Mental health nurses' measured attitudes to people and practice: Systematic review of UK empirical research 2000–2019

Accessible Summary What is known on the subject? Many studies have investigated the attitudes of mental health nurses towards a range of targets. These targets are person‐oriented (for example groups of people with a similar mental health diagnosis) or practice‐oriented (for example practices such as seclusion or restraint). It is thought that attitudes contribute to the practice of mental health nurses because research suggests attitudes have a role in shaping behaviour. What the paper adds to existing knowledge? To date, research about mental health nurses' attitudes has examined different attitudes in isolation from one another. By demonstrating a lack of connectedness across studies this paper highlights the need for new theory‐informed approaches to attitudinal research. By standardizing measurements across different studies this review demonstrates that the most negatively appraised attitudinal targets—indicated by large proportions of respondents who appraise negatively—concern people with diagnoses of borderline personality disorder, substance misuse, and acute mental health presentations. What are the implications for practice? Significant numbers of mental health nurses may have attitudes, especially towards people with borderline personality diagnoses and those who misuse substances, that may not be concordant with good practice. There is insufficient evidence about what the actual implications this has for practice because the body of relevant research lacks coherence, interconnectedness and a grounding in contemporary theoretical developments. Training programmes that focus on attitudinal change need to be more rigorously evaluated. Abstract Introduction Attitudes are considered integral to mental health nursing practice. Aims To comprehensively describe the (i) measured attitudes of UK mental health nurses towards people and practice; (ii) effectiveness of interventions to change attitudes; and (iii) relationships between their attitudes, other variables/constructs and practice. Methods Using systematic review methodology, multiple databases (CINAHL, Scopus, PsycINFO, Web of Science Core Collection, Google Scholar) were searched. Eligible studies involved measurement of UK‐based mental health nurses' attitudes with multi‐item scales. Studies were quality appraised, mean (SD) attitudinal data were standardized, and other results converted to standardized effect sizes. Results N = 42 studies were included. Negatively appraised attitudinal targets were people with a borderline personality disorder diagnosis, substance misuse, and acute mental health presentations. Educational interventions were associated with immediate increases in positive appraisals but sustainability was poorly evidenced. There was very limited study of attitude‐practice links. Discussion This review identifies priority attitudinal targets for action but also demonstrates that future work must consider the interconnectedness of attitudes and their relationship with practice. Implications for Practice Priority areas for consideration are attitudes to borderline personality disorder, substance misuse and mental health co‐morbidity. Addressing disparities between nurses' attitudes and those of service users is important. More robust research is required into the effectiveness of interventions to change attitudes and into attitude‐practice links.

. By "personal evaluation," it is meant that the subjects' response is positive or negative, such as like or dislike, or potentially like and dislike. Under this definition, attitudes differ from other constructs which lack this evaluative component. Examples include beliefs, which are expressions of agreement or non-agreement that a proposition is factually correct; perceptions, which are estimations of the extent, import or relevance of some construct; and knowledge, which is content from an organized body of facts that are generally held to be true and the response to which may be judged correct or incorrect.

| Why attitudes are important
Given the widespread conduct of, and broad range of targets of, attitudinal research in mental health nursing, one could assume that attitudes are important per se. Indeed, some studies simply assert that links between attitudes and practice are axiomatic. For example, Acford andDavies (2019: p. 1177) state the common sense view that nurses report attitudes towards patients with a personality disorder diagnosis which "are likely to have a detrimental effect on the therapeutic engagement and care these individuals receive." This may be true, but we consider it important to articulate the relevant theories about the underlying mechanisms behind such assertions.
Where they are voiced, studies of mental health nurses' attitudes invoke the importance of theory of planned behaviour (Ajzen, 1991) and theory related to stigma and its effects (Link, 1987).
While addressing somewhat different issues-the first aims to understand why people behave the way they do and the second why some groups are stigmatized-both theories are central to attitudinal research because both view the underlying processes involved as amenable to intervention. Either or both theories are explicitly or implicitly appealed to by researchers when outlining the rationale for their investigations. Ajzen (1991) posited that behaviour results largely from an individual's intention to behave in a certain way; however, that intention itself is influenced by key factors including her attitudes, perceived behavioural control (i.e., the extent to which she believes she has control over the behaviour), and subjective norms (i.e., the extent to which she perceives that others such as peers or managers also enact the behaviour). Further, these determinants of intended behaviour may also interact (e.g., attitudes partly influence perceived control and vice versa and so on). From this perspective, attitudes are important because they are linked with behaviour, opening up the possibility that behaviour itself might be susceptible to modification if they, and other contributing determinants such as knowledge, can be successfully addressed. As an illustration, Jansen et al. (2006) studied nurses' orientation to patient aggression using Ajzen's (1991) theory of planned behaviour as a theoretical framework. Theorizing that different approaches to the management of aggression across international boundaries might reflect the prevalent attitudes within nations, and thus partly determine the nature of actions required to successfully address behaviour, the authors studied those attitudes across five European countries including the UK. UK nurses were found to have significantly different attitudes to most other national groups in that they were more likely to view patient aggression as destructive, offensive and intrusive, and less likely to view it with tolerance. Of course, and as the authors allow, the study design used here does not allow for conclusions about the underlying reasons for the attitudinal differences found. Nevertheless, it demonstrates how the theory of planned behaviour has been used to justify and structure attitudinal investigation in mental health nurses.

