Participant flow and baseline data

Study results are presented across four chapters. This chapter focuses on participant flow (numbers assigned to experimental groups and analysed) and on baseline data. Chapter 8 presents the process evaluation findings on the implementation of the CLECC intervention. Chapter 9 presents findings on the feasibility of evaluating the effectiveness of CLECC using the planned CRT design, including the feasibility of older people and people with cognitive impairment participating, outcome measure performance, and assessment of bias in the trial. It also addresses the feasibility of estimating CLECC costs. Chapter 10 presents the results of outcome measurement in the pilot CRT.

The first part of this chapter describes the participation of individual ward teams (clusters) in the study and shares baseline characteristics of these teams, including ward specialty, staffing levels, skill mix, relational care, ward leadership and staff well-being. In the second part of the chapter, data are presented in relation to participant flow at the individual participant level in relation to each of the baseline data sets, and we include an overview of demographic characteristics for each data set. The chapter concludes with the findings from the baseline measures of quality of staff–patient interaction, patient-reported evaluations of emotional care and nurse-reported measures of empathy.

Recruitment and flow of ward teams (clusters)

Six ward nursing teams across two hospitals took part in the study, with three ward teams in each hospital. Ward leaders in all of the three nominated teams in Hospital A agreed to take part on behalf of their team. Two ward leaders in Hospital B agreed to take part, but a third ward leader nominated in Hospital B declined. Their team was about to embark on a quality improvement project and the leader was concerned that participating in the CLECC study as well would be too burdensome for the team. The matron concerned nominated an additional ward team caring for a high proportion of older patients, and this team’s ward leader agreed to take part. Each of the ward leaders was consulted about the prospect of randomisation to intervention or control conditions and no concerns were raised about this feature of study participation, indicating that the randomisation strategy was acceptable to participating staff. Thus, six ward teams entered the study and all remained in the study until all data collection was complete.

Both participating hospitals were NHS trusts located in separate urban areas in the same geographical region in England. Hospital A is a university hospital and foundation NHS trust employing between 7000 and 8000 staff and providing services to > 1 million people plus a wide range of specialist services to > 3 million people. Hospital B employs > 6000 staff and provides acute services to > 0.5 million people living locally and provides a smaller range of specialist services than Hospital A. The urban areas served by both hospitals have pockets of high deprivation. Life expectancy at birth is marginally higher than the national average in Hospital A locality and lower in the Hospital B locality. Hospital A locality has a lower percentage of the population than the English average who are classified as being from the white ethnic group and Hospital B locality has a higher than average white population.

Individual ward characteristics

Table 8 shows an overview of the characteristics of individual wards in the study, taken from the ward profiles completed by each of the ward leaders at baseline. Four wards were MOP wards and two were surgical wards. All the wards had a similar number of beds, with Hospital A wards having fewer beds than Hospital B wards. There was more variation between the wards in mean length of stay, ranging from 6 days on one of the surgical wards to 19 days on the other surgical ward. All four MOP wards had a similar mean length of stay (13–14 days). Planned full-time equivalent (FTE) staffing levels (RNs and HCAs) and nursing skill mix (proportion of registered nurses in total of RNs and HCAs) varied between the wards. The two surgical wards had higher planned staffing levels and skill mix than the MOP wards. In each ward in Hospital A, the numbers of actual staff in post were lower than planned, whereas for all wards in Hospital B, the staff numbers in post were higher than planned. All the wards used a mixture of short (approximately 8 hours) and long (approximately 12 hours) shifts for nursing work during the day.

There was substantial variation between the wards in the proportion of patients on the ward who ward leaders judged needed help with all of their activities of daily living (ADL), an indicator of patient dependency and thus need for help from nursing staff with these activities (0–45%). Wards in Hospital A had fewer beds and lower numbers of actual staff in post than Hospital B. Wards leaders in Hospital A also identified a higher proportion of patients needing help with all their ADL than Hospital B.

Ward leadership characteristics

All wards had a ward leader (senior sister) and deputy in post. Table 9 shows the length of experience for each ward leader in that role on that ward, as a ward leader overall and in any role on that ward. Wards A, D, E and F all had ward leaders with at least 9 years’ experience as a ward leader, whereas ward leaders for wards B and C were new or relatively new to the role. The ward leader for ward B took up post, their first at that level, at the beginning of the study. Only the ward F ward leader worked in a totally supernumerary role, with the others regularly or often being counted in the staffing numbers on a shift.

