Effective nurse-patient relationships in mental health care: A systematic review of interventions to improve the therapeutic alliance.

BACKGROUND: Therapeutic alliance is a core part of the nursing role and key to the attainment of positive outcomes for people utilising mental health care services. However, these relationships are sometimes difficult to develop and sustain, and nursing staff would arguably benefit from evidence-based support to foster more positive relationships.
OBJECTIVES: We aimed to collate and critique papers reporting on interventions targeted at improving the nurse-patient therapeutic alliance in mental health care settings.
DESIGN: Systematic literature review. DATA SOURCES: The online databases of Excerpta Medica database (Embase), PsycINFO, Medical Literature Analysis and Retrieval System Online (MEDLINE) and Cumulative Index of Nursing and Allied Health Literature (CINAHL) were searched, eligible full text paper references lists reviewed for additional works and a forward citation search conducted. REVIEW
METHODS: Original journal articles in English language were included where they reported on interventions targeting the nurse-patient therapeutic relationship and included a measure of alliance. Data were extracted using a pre-determined extraction form and inter-rater reliability evaluations were conducted. Information pertaining to design, participants, interventions and findings was collated. The papers were subject to quality assessment.
RESULTS: Relatively few eligible papers (n = 8) were identified, highlighting the limitations of the evidence base in this area. A range of interventions were tested, drawing on diverse theoretical and procedural underpinnings. Only half of the studies reported statistically significant results and were largely weak in methodological quality.
CONCLUSIONS: The evidence base for methods to support nursing staff to develop and maintain good therapeutic relationships is poor, despite this being a key aspect of the nursing role and a major contributor to positive outcomes for service users. We reflect on why this might be and make specific recommendations for the development of a stronger evidence base, with the hope that this paper serves as a catalyst for a renewed research agenda into interventions that support good therapeutic relationships that serve both staff and patients.

What is already known about the topic?

The therapeutic relationship between nursing staff and patients in mental health care is key to positive outcomes.

Therapeutic alliance is a multi-faceted concept that is valued by patients while being difficult for staff to develop.

What this paper adds

Interventions targeting the therapeutic alliance were identified; including training, psychological formulation, reflective groups, consultation and shared activity.

The review showed that none of these is as yet confirmed effective and that the methodological quality of the evidence base at present is predominantly poor.

Introduction

Nursing staff are the core of the caring profession and central to their role is the development of effective relationships with the individuals they support (Hoeve et al., 2014; Zugai et al., 2015). In the United Kingdom, engaging meaningfully with patients (rather than ‘doing to’) runs through the principles of the nursing profession (Royal College of Nursing, 2010); while in the United States, the interpersonal relationship is seen as foundational to a person-centred, recovery-oriented approach within mental health nursing (Kane, 2015). Internationally, the alliance features heavily in the remit of European and Australasian nursing associations, who highlight the need for human connection and the ability to demonstrate effective therapeutic relationships (Australian College of Mental Health Nurses, 2018; World Health Organisation, 2003).

The nursing relationship has been conceptualised as ‘a significant therapeutic interpersonal process [that] functions co‐operatively with other human processes that make health possible for individuals and communities’ (Peplau, 1988, p. 16). This therapeutic alliance—the relationship connecting professional and service user—is between one human and another, with a uniqueness of each dyad (Forchuk, 1995), which therefore requires renewed efforts at each new pairing. Nurses see the development of the alliance as requiring a convergence of interpersonal professional skills with personal life experience (Scanlon, 2006) and clients view it as life-sustaining in its ability to foster collaboration and a sense of being understood (O'Brien, 2001). The concept has recently enjoyed a renewed focus, with an emphasis on consumer-models that encourage personal recovery assisted by therapeutic alliance (Zugai et al., 2015).

Therapeutic alliance has the greatest impact on treatment outcomes for those with mental health difficulties, over and above the specific mode or model of intervention that is provided (Duncan et al., 2010; Martin et al., 2000; Messer and Wampold, 2002; Priebe and McCabe, 2006; Wampold, 2001). Emerging initially within the psychoanalytic discipline and later generalised to multiple therapeutic contexts, the concept of alliance has been defined as an agreement on goals, tasks and a therapeutic bond between therapist and client (Bordin, 1979). The phenomena can be measured using various tools from both the perspective of the professional and service user, with important differences between the two (Bachelor and Salamé, 2000; Fitzpatrick et al., 2005). In an age of increasingly remote therapeutic interactions, the alliance and concomitants of it are still seen as intrinsic to change, whether therapy is facilitated by phone (Mulligan et al., 2014), online dialogue (Cook and Doyle, 2002) or even fully-automated chat-bot (Fitzpatrick et al., 2017).