| Stigma-related theory
Stigma-related theory, growing from the work of Goffman (1963) and furthered considerably in the specific instance of mental disorder by Link et al. (1987) posits that stigma is an attribute, behaviour or reputation which is socially discrediting in a particular way: it causes an individual to be mentally classified by others as an undesirable, rejected stereotype rather than as an accepted ingroup member (Goffman, 1963). Elaborating on this, Link and Phelan (2001) proposed that stigma arises from a perfect storm where (i) individuals differentiate and label human variations; (ii) prevailing cultural beliefs tie those with attributes labelled as adverse; (iii) labelled individuals are placed in distinct groups which serve to distinguish "them" from "us" (i.e., those with and without the "adverse" attribute); and (iv) individuals in "othered" groups suffer loss of status and discrimination resulting from the prior process. Because labels are essentially viewed as social constructs stigma can in theory be successfully challenged through techniques aimed at redefining people's understanding of the attributes held by people which have been negatively appraised. An example of the use of stigma theory in the UK mental health nursing literature is Markham's (2003) account of attitudes to people with a label of borderline personality disorder.

| Pragmatic justifications for mental health nurse-related attitudinal research
More commonly, studies of mental health nurses' attitudes do not elaborate on underlying theoretical mechanisms; rather they simply point to existing empirical evidence of a link between attitudes and practice. In a study of attitudes towards suicidal behaviour, Anderson et al. (2000) point to evidence of an interaction between specific attitudes and the stigma of deliberate self-harm which jeopardizes the effectiveness of professional interventions. Lamph et al. (2018) highlight an array of evidence that suggests a link between healthcare professionals' attitudes and the experience of patients with a diagnosis of borderline personality disorder (e.g., Bodner et al., 2015). In terms of nurses' attitudes to their own practice, in rationalizing their investigation of mental health nurses' physical healthcare-related orientation, Robson and Haddad (2012) point to prior research linking attitudes to self-efficacy and engagement in these practices (e.g., Howard & Gamble, 2011).
Studies of mental health nurses' attitudes have been justified by claims that the attitudinal target in question is an increasing policy priority ; that in empirical studies mental health nurses have a demonstrated more negative appraisals than other professionals ; by a lack of knowledge about orientation towards a specific attitudinal target, for example, harm reduction approaches to self-harm (James et al., 2017); and that it is important to change attitudes or at least study the effect of interventions on them (e.g., Lavelle et al., 2017). In summary, it is widely held that attitudes are important in mental health nursing.

