Www.research.manchester.ac.uk



Therapeutic alliance in psychological therapy for people with schizophrenia

and related psychoses: a systematic review

L. Shattockᵃ, K. Berryᵃ*, A. Degnanᵇ, D. Edgeᵇ

ᵃDivision of Psychology and Mental Health; School of Health Sciences, University of Manchester, 2nd Floor, Zochonis Building, Brunswick Street, Manchester, M13 9PL, United Kingdom

ᵇDivision of Psychology and Mental Health; School of Health Sciences; University of Manchester, Room 3.306, Jean McFarlane Building, Oxford Road, Manchester, M13 9PL, United Kingdom

For publication in Clinical Psychology & Psychotherapy

* Corresponding author: School of Health Sciences, University of Manchester, 2nd Floor, Zochonis Building, Brunswick Street, Manchester, M13 9PL, United Kingdom (Email: Katherine.berry@manchester.ac.uk)

Abstract

Therapeutic alliance is a key predictor of therapy outcomes. Alliance may be particularly pertinent for people with schizophrenia as this group often have a history of interpersonal trauma and relationship difficulties including difficult relationships with mental health staff. This review aimed to determine: 1) the quality of therapeutic alliance between people with schizophrenia and their therapists; 2) whether alliance predicts therapeutic outcomes; and 3) variables associated with alliance. Databases were searched from inception up to April 2015. The search yielded 4,586 articles, resulting in 26 eligible studies, involving 18 independent samples. Weighted average client and therapist Working Alliance Inventory-Short Form (WAI-SF) total scores were 64.51 and 61.26, respectively. There was evidence that alliance predicts overall psychotic symptomatic outcomes and preliminary evidence for alliance predicting rehospitalisation, medication use, and self-esteem outcomes. There was evidence for specific client-related factors being linked to different perspectives of alliance. For example, poorer insight and previous sexual abuse were associated with worse client-rated alliance, whereas baseline negative symptoms were associated with worse therapist-rated alliance. Therapist and therapy-related factors, including therapists’ genuineness, trustworthiness and empathy were associated with better client-rated alliance, whereas suitability for therapy, homework compliance and attendance were associated with better therapist-rated alliance. Key clinical implications include the need to consider alliance from both client and therapist perspectives during therapy; and training and supervision to enhance therapist qualities that foster good alliance. Future research requires longitudinal studies with larger samples that include pan-theoretical, well-validated alliance measures to determine causal predictor variables.

Abstract word count: 248

Introduction

Therapeutic alliance is a predictor of therapy outcomes (Horvath et al. 2011; Martin et al. 2000) and is commonly conceptualised according to goal agreement, task agreement and therapeutic bond (Bordin, 1979). Alliance may be particularly relevant to people with schizophrenia and related psychoses as this group are likely to have early traumatic experiences impacting on adult relationships and difficulties with service engagement (Kreyenbuhl, Nossel, & Dixon, 2009; Varese et al., 2012).

Hewitt and Coffrey (2005) reviewed studies investigating alliance within nursing relationships for people with schizophrenia and report a definitive role of alliance in recovery from schizophrenia. However, this review did not provide detailed inclusion and exclusion criteria, nor did it systematically evaluate included studies using a quality appraisal tool. A second review that investigated alliance and outcomes in this client-group found some evidence for alliance predicting fewer hospitalisations, symptom reduction and improved functioning (Priebe, Richardson, Cooney, Adedeji, & McCabe, 2010). However, only one of the studies included in this review considered alliance in psychological therapy, with the remaining studies reporting on alliance within psychiatric or health settings.

There is evidence that psychological therapies are effective treatments for schizophrenia (e.g. Pharoah, Mari, Rathbone & Wong, 2010; Wykes, Steek, Everitt & Tarrier, 2008). Nevertheless, the size of the effect of therapy on outcomes are moderate, suggesting that further work is needed to understand predictors of better outcomes to enhance effectiveness (Turner, van der Gaag, Karyotaki, & Cuijpers, 2014). A growing number of studies have reported alliance as an important therapeutic variable for people with schizophrenia diagnoses due to its role in predicting outcome. Studies have also reported variables affecting the quality of the alliance that develops during therapy. It is important to understand what factors predict alliance so that these can be more closely targeted. For example, public-stigma and self-stigma (the internalisation of negative societal messages and stereotypes about mental health problems) are significant barriers to engagement in therapeutic interventions for people diagnosed with schizophrenia and may be key factors in determining the quality of therapeutic alliance (Vogel, Wade & Hackler, 2007; Wood, Burke, Byrne, Pyle, Chapman & Morrison, 2015; Pyle & Morrison, 2014).

Despite the growing number of studies that evaluate alliance in therapy for psychosis, the developing literature has not yet been synthesised. The aim of this paper is to systematically review studies investigating alliance in people with schizophrenia and related psychoses. Key objectives are to: (i) summarise average alliance ratings across studies, thereby providing a point of comparison for future alliance studies; (ii) examine whether alliance predicts therapy outcomes; and (iii) identify variables associated with the development of good quality alliance. The review also assesses the methodological quality of included studies and provides recommendations for future research.

