Involving carers in risk assessment: a study of a structured dialogue between mental health nurses and carers

Abstract

Background

Involving carers is a key priority in mental health services. Carers report the sharing of service users’ safety information by mental health nurses is problematic and seldom takes place.

Aims

The impact of an intervention on consensus between nurses and carers on perceptions of risk was investigated.

Methods

Carer–nurse risk consensus scores were measured pre- and post-introduction of a structured dialogue (paired t-test/ANOVA). Carer experience with involvement was surveyed pre-test (n = 60) and compared with the post-test intervention group (n = 32) (chi-square tests of linear-by-linear association).

Results

Consensus and perceptions regarding type and severity of risk did not change significantly for carers or nurses after engaging in a structured dialogue. Statistically significant differences were found with carers reporting higher levels of satisfaction with services in four out of six areas surveyed.

Conclusions

Findings provide support for increasing carer contribution to discussions regarding risk. Further work to embed carer involvement in clinical practice is warranted.

Introduction

A carer is anyone who cares, unpaid, for a friend or family member who due to illness, disability, a mental health problem or an addiction cannot cope without support (Carers UK, 2017). Carers have a crucial role to play in the delivery of high-quality mental health services (Happell et al., 2016). Best practice should see a collaborative approach to care between professionals, service users and carers (Worthington et al., 2013). Involving carers is a key part of international policy and guidance (Australian Government, 2010; US Substance Abuse and Mental Health Services Administration, 2009; UK Department of Health (DOH), 2010, 2014). Estimates suggest the 1.5 million carers of people with mental health issues in the United Kingdom contribute in excess of £26 billion to the UK economy (Carers UK, 2017).

Caring for a person with mental health difficulties can have a considerable effect on the wellbeing and mental or physical health of carers (Carers UK, 2017). However, involving carers may increase their satisfaction with services, reduce stress and burden, and provide them with a sense of empowerment (Campbell, 2004). Despite literature advocating carers could and should contribute to care, current levels of family and carer involvement remain tokenistic (Gee et al., 2016), variable, inadequate and aspirational (Stanbridge et al., 2009). Carers should be involved in the assessment and management of risk and service user safety (Clancy et al., 2014; National Patient Safety Agency, 2009; DOH, 2014). Yet serious incident investigations and media reports all too often highlight anger and recrimination from families or carers of mental health service users towards services for failing to listen to their concerns. By ignoring carer views or failing to inform them of significant events, their ability to provide support effectively post discharge compromises both their safety and that of the service user they support (Gray et al., 2008). Equally the involvement of service users in discussions regarding care and service user safety is not consistently undertaken within mental health care settings (Simpson et al., 2016).

When dialogue does take place between professionals and carers, it can be difficult, as professionals’ perception of what constitutes risk (Farrelly et al., 2015), how they prioritise risk (Ryan, 2002) and even the language they choose (Clancy et al., 2014) does not consistently reflect that used by carers. Whilst a number of different healthcare professionals may be involved in assessing risk and service user safety, in acute settings this dialogue is often the responsibility of nurses. Carers complain that nurses use technical, patronising or disempowering language (Chatzidamianos et al., 2015). Capacity for meaningful discussion is therefore limited leaving families and carers unable to communicate vital information which could provide a more comprehensive representation of the service users’ illness. Coffey et al. (2017) reported that nurses are sometimes reluctant to discuss risk with carers for fear of facing disagreement or causing upset, whereas Crombie et al. (2007) argued the main barrier to communication and carer involvement is conflict between service users’ right to privacy and confidentiality against carers’ need for information. Such dilemmas see nurses as well as other health care professionals being unclear about what can and cannot be shared and using this as a way to avoid engaging with carers (Gray et al., 2008).

There remains a gap between the rhetoric in policy, legislation and implementation of carer strategies (James, 2016). Policy and best practice states carer knowledge of service users should be used to build comprehensive and accurate pictures to inform assessment and support clinical management planning (Gray et al., 2008). However current levels of involvement offer limited opportunities for both service users and carers to manage safety (Coffey et al., 2017), and carer involvement continues to be poorly implemented and defined (Cree et al., 2015). Coffey et al. (2017) reported that service users and carers were often unaware that risk assessments had taken place or of what they contained, suggesting guidance surrounding the listening, obtaining and sharing of information has yet to be embedded into practice. In addition, the quality of risk assessments varies in depth and detail (Gilbert et al., 2011), and risk assessments are incomplete or even misunderstood (Cree et al., 2015).