| A note on terminology
While "attitude" is not a uniformly used construct in psychology (see e.g., Maio et al., 2018) an underlying assumption of the theories and other justifications outlined above is that attitudes contribute to real-world outcomes and are thus viewed as legitimate targets for interventions. The corollary of this is that, in addition to attitudes being inherently evaluative, there is also a desired direction such that attitudinal movement in the desired direction would, hypothetically, lead to changes in the associated real-world outcomes also in the desired direction. From this perspective, it may make sense to refer to a specific attitude as "positive" if its' location on a twodimensional measurement scale is in the portion past the midpoint oriented towards the desired end of that scale and to use that term synonymously with a value judgement such as "good" or, in comparison with measured attitudes of some other entity, "better." Similarly, appraisals towards the other scalar pole could be termed "negative" and used synonymously with evaluations like "bad" or "poorer." However, while this might seem appropriate for some attitudes it is not so straightforward. For example, it may be reasonable to assume that measured attitudes that are more endorsing or supportive of people with a personality disorder diagnosis are the desired attitudes because we would expect them to be associated with real-world outcomes like more respectful nurse-initiated interactions or reduced stigma for the individual. However, for some attitudes, this may not be so clear cut. For example, in the case of attitudes to restrictive measures such as physical restraint, we are faced with two problems. The first, easily solved, is that presumably attitudes more endorsing of physical restraint as an intervention are not the desired attitudes. We could therefore simply reverse the terminology by identifying the less endorsing end of the scale as the desired attitude and label scores falling on the associated part of the scale as positive attitudes. This is undeniably confusing. There is potential for it to be even more so when the desired direction of the real-world outcome is debatable. For example, attitudes to neuroleptic medication management have been conducted. There may not be an obvious desired direction of attitude. Do we require attitudes that endorse or reject their use? For these reasons, in this paper, we use language pertaining to attitudinal directionality with deliberate care. We use the terms "positive" and "negative" to indicate the direction of endorsement or rejection of the attitude construct under investigation rather than as a definitive indicator of the desired direction of that construct. We therefore eschew language indicating value judgements such as "good" or "bad" attitudes. When groups of individuals are compared then terms such as "more positive" are used to describe one group relative to another, and it should not be assumed that it indicates that either or indeed neither group are positive or negative in their appraisals overall.

| Contribution of this paper
Current thinking in the study of attitudinal change suggests three important contextual areas: first, attitudes change in individuals across time as part of human development; second, attitudes exist in, and change resulting from, the context of social relationships, especially contact with powerful communicators; third, they exist in their own socio-historical context and are subject to shaping by major events (Albarracin & Shavitt, 2018). We propose, therefore, that a viable contribution to the study of mental health nurses' attitudes can be made by examination of empirical studies from a specific temporal, geographical, and hence cultural, context. An attempt to identify and review the empirical literature on mental health nurses' attitudes in its entirety, and irrespective of context, would, in our view, be an unrealistically large task. While we do not rule out conducting increasingly broader reviews in future, we suggest that examination of mental health nurses' attitudes within a specific socio-geographicalhistorical context (UK 2000(UK -2019 will synthesize existing knowledge such that it can enlighten our understanding of contemporary attitudes in our own context. It is likely to enlighten understanding of current contemporary priorities and highlight knowledge gaps for future research. The overall aim of this review was to systematically identify and appraise studies published 2000-2019 which measured the attitudes (outcome) of UK mental health nurses (population) towards any relevant attitudinal target (exposure or study focus) either cross-sectionally or longitudinally (study type). Specific study objectives were to: 1. Identify what attitudinal targets mental health nurses appraise most positively and negatively and where the greatest polarizations lie.
2. Identify whether UK-based mental health nurses' attitudes differ significantly from those of any other group studied (e.g., other professionals, public, service users, and non-UK nurses).
3. Identify whether subgroups of UK-based mental health nurses (e.g., gender, experience, and age) differ in their measured attitudes.
4. Identify whether measured attitudes of UK-based mental health nurses change over time either in or out of the context of interventions to change attitudes. 5. Identify demonstrable associational or causal links between UKbased mental health nurses different measured attitudes and between their attitudes, other constructs and practices.

| ME THODS
This review uses a systematic approach to identification and appraisal of mental health nurse-related attitudinal research. For reasons discussed in the introduction, our review aims to address a specific socio-geographical-historical subset of all available research. We followed guidelines outlined in the Preferred Reporting Items for Systematic reviews and Meta-analyses (PRISMA; Moher et al., 2009) to structure the study. There was no published protocol but the aims, search strategy, inclusion/exclusion criteria and quality assessment were established prior to study conduct.

| Literature search strategy
We searched the CINAHL, Scopus (including Medline), PsycINFO and Web of Science Core Collection electronic databases in addition to Google Scholar. An example search is presented in Table 1. Two strategies were used. First, searching was undertaken on combinations of terms relevant to the population of interest ("mental health nurses," "psychiatric nurses" and "mental health practitioners") together with terms related to the outcome of interest, this being derived from combinations of terms related to "attitude" and a range of synonyms (e.g., "opinion," "belief," "perception" and "knowledge") and to "measurement" and a range of synonyms (scale, measure, inventory, checklist, questionnaire and so on). The focus or exposure of interest was captured using a comprehensive set of terms related to mental health and related issues. In a second stage of the search, we entered the names (and abbreviations/acronyms) of all measurement tools identified in stage one (plus other measures identified from a broader Google Scholar search) in combination with the name of the tool's first author (see Appendix S2 for a complete list of tools searched). In addition, we followed up references in selected papers and undertook separate searches of the Table of Contents of a number of specialist mental health nursing journals. All searches were conducted in August 2020. All searches included geographical terms (United Kingdom, UK, England, Scotland, Wales, Northern Ireland, Great Britain, GB) as limiters and searches were limited to English language publications post-1999. All search results were exported into EndNote X9 where duplicates were removed. The literature search strategy was devised by Author 1 and conducted independently by Author 1 (first two databases) and Author 2 (second two databases).