Traditionally, research has focused on understanding and improving the alliance between therapist and client; as part of one-to-one, psychotherapeutic interventions (Lambert and Barley, 2001; Martin et al., 2000). In contrast to these direct therapy roles, there are also staff members who adopt a care coordination role or act as a key worker; assessing, engaging and organising care with individuals (Burns, 2004; Simpson, 2005; Thurston, 2003). All of these roles inevitably involve the building and maintenance of an effective alliance and therefore research has shifted to include definitions and exploration of the relationship as built within these guises (Farrelly et al., 2014; Kirsh and Tate, 2006), where it remains correlated with outcomes (Cruz and Pincus, 2002; Howgego et al., 2003) and is developed in a range of settings.

As part of secondary care community services, nursing staff coordinate care and deliver brief therapies, which despite (or possibly because of) their short-term nature require the adept building of alliance. People supported by secondary care services are those experiencing severe mental health difficulties, often in the context of challenging relational and social circumstances. Here, nurses strive to develop mutuality, reciprocity, synchrony, and sense of belonging with their clients (Spiers and Wood, 2010) and clients value being known and related to as a person rather than service-recipient (Shattell et al., 2007), requiring a skilful use of the self (O'Brien, 2000). Service users value therapeutic relationships with care coordinators in community settings, and view these as central to recovery; over and above the role of specific care plans (Simpson et al., 2016). These complex processes are rendered difficult for staff members by burnout, struggles building engagement with patients and ineffective team-working (Koekkoek et al., 2011; Singh, 2000), with the nature of organisational structures and roles limiting the care that nurses can provide (Simpson, 2005).

The therapeutic role of nursing staff in mental health care is especially pertinent in settings such as inpatient wards, where patients interact with nurses for the largest proportion of time and the relationship with them is cited as key to therapeutic progression (Hopkins et al., 2009; McAndrew et al., 2014), with a perceived interplay between therapeutic relationships and the quality of care (Coffey et al., 2019) . Engagement in these challenging contexts requires a balance of approaches, the development of personalised understanding and use of the self to facilitate recovery-oriented growth of the patient (McAllister et al., 2019). However, how these specific competencies can be developed is not fully elucidated in the literature or supported by service structures. The alliance can be impeded by individual and organisational factors that leave it unseen and stifled in practice (Pazargadi et al., 2015), with nurses left to negotiate contradictory and challenging relational minefields (Cleary et al., 2012).

Despite the potential value and best efforts, attempts to develop strong nursing alliance in mental health care can be hampered by challenging settings where its development is impeded (McAndrew et al., 2014). Moreover, interactions between nursing staff and patients are often not supported or guided by the psychological theory of relationships; there is a substantial theory-practice gap (Cameron et al., 2005) that might leave both staff and service users vulnerable to relational difficulties and the consequent impact on wellbeing and outcomes. It seems that services, patients and staff place value on the therapeutic alliance, its core attributes have been explored and conceptualised, and yet theoretically-driven, evidenced systems that support its development and maintenance are lacking. With this deficiency in targeted support, staff members increasingly report feeling burnt-out as a result of managing complex and emotionally difficult relationships (Holmqvist, and Jeanneau, 2006; Nathan et al., 2007). This could lead to compassion sometimes waning when it is needed most (Lombardo, and Eyre, 2011; Ray et al., 2013) and staff retention proving extremely difficult, with subsequent strategic priorities to improve this in both the United Kingdom and internationally (Andrews, and Wan, 2009; European Commission, 2014; NHS, 2019; Parliament of Australia, 2002).

There has been some tentative progress in supporting effective relationships, with indications that clinical supervision can protect against staff burnout (Edwards et al., 2006) and psychologically-informed case discussions can enhance positive feelings towards service users and reduce staff self-blame (Berry et al., 2009). Team-based training to develop staff skills using psychological models can even improve patient engagement with the service (Caruso et al., 2013). However, there is no comprehensive, critical summary of interventions that have specifically targeted the element of treatment that we know is essential; the therapeutic alliance, for the group who potentially has the capacity and context to deliver compassionate, caring relationships; nurses.

