Although current definitions of compassion in nursing practice are imprecise and sometimes confused (see Chapter 1 ), there is intense interest in this problem both within and outside the profession of nursing. However, little is known about what strategies are effective in promoting compassionate care among nurses. To date there has been no rigorous critical overview of research into interventions designed to promote compassionate care among nurses in practice. This chapter aims to provide an overview of the evidence base on the evaluation of interventions for compassionate nursing care. It begins with an overview of qualitative research on compassionate care interventions. It then reports a systematic review of studies that evaluate the effectiveness of interventions to promote compassionate nursing care.
Qualitative research
Recent years have seen the use of qualitative research methods to underpin the development and evaluation of a number of interventions focused on improving compassionate care, or dignity in care, at the hospital ward level.30,34,36,47–49 Interventions developed and evaluated in this way have typically been faciliated by a senior nurse, using reflective learning, action research and/or appreciative inquiry to work with ward-based nursing staff (often using patient stories and/or observations of practice) to strengthen support for existing good practice and to make changes when needed. These interventions are often shaped by a ‘relationship-centred’ philosophy in which achieving the well-being of all groups (patients, staff, family carers) is seen as fundamental to high-quality care.41 They have used democratic and participatory processes involving patients, staff and sometimes family carers to articulate the patient’s needs and to shape the practice changes made.
The accompanying qualitative evaluations have provided important information about the processes of change and the factors enabling and inhibiting sustainable change. Some of these evaluations have reported concrete practice changes resulting from the intervention,36,47–50 while others report more variable success.30,34 For instance, Dewar and Nolan49 and Dewar50 used appreciative inquiry and action research to involve older people, staff and relatives in developing compassionate relationship-centred care on an acute hospital ward. Methods used included participant observation, interviews, storytelling and group discussions. The findings indicated the value of appreciative caring conversations between staff, patients and relatives enabling all parties to discover ‘who people are and what matters to them’49,50 and ‘how people feel about their experience’49,50, with this knowledge enabling them to ‘work together to shape the way things are done’.49 In the resulting model, Dewar and Nolan49 detail how older people, staff and relatives can work together to implement compassionate relationship-centred care.49 In specifying ‘how people can work together to shape the way that things are done here’, Dewar50 identified a number of important conditions for staff to feel able to express emotions, share experiences and ideas with each other, consider others’ perspectives, take risks, use ‘curious questioning’ to examine situations and challenge existing practice, all identified as important actions to support the delivery of compassionate care. These conditions included transformational leadership, the level of support received from colleagues and senior staff, a shared set of principles for caring, open dialogue within the team and opportunities in which people had permission and space to reflect. These conditions echo the findings from other research as conditions at the team level that can support high-quality care.26,27,51 Dewar50 reports how these conditions developed and how compassionate caring practices became embedded in the work of the team over the course of the year-long project, providing valuable evidence that change of this kind is possible.
However, Dewar’s project took place over the course of 1 year on an already high-performing ward with a strong leader.50 The findings informed development work across a wider Leadership in Compassionate Care project implemented in a number of settings, but evaluation of the impact of these strategies elsewhere does not report the influence of the ward climate or programme length on outcomes, so evidence is lacking that such strategies can be universally effective regardless of work team context.52 In a contrasting study to Dewar’s that explored the use of discovery interviews with older hospital patients as a way of improving dignity in care, Bridges and Tziggili30 found that ward teams required strong and consistent leadership and intense preparation before they were able to hear and respond to patient stories about care. Both organisations involved in this dignity project experienced significant delays in the progress of the project and limitations in its impact because of a lack of leadership at the ward level and a lack of preparedness of the ward teams to engage in responding positively to patient feedback. One ward team with a strong leader was able to successfully engage with the patient stories, but only after some months of team preparation. These findings indicate that, while some wards may be ready to engage in programmes, such as Dewar’s, others could benefit from a period of groundwork in which leadership and mutually supportive team practices are established.
The evaluative focus of these studies is the mechanisms for change used, particularly the democratic and participatory processes that involve patients, staff and sometimes family carers in articulating the patients’ needs and shaping the practice changes made. These qualitative accounts often provide a fuller picture of the interventions deployed than the studies reviewed below, and often include an analysis of the enablers and barriers to change. However, they do not examine in depth the process of implementation itself and so fail to systematically identify the contexts in which successful implementation is more likely or, where contexts are not receptive, how resources, relationships and norms in the wider system may need purposeful restructuring to support implementation and sustain longer-term change.53,54 In addition, as would be expected with a qualitative approach, there is only weak objective evidence of effectiveness of the interventions deployed in these studies in relation to impact on patient outcomes.