| Inclusion and exclusion criteria
For inclusion, papers must have described an empirical study conducted with a sample comprising or including UK-based registered mental health nurses and published in peer-reviewed journals post-1999. Studies of nursing students or solely of those working with forensic or older age/dementia populations were not included. Details of the application of inclusion criteria are presented in the PRISMA flow diagram (see Figure 1).
Studies were eligible for inclusion if they were designed in whole or part to measure attitudes. As an operational definition, we used that of the American Psychological Association (APA, 2022 see Introduction). We included studies using scales that self-defined as "attitudinal" (e.g., Attitudes to Personality Disorder Questionnaire) but scales or subscales were not included where they were not self-defined as attitudinal and were judged to measure knowledge, self-efficacy or confidence, or any other dimension which we did not believe fit the APA definition. We operationalized "measure" as the use of a multi-item (2+) scale with a summary score because there was no way to reasonably interpret studies which reported responses to questionnaires solely on an item-by-item basis. However, while we extracted information about the robustness of reported measures (e.g., internal reliability) and used it to inform our evidence synthesis, we did not exclude any study based on sub-acceptable threshold scores.

| Study quality/risk of bias
Studies were assessed against quality criteria adapted from two sources (Greenhalgh, 2006; University of York Centre for Reviews & Dissemination, 2008). Study quality and risk of bias were assessed by Author 1 with Author 3 independently conducting assessment of 50% of studies. Assessment considered categories related to study aims, sampling, questionnaire development and measurement, generalizability and funding (see Appendix S1).
Discrepancies were discussed until consensus was achieved.
Study quality assessment generated one or more overall quality rating for papers or groups of papers associated with a unique sample; this was because each attitudinal scale used with a sample was assessed on its own merits. All papers were included in the review regardless of quality.

| Data extraction
We extracted information about study setting, sample (number and proportion of mental health nurses), the attitudinal construct investigated, the attitudinal and other tools used (subscale names/ descriptors, number of items, and information about the internal reliability, test-retest reliability or external validity of the tool based on the sample under investigation or cited in previous literature).
Where possible, study results were extracted. Mean (M) and standard deviation (SD) scale/subscale scores were converted to a standardized M (SD) score and plotted in order to help gauge the sample's absolute level of attitudinal orientation, that is, the proximity of the mean to the scale midpoint and dispersal (SD) around the mean. As per our "note on terminology" (see above), we use the terms "positive" and "negative" to indicate level of endorsement of the scale F I G U R E 1 PRISMA Flow chart diagram items rather than as an indicator about the desired direction of the construct. There was one exception: the Self-Harm Antipathy Scale (Patterson et al., 2007a(Patterson et al., ,2007b self-evidently measures endorsement of statements which express feelings of aversion. In this instance, a pragmatic decision was made to simply reverse scores such that endorsement of such feelings was rated as a negative appraisal and rejection as positive appraisal. Where possible all inferential tests for difference or correlation were converted to a standardized effect size (Cohen's d) using an online converter (Wilson, 2001) in order to inform interpretation of the magnitude of differences or relationships between variables. We used the heuristic thresholds of d = 0.2 (small), 0.5 (moderate) and 0.8 (large) to interpret effect sizes.
Data extraction was conducted by Author 1 while Author 3 independently extracted 50% of data to check for accuracy.  Table 2). Studies covered person-oriented attitudinal targets including people in specific diagnostic categories (personality disorder k = 9; severe mental illness k = 4) and behaviours (aggression k = 10; self-harm and suicidal behaviour k = 4, substance misuse k = 2); and practice-oriented targets (attitude to contain- and five high quality (9 or above).