Aims

In conducting this review, we systematically collated and critiqued studies reporting on interventions targeting the therapeutic relationship between nursing staff and service users in mental health settings. We aimed to answer some key questions; what intervention methods have been tested, in what clinical contexts and with which groups, what outcome measures were utilised, what effects were demonstrated and what was the quality of the methods used.

Method

Protocol and registration

The review protocol was pre-registered and is available online within the international prospective register of systematic reviews (PROSPERO) under the registration number CRD42018111022.

Criteria for inclusion

Papers were considered eligible for inclusion if they: were published as an original, peer-reviewed journal article; written in English; included an intervention aimed to improve the therapeutic relationship between nursing staff and user/s of psychiatric or mental health services; included analysis of the impact of the intervention utilising a standardised measure of alliance. The review includes studies from any mental health service context (such as community, inpatient), client groups of any age, with any mental health diagnosis or need and interventions targeted at the relationship with either qualified or non-qualified nursing staff. Where the staff group incorporated other disciplines, these findings were still eligible if that included nurses in the overall sample.

Search methods

The current review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (Moher et al., 2009). Search terms were generated through review of empirical and opinion pieces regarding alliance and systematic review papers of alliance (for example, Elvins, and Green, 2008; Flückiger et al., 2012; Horvath, and Luborsky, 1993; Priebe, and McCabe, 2006; Zaitsoff et al., 2015), scoping literature searches and consultation with experts in the field. A systematic search of Excerpta Medica database (Embase), PsycINFO, Medical Literature Analysis and Retrieval System Online (MEDLINE) and Cumulative Index of Nursing and Allied Health Literature (CINAHL) databases was conducted on 05/07/18 and then updated on 04/04/19 using the following search strings: (nurse* OR nursing OR staff) AND (alliance OR relationship*) AND (mental OR psychiatry*) AND (improve* OR interven* OR change OR support). The reference lists of eligible papers were also consulted for any additional studies and a forward citation search undertaken.

Data collection and analysis

All potentially eligible records were imported into Endnote reference management software package (Version 8) and duplicate references identified and deleted. One reviewer screened titles and abstracts for relevance, using the inclusion criteria set out above and alongside regular discussion with the research team. Another independent reviewer blindly assessed 50% (randomly selected) of the full texts against the inclusion criteria demonstrating 88% agreement, with any remaining disagreements resolved through discussion with the project team, resulting in full agreement. Data extraction was guided by a pre-specified data extraction sheet detailing key features of the study: sample, setting, design, intervention, outcome measure, analysis, effect size, limitations. Authors were contacted where effect size data was not available in the original paper. The review used a narrative synthesis approach, whereby an attempt was made to go beyond a description of the studies to explore relationships within and between them (Popay et al., 2006).

Assessment of methodological quality

In order to evaluate the methodological rigour of the studies included and therefore to inform the critical synthesis of the findings produced and subsequent recommendations, the papers were assessed using a standardised quality tool. There is a distinct lack of assessment tools that fulfil both the need to be demonstrably reliable and valid and also appropriate for use with a range of study designs. Based on a previous review of quality assessment tools (Deeks et al., 2003), the Effective Public Health Practice Project tool (Effective Public Health Practice Project, 1998) was selected as one that could offer valid (Thomas et al., 2004), reliable (Armijo-Olivo et al., 2012) and flexible appraisal of varying study designs. All of the eligible papers were assessed for quality by the first author and blind second-rated, with 86% agreement demonstrated. Initial ratings from the first author were reviewed collaboratively with the second rater and a decision was made to retain these scores.

Results

Flow of records

The flow of records through the review process can be seen in Fig. 1, in line with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines (Moher et al., 2009). The search resulted in a total of eight papers which met inclusion criteria. These are summarised in Table 1.

Fig. 1

Flow of records

Table 1

Summary of papers and quality ratings.