It remains unclear what joint risk-assessing interventions would look like and how these would be delivered in practice (Worthington et al., 2013). How this gap might be effectively tackled is unclear, as currently those responsible for assessing risk in acute settings, e.g. nurses, lack the skills, training and confidence necessary to work with carers effectively (Stanbridge et al., 2013). This paper describes an intervention undertaken to promote a structured dialogue between nurses and carers regarding concerns about service user safety and risk. The aim of the study was to investigate the impact of increasing the level of contribution that carers make to risk assessments. Specifically:

1. What impact does involving carers in risk assessment have on the consensus between carers and nurses about the areas and levels of risk a service user presents with? and

2. What impact does involving carers in risk assessment have on their levels of satisfaction around their involvement with risk assessment and services?

Design

The study had a pragmatic non-randomised design and used two separate groups of carers. It was not practicable to randomise carers, as once intervention training had been delivered to nursing staff, practice may have changed.

Methods

Setting and sample

The host organisation was a healthcare provider in the north of England delivering mental health services to a mixed urban and rural population across a number of inpatient units for service users aged between 18 and 65 years old. Units were located at separate geographical sites, but all were subject to the same policy and procedural guidance.

Carers were all supporters of adults with a range of psychiatric disorders. Group A consisted of 60 carers of service users admitted to one of the units between November 2014 and April 2015. Group B consisted of 36 carers of service users admitted between June 2015 and April 2016. As per standard practice, at the point of admission, the care team established whether a service user had a family member, friend or carer involved in their care. Only carers of service users who had signed a ‘Permission to Share Information’ form (a standard document used within the service and filed within their clinical records) were invited to participate in the study. Age parameters were not set for carer participants as demographic details are not collected by the service.

Data collection

Carer Experience Data Capture Survey

As no validated questionnaire to measure carer satisfaction with involvement in risk discussions was identified, the Carer Experience Data Capture Survey was developed by the research team and approved as appropriate by independent carers not connected to the study. This was not however subject to testing for reliability and validity. The six-question survey asked carers to rate satisfaction with their level of involvement in discussions about risk and overall satisfaction with the service using a 5-item Likert rating scale with the options: “A lot”, “Quite a bit”, “Moderately”, “A little”, “Not at all” (see Table 2).

Table 2.

Carer experience data capture survey results.

Question	Group B Percentage satisfied (number/number responding)	Group A percentage satisfied (number/number responding)	Chi-square test value for linear-by-linear association (df = 1 in all cases)	p-value
Q1. Did you get the opportunity to say what you were worried about?	85.3% (29/34)	58.4% (35/60)	15.8	<0.001
Q2. Did the care team share what they were worried about?	78.8% (26/33)	62.7% (37/59)	1.65	0.202
Q3. How involved did you feel in the decisions made about risk?	78.8% (26/33)	72.9% (43/59)	0.104	0.764
Q4. How satisfied were you that they focussed on the right area of risk?	85.2% (29/34)	53.7% (29/54)	14.2	<0.001
Q5. How satisfied were you that the care team understood the severity of the risk?	84.8% (28/33)	40.0% (20/50)	20.8	<0.001
Q6. Overall how satisfied were you with the service?	88.2% (30/34)	46.6% (27/58)	26.3	<0.001

Likert scale responses for each question were labelled “A lot”, “Quite a bit”, “Moderately”, “A little”, “Not at all”. Percentages reported above are the percentages of respondents who gave one of the first three responses.

Group A

Group A carers only completed the Carer Experience Data Capture Survey. Eligible carers were identified from clinical notes by the research team and posted a letter between 7 and 10 days post admission giving details of the survey. They were contacted 2 to 3 weeks later and invited to complete the survey over the telephone. Data provided a baseline for carer satisfaction regarding the experience of supporting a service user admitted to the local service.