| Person-oriented attitudes
Person-oriented attitudes were measured in 24 samples (see Table 2). For these studies, an acceptable or near acceptable internal reliability coefficient was usually reported for the sample or cited from prior research. All other standardized standard deviations were dispersed across the midpoint to some extent. Means with the widest SDs, suggesting more polarized appraisals, were as follows: the extent to which self-harm represents an intention of manipulation (Patterson et al., 2007a(Patterson et al., ,2007b, the level of optimism reported by mental health nurses about people with a diagnosis of borderline personality disorder (Markham, 2003), and the amount of enthusiasm reported for working with people with personality disorder .

| Objective 2: Attitudinal differences between UK mental health nurses' and other groups
Other findings with contextual quality and effect size summaries are presented in Table 3. Evidence of differences between UK mental health nurses' attitudes and other groups were sparse but of interest. Markham (2003)  Cross-sectional survey Self-cutting management Self-harm SHAS (Patterson et al., 2007a). Low internal reliability and new analysis resulted in 3-subscales explaining 62.9% of variance: (i) perceived manipulative functionality; (ii) perceived positive functionality; (iii) freedom of choice  Cartwright, 1980).

34-items. Single scale.
"a low total score represents a positive therapeutic attitude and a high total score represents a more negative therapeutic attitude" (Watson et al., 2003).  (Maslach et al., 1996).  (Mistral et al., 2002). Based on the AAPPQ (Cartwright, 1980)*. Relevant subscales: (i) feelings of skill and knowledge adequacy; (ii) self-esteem in this work; (iii) willingness to work with these patients; (iv) satisfaction; (v) a right to work with these patients; (vi) role support; and (vii) general self-esteem.

TA B L E 2 (Continued)
completer analysis highly susceptible to bias. Elsewhere, when attitudinal measures were secondary outcomes in trials (e.g., Bowers, Brennan, et al., 2006;Bowers et al., 2015), there was no significant change despite significant improvements in primary outcomes including measures of conflict and containment.

| Objective 5: Links between different attitudes or between attitudes and other constructs
The MBI depersonalization subscale was negatively correlated with tolerance of aggression (Whittington, 2002; large effect size), though this relationship was not detected in a different sample (Whittington & Higgins, 2002). Hannigan et al. (2000) found depersonalization to be linked to measures of general health (small effect), self-esteem, coping and stress (all moderate effect size), and Laker et al. (2019) reported an association with a lack of self-reported efficacy to effect change. Hosie and Dickens (2018) reported correlations between scales of an amended Self-Harm Antipathy Scale and their Attitudes to Self-cutting Management (ASc-ME) scale suggesting a relationship between lower antipathy and approval of supportive and harmreducing self-harm management techniques (providing first aid kit, giving advice, remaining present during a cutting episode and providing sterile cutting implements). Finally, Bowers, Whittington, et al. (2006) reported significant correlations between subscales of the Attitudes to Personality Disorder Questionnaire (APDQ) and data-derived factors relating to ward conflict (associated with low "security," security provision (low APDQ "acceptance" and observation (low "purpose" and low "enthusiasm").

Within/between group difference
Change over time

Relationship with other measures
Person-oriented attitudes  Bowers, Whittington, et al. (2006) [M] Dataderived factors relating to ward conflict associated with APDQ low "security"; security provision with low APDQ "acceptance"; and observation with low APDQ "purpose" and "enthusiasm" (ES not calculable) Borderline personality disorder MHNs rated BPD vignette more poorly than a schizophrenia vignette on social distance, dangerousness and treatment optimism while healthcare assistants did not ES+++ (Markham, 2003 [M]) Dickens et al., 2018; [M] Short-term positive change following educational intervention. ES +++ for "treatment characteristics," "perception of suicidal tendencies" and "negative attitudes." Evidence for sustained change hampered by low retention Hosie and Dickens (2018) [H]correlation between supportive attitude to harm reduction techniques and more positive self-harm antipathy rating Suicidal behaviour/ self-harm MHNs less antipathy to self-harm than those with general nursing-only qualification ES+++ [M] Self-harm antipathy positive changes in short term (Patterson et al., 2007b).

Within/between group difference
Change over time

Relationship with other measures
Practice-oriented attitudes Self-harm management Hosie and Dickens (2018)  -Attitudes to containment measures not a significant independent predictor of all selfharm or moderate self-harm. A data-derived factor comprising variables related to drug/ alcohol use and absconding was associated with greater acceptability and safety for patients of containment measures (Bowers, Whittington, et al., 2006  Within/without group difference, comparisons between sample sub-groups or between samples and others.

