Participants
We were interested in all studies that provide evidence on how to improve TC with BME psychiatric patients in the setting of specialist psychiatric care. Specialist psychiatric care is delivered by many professional disciplines. We adopt the term psychiatric services rather than mental health services as the latter is a very broad term that includes public health and social interventions in the community, for example housing and other services, that were not within the scope of the commissioned review.
Key populations included all age groups (young people, adults and the elderly) and all ethnic groups known to be prominent in health-care settings in the UK, i.e. people from Indian, Pakistani, Bangladeshi, Sri Lankan, black Caribbean, black British, black African, Irish and Chinese backgrounds. We used these terms in the search strategy but complemented them with other terms in order to identify the broadest literature of relevance to the UK. Although commissioned as a review of BME groups, our analysis specifies the groups of importance to the UK setting and reveals the ethnic and cultural specificity of interventions.
Although originally commissioned to include only UK studies of the larger ethnic groups, we amended this aspect of the protocol during the study wherever relevant as useful information might otherwise be overlooked, for example if a study included a diaspora population of relevance to the UK (e.g. east Europeans) or if an intervention was judged to be transferable to the UK (e.g. an intervention for African Americans). However, we included such studies only where the other inclusion criteria were met.
During the review we came across some studies located at the interface of specialist psychiatric care and other sectors of care, but requiring specialist input; we included those studies in the review.
Interventions
At the outset, we defined TC as:
[A]ny conversation (face-to-face or technology-assisted) that is undertaken using a pre-defined model that seeks to improve understanding, engagement and therapeutic outcomes. For communication in health care to be therapeutic, it must involve a relationship and exchange of ideas between a patient and professional helper, be patient centred and engaging in order to influence the patient’s emotional world, and directed by the professional using expertise and skill. Therapeutic communications include all interactions that enable people in distress to resolve conflicts, divergent expectations, traumatic histories and adverse life events, and to take up and to overcome distress and also take up offers of help.
Bhui et al.46
In this review we were specifically interested in all interventions seeking to improve TC with BME patients receiving psychiatric care, and expected a broad range of such interventions, for example conflict resolution, cultural consultancy, cultural competence and others as yet undefined.
These interventions might be aimed at either individuals or populations.
Care could be delivered by psychiatrists, general practitioners (GPs), psychologists, nurses or any other professional as long as it was located in specialist psychiatric settings.
From existing knowledge of this field, interventions to promote TC include those that:
employ mediation to enhance mutual understanding and to improve engagement with care
seek to manage divergent views, conflict and differing explanatory models and illness perceptions through negotiation and mediation
include narrative-based interventions (i.e. that place the service user and patient perspectives at the heart of consultation, assessment and treatment)
employ cultural consultation as a process of gathering narratives
apply cultural competence interventions focused on communication
any other new method or process for improving TCs that is not captured by the above, but is suited for BME populations in psychiatric care.
Any of the above processes could be delivered face to face or through two-way real-time communication technologies (e.g. NHS Direct or other support systems, telemedicine or e-mail).
We did not review the literature on interventions that are considered to be (generic) TCs themselves, such as psychological therapies or music therapies, unless the research evidence focused on interventions that might improve TCs with the specified target group(s), while meeting the other inclusion criteria (see Box 2).
Inclusion and exclusion criteria Studies that report evaluations of:
Furthermore, given the evidence base already available,13 we did not include studies that were purely testing models of translation or interpretation as an intervention. The reasons for this were twofold. First, there is already a concurrent enquiry into interpretation, translation and language support (ITALS) in mental health care;12 and, second, without a control group who are denied access to an interpreter (which itself would be deemed unethical), it is difficult to determine any effect on outcome that is produced by the interpretation as distinct from the effect of the psychiatric intervention of which the interpretation is a component.
Review procedures and processes
A systematic review was carried out in accordance with the methods outlined in guidance issued by the Centre for Reviews and Dissemination. The full peer-reviewed protocol submitted in the funding application has been published in an open access journal for scrutiny.46 During the course of the review, the original protocol was changed in two ways, as indicated in Participants. First, we included studies from other countries of interventions that showed evidence of transferability to the UK and the ethnic groups were relevant to the UK (e.g. by being from the same diaspora), if other inclusion and exclusion criteria were met. Second, we discovered many studies that seemed not to include evaluations but only to describe an intervention; these were assessed more carefully for evaluative statements and conclusions to create a category of near misses (A– rather than A+).