Quality of care

Table 10 shows the results of the MAQC completed through written matron’s survey. A completed matron’s assessment was received for each ward (n = 6). Higher scores indicate more favourable ratings. The wards have similar scores across both subscales with the exception of ward F, which had markedly lower scores for both subscales.

Relational care

A sample of nursing staff (n = 29) from all wards was interviewed using qualitative methods to ascertain their understanding of relational care on the ward where they worked. There was no significant difference between wards and hospitals in the way that nurses generally described relational care. The term compassion was used with examples given of interactions that staff perceived as conveying compassion in their care delivery, for example offering a relative a cup of tea to make the patient more relaxed because patients see we are looking after their family as well (N002 staff nurse). There was a recognition among nurses of interactions that patients do not perceive as compassionate, for example getting on with a nursing intervention with a patient without first introducing yourself and what you intend doing (N001 staff nurse). Staff explained that this happened because of time pressure to complete all expected nursing activities. Nurses who admitted to not always introducing themselves or actively getting to know the patient perceived these episodes of care as lacking compassion. Time was reported as a significant factor in delivering compassionate care, with staff challenged by patients who appear to need more time than others to express their needs or concerns. However, staff felt that compassion can be transmitted even in few words, for example when someone is clearly distressed because they have been incontinent, taking care with tone of voice, saying ‘it’s OK’ (N011 staff nurse).

Ward leaders reported incorporating discussion about compassionate care delivery with their ward team. They reported that a number of wards had patients who needed individual care, patients who tried to walk but, because of cognitive impairment, were unaware of their physical limitations to do this effectively and safely, and so their care required an increase in the usual staff complement. Agency staff were employed to provide this individual care and ward staff reported seeing them work without compassion, for example sitting and watching the patient rather than interacting with them. Ward leaders reported actively engaging with all nursing staff who interact with patients on the ward, including agency staff, to promote staff expectations of compassionate care delivery.

Care on the ward was reported not to be focused solely on medical management of a patient but also on the incorporation of other interventions seen as important to an individual’s well-being. Staff saw their role as offering choice to patients and this was reported to be led and supported by ward leaders, for example by offering the use of a dining table for patients who are able with or without assistance to get there (N034 ward leader). Relational care was talked about as an important dimension to the nursing role, and this was reflected in stories about connections developed between individual patients and individual staff members, and the sadness felt by staff when these patients died. This was especially the case when the staff member felt that they had been successful in providing good relational care, and that the relationship had been strong. All of the ward leaders considered role modelling and educating the nursing team to deliver good relational (compassionate) care to be an important part of their role.

Nurses on all wards shared an understanding of good relational care and felt that the general public perceived nurses as being less compassionate than they used to be. Examples were readily available of what constituted good relational care, as well as the barriers to and facilitators of the achievement of this standard. There was consensus that the degree to which care was compassionate varied, but that it was not ward specific, rather it was both time and staff dependent. Nursing staff were aware of a variety of initiatives that had been put in place to address the perceived lack of compassion in care delivery, citing the Friends and Family Test119 most frequently. Staff generally were keen to find or be offered strategies to assist them in delivering good relational care more consistently. The following quotation reflects what ward leaders expected and what nurses expected to be able to deliver:

I’m expecting them to look after the whole person, see what their needs are, help them with those needs and make them feel like the most important person they’re dealing with at that point in time.

Staff well-being

Table 11 displays findings from the baseline ward profile completed by the ward leader in relation to indicators of staff well-being. All the wards had new staff members joining over the previous 6 months, with wards D and E having the highest number of new starters. Sickness absence rates varied considerably, from 0.6% on ward A to 10.7% on ward D. All the wards used bank or agency staff regularly, at least several times per week. On ward A, staff very rarely missed their breaks, whereas missing breaks happened more frequently on the other wards and on a daily basis on ward C.

Table 12 indicates the degree of burnout taken from baseline nursing survey responses (n = 91). Higher mean emotional exhaustion and depersonalisation scores, and lower personal accomplishment scores, indicate less favourable conditions and higher burnout. Wards A and E had a mean emotional exhaustion score lower than the overall mean, and wards A, D and F had a lower mean score for personal accomplishment. Wards A and D had a higher personal accomplishment mean score.