Review methods
The remainder of this chapter reports a systematic review of studies that evaluate the effectiveness of interventions for compassionate nursing care, using the four key components of the compassion narrative identified in Chapter 1 to provide an operational definition. The objectives of the review were:
- to systematically identify, analyse and describe studies that evaluate interventions for compassionate nursing care
- to assess the descriptions of the interventions for compassionate care used, including design and delivery of the intervention and theoretical framework
- to evaluate the nature and strength of evidence for the impact of interventions.
The review was conducted, guided by the Cochrane Collaboration methods55 to assure comprehensive search methods and systematic approaches to analysis of the review materials. Sections of this review report are reproduced or adapted with permission from Blomberg et al. 16 and with permission from Elsevier. © 2016 The Authors. This is an open access article under the CC BY-NC-ND 4.0 license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Search strategy
A systematic search for primary research evaluating compassionate care interventions was undertaken on CINAHL, MEDLINE and The Cochrane Library databases (including the Cochrane Database of Systematic Reviews, Database of Abstracts of Reviews of Effectiveness, CENTRAL register of controlled trials, Health Technology Assessment Database and Economic Evaluations Database) in June 2015. No date limits were applied to the searches conducted.
Terminology in relation to compassionate care is problematic and as noted above, there is no one agreed definition of compassionate care. Instead, a number of terms are used interchangeably and inconsistently across the health-care literature. A broad and inclusive approach was therefore used in preliminary searches to scope and map the field. As many terms relating to compassionate care were identified and used as possible, but with a focus on identifying studies that reflected one or more of the key components of compassionate care outlined above. Through this mapping, relevant keywords were identified (e.g. professional–patients relations, dignity, person-centred care, relationship centred care, empathy, compassion, caring, and emotional intelligence). These keywords were used in final searches. Terms related to compassion were combined (AND) with terms related to relevant methods and occupational groups. Relevant index terms were included, which varied across databases (see Appendix 1 for MEDLINE and CINAHL search strategies). Although no additional searches for unpublished (so called ‘grey’) literature were conducted, the sources used do index PhD theses (CINAHL) and some conference abstracts (CIHAHL, The Cochrane Library). Searches were limited to the English language.
Selection
An adapted PICO (population, intervention, comparison and outcome) framework was used to guide study selection.56 We included primary research studies comparing the outcomes of an intervention designed to enhance compassionate nursing care (in any setting to any client group) with those of a control condition. Eligible designs were randomised controlled trials (including CRTs) or other quasi-random studies, interrupted time series and before-and-after studies (controlled or uncontrolled). Studies were excluded if they were focused exclusively on students, or if interventions were not directed at changing nursing staff behaviour.
The lack of conceptual clarity about compassion in the literature necessitated an inclusive approach to studies that were not necessarily labelled as addressing ‘compassion’. We developed selection criteria based on the four elements of the compassion narrative described above (moral attributes of a ‘compassionate’ nurse including empathy, nurses’ situational awareness of vulnerability and suffering, nurses’ responsive action aimed at relieving suffering and ensuring dignity, and nurses’ relational capacity) so that studies were included if they met one or both of the following criteria:
- explicit goal of the intervention was stated as improving compassionate nursing care (or a closely related construct, i.e. dignity, relational care, emotional care) (through addressing nurses’ moral attributes, situational awareness, responsive action and/or relational capacity) and/or
- primary outcomes that assessed or evaluated either nurses’ self-reports of compassion and/or ability to deliver compassionate care (moral attributes, relational capacity), and/or observed quality of interactions or other measure of compassion (situational awareness, responsive action), including patient reports of experienced compassion or a closely related construct.
The titles and abstracts from the search were screened against the inclusion criteria independently by individual review team members (JB, KB, PG and YW; see systematic review team members listed in Acknowledgements ). During the screening process, frequent meetings were held by the research team in order to compare independent selections, resolve disagreements and make decisions. On independent rating (i.e. before discussion) reviewer pairs achieved between 80% and 90% agreement. In most cases of disagreement papers were excluded after discussion. Full-text papers were retrieved for all papers that screened positively in the first stage or about which a clear decision could not be taken (due to lack of information). Each full-text paper was reviewed independently by two team members followed by a decision to include or exclude in the final review. These reviews were followed by further team discussion to finalise inclusion into the data set. The reference lists of full-text papers included were scanned for further items. The search and selection process is summarised in the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow chart (Figure 2). The number of duplicates removed was not recorded.
FIGURE 2 Flow diagram for systematic review searches. Adapted from Blomberg et al.16 with permission from Elsevier. © 2016 The Authors. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). (Figure not reproduced here; see the original article.)