Author, year	Design	Sample and setting	N	Nature of intervention	Primary outcome	Summary of findings	Effect size (Cohen's D)*	EPHPP quality rating⁎⁎
Berry et al., 2012	Pre-post intervention, no control group	Adult low secure psychiatric inpatient unit, NHS, UK.Nursing staff employed at least three months, 10 h contact with patients per week, including day shifts.	25 baseline (60% mental health nurses); 13 with both intervention and follow up measures	Three hour workshop by clinical psychologist, eliciting psychological factors to understand patients.	WAI-short form Staff report	No significant change in WAI. Rated as helpful, relevant by attendees. Low take-up (42% of baseline).	0.28 (not in favour of intervention)	A Selection 2B Design 2C Confounders 3D Blinding 3E Data 1F Withdrawal 3GLOBAL Weak
Berry et al., 2016	Single-blind cluster RCT. Intervention plus TAU versus TAU alone	Adult psychiatric rehabilitation units, UK NHS and private. Nursing staff on units, at least three months experience, including day shifts.	85 baseline (42% mental health nurses), 74 at follow up.	24 1 h sessions per unit over six months. Facilitated by clinical psychologist. Formulation-based with written report and care planning.	WAI- short form (staff and patient report), completed in relation to key worker	Intention to treat analysis controlling for clustering. T-tests showed no significant difference for staff or patient reported outcomes. 87% completed follow up.	0.45 staff (not in favour of intervention)0.92 patient	A Selection 2B Design 1C Confounders 1D Blinding 2E Data 1F Withdrawal 2GLOBAL Strong
Byrne and Deane, 2011	Pre and post, no control group	Adult community services for severe mental illness, Australia. Workers (including nurses), at least 12 month experience, therapeutic relationship at least 12 months.	46 clinicians (‘mostly nurses’)	Three day workshop on medication alliance training, clinician attitudes addressed. Six and 12- month follow up.	WAI- short form, clinician rated only.	72% remained until completion. Significant change in WAI between baseline and six mth.	0.53	A Selection 2B Design 2C Confounders 3D Blinding 3E Data 1F Withdrawal 2GLOBAL Weak
Carpenter et al., 2007	Quasi-experimental design. Results compared between training cohort and cohort from another setting, also pre-post course evaluation.	Staff training context with follow up of outcomes with patients in community, mostly with psychotic disorders, NHS, UK.	36 trainees (60% nursing)	Training in psycho-social interventions	Non-validated measure of relationships between service user and trainee, service-user rated.	No significant differences between time 1 and 2 (6 months). Some differences between those in training cohort and those not.	Insufficient data to report.	A Selection 2B Design 2C Confounders 3D Blinding 3E Data 3F Withdrawal 3GLOBAL Weak
Kellett et al., 2019	Two nested studies: pre-post one group design with three sites (1) and case series (2)	Secondary Care Community Mental Health NHS-based services (UK)	58 staff-patient dyads in study 1, five dyads in case series (study 2). Hard to engage clients not receiving individual therapy.	Cognitive analytic consultancy delivered by accredited practitioners to staff-patient dyads. Five sessions (typically lasting 1 h)	Working alliance inventory: staff and patient long form in study 1 (NB only one site used this with n = 12) and short form (Staff and patient versions) in study 2.	No significant difference in study 1. Client-rated WAI improved significantly between intervention and follow-up (large effects size). No significant difference for staff- rated.	Study 1client = 0.22 and staff = 0.23Study 2(case series)client = 1.11 staff = 0.29	A Selection 2B Design 2C Confounders 3D Blinding 3E Data 1F Withdrawal 3GLOBAL Weak
Molin et al., 2018	Pre-post visual display of findings, no control group.	Three adult psychiatric clinics in three hospitals in Sweden.	50 staff members across three ‘systems’. 70% nurses in training or registered nurses.	‘Time together’ intervention, focusing on protected time for shared activities between staff and patients.	Caring professional scale completed by patients	No reported effects on quality of interactions. Statistics not reported.	Insufficient data to report.	A Selection 2B Design 2C Confounders 3D Blinding 3E Data 1F Withdrawal 3GLOBAL Weak
Moreno-Poyato et al., 2018	Participatory action research. Quasi-experimental pre-post design.	Adult acute psychiatric hospital, Barcelona, two sites	26 nurses, four dropped outPermanent staff, at least 21 h per week.	Ten months. Individual interactions, reflective groups, scientific texts.	WAI-short form, clinician-rated	Significant difference between groups post-intervention Wilcoxon signed rank	1.14	A Selection 2B Design 2C Confounders 1D Blinding 3E Data 1F Withdrawal 2GLOBAL Moderate
Stringer et al., 2015	Comparative multiple case study design	Adult borderline personality disorder services, community. Netherlands. Caseloads of nurses approached in random order.	Ten nurses for experimental condition and five from the control condition.	Collaborative care programme versus care as usual. Allocated by patient, not nurse.	STAR- scale to assess therapeutic relationships	No significant effect of intervention.	0.13 at time 2 (nine months)	A Selection 3B Design 2C Confounders 3D Blinding 3E Data 1F Withdrawal 3GLOBAL Weak