Table 13 shows the proportion of staff on each ward experiencing burnout calculated from nursing survey responses (n = 91). Emotional exhaustion varied from 27% of 22 staff on ward E to 50% of 18 staff on ward C. Depersonalisation ranged from 0% of 12 staff on ward A to 40% of five staff on ward B. Personal accomplishment ranged from 10% of 21 staff on ward F to 22% of 18 staff on ward C.

Other questionnaire results

Baseline and follow-up results for all questionnaire data are reported in Appendix 8 .

Individual participant flow

This section and the accompanying CONSORT diagram (Figure 7) focus primarily on individual participant flow in relation to the pilot CRT. Later, in Nursing staff qualitative interviews , but not reflected in the CONSORT diagram in Figure 7 , data are also shared about participant flow in relation to nursing staff in the qualitative elements of the study.

FIGURE 7 The CONSORT flow diagram for pilot cluster CRT. Reproduced from Gould et al. 2018.120 © Article author(s) (or their employer(s) unless otherwise stated in the text of the article) 2018. All rights reserved. No commercial use is permitted unless otherwise expressly granted. This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/. (Figure not reproduced here; see the original article.)

All four wards randomised to the intervention went on to receive the intervention. Figure 7 shows the flow of clusters and participants through the pilot CRT. Randomisation took place after baseline data collection, but the results are presented by allocation for baseline and follow-up data to enable comparisons between groups to be made.

A number of data from nursing staff, patients and visitors were gathered at baseline and repeated at follow-up to enable study outcomes to be assessed. Follow-up was at the cluster level rather than at the individual participant level. For instance, eligibility of patients and visitors was dependent on their being an inpatient during both assessment periods, which was an unlikely scenario given that they were 1 year apart. In addition, some turnover of staff employed on the ward teams was anticipated. Therefore, to maximise recruitment at each assessment point, recruitment for baseline and follow-up assessment periods was independent.

Observations

Recruitment was at the individual patient level for observation of staff–patient interactions. It was possible for individual patients to be involved in more than one observation session and the longer their stay on the ward, the more likely it was that they would be approached more than once to consent to their care being observed (provided they had consented on previous approaches). In total, 278 individual patients were approached, with a total number of 362 approaches. Observation recruitment data in Figure 7 reflect the number of approaches rather than the number of individual patients.

The overall recruitment rate to baseline observations was 97%, that is, eligible patients agreed to take part in 152 out of 157 approaches. These rates were similar between intervention and control wards (96% vs. 98%). At baseline, patients consented to participate on 152 occasions. This corresponds to 123 individual patients because some were approached and consented more than once. Of the 152 occasions on which patients were consented they were observed 133 (88%) times. Some patients were consented and not observed because they were no longer available or eligible once observation was due to start.

Similar patterns were evident in the recruitment data for follow-up assessment, although recruitment rates were slightly lower. Overall recruitment to observations at follow-up was 90% (157 of 175 approaches). These rates were similar between intervention and control wards (90% vs. 88%). At follow-up, on 157 occasions, patients consented to participate. This corresponds to 114 individual patients as some patients consented more than once. Of the 157 occasions on which patients were consented, they were observed 140 (89%) times. Again, some patients were consented but not observed.

Across both assessment periods, in 93% (i.e. 309 out of 332) of approaches to eligible patients inviting them to participate in observations, patients agreed to take part, indicating a high acceptability. Recruitment rates at baseline were slightly higher than rates at follow-up. Overall recruitment rates were the same between intervention and control wards (93% vs. 93%).

Reasons recorded for patients declining participation included ‘not feeling up to it’ (17%, n = 4), ‘too unwell’ (4%, n = 1) and ‘no reason’ (8%, n = 2). No specific reason was recorded for 70% of approaches (n = 16). In 17% of approaches (i.e. 63 out of 362) the patient was approached and then assessed as not having the capacity to make the decision to take part in the research. In 67% (i.e. 42 out of 63) of these cases, researchers were able to contact a consultee for advice and in 100% of these cases the consultee advised that the patient should participate.

The mean age of patients observed was 82 years (84 years in the intervention group and 77 years in the control group). Patients in the control group were, on average, 7 years younger than patients in the intervention group, and this difference was statistically significant (p < 0.001). Most patients were female (77%) and 25% had evidence of cognitive impairment (31% at baseline and 19% at follow-up). There were no overall differences in gender and cognitive impairment between experimental groups.