Group B

Data collection from the intervention Group B commenced in June 2015. The intervention required a member of nursing staff to engage the carer(s) in a structured dialogue focussed upon their concerns about the risk and safety of the service user they supported. Recruitment was undertaken by nursing staff as they assessed capacity of service users and had carer contact. All eligible carers had the chance to participate. Service users were approached and consented for permission to approach carers. Carers were then provided with an information sheet to explain the purpose of the study and invited to participate in the structured dialogue. Group B carers completed the intervention at a time convenient for them and with a nurse on the ward where the service user they supported was receiving care. Two to three weeks later they were contacted by telephone by a member of the research team and invited to complete the Carer Experience Data Capture Survey.

The structured dialogue intervention

The intervention consisted of a face-to-face structured dialogue based on the GRIST (Galatean Risk and Safety Tool) (https://www.egrist.org/) web-based decision support system collecting contextual and historical data, in addition to the immediate risk history and current behaviour. GRIST was developed in the UK and combined both clinical and actuarial approaches to risk assessment. It contained six areas of risk: suicide, self-harm, harm to others/damage to property, self-neglect, risk to dependents and vulnerability. Although it has not been subject to independent evaluation, GRIST was one of three multiple risk tools recommended in Best Practice in Managing Risk (DOH, 2007). Furthermore, the authors claim it demonstrated outcome validity (www.egrist.org/sites/egrist.org/files/grist-summary_0.pdf). GRIST was chosen, as this was the risk assessment tool that clinicians were required to use in the host organisation. Practice as usual was for GRIST to be completed for each service user by a nurse on referral to the service. This should have been updated to reflect significant events or new information. Professionals completing GRIST rated each area of risk using a numerical scale (0–10) for each concern. For the purposes of this research, the numerical scale (0–10) on the GRIST was replaced with an 11-point visual scale ranging from green through orange and red. Green represented no risk (0), orange a mid-point (around 5 or 6) and red extreme risk (10). Carer advisers to the project suggested a visual scale would be more acceptable and easier for carers to use during the structured dialogue between nurse and carer and approved its inclusion for use.

Nurses involved had all received risk assessment and GRIST completion training as part of their clinical role independently of this study. Family and carer engagement training was not provided by the researchers, although it was on offer to nurses via their organisation. In preparation for conducting the intervention, in May and June 2015, 41 nursing staff from the participating units attended a 1-hour training session delivered by two members of the research team experienced in delivering clinical practice-based training. Training focussed on guiding staff through the completion of the key worker/carer interview recording sheet. Detailed instructions for completion of this recording sheet were provided in the form of a script to guide the discussion between nurse and carer to promote consistency. A member of the research team assessed each nurse’s understanding of the intervention and ability to follow the script using role play at the end of the training session. The ability of the nurses to engage with carers was subject to their ‘practice as usual’. As standard, practice nurses within the service were expected to identify, liaise with and signpost carers for further support. All nurses trained were deemed competent to deliver the structured dialogue. A member of the research team also observed the first intervention undertaken by each nurse to offer support with completion of the interview recording sheet and offer the nurse feedback on performance.

The nurse instigated the intervention following the format shown on the key worker/carer interview recording sheet. This started with a general question, ‘What three things keep you awake at night' and then proceeded with a discussion focussing on each of the six areas of risk. The GRIST risk rating undertaken by the nurse on admission was used as the pre-discussion nursing score and provided a starting point for each discussion. For each area of risk, the nurse would share the pre-discussion nursing score and ask the carer to indicate and record on the visual scale the pre-discussion carer score. A dialogue would then take place. The carer was then asked to indicate their post-discussion score. Changes to the nurses’ perception of the risk posed by the service user as a consequence of the discussion with the carer were recorded as per standard practice when new information was received. Three days after the intervention, the service users’ GRIST was accessed by the research team and used as the post-discussion nurse score to assess if the nurse had updated the risk assessment held by the service.

Outcome measures

The primary outcome measure was the carer and nurse pre- and post-discussion scores recorded on the interview recording sheet (for Group B only), relating to their perceptions of risk across the six areas of the GRIST.