In favour of intervention unless otherwise specified.

1 = Strong, 2 = Moderate, 3 = Weak.

Methodological quality

Table 1 provides the overall quality rating for each paper, based on the constituent ratings of the Effective Public Health Practice Project tool and its guideline procedure. Full details of the ratings are available from the first author. As is evident, six of the eight studies were rated as weak methodologically. Berry et al. (2016) was the only study that met ‘strong’ design criteria, while Moreno-Poyato et al. (2018) was deemed moderate. The studies were predominantly down-rated due to lack of randomised designs, blinding, confounder control and full reporting of retention information.

Participant and settings

The eligible papers included interventions delivered in a range of settings (inpatient and community services, acute, rehabilitation), in a number of different countries (United Kingdom, Australia, Sweden, Spain, Netherlands) for individuals with a range of conditions under the umbrella of severe mental health problems, including psychosis and difficulties associated with a diagnosis of personality disorder. As per the inclusion criteria, all studies targeted the relationship between mental health nursing staff and service users, although there were a range of additional criteria. The range of settings and participants is reflective of the variety of roles and services within which the therapeutic alliance is fostered.

Interventions

The interventions varied in scope, focus and theoretical underpinning. Berry et al. (2012, 2016) drew on psychological formulation (allied mainly to a cognitive behavioural model) to foster understanding of drivers for patients’ behaviour and clinical presentation within staff group workshops. Carpenter et al. (2007) offered group training in psycho-social interventions that included focus on specific models and also core values of therapeutic engagement. Kellett et al. (2019) utilised a consultancy model provided to pairs of staff members and service users, with cognitive analytic therapy as the underlying approach. Molin et al. (2018) focused on the structure of ward timetables, creating space for time spent together between nursing staff and patients, providing opportunity for joint activities and meaningful engagement. The authors involved staff on each ward in the precise implementation structure to provide a good fit to routine ward activities and preferences. Moreno-Poyato et al. (2018) utilised a participatory action research approach whereby the nurses involved in the study generated intervention elements in a bid to reach best practice, which ultimately involved daily interactions with individual patients, reflective groups for staff and study of scientific texts as selected by staff. Stringer et al. (2015) adopted a broad-based collaborative care programme that consisted of elements including understanding (via timeline work), structural changes (teams, treatment plans), specific interventions (problem solving) and psycho-education. The primary focus of Byrne and Deane's (2011) intervention was to improve medication adherence, with the alliance targeted as a mediating variable, involving modifying clinician beliefs about non-adherent patients that can pose barriers to the relationship and support for adherence.

Outcome measurement

The majority of the studies (Berry et al., 2012, 2016; Byrne and Deane, 2011; Kellett et al., 2019; Moreno-Poyato et al., 2018) utilised the Working Alliance Inventory (WAI; Horvath and Greenberg, 1989) as the primary alliance outcome, reflecting its prominence in the literature, although only Berry et al. (2016) and Kellett et al. (2019) used both the patient- and clinician-rated version, with the rest opting for the latter only. Others (Molin et al., 2018; Stringer et al., 2015) used the Caring Professional Scale (Swanson, 2000), as rated by service users and the scale to asses therapeutic relationships (STAR; McGuire-Snieckus et al., 2007), respectively. Carpenter et al. (2007) relied on a non-validated but user-defined outcome scale (Barnes et al., 2000), which included numerous items relevant to the alliance; including involvement, listening and understanding.