Quality assessment
In order to effectively represent the variation in study quality evident in findings from the preliminary mapping phase, and to properly reflect the strength of evidence, we undertook a simple grading in order to categorise the strength of the underlying design of studies we retrieved.57 Because of heterogeneity of study design identified in early scoping work, a rating of strong, medium or weak quality was allocated to each study depending on where the study design sat on the hierarchy of evidence for effectiveness in tandem with an assessment of its design and execution. The method selected was in line with the GRADE (Grading of Recommendations Assessment, Development and Evaluation) system as used by Cochrane for rating evidence to guide a broad assessment of individual study quality.57,58 Studies were rated as high quality where outcomes were compared between treatment (intervention) and control groups, where allocation to groups was random, and where equivalence between groups was explicitly demonstrated. Study designs included here were randomised controlled trials (RCTs) and CRTs which met these conditions. Studies were rated as medium quality where outcomes were compared between intervention and control groups, and where equivalence between groups was demonstrated, but where other methodological issues weakened the design, for instance non-random allocation to groups or small sample size. Study designs included here were CRTs with small numbers of clusters (for instance, n = 2) and controlled before-and-after studies with non-random allocation to groups. Uncontrolled before-and-after studies were rated as low quality as were other studies where other significant methodological shortfalls weakened claims of demonstrating effectiveness (e.g. controlled before-and-after studies where equivalence between groups is not demonstrated). These quality assessments were made by individual members of the review team, and checked with one other team member’s ratings until consistent ratings were achieved.
An evaluation of quality of description of the intervention was also performed for each included study. The material used as the basis for this evaluation was the information provided in the paper about the intervention in addition to further information about the intervention accessed from sources referenced within the original paper. We did not otherwise seek out information about the intervention, wishing to test the extent to which the original paper and its referenced sources provided sufficient information to enable the intervention to be replicated. Each study was analysed against the criteria for description of group-based behaviour change interventions devised by Borek et al. 59 This framework provides a checklist for assessing the reporting of behaviour change interventions against 26 criteria covering intervention design, intervention content, participants and facilitators. Intervention design features assessed included intervention development methods; setting; venue characteristics; number, length and frequency of group sessions; and period of time over which group meetings were held. Intervention content assessed included change mechanisms or theories of change, change techniques, session content, sequencing of sessions, and participants’ materials activities during sessions and methods for checking fidelity of delivery. Participant features assessed included group composition and size, methods for group allocation, and continuity of group membership. Facilitation features assessed included number of facilitators; facilitator characteristics and preparation including professional background, personal characteristics, training in intervention delivery and training in group facilitation; continuity of facilitator’s group assignment, facilitator’s materials and intended facilitation style. These assessments were conducted by one team member, and supplemented and refined in discussion with other team members.
Data analysis
A qualitative analysis was conducted across the different interventions reported to describe intervention types and contexts, and mechanisms for change. This analysis was conducted in smaller groups in the review team but further enriched through discussion of process and emerging findings among all group members.
Data were extracted for each study by Jackie Bridges and Karin Blomberg, including study design, sample and settings, summary details of intervention, outcomes and measurements, results and process issues. Results were tabulated and used to generate summary descriptions across key characteristics. Heterogeneity of studies in terms of interventions, methods and outcomes meant that a meta-analysis was not warranted, and so a more descriptive approach was merited. We considered the potential to pool studies using standard mean differences for measures, but this method requires that the instruments are measuring the same underlying construct and that the interventions have common mechanisms, but this was not clearly the case. The main intervention types were agreed through team discussion, as were key outcome types. Findings on effectiveness of individual interventions were plotted against key outcome types, and this was used as the basis for an analysis of evaluation strategies by intervention type and strength of evidence of effectiveness across intervention type and across the field as a whole. We recorded and tabulated both the direction of differences between groups (where reported) and statistical significance of differences. For controlled before and after studies, where there was no test of between group differences or groupby time interaction, this was categorised as a non-significant difference irrespective of a significant within-group difference. To inform the design of a future evaluation, we undertook a descriptive analysis of feasibility findings and other limitations identified in the medium- and high-quality studies we included.
Review findings
The review findings are presented here to address each of the review objectives in turn. First, we describe study characteristics to give an overview of studies that evaluate interventions for compassionate care. Second, we present an assessment of the quality of reporting of the interventions in the included studies, including their theoretical foundations. Third, we present evidence of effectiveness of the interventions in the included studies and analysis of the quality of that evidence.
Study characteristics
The final data set comprised 24 studies reporting 25 interventions. Twenty-two studies were published in journals60–81 and a further two were doctoral theses.82,83 Three types of intervention were identified. Staff training interventions (n = 10) focused on the development of new skills and knowledge in nursing staff, such as a training course in empathic skills communication. Care model interventions (n = 9) focused on the introduction of a new care model to a service, such as person-centred care. Nurse support interventions (n = 6) focused on improving nursing staff support and well-being through, for instance, the provision of clinical supervision.