Data sources and search strategy for published literature
The following databases were searched: MEDLINE, PsycINFO, EMBASE, Applied Social Sciences Index and Abstracts (ASSIA), Cochrane Database of Systematic Reviews, The Campbell Collaboration, ACP (American College of Physicians) Journal Club, The Cochrane Central Register of Controlled Trials, Cochrane Methodology Register, Allied and Complementary Medicine Database (AMED), Cumulative Index to Nursing and Allied Health Literature (CINAHL), British Nursing Index, Health Management Information Consortium, Social Science Citation Index (SSCI), Social Care Online (www.scie-socialcareonline.org.uk) and NHS Evidence collection on ethnicity and health. We also searched university databases for Doctor of Philosophy (PhD) theses (ProQuest assisted) and Master of Science theses from specialist centres on ethnicity and health.
The terms provided by Anderson et al.47 to understand culturally competent care were adapted as a framework from which to generate search terms that might help identify publications on the effectiveness of interventions. Search strategies were constructed around BME groups using search terms refined through a number of systematic reviews29,48,49 and terms related to key descriptors of TCs: (1) aspect of TC; (2) types of mental disorder; (3) the professionals involved; and (4) aspects of clinical success.
These terms were tested and then applied to the electronic databases from database inception to 4 April 2012. All searches were rerun in early January and February 2013. The search strategy was designed to identify a broad range of literature on TC with BME patients and staff working in specialist psychiatric services (see Appendix 1 for the final search strategy). This strategy evolved as it was iteratively tested using a range of keywords and refined during pilot searches to provide a maximum yield.
The remit of the current systematic review is necessarily very broad: interventions for improving TC might be direct (i.e. between patient and professional) or indirect (through structural modifications to services to create more space for conflict to be resolved). Studies in both community and institutional settings were eligible. Studies of interventions designed to improve TC in our target population might not have been explicitly labelled as studies of TC. Indeed, relevant studies could be classified in many ways, for example as studies of communication, cultural competence or awareness, or just training to improve outcomes in ethnically diverse areas. Therefore, an inclusive strategy was felt most appropriate to capture all potentially relevant literature; specificity was sacrificed to maximise sensitivity.
The search strategy was iteratively developed to test its sensitivity to capture gold standard papers that were known to the research team. The final search strategy was settled following further discussion with the review management group (all investigators, listed as authors of this review) and carer representatives (Patrick Vernon/Afiya Trust). The complexity of the concept of TCs and the range of BME groups that might be relevant required us to develop this block search approach to ensure we captured all possible papers. The initial search strategy was trialled on different databases to assess the number of hits and relevant papers generated. We found a separate search strategy was needed to identify papers that included dementia, as these seemed not to be picked up by the generic search strategy. An example of the blocks of terms entered is shown in Box 1.
Terms used in the search strategy (BME or black ethnic minorit* or black minorit* ethnic*).mp.
Hand searches of the following journals were completed for the time period April 2007 to May 2012: Transcultural Psychiatry; Culture, Medicine and Psychiatry; International Journal of Social Psychiatry; Journal of Cross-Cultural Psychology; Ethnicity and Health; Ethnicity and Disease; and Diversity in Health & Care. In addition, two special issues of journals were also screened:
Psychotherapy:
Theory, Research, Practice, Training, a special issue on culture, race, and ethnicity in psychotherapy, Volume 43, pp. 377–560, winter 2006.
50 Journal of Counselling Psychology, Volume 58, pp. 457–646, October 2011.
51
Grey literature: data sources and search strategy
An important body of relevant evidence was expected in the grey literature: unpublished reports and papers containing practice and community-based information on interventions. For this material, standard database searches were replaced by a variety of strategies: hand-searching more recent issues of journals on ethnicity and health (those that had appeared in the last 10 years), and journals on communications; cascade-searching; and searching specialist collections at the Centre for Evidence in Ethnicity, Health and Diversity, The King’s Fund, the NHS library on ethnicity and health, National Institute for Health Research (NIHR)’s Health Technology Assessment (HTA) programme, the National Institute for Health and Care Excellence (NICE), the Royal College of Psychiatrists and the Medical Foundation for the Care of Victims of Torture. We also made use of various web-based resources [e.g. Google (Google Inc., Menlo Park, CA, USA), NHS Evidence, JISCMail] to search for reports that were not published in conventional research or professional journals and research in progress.