TA B L E 3 (Continued)
4.2.1 | Objective 1: Positive, negative and polarized appraisals Three attitudinal targets were rated at least one standard deviation below the scale midpoint (See Figure 3), these were the measures "refusing treatment to a person who has self-harmed" and "giving inappropriate treatment to a person who has self-injured" on the ASc-ME scale (Hosie & Dickens, 2018) and "mechanical restraint" on the Attitudes to Containment Measures Questionnaire (ACMQ; Bowers et al., 2004). Targets rated by nurses at least one standard deviation above scale midpoint were eight items of the ASc-ME including "care planning," "suggest distraction techniques," "providing advice on wound care," "provide first aid kit" and "offer PRN." For general containment, IM medication, constant observation and manual restraint all achieved high ratings in two studies each (Pettit et al., 2016;Whittington et al., 2009) while seclusion was rated so highly only in the latter study. Elsewhere, mental health nurses rated towards the positive end of the scale in relation to neuroleptic treatment (Harris et al., 2007), three aspects of physical health care for people with mental illness (Robson & Haddad, 2012) and ECT (Wood et al., 2007). Attitudinal targets with wide standard deviations suggesting greater range of orientation were those related to physical deterioration in the context of the management of medical emergency in mental health settings (Lavelle et al., 2017).

| Objective 2: Attitudinal differences between UK mental health nurses' and other groups
Other results are summarized in Table 3. There were important differences between mental health nurse raters of both the ACMQ and ASc-ME and patients or previous service users. Whittington et al. (2009)  nurses rated nine of seventeen techniques more positively than prior users of services, mostly the least restrictive ones but also seclusion and physical restraint (effect sizes ranged from small to large). Other items were rated similarly or, in the case of "refusing treatment," the effect size was small. Comparisons with other groups showed more positive orientation to mental health nurse prescribing among nurses than psychiatrists (Patel et al., 2009) For the other practice-related attitudes, within-group analyses by Harris et al. (2007) suggested there may be professional and experiential associations with attitudes to maintenance neuroleptic treatment where community mental health nurses were significantly more positive than ward-based mental health nurses; and to physical health care of mental health patient, an additional general nursing qualification was associated with confidence and smoking disapproval (and mental health nurses with past 5-year training in physical healthcare had a higher total PHASe score than those without.

| Objective 4: Attitudinal change
There was little evidence that tested whether practice-oriented attitudes changed over time; Lavelle et al. (2017) examined change associated with simulation-based physical healthcare training and reported significant (small effect size) increase medical emergencyrelated attitudes in total attitudes score following baseline and intervention.
4.2.5 | Objective 5: Links between different attitudes or between attitudes and other constructs Multilevel modelling did not find ACMQ to be significantly associated with either all self-harm or moderate self-harm (Bowers et al., 2008). A data-derived factor comprising variables related to patient drug/alcohol use and absconding was associated with greater ACMQ acceptability and safety for patients (Bowers, Whittington, et al., 2006).

| DISCUSS ION
This review has unified a sizeable but hitherto disparate literature about the measured attitudes of mental health nurses in the UK.
Given that attitudes are generally considered to be specific to particular socio-geographical-historical contexts then it makes sense to analyse a meaningful subset of contemporary studies, in this case from the UK 2000 to present. While there have been few crosscultural studies, it is instructive that investigations into aggressionrelated attitudes across Europe (Jansen et al., 2005) and between UK and China (Whittington & Higgins, 2002) provide evidence for namely as about either specific groups of people or about specific aspects of practice. We had five specific objectives and we discuss each in turn.