Reports reflect a range of study settings including hospital (n = 14),60,61,63,64,67,68,73,74,76,77,79,81–83 care/nursing homes (n = 6),65,66,69–72 other community settings (n = 3)62,75,78 and one study that used a range of health and social care settings (n = 1).80 All but one of the staff training studies was conducted in a hospital setting, and six out of eight care model interventions were conducted in care home settings. Nurse support intervention studies were conducted in hospital settings (n = 3), district nursing services (n = 1), hospice at home (n = 1) and outpatient oncology service (n = 1). Eleven studies were conducted in the USA, with the other studies conducted in a range of other countries mostly in Europe but also including Australia, Canada, China and Turkey.
Study participants included nurses, nurse managers, patients and relatives. To evaluate the effect of the interventions a range of measurements was used, mainly self-reported instruments, but the effect was also proxy rated by researchers and using instruments based on researcher assessments of verbal communication and interaction. The outcomes measured in the studies varied widely, but could be classified into three types: nurse-based, quality-of-care and patient-based outcomes.
A table for each intervention type providing summary individual study characteristics and findings can be found in Appendix 2 .
Quality of intervention reporting
Three types of intervention were identified: staff training, care model and nurse support. Interventions varied considerably in the extent to which they drew on an explicit theoretical foundation. Staff training interventions comprised training on verbal interactions, communication, communicating about spirituality and spiritual care, and empathy. Only four staff training interventions in the included studies had an explicit theoretical base. These were solution-focused brief therapy,60 relationship-based care model/caring theories,61 reminiscence theory and adult learning theory,83 and the Tibetan Buddhist tradition.62 Some interventions drew on definitions of particular concepts, such as empathy63,64,82 and caring behaviours.81 Other studies lacked an explicit theoretical foundation, referring only to results from previous research studies.
By contrast, all interventions introducing and testing a new care model were underpinned by an explicit framework. Most used theories or models developed in caring and nursing, except for one study using the International Classification of Functioning, Disability and Health 84 as the basis for an intervention to promote patient-centred communication with those living with aphasia/communication impairments.65 Frameworks emphasised the person-centred care/environment/nursing,66–68 the relationship between nurses and patients69–71 or dignity in care.72
Nurse support interventions were based on reducing compassion fatigue, burnout and/or secondary traumatic stress;73,74 and/or bolstering personal resources, such as compassion satisfaction, resiliency, empathy or sense of coherence.73–75 Three were based on mindfulness theory.76–78
Reviewer ratings of the quality of intervention reporting in each study against each item in the Borek et al. 59 frameworkfor description of group-based behaviour change interventions are displayed in Tables 2 and 3 . As is evident, the reporting of the interventions varied across all intervention types but was generally weak, with no intervention reports meeting all of the criteria deemed necessary for full intervention reporting. The design and the content of the interventions tended to be better described than details of the participants and the facilitators of the interventions. Overall compliance for intervention design reporting was 52% of criteria (shown in Table 3 row labelled ‘average % compliance by aspect of reporting’). The intervention design item with highest compliance (inclusion of details of the length of training sessions) was included in 73% (n = 16) of the 22 studies applicable here. The lowest was a specification of venue characteristics (n = 4, 17%).
| Intervention type | Study (first author and year) | Intervention | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Design | Content | ||||||||||||||
| Intervention source or development methods | General setting | Venue characteristics | Total number of group sessions | Length of group sessions | Frequency of group sessions | Duration of the intervention | Change mechanism or theories of change | Change techniques | Session content | Sequencing of sessions | Participants’ materials | Activities during the sessions | Methods for checking fidelity of delivery | ||
| Training | Ançel 200663 | N | Y | N | Y | Y | Y | Y | Y | Y | Y | N | N | Y | N |
| Boscart 200960 | N | Y | N | N | Y | Y | N | N | N | Y | Y | Y | Y | N | |