Four of the eight included studies reported no statistically significant difference as a result of the intervention in terms of assessed therapeutic alliance (Berry et al., 2012, 2016; Molin et al., 2018; Stringer et al., 2015). Molin et al. (2018) did not provide data to facilitate a more nuanced interpretation. Berry et al. (2016) findings are intriguing as staff-rated alliance was better for the control than intervention group (similar to Berry et al., 2012), whereas client-rated alliance delivered a large effect size in favour of the intervention, which was presumably then diluted by a more rigorous analysis method in their strongly rated study. Stringer et al. (2015) showed a small effect in favour of the intervention group at nine months, although crucially, both groups declined in reported alliance through the course of the intervention period, leaving the findings difficult to interpret and confounded by a weak study design.

Carpenter et al. (2007) did not report longitudinal statistical improvement but did find a difference between the intervention and control groups in terms of user-rated involvement with staff, although sufficient data was not available to comment on the size of this effect and the weak study quality rating detracts from the significance. Byrne and Deane (2011) reported a statistically significant change in Working Alliance Inventory scores between baseline and six months following the alliance intervention with a medium effect size, although no improvement at 12 month follow-up and concerns raised by the weak study quality rating. Moreno-Poyato et al. (2018) demonstrated a significant difference in post-intervention alliance scores between the active and comparison group, with a median difference of 7 scale points, and large effect size. This study was one of only two in the review to score better than a weak rating in terms of quality assessment, placing more weight on the potential effectiveness of their collaboratively developed intervention. Kellett et al. (2019) demonstrated statistically significant change and a large effect size improvement for client-rated alliance in their case series study, but not for staff-rated alliance or for their larger (n = 12 staff-client pairings) pre-post service evaluation. The weak study rating also means that these findings should be interpreted with caution, despite the encouraging ratings from service-user perspectives.

Discussion

We set out to collate and synthesise information pertaining to interventions that aim to support effective therapeutic alliance between nursing staff and users of mental health services, which has been shown to be central to positive outcomes in both psychotherapy and broader engagement contexts (Cruz and Pincus, 2002; Duncan et al., 2010; Howgego et al., 2003; Martin et al., 2000; Messer and Wampold, 2002; Priebe and McCabe, 2006). The central finding of note, in the authors’ view, is the dearth of studies in this area. It seems surprising that an element of care that is so intrinsic to both patient progress and clinician role does not have a robust evidence base on which to draw in order to provide effective foundations for its development and maintenance. It might be that this has emerged from an assumption of alliance building and maintenance as an implicit ability rather than a skill to be honed and scaffolded, or of the difficulty in addressing a complex issue with a readily-evaluated design.

The evaluated interventions adopted a range of methods, including psychosocial approaches (Berry et al., 2012, 2016; Carpenter et al., 2007; Kellett et al., 2019), those targeting specific clinician attitudes (Byrne and Deane, 2011) and those derived from action research, where the intervention content was collaboratively developed with staff in an attempt to bridge the gap between current and best practice (Moreno-Poyato et al., 2018). The lack of a coherent, shared theoretical underpinning might contribute to the deficiency of robust research and consistent evidence base. The interventions largely consisted of group-based programmes, whereas the relationship that is predominantly measured is a dyadic one between one nurse and one service user; this discrepancy might be contributing to lack of consistent positive change. Kellett et al. (2019) did demonstrate change when the dyadic relationship was specifically targeted and measured as such in the context of a dynamic consultation approach rather than workshop-based training. It is also not entirely clear what level of public (staff and/ or service user) involvement there was in the development of the intervention packages, which might have enhanced the feasibility, acceptability and ultimately, impact (Brett et al., 2014).

Data gathered by the current review do not permit recommendations in terms of effective interventions and their theoretical underpinnings. The evolution of the concept of the therapeutic alliance, from one of dyadic and confined psychotherapeutic engagement to one that often spans relationships within complex team structures, with multiple professionals at once and in varying roles outside of the narrow remit of formal therapy no doubt contribute to the difficulty in developing interventions that are sufficiently targeted yet appropriately flexible. Given the nuanced conceptualisation of the core elements of therapeutic relationships between different settings (Cleary et al., 2012; McAllister et al., 2019; O'Brien, 2000; Pazargadi et al., 2015; Shattell et al., 2007; Spiers and Wood, 2010), it might be that the development of an alliance intervention needs to draw from both theoretical understanding, alongside setting-based contextual needs, which will need to be explored and supported in situ.

