapist disclosures was related to therapeutic alliance, patient self-disclosure, shame, and eating problems, reliability analyses were conducted to assess the viability of aggregating items pertaining to personal and immediacy disclosures into two separate scales. Reliability analysis supportedaggregatingthe personal disclosure helpfulness items (Cronbach’s alpha .84; all corrected item-total correlations >.45). Given not all disclosures were applicable, a mean helpfulness score was derived by dividing total helpfulness score by the number of applicable items. Reliability analysis did not support aggregatingthe helpfulness of perceived therapist immediacy disclosure items. Therefore, these three items were used separately in further analyses.

Correlation analysis (Table 5)

All three helpfulness of therapist immediacy disclosure items (not shown in Table 5) were correlated positively with perceived therapeutic alliance, with moderate to large effect sizes: negative feelings (r=.45, p=.001, N=48); positive feelings (r=.35, p<.01, N=100) and mistakes made (r=.41, p=.002, N=56). Perceived helpfulness of therapist personal disclosures was also significantly positively correlated with therapeutic alliance. Therapeutic alliance was positively correlated with patient self-disclosure. Patient self-disclosure was negatively correlated with all indices of shame-proneness. Total shame was positively correlated with EAT scores. All effect sizes weremoderate to large.

INSERT TABLE 5 ABOUT HERE

Serial multiple mediator analysis

Hayes’ (2013) PROCESS Model 6 was used to test the putative causal pathways indicated in Figure 1. Two analyses were run with the antecedent X variable beingeither helpfulness of personal therapist disclosures, or helpfulness of therapist disclosure of positive feelings. The other two immediacy items were not used because the subsample N was very small (48 for disclosure of negative feelings and 57 for mistakes made). In addition to the direct effect of therapist personal disclosure on eating problem severity controlling for therapeutic alliance (TA), patient self-disclosure and shame (c1), seven indirect effects were tested as depicted in Figure 1. Of particular interest was the indirect effect: therapist personal disclosure to TA to patient self-disclosure to shame to eating problem severity (a1d21d32b3). Bias-corrected bootstrap confidence intervals were calculated to test indirect effects based on 20,000 bootstrap samples.

INSERT FIGURE 1 ABOUT HERE

Helpfulness of therapist personal disclosure.Serial multiple mediation analysis indicated that specific indirect effectof interest (a1d21d32b3)was significantly negative. The bias-corrected bootstrap confidence interval for the indirect effect was entirely below zero (-1.798 to -0.129). The regression coefficients (Tables 6) indicate the greater the perceived helpfulness of therapist personal disclosure the stronger the therapeutic alliance (9.127), the stronger the therapeutic alliance the greater the patient self-disclosure (0.013), the greater the patient self-disclosure the lower the shame (-0.400), and the lower the shame the lower the EAT score (11.595). There was no evidence that helpfulness of therapist personal disclosures was related to EAT scores independent of alliance, self-disclosure or shame (c’ = 2.670, p = .315). There was no evidence of other indirect effects as the bootstrap confidence intervals spanned zero.

INSERT TABLE 6 ABOUT HERE

Helpfulness of therapist disclosure of positive feelings.Results were similar to those for therapist personal disclosures. Indirect effect of interest (a1d21d32b3) was significantly negative. The bias-corrected bootstrap confidence interval for the indirect effect was entirely below zero (-1.510 to -0.136). The regression coefficients (Tables 7) indicate the greater the perceived helpfulness of therapist disclosure of positive feelings the stronger the therapeutic alliance (4.796), the stronger the therapeutic alliance the greater the patient self-disclosure (0.023), the greater the patient self-disclosure the lower the shame (-0.400), and the lower the shame the lower the EAT score (11.403). There was no evidence that helpfulness of therapist disclosure of positive feelings was related to EAT scores independent of alliance, self-disclosure or shame (c’ = -0.356, p = .825). There was no evidence of other indirect effects as the bootstrap confidence intervals spanned zero.