Search strategies for these grey literature databases were also derived by an information scientist and researcher. They built on the terms used for searching the published literature (see Appendix 1). An initial stage consisted of testing several key terms and search strategies, followed by filtering by eye and then iterative refinement of the original searches.
Doctor of Philosophy or Doctor of Medicine theses
A search was undertaken of all dissertations and theses accepted for higher degrees by universities in Europe and North America up to February 2013. Examination of titles and abstracts (where available) identified six documents of potential importance. All were ordered; however, three had been lost by the originator universities52–54 and could not be assessed further. The other three were obtained,44,55,56 but the studies did not meet the criteria for inclusion (see Appendix 2 for details).
Conference papers
Conference papers were identified through key term searches of the ProQuest Conference Papers Index from June 2004 to February 2013. A total of 138 conference papers were identified and, following examination, 16 were selected as potentially relevant (see Appendix 3 for details). None of these papers, however, was included in the review after the full text had been read.
Bibliographies
Bibliographies of peer-reviewed articles short-listed for the main review were searched to identify any grey literature references. In total, 50 grey literature references were identified via this route; following examination, 32 were selected as likely to be relevant and a further 19 as possibly relevant (see Appendix 4 for details), but on testing against the inclusion and exclusion criteria, none was included in the review.
Websites and other electronic sources
Various electronic sources were searched using predefined search terms. Search strategies followed particular threads iteratively. A broad range of websites was searched systematically using key terms (see Appendix 5). Items identified were examined and short-listed by two reviewers. UK material was separated from non-UK material. A total of 97 items were short-listed: 86 from the UK and 11 from elsewhere (see Appendix 6). None was included in the final review.
Two further electronic sources were examined:
NHS Evidence: 380 items were identified; 34 were short-listed as relevant (see
Appendix 7).
JISCMail archive: 29 items were identified and five were short-listed as relevant (see
Appendix 8).
None of these entered the review after examination of the full text.
Other grey literature databases were considered but not searched, including the System for Information on Grey Literature (SIGLE), which was last updated in 2005, and the British National Bibliography for Report Literature, which ceased in 1998.
Research databases
Websites of research funding bodies were searched to identify projects in progress or those that had been completed. Intervention trials were distinguished from other studies. Potential items were examined and short-listed by two reviewers. This produced the following items for further examination:
UK Clinical Research Network (UKCRN) Study Portfolio: 15 projects of potential interest were short-listed; these include seven intervention trials, two of which focused on effective patient–clinician communication, but these
specifically excluded non-English speakers (see
Appendix 9).
NIHR Evaluation, Trials and Studies Coordinating Centre (NETSCC): 15 projects of potential interest were short-listed (see
Appendix 9).
Research councils: no studies were identified on the Medical Research Council (MRC) website but five projects were short-listed from the Economic and Social Research Council (ESRC) website (see
Appendix 9).
Following assessment of the full text, these searches did not yield studies beyond those identified in the published literature. Any completed project was tracked to identify outputs.
Economic literature
A series of systematic MEDLINE searches was undertaken by the information scientist to identify economic materials that were potentially relevant to interventions (see Appendix 10). Abstracts, were assessed and a total of 21 items were short-listed for full examination (see Appendix 10), as a result of which two studies with economic data were entered into the review; however, both had already been identified, one from the published literature search57 and one from the grey literature search.58
Survey of experts
We consulted with experts, both those who are known in the field and those identified by asking service users about who they consider to have expertise in this area. We identified key researchers who had studied TCs. The applicants and collaborators drew on their networks in the UK, the European Union (EU) and beyond. Community groups and charities were also contacted to identify materials in community-based collections. All individuals so identified were invited to comment on omissions in the searches and to put forward candidate papers and to volunteer research work that was unpublished or in progress.