| Glembocki 201061 | N | N | N | Y | Y | Y | N | Y | Y | Y | N | N | N | Y | |
| La Monica 198764 | Y | N | N | N | N | N | N | Y | Y | Y | N | N | Y | N | |
| Langewitz 201079 | Y | N | N | Y | Y | Y | Y | Y | N | Y | Y | N | Y | N | |
| Puentes 199583 | N | Y | Y | Y | Y | N | N | Y | Y | Y | Y | Y | Y | N | |
| Searcy 199082 | N | Y | N | Y | Y | Y | Y | Y | Y | Y | N | Y | Y | N | |
| Taylor 200980 | Y | Y | NA | NA | NA | NA | N | Y | N | Y | N | Y | N | N | |
| Wasner 200562 | Y | N | N | Y | Y | Y | Y | Y | Y | Y | N | N | Y | N | |
| Yeakel 200381 | N | N | N | N | N | N | Y | N | N | N | N | N | Y | N | |
| Care model | Brown Wilson 201369 | Y | Y | N | Y | Y | N | N | Y | Y | Y | Y | N | Y | N |
| Chenoweth 201466 | N | Y | N | N | N | N | N | N | N | Y | N | N | N | N | |
| Finnema 200170 | N | Y | N | Y | Y | Y | Y | N | N | Y | Y | N | Y | N | |
| Ho 201672 | N | Y | N | N | N | N | N | Y | Y | Y | N | N | N | N | |
| McCance 200967 | Y | Y | N | N | N | N | N | N | N | N | N | N | N | N | |
| McGilton 200371 | Y | Y | N | N | Y | N | Y | Y | Y | Y | Y | Y | N | N | |
| McGilton 201165 | N | N | N | N | Y | N | Y | Y | Y | Y | N | N | Y | N | |
| Pipe 201068 | N | Y | Y | NA | NA | NA | N | Y | N | NA | NA | Y | NA | N | |
| Nurse support | Flarity 201373 | Y | N | N | Y | Y | Y | N | Y | N | Y | Y | Y | Y | N |
| Gauthier 201576 | Y | Y | Y | Y | Y | Y | Y | Y | N | Y | Y | Y | Y | N | |
| Horner 201477 | N | Y | N | Y | Y | Y | Y | N | N | Y | N | N | Y | N | |
| Palmer 201078 | N | N | N | N | N | N | Y | N | N | N | N | N | N | N | |
| Pålsson 199675 | N | Y | N | Y | Y | Y | Y | Y | N | Y | Y | N | Y | N | |
| Potter 201374 | Y | Y | Y | Y | Y | Y | Y | Y | Y | Y | N | N | Y | N | |
| Compliance (%) | 42 | 67 | 17 | 59 | 73 | 55 | 54 | 71 | 46 | 87 | 39 | 33 | 70 | 4 | |
| Average compliance by aspect of reporting (%) | 52 | 50 | |||||||||||||
| Intervention type | Study (first author and year) | Intervention | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Participants | Facilitators | Compliance (%) | Average compliance by intervention type (%) | ||||||||||||
| Group composition | Methods for group allocation | Continuity of participants’ group membership | Group size | Number of facilitators | Continuity of facilitators’ group assignment | Facilitators’ professional background | Facilitators’ personal characteristics | Facilitators’ training in intervention delivery | Facilitators’ training in group facilitation | Facilitators’ materials | Intended facilitation style | ||||
| Training | Ançel 200663 | Y | N | Y | Y | N | N | Y | N | Y | N | N | Y | 58 | 45 |
| Boscart 200960 | Y | NA | NA | NA | Y | Y | Y | N | Y | N | Y | N | 57 | ||
| Glembocki 201061 | Y | N | N | N | Y | N | Y | N | Y | N | N | N | 42 | ||
| La Monica 198764 | Y | N | Y | Y | N | N | N | Y | N | Y | N | Y | 42 | ||
| Langewitz 201079 | Y | N | N | N | Y | N | Y | N | N | N | Y | N | 50 | ||
| Puentes 199583 | Y | N | NA | NA | N | NA | N | N | N | N | Y | Y | 57 | ||
| Searcy 199082 | Y | Y | N | N | N | N | N | N | N | N | N | Y | 50 | ||
| Taylor 200980 | Y | NA | NA | N | NA | NA | NA | NA | NA | NA | NA | NA | 50 | ||
| Wasner 200562 | Y | NA | N | N | N | N | N | N | N | N | N | N | 40 | ||
| Yeakel 200381 | Y | N | N | N | N | N | N | N | N | N | N | N | 12 | ||
| Care model | Brown Wilson 201369 | N | N | N | N | N | N | N | N | N | N | N | Y | 38 | 33 |
| Chenoweth 201466 | Y | Y | N | N | Y | N | Y | N | Y | N | N | N | 27 | ||
| Finnema 200170 | Y | N | N | N | N | N | N | N | N | N | N | N | 35 | ||
| Ho 201672 | Y | Y | N | N | N | N | Y | N | N | N | N | N | 27 | ||
| McCance 200967 | N | N | N | N | N | N | N | N | N | N | N | N | 8 | ||
| McGilton 200371 | Y | Y | N | N | Y | Y | Y | N | Y | N | Y | N | 62 | ||
| McGilton 201165 | Y | N | N | N | Y | N | Y | N | N | N | N | N | 35 | ||
| Pipe 201068 | N | NA | N | N | NA | NA | NA | NA | NA | NA | NA | NA | 36 | ||
| Nurse support | Flarity 201373 | Y | N | N | N | Y | N | N | N | N | N | N | N | 42 | 46 |
| Gauthier 201576 | Y | Y | N | N | Y | Y | Y | N | N | N | N | N | 65 | ||
| Horner 201477 | Y | N | N | N | N | N | Y | N | N | N | N | N | 35 | ||
| Palmer 201078 | Y | N | N | Y | N | N | N | N | N | N | N | N | 12 | ||
| Pålsson 199675 | Y | Y | Y | Y | Y | N | Y | N | Y | N | N | N | 62 | ||
| Potter 201374 | Y | N | N | N | Y | N | Y | N | Y | N | N | N | 58 | ||
| Compliance (%) | 88 | 30 | 14 | 18 | 45 | 14 | 55 | 5 | 32 | 5 | 18 | 23 | |||
| Average compliance by aspect of reporting (%) | 37 | 29 | |||||||||||||