Method

Search Strategy

The databases Medline, Web of Science and PsycINFO were searched from inception to April 2015 using the following terms:

1. (psychotic OR schizo*OR psychos*s) OR (chronic* OR serious* OR sever*) NEAR/3 (mental*) NEAR/3 (ill* OR disorder*)

AND

2. (therap* OR working* OR helping*) NEAR/2 (alliance* OR relation* OR process*) OR (staff* OR professional*) NEAR/2 (client* OR patient) NEAR/3 (alliance* OR relation* OR process*)

This search resulted in 6,980 citations leaving 4,586 citations when duplicates were removed. One-third (n=1,520) were independently screened at title level by the lead author and a postgraduate student. Following high levels of agreement (98% of cases; k=.77), remaining citations were screened by the lead author leaving 160 that were all screened at abstract level by the lead author and a postgraduate student (91% of cases; k=.79). The lead author screened the remaining 55 citations at a full-text level against specified inclusion criteria. Inclusivity was discussed with the research team, resulting in 26 included articles (see Figure 1).

Inclusion Criteria

The inclusion criteria were:

i) sample with non-affective psychosis

ii) a validated measure of alliance between client and therapist

iii) clients receiving psychological therapy

iv) English language

v) peer-reviewed.

Criterion (i) included samples with severe mental illness (SMI) where >60% had non-affective psychosis. Preliminary searches revealed that studies with SMI samples generally included a proportion of people with non-affective psychosis, thus excluding these studies would limit the evidence available to review. One study (Moran et al., 2014). was excluded as diagnostic information could not be provided by the authors.

For criterion (iii) ‘psychological therapy’ was defined as “…meeting with a therapist (a healthcare professional competent in giving psychological therapy to people with psychosis or schizophrenia) to talk about your feelings and thoughts and how these affect your behaviour and wellbeing” (NICE, 2014). When it was unclear if this criterion was met, the authors were contacted. Four out of six authors responded with information that informed the exclusion criteria - samples included assertive outreach (e.g. Cunningham, Calsyn, Burger, Morse, & Klinkenberg, 2007) and vocational rehabilitation (e.g. Catty et al., 2011).

Data Synthesis

Marked heterogeneity in the methodology of studies and types of relationships measured meant that a meta-analysis was not appropriate. Instead, a narrative synthesis of the literature (Mays, Roberts, & Popay, 2001) was conducted, reporting effect sizes of individual studies where available. The effect size of studies that investigated whether alliance predicted therapeutic outcomes (both symptomatic and other outcomes) are documented in Table 3. If the effect size was not reported, but the β statistic was, the authors adopted Peterson and Brown’s (2005) formula: r = β + .5ƛ, where ƛ=1 on occasions when β is a positive value, in cases where β value was ±0.5 to calculate the effect size. In cases where it was not possible to report the effect size, the results of other relevant statistical tests were reported in Table 3.

Quality Assessment

Methodological quality of studies was assessed using the Effective Public Health Practice Project (EPHPP) Quality Assessment Tool for Quantitative Studies (Thomas, 2003). The tool has good reliability and validity (Armijo-Olivo et al., 2012; Thomas et al., 2004). Consistent with other systematic reviews involving predominantly non-randomised studies (e.g. Safavi, Berry, & Wearden, 2015), the tool was amended to include five relevant rating domains: (i) Selection Bias; (ii) Confounders; (iii) Data Collection Methods; (iv) Withdrawals and Drop-outs; and (v) Analysis (two of the four items). The original version of EPHPP does not include the Analysis domain in final ratings and for consistency this was also excluded from the adapted version. The lead author and a postgraduate student independently rated all papers, with substantial agreement found for overall ratings (92% level of agreement; k =.781).

INSERT FIGURE 1

Results

Study Characteristics

Table 1 presents study characteristics and key findings. Five studies were purely cross-sectional. There were two types of longitudinal designs which were termed ‘alliance baseline’ and ‘alliance outcome’ to differentiate between them. There were eight ‘alliance baseline’ studies which measured variables prior to therapy and measured alliance early (e.g. session 3) in therapy. These studies are discussed alongside the cross-sectional studies as they focus on factors affecting the development of alliance. There were thirteen ‘alliance outcome’ studies which measured alliance and/or other variables at multiple time points over the course of therapy. These latter studies are discussed in a separate section (titled ‘Relationship between Alliance and Outcome’) to described role of alliance in predicting therapy outcomes.

Alliance was most frequently assessed (n=11) using the Working Alliance Inventory-Short Form (WAI-SF; Tracey & Kokotovic, 1989). It was assessed from different perspectives including: client and therapist-rated (n=21), client-rated (n=3), therapist-rated (n=1) and observer-rated (n=1). Therapy was delivered in individual (n=23), group (n=2) and family (n=1) settings. Eight pairs of studies used participants drawn from the same sample (i.e. full dataset or subsample from a larger trial), resulting in 18 independent samples across 26 studies.

INSERT TABLE 1 HERE

Quality Assessment

The results of the quality assessment are summarised in Table 2. Most studies applied additional selection criteria to secondary data (e.g. only including participants who had at least three measures of alliance) and were therefore rated ‘moderate’ for selection bias. Ten studies received a ‘weak’ rating of selection bias due to lack of recruitment and selection detail (n=4), no details of trial given (n=2), participants self-referred (n=1) or less than 60% of selected individuals consented to participate (n=3). Most studies (n=19) considered confounders either in the design and/or analyses, so were given ‘moderate’ or ‘strong’ ratings for this criterion. Seven studies received a ‘weak’ rating for confounders as they did not report controlling for confounders in their design or analyses. ‘Strong’ ratings were given to 22 studies for data collection methods, as they used reliable and valid measures. Withdrawal and drop-outs was not applicable to the cross-sectional studies, but eight longitudinal studies received a ‘weak’ rating on withdrawal and drop-outs due to lack of detail (n=7) or because ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download