An online questionnaire59 was developed and a personal invitation sent to 37 experts in the field and to 75 organisations for circulation to their members (for list see Appendix 11). The questionnaire aimed to identify any additional grey literature, such as reports on projects, research in progress and resources or toolkits for mental health professionals (see Appendix 12). A reminder was sent to all non-responders after 2 weeks. This resulted in a total of 60 replies, of which 30 provided details of available materials. Finally, a request was posted on the ‘Minority-Ethnic-Health’ JISCMail discussion group with apologies for any cross-posting. This generated a further 10 replies, but no additional materials were identified. In total, these 70 respondents identified 30 separate grey literature items; 12 items were identified as relevant (see Appendix 13) but, after reading the full text, none was deemed to report sufficient evidence to meet our inclusion criteria.
Selecting appropriate sources
All databases searched (and number of hits retrieved from each) are shown in .
Databases searched and hits
Citations in the scientific literature were downloaded into an EndNote (version X5; Thomson Reuters, CA, USA) library. The relevance of all papers was assessed against the pre determined inclusion and exclusion criteria (Box 2) by two researchers, who inspected all the titles and abstracts. Each reviewer worked independently. Forward and backward citation tracking complemented the database searches.
Full-text manuscripts of any titles/abstracts were obtained if these met the inclusion criteria or if there was uncertainty when reviewing the title and abstract. Any discrepancies were resolved by consensus and, if necessary, a third reviewer was consulted. The papers were classified into one of three groups: A+, A– and B. Only A+ papers entered the review.
The A+ papers met the original inclusion criteria and were entered into the review.
The A– papers:
Were descriptions of an intervention or elements of an intervention with no evaluation data.
Reported on studies that included patients from a group not relevant to the UK and held no immediate lessons for UK ethnic groups, for example they examined issues stemming from aboriginality or indigenousness. Transferability of the intervention to the UK context was a consideration.
Quality assessment
Core quality criteria
Different quality assessment tools had to be used because there were multiple research designs in the studies selected for review. Particular core criteria, though, were assessed for all studies. These were (1) the clarity with which the intervention was described as improving TC directly, by inference only, or not at all; (2) whether or not the outcomes (e.g. alliance, reduced conflict, greater trust) of a change in TC were directly measured using a reliable and valid scale; and (3) whether or not the ethnic groups would be of relevance to the UK and described in a manner consistent with a specific classification scheme for ethnicity (not just ‘race’). The last criterion was later relaxed where studies were of minorities and there were lessons that could be applied to ethnic minority groups in the UK. Core criteria scores ranged from 2 to 12 [scores of 0 on the intervention to improve TCs (1 above) and on outcome (2 above) would have led to exclusion of the study].
If there was an economic evaluation within a trial, this was separately scored. In addition, a quality-rating schema suited to different study types was used. This schema covered randomised controlled trials, case–control studies, observational quantitative studies, case studies and case series, and qualitative studies. The sources of our scoring scheme are presented in the following six sections. Copies of the full scoring scheme are presented in Appendix 14.
Quality criteria for randomised controlled trials
Through a brief review of the literature, Moncrieff et al.76 developed a tool to assess the design of randomised controlled trials. Fifteen items could score 0, 1 or 2 (range of total score 0–30). Checklist items relate to the appropriateness and adequate description of the hypotheses, study design, intervention, main outcomes and methods of analysis. The checklist demonstrated good inter-rater reliability, and correlations between the three raters in the validation paper were high (r = 0.75–0.86).
Quality criteria for non-randomised observational quantitative studies
For the quality assessment of case–control or cohort studies, we used the evaluation of non-randomised observational studies by Deeks et al.77 From the recommended scales, we selected that created by Reisch et al.78 because it considers important confounding factors and differences between groups prior to the intervention, it has a good case-mix adjustment and it is a validated numeric scale (scores 0–34).
Quality criteria for case series
Through a brief literature review, we identified the NICE criteria for the assessment of case series.79 However, with this quality assessment tool alone it was very difficult to discriminate between two studies that were similar in design or execution but that differed in the importance of the findings and their implications for practice. Furthermore, important characteristics that might better reflect quality in a case series – like length of follow-up and loss of clients over time – are not mentioned.