For intervention content, highest compliance was reported for session content (n = 20 of the 21 applicable studies, 87%) and lowest for participants’ materials (n = 8, 33%). Overall compliance for this aspect of intervention reporting was 50% of criteria. For reporting of participants, highest compliance was for description of group composition (n = 21, 88%) and lowest for continuity of participants’ group membership (n = 3, 14% of 21 applicable studies). Overall compliance for this aspect of intervention reporting was 37% of criteria. For reporting of facilitators, highest compliance was for reporting facilitators’ professional background (n = 12, 55% of 22 applicable studies) and lowest was for facilitators’ personal characteristics and training in-group facilitation (both n = 1, 5% of 22 applicable studies). Overall compliance for this aspect of intervention reporting was 25% of criteria. On average, individual study compliance with the criteria was 42%, ranging from 8% to 65%. Of intervention types, care model interventions tended to be less well described than other types (average of 33% compliance).
Evidence of effectiveness
This section presents findings on the quality of evidence of effectiveness of the interventions in the included studies. Overall, methodological quality was low. Most studies either did not randomise to the groups and/or did not demonstrate equivalence between groups, weakening confidence in the findings. Only two studies were assessed as high quality and four as medium. The remaining 18 studies were assessed as low quality. Most studies (n = 16) were uncontrolled before-and-after studies. Four studies were before-and-after studies with separate intervention and control groups.71,75,77,82 Four studies used a randomised controlled design. Three used a cluster randomised trial design, with clustering at unit or institutional level.64,66,70 A further study was controlled but only included a post-test measure.83
Of the 24 studies, only eight studies included more than 100 participants. The largest sample included 115 nurses and 656 patients in an evaluation of an empathy-training programme.64 The smallest sample included nine nurses in an evaluation of mindfulness based cognitive therapy for district nurses working with women with newly diagnosed breast cancer.78 The number of clusters in controlled studies ranged from two to 38. Of the studies with experimental or quasi-experimental design,64,66,70,71,75,82 just one66 reported powering of sample size, but was not explicit about which outcome measure was the primary one used for these calculations.
Table 4 provides an overview of results from the individual studies against the range of outcomes used. Eighteen different types of outcomes were reported. For simplicity and brevity results for multiple measures using the same instrument or different instruments measuring same phenomena have been grouped together and treated as one. Across all studies and all outcome types results for 67 outcomes are reported. Further information on effect sizes is displayed in Appendix 2 (see Tables 41–43 ).
| Study (first author and year) | Study quality | Outcome type | |||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Nurse | Quality of care | Patient | |||||||||||||||||
| Empathy | Compassion | Burnout | Stress | Mindfulness | Job satisfaction | Caring | Attitude | Other well-being | Quality of interactions | Quality of relationship | Patient centredness | Continuity | Quality of care | Satisfaction/experience | Behavioural (agitation) | Quality of life | Mood/well-being | ||
| Training intervention | |||||||||||||||||||
| La Monica 198764 | Medium | – | – | ▲ | |||||||||||||||
| Searcy 199082 | Medium | – | – | – | |||||||||||||||
| Ançel 200663 | Low | ▲ | |||||||||||||||||
| Boscart 200960 | Low | ▲ | |||||||||||||||||
| Glembocki 201061 | Low | ▲ | |||||||||||||||||
| Langewitz 201079 | Low | ▲ | |||||||||||||||||
| Puentes 199583 | Low | ▲ | ▲ | ||||||||||||||||
| Taylor 200980 | Low | ▲ | ▲ | ||||||||||||||||
| Wasner 200562 | Low | ▲ | ▲ | ▲ | ▲ | ▲ | |||||||||||||
| Yeakel 200381 | Low | ▲ | ▲ | ||||||||||||||||
| Care model intervention | |||||||||||||||||||
| aChenoweth 2014 (single)66 | High | – | ▲ | △ | △ | ||||||||||||||
| aChenoweth 2014 (combined)66 | High | ▲ | ▽ | △ | △ | ||||||||||||||
| Finnema 200170 | High | ▲ | |||||||||||||||||
| McGilton 200371 | Medium | ▲ | ▲ | ||||||||||||||||
| Brown Wilson 201369 | Low | △ | △ | △ | △ | ||||||||||||||
| Ho 201672 | Low | ▲ | |||||||||||||||||
| McCance 200967 | Low | ▲ | ▲ | ||||||||||||||||
| McGilton 201165 | Low | ▲ | ▲ | △ | ▽ | ||||||||||||||
| Pipe 201068 | Low | ▲ | |||||||||||||||||
| Nurse support intervention | |||||||||||||||||||
| bPålsson 199675 | Medium | – | – | – | |||||||||||||||
| Flarity 201373 | Low | ▲ | ▲ | ▲ | |||||||||||||||
| Gauthier 201576 | Low | △ | – | ▲ | △ | △ | |||||||||||||
| Horner 201477 | Low | – | – | – | – | △ | |||||||||||||
| Palmer 201078 | Low | △ | △ | △ | |||||||||||||||
| Potter 201374 | Low | △ | △ | ▲ | – | ||||||||||||||