Patient questionnaires

Potentially eligible patients (n = 621) were approached and invited to complete a written questionnaire, with researcher help if preferred, resulting in the completion of 354 questionnaires. Patients were approached having been screened as potentially eligible but, once approached, some patients (21%, 57 out of 274 approaches at baseline across six wards) were assessed as not meeting the inclusion criteria and were excluded from the study. A further 44 patients approached at baseline declined to take part. The overall recruitment rate to the baseline patient questionnaires was 80% (i.e. 173 out of 217 eligible patients agreed to take part). Recruitment rates were similar between intervention and control wards (78% vs. 83%). At baseline, questionnaires were completed for 97% (i.e. 168 out of 173) of the patients who consented.

Similar patterns were evident in the recruitment data for follow-up assessment. On approach, 100 patients in 347 approaches were assessed as not the meeting inclusion criteria. A total of 61 patients declined to take part. Overall recruitment to patient questionnaires at follow-up was 75% (i.e. 186 out of 247 eligible patients). Recruitment rates were similar between intervention and control wards (76% vs. 74%). At follow-up, a questionnaire was returned for 100% of the 186 patients who consented.

Across both assessment periods, 77% (i.e. 359 out of 464) of eligible patients agreed to take part in the questionnaire survey, indicating good acceptability. Although follow-up recruitment rates were slightly lower than at baseline, the number of patients approached was higher, so more patients were recruited overall to complete a questionnaire at follow-up than at baseline. Overall recruitment rates were similar between intervention and control wards (77% vs. 78%).

The most frequent reasons recorded for patients declining participation in the questionnaire survey were ‘tired’ (40%, n = 12) and ‘questionnaire too difficult’ (10%, n = 3). The most frequent reasons recorded for excluding patients were not having capacity (43%, n = 48) and being ‘very cognitively impaired’ (29%, n = 32).

All returned questionnaires were included in analyses. Most patients who completed questionnaires were female (70%), aged > 70 years (83%) and of white British ethnicity (97%). In total, 61% had other illnesses apart from the reason for hospital admission, 68% needed help with ADL while in hospital and 78% had been in hospital for > 3 days. Of all the patient questionnaires returned, 12% were completed by patients with cognitive impairment.

There were a small number of differences (not statistically significant) between intervention and control group patients who completed questionnaires at baseline (Table 14). At baseline, all of the intervention ward patients who completed a questionnaire were aged ≥ 61 years, whereas the control group, although including mostly patients aged ≥ 61 years, also included 13 people (22%) who were aged 31–60 years. At baseline, 26 (43%) of the control group were male compared with 25 (25%) of the intervention group. At baseline, 30 (53%) of the control group patients identified that they had other health conditions apart from the one that had brought them into hospital, compared with 62 (70%) of intervention ward patients. Twenty-eight (48%) of the baseline control group patients said that they needed help from others with ADL in hospital, compared with 74 (78%) of the intervention group. These differences about needing help from others were similar at follow-up. For the other responses identified here, there were no marked differences between groups at follow-up.

Visitor questionnaires

Visitors on the ward were approached and invited to complete a written questionnaire (n = 219), with researcher help if preferred. This resulted in completion of 176 questionnaires over both periods.

Of those approached across the six wards at baseline, six people declined to take part. The overall recruitment rate to baseline visitor questionnaires was therefore 94% (i.e. 101 out of 107 visitors agreed to take part). This rate was similar between intervention and control wards (93% vs. 97%). At baseline, questionnaires were completed for 88% (i.e. 89 out of 101) of the visitors who consented.

Similar patterns were evident in the recruitment data for follow-up assessment. A total of 13 visitors declined to take part. Overall recruitment to visitor questionnaires at follow-up was 88% (i.e. 99 of 112 eligible patients). These rates were similar between intervention and control wards (87% vs. 90%). At follow-up, a questionnaire was returned for 88% (i.e. 87 out of 99) of the visitors who consented.

Across both assessment periods, 91% of visitors approached (i.e. 200 out of 219) agreed to take part, indicating a high acceptability. Recruitment rates at baseline were slightly higher than rates at follow-up. Overall recruitment rates were similar between intervention and control wards (90% vs. 94%).

All returned questionnaires were included in analyses. Table 15 shows the characteristics of visitor questionnaire respondents. The mean age for visitors who completed the questionnaires was 62 years and 63% were female. Most (98%) were of white British ethnicity. In total, 25% were either the spouse or the partner of the patient, and 42% were a son or daughter of the patients. A total of 58% identified themselves as a carer to the patient. At follow-up, 43% of intervention group visitors were visiting a patient with a dementia diagnosis, compared with 18% of control group visitors, and this was the only statistically significant difference between the two groups.