A previous HTA report indicated how difficult it is to use a specific system of quality ratings for case series, given that so little methodological research on quality ratings has taken place.80 Consequently, we adopted the scoring system developed by the Canadian Institute of Health Economics (IHE).81 This comprehensive 18-item checklist is based on quality criteria for assessment of case series from the Centre for Review and Dissemination. The inter-rater reliability of the checklist was based on the three reviewers, and has high kappa values. The IHE Delphi panel did not develop a scoring system for the checklist. According to the report, the quality of a study was assessed by counting the number of ‘yes’ responses to different criteria in the checklist. A study with 14 or more ‘yes’ responses was considered of acceptable quality. We decided to add the scoring system to the checklist; hence, a ‘yes’ to a criterion in the checklist would qualify for a score of 1 and ‘no’ would score 0 (scores range from 0 to 38).
Quality criteria for case studies, qualitative studies and studies from the grey literature
We chose the National Centre for Social Research (NATCEN) quality assessment criteria.82 These criteria not only concentrate on the methodological quality of qualitative studies, but also highlight its conceptual quality. The NATCEN tool was based on 29 sets of previously suggested assessment criteria and consists of 18 appraisal questions underpinned by four guiding principles. We allocated a mark for each question asked and each of the items that might be endorsed to indicate quality and so the scale offers a range from 0 to 87.
Economic studies
These were rated 1–4 on the basis of the type of economic analysis. Cost-effectiveness studies scored 4, impact of interventions and cost–benefit studies scored 3, an intervention being costed scored 1, or the benefits being considered in terms of finances scored 1. A 0 was scored if there was no economic evaluation.
Overall quality score
The approach taken to quality rating was to use the core criteria and add the specialised criteria according to study design, and for the trials to include the economic score. For all other study types, the core and specialised criteria were used. The scores were summed and presented as percentages of the maximum score for each of the core items, the aggregated score of the specialised items and then as a total overall quality score. These were then categorised into low, medium or high quality on the basis of percentage of the maximum score (i.e. < 33% low, 33–66% medium and > 66% high). These are presented in – to give an overall visual impression of the quality ratings according to study design, including how key elements of quality were rated, rather than relying only on a single total score. In undertaking the scoring, details about the methodological and design issues were evaluated using the scoring schedules. Specific methodological strengths or weaknesses were considered in the synthesis.
Quality assessments of case studies and qualitative studies
The B papers were:
reviews with no primary data
descriptions of potential interventions on theoretical grounds, but without an evaluation; often these were case studies for teaching purposes, without an evaluative and critical element or conclusion.
Quality assessment of quantitative studies
Quality assessment of case series
Studies included
The database searches yielded 7329 hits, and 3733 records were found to be potentially relevant after removing duplicates. An extensive search for grey literature yielded 608 sources, including six PhD theses, two of which could not be located from the original universities. shows the selection of papers at each stage of the review.
The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) diagram showing selection of papers at each stage.
A total of 21 publications were deemed to be relevant and met all the inclusion criteria. These comprised 12 trials52,57,60–69 one of which was from the grey literature,52 two observational studies,70,71 three case series58,72,73 (one of which was from the grey literature58 and one of which had a qualitative component that was separately extracted72), two qualitative studies40,74 (which included a case study in each) and two pure case studies.14,75
Methods of analysis and synthesis
The findings are presented below in two groups (trials and non-trials) given that within the hierarchies of evidence it is the trials that provide the most definitive evidence of effectiveness.83 We set out the interventions, study design, ethnic groups and service setting, and outcomes. Given the diversity of study settings, interventions and outcomes, the studies were not suitable for a meta-analysis and thus the data were subjected to a narrative synthesis. Popay et al.84 define narrative synthesis as:
[A]n approach to the systematic review and synthesis of findings from multiple studies that relies primarily on the use of words and text to summarise and explain the findings of the synthesis. Whilst narrative synthesis can involve the manipulation of statistical data, the defining characteristic is that it adopts a textual approach to the process of synthesis to ‘tell the story’ of the findings from the included studies. As used here ‘narrative synthesis’ refers to a process of synthesis that can be used in systematic reviews focusing on a wide range of questions, not only those relating to the effectiveness of a particular intervention.