Studies of similar intervention types tended to use similar outcome types. Nurse support intervention studies primarily measured nurse-based outcomes. No nurse support studies used quality-of-care outcomes, and just one study used patient-based outcomes. In contrast, care model intervention studies primarily used outcomes related to quality-of-care and patient-based outcomes, but use of nurse outcomes was less common. Training intervention studies used the widest range of outcome types. Although the majority used nurse-based outcomes a small number drew on quality-of-care and patient outcomes.
Nineteen studies (79%) reported a significant positive difference in one or more outcomes (i.e. a beneficial effect of the compassionate care intervention). Only five (21%) of the 24 studies reported no significant difference in any of the outcome types measured. Of the 67 outcome types assessed across all studies, 32 (48%) showed significant positive effects for the intervention, with a further 18 (27%) showing positive but non-significant results. There were no significant negative differences and only three non-significant negative results. Patient outcomes were less likely to show significant differences, with only 5 out of 17 (29%) showing statistically significant differences.
Studies of low methodological quality were more likely to report outcomes in favour of the intervention, with low methodological quality studies reporting a mean of 92% of outcomes in favour of the intervention (significant + non-significant positives), whereas higher-quality (medium, high) studies report 55% of outcomes in favour of the intervention. Although an average of 76% of outcomes reported in studies of training interventions showed a statistically significant benefit, only 21% of outcomes for nurse support interventions were significant. Crucially, no intervention has been evaluated more than once.
Effects on patient-based outcomes
Six care model intervention studies reported patient-based outcomes. Of these, three showed statistically significant effects on a patient-based outcome. Of these, one was rated as high quality. In their CRT with 38 nursing homes, Chenoweth et al. 66 reported that the person-centred care intervention had a significant positive effect on reducing patient agitation, but the combined intervention (person-centred care plus person-centred environment) reported in the same study showed a non-significant effect of increasing patient agitation. This study fared poorly in terms of reporting of intervention description, meeting only 27% of criteria.
Three training intervention studies reported patient-based outcomes and, of these, two showed a significant positive effect. One medium-quality study reported significant positive effects on patient anxiety64 and one low-quality study reported a non-significant positive difference in patient satisfaction.81 A low-quality nurse support intervention study reported a non-significant improvement in patient satisfaction.77
Effects on quality-of-care outcomes
Four60,79,81,82 training intervention and five65–67,70,71 care model intervention studies (one of which66 reported two interventions) examined effects on quality-of-care outcomes. Of these, eight60,65–67,70,71,79,81 reported a statistically significant improvement in one or more outcomes. The combined person-centred care model intervention reported by Chenoweth et al. 66 was associated with a significant improvement in quality of interactions, but, although this finding is from a high-quality study, conclusions are tempered by the lack of intervention description noted above. In a CRT rated as high quality,70 the authors reported a significant change in one dimension of quality of care following implementation of emotion-oriented care in nursing home settings, but the intervention description met only 35% of the criteria. A medium-quality evaluation of a relationship-enhancing programme of care in nursing homes71 reported significant improvements in relational care, care providers’ relational behaviour and continuity of care. A medium-quality evaluation of empathy training for hospital nurses82 found no difference in interpersonal support. Other improvements in quality-of-care outcomes were reported in a range of low-quality studies.60,65,67,79,81
Effects on nurse-based outcomes
Seven61–64,80,83 training, six73–78 nurse support and three65,67,69 care model intervention studies examined effects on nurse-based outcomes and, of these, 1061–63,65,73,74,76,80,83,98 reported a significant improvement associated with the intervention. All of these 10 studies were rated as low quality. Three medium-quality studies64,75,82 investigated nurse-based outcomes but none showed significant differences. No high-quality studies reported on nurse-based outcomes.