As the methodological quality assessment indicates, the studies were predominantly under-controlled, with a lack of randomised designs and matched comparison groups or controls. This limits conclusions that can be drawn even from those studies where positive outcomes were identified. The relatively small samples sizes might also have limited statistical power to detect differences, and those studies that showed promise in terms of effect sizes will need to be replicated with more adequate samples and appropriate statistical control. Few studies checked the validity of their intervention with any kind of fidelity tool (Santacroce et al., 2004). It is therefore also possible that negative outcomes were due to a dilution of or divergence from the intervention protocol, rather than necessarily an intrinsic limitation of the active elements. The lack of statistical control also highlights difficulties central to the issue of intervention design and evaluation in this field. As the alliance is conceptualised as a reciprocally developed, genuine human connection, then how can this be reliably supported and measured between different pairs of individuals, where the goals, needs and definition of the relationship will be inherently idiosyncratic? The current review therefore also points to the need for a range of evaluation methods, alongside the development of controlled trials, which can adequately capture the key elements of the alliance both in terms of supporting its development and evaluating its progress.

Most studies chose to measure outcome using the working alliance inventory. Although fairly ubiquitous, this might be somewhat problematic. The Working Alliance Inventory is a tool designed for use in the context of therapy that does not fully converge with the nature of a nurse–patient relationship, which might often be more accurately conceptualised as a therapeutic key worker role, where therapeutic conversations might be had in the context of a broader care-coordination or care-planning role (Burns, 2004; Simpson, 2005; Thurston, 2003). Thus, the relationship of interest and its quality might not be accurately evaluated by the use of Working Alliance Inventory. In terms of the source of the ratings, two studies that reported mixed findings (Kellett et al., 2019; Berry et al., 2016) demonstrated larger effect sizes for the client perspective as compared to the staff member. This underscores the importance of considering multiple sources but also the need to elaborate on the best method for determining good relational outcomes, especially since client-rated alliance is the better predictor of therapeutic progress (Bachelor, 1991; Fitzpatrick et al., 2005)

Strengths and limitations of the review

We aimed to provide a comprehensive and critical summary of current evidence pertaining to intervention to support nursing staff in their therapeutic relationships with people utilising mental health services. The systematic, thorough nature of the review offers a contribution to an important area of literature hitherto unreported. The small sample of studies is reflective of the field as it stands while also limiting the utility of the review and it will be important to repeat the process as the evidence based develops, incorporating more nuanced analysis of the findings and methods and robust meta-analysis or meta-synthesis, where the nature of the interventions delivered and their qualitative impact could be more fully explored. Our decision to include both qualified and non-qualified staff interventions, and to include studies whose sample include nurses alongside other staff groups was pragmatic in nature to ensure any relevant intervention was highlighted, particularly in a limited evidence base. However, future work might wish to consider how to delineate samples more precisely and thus provide evidence best-suited to the target discipline. This review has focused on interventions that specifically target the therapeutic alliance as a core aspect of effective mental health care. It is acknowledged that the alliance is part of the approach of other modality-specific therapies, such as cognitive behavioural therapy, dialectical behaviour therapy or cognitive analytic therapy (Bennettet al., 2006; Bedics et al., 2015; Gilbert, and Leahy, 2007). Therefore, synthesising findings from modality-specific intervention research, where the alliance might be evaluated as a mechanism of change, with the current, more focused conceptualisation might be warranted.

Clinical implications and recommendations for future work

The nature of the evidence base as it stands renders it difficult to make clear clinical recommendations in terms of how nursing staff should be best supported to develop and maintain effective therapeutic relationships with the individuals they work with in the settings they operate. There is some indication that interventions targeting clinician attitudes and reflective capacity, relational understanding and dynamic interactions might be helpful, although further work is needed. There is also a relative absence of service user and clinical staff involvement in the development of the reported interventions, which might limit their acceptability and feasibility. This paper, therefore, should serve as an impetus to develop a clear trajectory of research endeavours that build on the current findings and creates a more robust body of work incorporating the following key elements: i) intervention based on strong theoretical underpinnings and service user and clinician involvement; ii) methodologically sound studies; iii) assessment of fidelity to the intervention model; iv) targeting and evaluation of the alliance aligned with its conceptualisation as a dyadic, mutual, professional relationship outside the specific bounds of psychotherapy.