| Positive, negative and polarized appraisals
We identified studies of mental health nurses' attitudes related to seven types of person-oriented targets and five types of practice-  Dickens et al., 2018); suicidal behaviour as a "moral evil" (Anderson et al., 2000); substance misuse (permissiveness; treatment intervention orientation and treatment optimism subscales; Richmond & Foster, 2003); and acute mental health (care or control and therapeutic perspective subscales; Baker et al., 2005). Attitudes to personality disorders were also polarized in studies by Markham (2003) and Bowers, Brennan, et al. (2006) and to self-harm (Patterson et al., 2007a(Patterson et al., ,2007b suggesting significant numbers with more negatively oriented evaluations. From an evidence-based perspective, these are the areas that should be prioritized in terms of attitude improvement and in terms of further exploring whether and how attitudes and behaviour are linked phenomena. Importantly, a review of research about the experiences of service users with a diagnosis of borderline personality disorder and their families and carers suggests that they commonly perceived staff in mental health services as judgemental and to hold negative attitudes (Lamont & Dickens, 2019). Similar findings are also prevalent in reviews of studies of the experiences of people who self-harm (Lindgren et al., 2018) and of people who have spent time in acute mental health care (Schmidt & Uman, 2020). This suggests that, at least in these areas, there is congruence between mental health nurses' measured attitudes and service users' experiences; moreover, that congruence is negative. While these reviews are all international in scope, all of them reference evidence from the UK suggesting that there is little reason to suspect that UK-based service users' experiences are more positive than anybody else. In relation to people who use mental health services and who hold a comorbid substance misuse diagnosis, similar findings of negative experience are not widespread and it has been suggested that it may be that, in this domain, it is a lack of mental health professional substance use disorder-specific training that is problematic (Priester et al., 2016).

| Attitudinal differences between UK mental health nurses and others
There was little evidence that UK mental health nurses' attitudes differed significantly from other professional or occupational groups. They were more positively oriented to nurse prescribing than psychiatrists (Patel et al., 2009) though this may say more about psychiatrists than nurses. Mental health practitioners including nurses were less negatively oriented to self-harm behaviour than a sample of general health qualified practitioners (Patterson et al., 2007a(Patterson et al., ,2007b, which has a positive side but does not negate the apparently polarized attitudes of mental health practitioners in the same study and discussed above. UK mental health nurses were less positively oriented to aspects of patient aggression including offensiveness, destructiveness protectiveness and intrusiveness compared with most other European countries' nurses (Jansen et al., 2006). Research from outside of the UK but using the same Attitudes Towards Aggression Scale has suggested that mental health nurses measured attitudes to aggression are largely personal and idiosyncratic rather than clustered by demographic characteristics or by ward (Laiho et al., 2014). The authors suggest as a result that interventions to change attitudes to aggression at, for example, ward level through a culture alteration programme is likely to be less useful than interventions that target the individual. However, UK nurses' attitudes were not alarming in themselves and the practical import of differences between nations is unknown given a lack of studies examining relationships between these attitudes and actual practice.
Of greater salience was findings about the differences between mental health nurses' attitudes and those of people who use or have used services. In relation to the use both of containment measures (Whittington et al., 2009) and measures to manage self-cutting (Hosie & Dickens, 2018), nurses were significantly more positive than service users for most types of intervention. In  Bowers, Whittington, et al. (2006)

| Attitudinal change
Most research examining change over time in the context of interventions has been limited by a failure to retain participants beyond the immediate end of intervention assessment and thus demonstrate that any change is sustained. This was the case for intervention targeted at borderline personality disorder-related attitudes (Acford & Davies, 2019;Davies et al., 2014;Dickens et al., 2018;Ebrahim et al., 2016;Lamph et al., 2018), self-harm antipathy (Patterson et al., 2007b), medical emergency-related attitudes (Lavelle et al., 2017) and schizophrenia-related attitudes (Bradshaw et al., 2007). Interestingly, in Bowers, Brennan, et al. (2006), Bowers et al.'s (2015, Safewards studies there was no change in measured personality disorder-related attitudes over time despite more tangible study outcomes related to reduced occurrences of conflict and containment. We do, however, query whether the tool used in these studies is actually an attitudinal measure (see "relationship between attitudes and between attitudes and other constructs" below). These findings suggest that evaluation studies where a target outcome is attitudinal need to be more rigorously conducted to demonstrate effectiveness beyond the immediate post-intervention measurement.
Further, given Bowers, Brennan, et al. (2006), Bowers et al. (2015) findings that the mechanisms and relationships between education and attitude change need considerable further study to determine whether such interventions are even likely to be useful.