The most frequent reasons given by visitors declining participation were ‘had already filled in a form that day’ (16%, n = 3) and ‘felt unable to answer accurately’ (16%, n = 3).

Nursing questionnaires

All RNs and HCAs employed to work on the participating wards at the time of data collection (n = 496) were given a questionnaire to complete. This resulted in the completion of 178 questionnaires.

Of the 249 questionnaires distributed across the six wards at baseline, 158 were not returned. The overall recruitment rate to baseline nursing questionnaires was therefore 37% (i.e. 91 returned out of 249). Baseline return rates were lower on intervention wards than on control wards (31% vs. 48%). At follow-up the response rate of 35% (i.e. 87 questionnaires returned out of 247 distributed) was similar to baseline. Follow-up return rates were similar between intervention and control wards (33% vs. 39%).

Across both assessment periods 36% of questionnaires (i.e. 178 out of 496) distributed were returned completed. Overall return rates were lower on intervention wards than on control wards (32% vs. 43%).

All of the returned questionnaires were included in the analysis. As illustrated in Table 16 , respondents represented a range of ages, ethnic groups, job roles/bands and years of experience.

Nursing staff qualitative interviews

All nursing staff on all participating wards were invited to participate in qualitative interviews, in addition to the two PDNs, two senior trust nurses and three matrons. Intervention ward staff and PDNs were invited to interview on three occasions (period 1, period 2 and period 3), and control ward staff were invited on two occasions (period 1 and period 3). At the final interview round (period 3), all three matrons overseeing the participating wards and two further people in senior trust nursing roles in the trusts were invited.

In total, 59 interviews were conducted, over three rounds, with 33 people. In total, 17 people were interviewed once, six were interviewed twice and 10 were interviewed three times. Eleven people who were interviewed once (at either period 1 or period 2) could not be contacted again or declined to participate in a further interview round, and one person who was interviewed twice did not take part in the third interview round. Six people who dropped out were HCAs and six were RNs.

Two senior trust nurses, two PDNs, 21 intervention ward staff and eight control ward staff were interviewed. All staff levels were represented at each interview round. All intervention wards were represented at each interview round and both control wards were represented at the first and third rounds.

All but one of the interviewees were female. In total, 11 interviewees were band 5 staff nurses, 10 were HCAs, six were ward leaders, two were deputy ward leaders, two were PDNs and two were senior trust nurses. A range of ages was represented, from the ‘29 years or under’ to the ‘60–69 years’ age band. RNs (including PDNs and senior trust nurses) had between 2 and 26 years’ experience as a RN, with a mean length of experience equalling 11 years. HCAs had between 2 and 35 years’ experience as a HCA, with a mean length of experience equalling 14 years.

Table 17 shows the characteristics of the nursing staff interviewed, not including the PDNs or the senior trust nurses. No significant differences between control and intervention ward staff were recorded.

Baseline measures

This subsection reports the baseline scores for each ward on the study’s core outcomes: quality of staff–patient interactions, patient evaluation of emotional care and nursing self-reported empathy. It also compares these outcomes by experimental group in the study, that is, intervention or control.

Quality of staff–patient interactions

The observed quality of staff–patient interactions using the QuIS is shown per ward in Table 18 . All wards were observed for the same total time (20 hours each) balanced across days of the week (Monday to Friday) and times of day (08.00 to 22.00). The number of interactions across this 20-hour period varied between wards from 201 on ward F to 322 on ward B (see Table 18 ). The distribution of scores across the five available categories was broadly similar between wards, with most interactions (73% overall) scoring either positive social (13%) or positive care (60%). A total of 17% of interactions were rated neutral, 6% were rated as negative restrictive and 4% were rated as negative protective.

Table 18 also shows the number and proportion of all negative interactions, that is, the sum of negative protective interactions and negative restrictive interactions. The proportion of negative interactions ranged from 2% on ward F to 18% on ward C.

In total, 40% of patients observed for the full planned 2-hour period had one or more negative interactions, but the proportion on individual wards varied from 13% on ward F to 71% on ward A (see Table 18 ).