The secondary outcome measure was carer experience data collected from both Group A and Group B carers 2 to 3 weeks post admission or participation in the intervention.

Statistical method

For the purposes of analysis, the visual scores were converted back to their numeric equivalent. An effect size of 0.4 was anticipated for the change in mean totals for the primary outcome measure (GRIST Consensus Scores) after discussion and a before–after correlation of 0.5 was anticipated. A paired sample t-test of the before and after means of the primary outcome measure required 52 participants to achieve 80% power using a 5% significance level for testing (G Power, Version 3.1.9.2).

The absolute value of the difference between carer’s and nurse’s total GRIST pre-discussion scores was calculated to give the pre-discussion discrepancy score. This calculation was repeated on the total GRIST post-discussion scores to give the post-discussion discrepancy score. Differences between the two discrepancy scores (pre-score minus the post-score) were then calculated and significance was tested using a paired t-test. The changes were analysed by one-way ANOVA to test for any differences between the four locations (clinical units) in the host organisation used for the study. Bootstrapped p-values and confidence intervals were used as a precaution in case of violation of the standard assumptions for the paired t-test and one-way ANOVA.

Carer satisfaction responses for each of the six questions asked were summarised descriptively with the figures reflecting the percentage and number of carers giving a positive response. Differences in satisfaction between the intervention (Group B) and the comparator group of carers (Group A) were tested using chi-squared tests of linear-by-linear association.

Results

Clinical pressures and the withdrawal of the GRIST as the preferred risk assessment tool of use within the host organisation impacted upon recruitment so only 36 Group B carers participated in an intervention rather than the required 52. In addition, missing data on some components of the GRIST for either the carer or the nurse pre- or post-discussion resulted in 32 sets of carer–nurse pairs being available on which to calculate GRIST discrepancy scores (see Table 1).

Table 1.

Paired t-tests for GRIST carer–staff consensus scores.

GRIST component	Mean change in discrepancy score	Standards error of mean change	95% confidence interval for mean change	t	p-value
Suicide	0.156	0.404	(−0.625,1.03)	0.386	0.674
Self-harm	0.781	0.470	(−0.094,1.75)	1.66	0.105
Harm to others or  damage to property	0.188	0.427	(−0.656,1.03)	0.439	0.679
Self-neglect	0.469	0.435	(−0.374,1.31)	1.08	0.308
Risk to dependents	0.281	0.346	(−0.375,.969)	0.814	0.459
Vulnerability of service user	0.594	0.386	(−0.156,1.34)	1.54	0.152

n = 32 for all GRIST components; positive mean change for a component indicates that carer and staff converged in assessment following the intervention. Bootstrapped p-values and 95% CIs based on 1000 bootstrapped samples are presented.

The mean pre-discussion discrepancy score was 10.7 (SD = 9.81), while the mean post-discussion discrepancy score was 9.56 (SD = 7.69). Convergence between carer and nurse perceptions of risk would be indicated by positive changes to discrepancy scores. The mean change in discrepancy score was 1.16 (SD = 1.29) which was not statistically significant (t = 0.891, df = 31, p = 0.391, 95% CI (−1.19, 3.88), effect size = 0.158). It had been anticipated that there would be convergence between carers and nurses in their perceptions of risk following their discussions; however, the results did not support this assumption as carer–nurse consensus regarding overall risk did not change substantially. No significant differences between the four individual inpatient units were found from the one-way ANOVA (F(3,28) = 1.00, p = 0.407), although it should be noted that three of the four units contributed very few carer–nurse pairs due to staff turnover and resourcing issues. Mean changes for each of the six specific risk areas of the GRIST in carer–nurse discrepancies were in a direction consistent with increased convergence but again were not statistically significant. The effect sizes for convergence in GRIST scores between carers and nurses undertaking the intervention pre- and post-discussion were not significant in any of the risk areas measured, although risk related to self-harm showed the greatest movement in terms of consensus, with p = 0.105.