| Relationship between attitudes and between attitudes and other constructs
There has been limited research into this aspect. Specifically, personality disorder-related attitudes were found not to be related to ward-level self-harm rates in Bowers et al.'s (2008) highly powered study across 136 acute mental health wards. While this might superficially suggest that attitude-practice links are not worth pursuing in future research studies, we note that the finding was only true in relation to the APDQ, a measure which, unusually for an attitude scale, captures respondent reports of the frequency of their affinity with items rather than their degree of positive/ negative evaluation. Hence, had we been rather stricter in our inclusion criteria, studies using this tool would have been ineligible.
Attitudes to Containment Measures Questionnaire scores were also found not to be associated with self-harm rates in Bowers et al.'s (2008) study, again suggesting that attitudes may not be worth further investigation. However, this may not be the case since the ACMQ is used to investigate measures for disturbed behaviour in particular and not for self-harm specifically. Further investigations are therefore required to determine whether attitudes to managing self-harm are associated with its incidence.
However, this also speaks to the question of whether extinction of self-harming behaviour rather than harm reduction should be the measure of success of any practice innovation (James et al., 2017).
Further, research is required into differences between attitudes to managing self-harm specifically with coercive measures and whether these differ from their use for violent and aggressive behaviour. It may be that attitudes other than those more relevant to personality disorder or containment measures play a role in selfharm rates and these should be investigated more thoroughly. As examples, items contributing to relevant scales on the Attitudes To Acute Mental Health Scale (ATAMHS; Baker et al., 2005) "care or control" (sample item "members of society are at risk from the mentally ill") and "therapeutic perspective" ("psychiatric patients are generally difficult to like") do not ostensibly sound compatible with contemporary notions of recovery-oriented care.
Finally, attitudes as behavioural manifestations of selfreported depersonalized responses to service users were associated with poorer general health, self-esteem, stress (Hannigan et al., 2000) and reduced self-efficacy to effect change (Laker et al., 2019). There were contradictory findings about whether they were associated with tolerance of aggression (Whittington, 2002;Whittington & Higgins, 2002). The former findings suggest that interventions to change attitudes might have associated benefits in the domains of stress and self-esteem or indeed vice versa.
There is some evidence that mindfulness-based approaches are effective in reducing stress in mental health professionals (Rudaz et al., 2017).

| S TRENG TH S AND LIMITATI ON S
This review has used a number of techniques which add new insight to the collection of work both in terms of the relative import of the various studies and the unified body as a whole. First, extraction and standardization of means from different measurement tools allow a clear picture to be drawn of what the literature tells us about the measured attitudes of UK mental health nurses.
Second, the calculation of standardized effect sizes where possible for all correlations or differences also informs interpretation of the import and relevance of individual studies. Third, the quality assessment of studies has facilitated the weighting of evidence.
Fourth, and finally, the integration of these study findings in itself highlights the hitherto lack of such an integrated approach. The review highlights that studies have not examined relationships between attitudes and practice. Key contributions of this review, therefore, are the highlighting of the lack of connectedness between investigation of different attitudes and the lack of studies which take the next step and examine relationships between attitudes and behaviour and/or practice. In short, mental health nursing research has fallen behind contemporary theoretical developments in attitudinal research such as the causal attitude network model (Dalege et al., 2016) which conceptualize clusters of similar attitude types and provide testable models of those structures thus having the potential to inform causal attributions. One immediate implication of this is that new mental health nursing research about attitudes needs to consider using designs based on more contemporary theoretical approaches.
This review concentrates solely on one socio-geographicalhistorical subsample of available attitudinal research, namely that conducted in the UK and from the past 20 years only. Further, studies rarely report mental health nurse data separately from those of other healthcare professionals yet we have included studies which contain only a proportion of mental health nurses.
We have concentrated on studies where attitudes are measured using scales. As a result, we excluded a number of studies where questionnaires were administered but results were reported on an item-by-item basis. Similarly, we have not included data from qualitative studies. It may be possible for future reviews to widen the socio-geographical-historical scope and the type of studies included. Finally, we only included studies of constructs that met our operational definition of "attitude" and other constructs such as "beliefs," "opinions," "perceptions" and "ideologies" also warrant attention.

| CON CLUS ION
Attitudinal research in UK mental health nursing has proliferated in the last two decades. There is some good evidence that many nurses make negative attitudinal appraisals about personality disorder and substance misuse. However, the extent and importance of this are somewhat shrouded by a lack of connectedness in current approaches to mental health nursing attitudinal research. What is now needed is a focus on a more integrated approach to study using new and innovative techniques based on contemporary models of attitudes.

| RELE VAN CE S TATEMENT
Attitudes are considered integral to good mental health nursing practice. There has been considerable research about them but the work is disparate and lacks a joined-up approach. The current paper takes a broader approach and examines the attitudinal literature as a body of work rather than looking at single attitudes. In doing so it highlights where the evidence lies in terms of priorities for future work and identifies the need for new approaches that consider the interconnectedness of attitudes and the links between attitudes and practice.

E TH I C S S TATEM ENT
The study is a review of existing literature and did not require ethics review.