Table 19 shows the quality of interaction by experimental group. Distribution across the QuIS categories is similar for intervention and control wards. Table 19 confirms that the proportion of negative interactions is the same (10%) for intervention and control wards. Of all patients observed, an average of 8% of interactions per patient were negative, and these proportions were similar across intervention (8%) and control (7%) wards. Of patients observed for the full 2-hour period, 45% of intervention ward patients had one or more negative interactions in a 2-hour period, compared with 30% of control ward patients.

In addition to tests of effectiveness, we also used baseline QuIS data to quantify and characterise staff–patient interactions and to identify the factors associated with negative interaction ratings. Findings are shown in our published paper displayed in Appendix 7 .118

Patient evaluation of emotional care

Patient evaluations of emotional care in each ward using the PEECH tool, administered through written patient surveys, are shown in Tables 20 and 21 . Higher scores indicate more favourable ratings. Table 20 displays mean and median PEECH scores for each subscale and total by ward. Scores tended to be broadly similar between wards for each subscale, with the connection subscale consistently scoring lower than the other three subscales on all wards. Total mean PEECH scores (sum of the mean of all items) ranged from 44.5 on ward C to 52.9 on ward E (out of possible total of 66). The distribution of scores was also similar between the two experimental groups (see Table 21 ).

An analysis was also performed of frequency of low PEECH scores, that is, when patients scored on average 2 or below (Table 22). Ward D had the highest proportion of patients with lowest scores for each subscale. No one ward consistently had the lowest proportion of patients with low scores for each subscale.

Table 23 indicates similar frequencies of patients with low scores for each subscale between intervention and control wards.

Nursing empathy

Levels of self-reported empathy using the JSE from the nursing survey varied across the individual wards, with wards B and F scoring lower than the mean for all wards, and wards C, D and E scoring higher (Table 24). Higher scores indicate higher empathy.

Table 25 indicates that intervention wards at baseline had a higher mean and median empathy score than control wards.

Chapter summary

Six out of seven ward teams invited to take part participated in the study and all remained in the study until data collection was complete. Ward teams were randomised to intervention (n = 4) or control (n = 2), and staff were amenable to the prospect of randomisation to either experimental condition.

This chapter has described the six participating wards at baseline assessment across a range of characteristics including specialty, patient dependency, staffing, ward leadership, ward climate, staff well-being, care quality, patient safety and core study baseline measures. The findings reflect a range of ward contexts, with similarities across some dimensions, such as bed numbers. Some differences between wards were evident, including staffing levels and duration of ward leadership. Matrons’ assessments of ward quality indicate that for all but one low-scoring ward, matrons assessed the care quality as good but not of the highest possible quality. Our qualitative interviews with staff did not identify particular differences between the wards, and, although all talked about the value of relational care, staff also reflected that limited time with individual patients constrained their capacity to be compassionate. Most staff–patient interactions observed were rated as positive. However, a proportion of interactions on all wards were rated negatively. In total, 40% of patients observed for 2 hours were rated as having one or more negative interactions during the 2-hour period. This proportion varied widely between wards (with intervention wards having a higher proportion than control wards) but indicates room for improvement on all participating wards. Staff survey results indicated that staff on all the wards showed signs of burnout; this was the case for over one-third of respondents. However, low response rates reduce certainty in the representativeness of these proportions in the staff group as a whole. Some individual staff on all the wards rated their empathy levels as low, others as high, and mean empathy levels for ward teams varied, with intervention wards having higher reported empathy. Low staff survey response rates suggest that these results need to be treated with particular caution. Overall, although these findings indicate some differences between the wards at baseline, there are also some remarkable similarities, particularly in relation to the quality of relational care, which is largely positive, yet with clear scope for improvement, and in relation to the presence of burnout in the staff group.

Although we were not successful in recruiting high proportions of nursing staff to take part in the written survey, the staff that did take part represented a wide range of characteristics in terms of work role, seniority, professional experience, age and ethnic group. Recruitment of patients and visitors was more successful, with the observations of staff–patient interactions attracting a particularly high proportion of recruits. Although not as high, recruitment to patient survey was still at a good level. Most patients involved in observations and the survey were older and female. One-quarter of patients who were observed had evidence of cognitive impairment, compared with 12% of questionnaire respondents. A high proportion of visitors approached to complete a questionnaire agreed to do so, with an average age of 62 years. The findings also indicate that the characteristics of participating staff, patients and visitors were equivalent across the experimental groups.