It is part of a larger review process that includes a systematic approach to searching for and quality[-]appraising research[-]based evidence as well as the synthesis of this evidence.84
Narrative synthesis includes the following elements: textual description, tabulation, grouping and thematic analysis. We undertook this process and by thematic analysis contrasted, within- and non-trial designs, elements of intervention, the outcomes used in studies, and direct or indirect measures of effectiveness, taking account of the elements contributing to the quality score and the perspectives of patients and carers who reviewed the emergent evidence (see Patients’ and carers’ views).
Data extraction
Quantitative studies
An electronic version of a quantitative data extraction form was circulated to all members of the review group for comments and revision as appropriate. The two main reviewers used an Excel (Microsoft Corporation, Redmond, WA, USA) spreedsheet to help ensure the consistency of the extracted data by allowing only certain types of options to be entered in any one field, thus ensuring that all data were categorised in a similar way. In the first stage, detailed information relating to study methods was extracted concerning the interventions (e.g. whether the effect of the change in TC was inferred or directly measured), size of the population, ethnic group, setting, timeline of intervention and of follow-up, and the outcomes that were measured (including the tool used to measure each outcome).
The second stage involved more detailed extraction of appropriate numerical data for all studies categorised as either randomised trials or quasi-experimental designs. This too was recorded in an Excel spreadsheet. The effect sizes of the various studies were extracted to judge whether or not a meta-analysis could be performed.
Qualitative studies
A similar pattern was followed with the qualitative studies. Data were extracted into a summary table in Excel. Data extracted comprised the concepts identified and evaluation methods of studies and narrative findings.
Patients’ and carers’ views
A key element of the study was to ascertain user views from the outset; in the design and operation of the literature review, as well as in the evaluation of the interventions and themes identified as having potential to improve TC. In other words, we asked service users and their carers what they regarded (or would regard) as TC and what they would see as better practice. The involvement of minority ethnic users (i.e. BME people currently using health services for mental health needs, or significantly involved as carers of such patients) was operationalised by a collaboration with the national charity The Afiya Trust (www.afiya-trust.org) and its links to the National Black Carers Workers Network (NWBCN), CatchAFiya, and the National Survivor User Network, which support and enable the voices of such individuals through a panel of users who have had basic training in self-representation and some experience of supporting similar research and service development.
The Afiya Trust circulated a call for expressions of interest and selected a panel of eight people who had relevant experience and covered a range of ethnicities and mental health needs. The panel members, chaired by an Afiya Trustee, then worked closely with the research team in advising on the literature search, considering the emergent themes and assisting with the rating, validation and dissemination of the emergent models of good practice.
The core user panel comprised two men and six women. Three self-identified as being of African/Caribbean/black British origin and five were of South Asian background, with Hindu, Sikh, Muslim and other religious affiliations, and from a variety of the major South Asian national/linguistic groups. Three had experience as carers of mental health service users (some having also been patients themselves) and most had also some experience of working with community-based third-sector support groups.
Initially, a series of three user workshops were held, to elicit key themes and to discuss perspectives on TC as a process and what issues were seen as likely to be of concern. The panel’s views were then compared and contrasted with those emerging from the review of the published scientific literature and the grey literature. This background exploration was important for the patients and carers in order to prepare themselves to consider the outputs from the review, so that they were more active in deliberating the value of the interventions identified in the review.
The interventions or other solutions identified by the literature review were grouped and presented as model responses, using a technique of vignettes that included quotes from the publications and a brief description of key elements. A series of Delphi technique consultation rounds were engaged in using e-mail and workshops to highlight the most significant issues and to draw up a list of priorities for desired changes or best practice, as well as a list of issues of concern, responding to the research team’s initial questions and planned recommendations. Although overall agreement did emerge, there was a degree of variation between panel members, and many of the model interventions were ranked as ‘most desirable’ by some members but ‘least important’ by one or more others, illustrating the heterogeneity of need and the requirement for a variety of options to be available to service providers and users.
The patient and carers commented on the proposed interventions and ranked them as high or low priority, and these judgements were included in the synthesis alongside the quality of design and methodological issues, and the strength of the findings.