Within-person changes in the GRIST component ratings pre- and post-intervention were also examined for carer and nursing groups. Most people, both carers and nurses, did not change their ratings pre- to post-intervention for any of the areas of risk. However, there was a lack of post-intervention nurse scores, with only 25% of risk assessments being updated within 72 hours of the dialogue.

Carer experience data (see Table 2) generally showed that a greater percentage of carers who participated in the intervention reported higher levels of satisfaction - as opposed to Group A carers who had received service as usual. Group B carers demonstrated higher levels of satisfaction across all six questions, four of these areas being statistically significant. Overall satisfaction with the admission was higher, increasing from 41.6% to 88.2%. Furthermore, a total of 85.3% of carers in Group B were satisfied that they had been given the opportunity to discuss their concerns, which was 26.9% higher than the score recorded for Group A. Q4 and Q5 related to carer satisfaction that nursing staff focussed upon the right area of risk and understood the severity of the carers’ concerns. Carer satisfaction that nurses acknowledged the severity of the risk was rated at 85% for carers in Group B, more than twice that of Group A. Higher levels of carer satisfaction regarding nursing teams sharing information and carers feeling involved in the decision-making process was insignificant, with more than 20% of carers not responding positively.

Discussion

This study demonstrated that carers who took part in the structured dialogue were more satisfied overall with their experience of supporting a service user through an inpatient admission than those in the comparator group. In addition, carers given the opportunity to discuss risk expressed higher levels of satisfaction at being able to voice their concerns in terms of both the area and severity of risk. This supports evidence suggesting carers place great importance on being heard (Worthington et al., 2013) and being given the opportunity to contribute to care (Cree et al., 2015). Furthermore, as with carers who were involved with care planning and supporting a service user (Simpson et al., 2016), carers in this study were not convinced they were equal partners and did not feel any more significantly involved in decisions. It is difficult to say with any certainty if individual carer dissatisfaction was due to criticisms around the level or quality of contact, or feelings that nursing staff misunderstood or did not take into account what they said.

The provision of a structured dialogue did not provide an overall statistically significant difference in changes to perceptions of risk for either carers or nurses across a range of areas of risk. In fact, neither group substantially altered their ratings across any of the risk domains. Therefore, despite delivery of an intervention that aimed to increase discussion and sharing of views on risk, neither group changed their perceptions as a consequence of this discussion. Within-person changes were also minimal. These findings may suggest that nurses did not gain new information, did not listen to carers or failed to update risk assessments; however, this is conjecture as further follow up with staff was not conducted. The lack of updating or change to post-discussion nurse scores does mirror findings by Gilbert et al. (2011) who found mental health inpatient nurses unexpectedly record a low number of risk assessment reviews.

As found by Littlechild and Hawley (2010), carers and nurses also maintained conflicting views about the seriousness of risk. Such divergence potentially supports concerns that the language of risk used by nurses differs from that of carers, affecting the quality of any discussions (Clancy et al., 2014). However, some convergence between pre- and post-scores was evident in relation to self-harm, which again matches other findings that carers place greater emphasis on service users’ risk to self rather than risk to others (Coffey et al., 2017). Discussion in this area may have raised concerns for nurses or offered reassurance for carers demonstrating dialogue may be useful.

This study was subject to the difficulties of delivering ‘real world research' in busy healthcare and clinical environments (Gray, 2017). The difficulties encountered in this study could be attributed to a number of factors but are also likely to be relevant to the establishment of successful carer engagement strategies.

Firstly, Ulrich et al. (2012) highlighted the competing tensions faced by nurses tasked with undertaking research whilst engaged in clinical practice. Likewise, despite recognising the importance and benefits of carer engagement, the nurses in this study appeared to struggle to prioritise the research – although 41 members of nursing staff were trained, only 21 of them completed an intervention. There was also a marked difference in recruitment between the units. Local conditions, resources and priorities may have affected the likelihood nurses would participate, as nurses on the unit completing the greatest number of interventions had a more stable staff group and generally a higher nurse to service user ratio. Therefore, organisations serious about involving carers should ensure services are adequately resourced if collaborative risk assessment and successful carer engagement strategies are to be achieved.

Secondly, recruitment in the comparator group was undertaken in a systematic manner from clinical notes by the research team and was less subject to selection bias. Whereas for the intervention, nursing staff controlled access to eligible carers, citing barriers including service user capacity and consent to involve carers. This is consistent with evidence suggesting that nurses do not routinely reciprocate in information sharing (Rapaport et al., 2006); they use confidentiality as a barrier and withhold details about significant events or issues of safety (Gray et al., 2008). Nursing staff may have been more willing to recruit carers with whom they had existing positive working relationships or felt more at ease with when faced with competing demands for their time and energies (Cleary, 2004). Subject to such conditions, the sample recruited to participate in the intervention may be less likely to disclose new information to nurses in dialogue, and their existing satisfaction with the service could possibly be higher. Likewise, it is possible that carers viewed by nurses as challenging or difficult were less likely to be recruited. They may have information not known to nurses or have less favourable perceptions of the service. The ways in which nurses can be supported to initiate and maintain contact with more challenging and difficult to reach carers warrant further enquiry. Although this study focussed on discussions in relation to risk and service user’s safety, it has highlighted a number of issues relevant to how nurses communicate with and engage carers to exchange information more generally.

Thirdly, operational changes within the organisation saw the GRIST replaced with another risk assessment tool before the required number of interventions to provide meaningful analysis was reached. Although the intervention could still have been delivered, the research team decided it was unlikely that pre- and post-discussion nursing scores would be recorded, therefore recruitment ceased. The lack of evidence regarding the validity and reliability of the GRIST must be acknowledged. In addition, the Carer Experience Data Capture Survey was not subject to robust testing and development.

Finally, this was a small-scale pragmatic study conducted with inpatient nurses and two independent groups of carers within one service. Consequently, findings are not necessarily generalisable but may be of interest to similar settings and groups of carers.

Conclusion

In this study, carers who were engaged in a structured dialogue focussed on risk reported higher levels of satisfaction than those who were not. However, carer–nurse consensus regarding the type and severity of risk was not achieved. Also, despite findings demonstrating carers and nurses could come together to discuss their concerns about the planning and management of safety, whether or not communication actually takes place may be at the discretion of the nurses rather than the right of the carer. Thus, there would be benefit in finding ways to help nurses to navigate situations in which service users’ rights for confidentiality conflicted with a carer’s right for support. Nurses should be encouraged to engage proactively with all carers. Findings provide support for increasing carer contribution in discussions regarding risk and suggest further work investigating the use of structured approaches to embed carer involvement in clinical practice is warranted.

Key points for policy, practice and/or research

• Carers expressed increased levels of satisfaction when engaged in a structured dialogue focussed upon risk and safety issues.

• Although carers given the opportunity to discuss their concerns felt listened too, neither carers nor nurses changed their views about which type of risk or what level of risk service users posed.

• It is not clear what takes place when nurses and carers discuss risk. There is a need for further research to aid understanding about these interactions.

• Further work to identify strategies to engage carers in discussions about service user safety are warranted.

Biography

Haley Jackson is a Research Nurse working within the Inpatient Service at Humber Teaching NHS Foundation Trust. She is a PhD student at Leeds University and engaged in work around patient safety.

Jane Wray qualified as a registered nurse in Leeds and after a period in clinical practice began her academic research career in the field of learning disabilities, working with children, their families and carers. She is currently Director of Nursing at the faculty of Health Sciences at the University of Hull.

Eric Gardiner has worked at the University of Hull since 1990 and has collaborated extensively with health researchers from a variety of disciplines, leading to more than fifty publications.

Tracy Flanagan is the Deputy Director of Nursing at Humber Teaching NHS Foundation Trust. She has presented her research at national and international conferences and co-authored a book based around the culture and context of acute inpatient care.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Ethics

Approval for the study was gained from the Yorkshire and Humber-South Yorkshire NRES Committee 15/YH/0033. Participation was voluntary but participants had to be fluent in English. Written informed consent was obtained from service users and carers. Participants could withdraw their data from the research study up until the point at which data were prepared for publication. However new information relating to safety would be retained for the purposes of ongoing risk assessment by the service.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was funded by a grant from Hull Clinical Commissioning Group.