Feasibility findings
Findings from our analysis of feasibility issues and limitations documented in the reports of high and medium-quality studies (n = 6) are summarised here. The included studies were either experimental in the form of a cluster RCT or quasi-experimental with before-and-after measurements of intervention and control groups, but no randomisation to groups. Papers varied in the feasibility findings they reported and in the limitations identified but all were able to identify where improvements could be made in future research of this kind.
Two studies64,82 in this subset were conducted in a hospital setting, both single-site hospital US settings. La Monica et al. 64 conducted a cluster RCT in four cancer units (two medical and two surgical) to determine the impact of a nurse empathy training programme on patient outcomes (anxiety, depression, hostility, satisfaction with care) and nurse outcomes (nurse empathy). The study was not focused on older people, and patients too ill or confused to complete the questionnaires were excluded. Baseline data were gathered over a 4-week period, followed by a 4-week empathy training delivery period, followed by a 4-week follow-up assessment period. La Monica et al. 64 reported that patients were not admitted for long enough to take part in both assessment periods of the study. Patient participation rate was reported to be 73% with reasons for non-participation including not feeling well enough, having a conflict with a treatment or personal schedule, being reluctant to rate the nurse, and generally not being interested. The research team also identified a number of issues with the outcome measures involving rating nurse empathy and satisfaction with care. The team noted that patients consistently rated nurses’ empathy higher than nurses rated their own empathy, and speculated that it may be psychologically threatening for patients to rate nurses and nursing care, particularly while still in need of care.64 In addition, at baseline and follow-up in both experimental groups, nurse- and patient-rated empathy scores were close to maximum, implying a ceiling effect.
Searcy82 conducted a before-and-after study with an intervention group (one ward) and a control group (one ward) to determine the impact of an empathy education programme for hospital nursing staff on patient satisfaction with care, including perceptions of interpersonal support. Baseline data were gathered over a 6-month period, followed by a 2-week training period (consisting of two 1-hour sessions), followed by a 6-month follow-up assessment period. The patient survey was mailed 1 week after discharge to all patients discharged from the two participating units. All patients were adults and no exclusion criteria, such as dementia, were reported.82 Survey response rate was reported as approximately 25%. Searcy82 reported that baseline ratings were high, implying a ceiling effect to the chosen measure, and also found that older patients rated higher satisfaction than younger patients.
Three studies66,70,71 in this subset evaluated the impact of care model interventions in care home settings in Australia, the Netherlands and Canada. All were focused on interventions targeted at improving care for older residents, with two particularly focused on dementia care.66,70 One study involved different sets of residents in questionnaires and in observations.71 The questionnaire subsample included people who were medically stable, able to understand English and cognitively able to provide consent and respond to questions.71 Just five out of the original subsample (n = 50) did not participate in the follow-up assessment period (10 months after intervention start), suggesting a relatively healthy subsample.71 The other subsample included people who required moderate to high levels of assistance with personal hygienic care. Residents recruited to this subsample were included in observations of care carried out at baseline and follow-up.71 However, whether or not the subsamples in the two time periods were independent of each other was not reported, so attrition rates cannot be established.71 The two other care-home studies reported high attrition rates.66,70 Chenoweth et al.66 retained 36 out of 38 nursing home clusters, and 305 out of 601 nursing home residents remained in the study over 8 months. Most (i.e. 216 out of 296) attrition was attributable to residents dying. Finnema et al. 70 reported that 132 eligible nursing home family members completed a second questionnaire 8 months following completion of a first questionnaire by 241 family members. Forty-four people were not included in the second round because their relative had died.70
Two of the care home studies used observations of care but neither study reported feasibility issues with the observations.66,71 Feasibility issues were, however, more apparent when self-report and/or proxy questionnaires were used, especially when cognitive impairment was more severe. As noted above, McGilton et al. 71 limited questionnaire distribution to relatively healthy residents and did not report any feasibility issues. Chenoweth et al. 66 used a range of instruments to measure resident well-being, and concluded that a lack of association between the intervention and quality of life, and depression, may be attributable to the difficulty in measuring these constructs through self-report and/or proxy in this population. Finnema et al. 70 invited family members to complete a newly developed questionnaire but not all items were completed, suggesting a need for further development and piloting.
The final study in this subset investigated the impact of clinical supervision (a form of nurse support) on burnout, empathy and sense of coherence in district nurses in Sweden.75 Improving the care of older people and/or people living with a cognitive impairment was not a stated focus of the study. Measures used were all deployed through a written questionnaire completed by staff. In total, 33 out of 39 district nurses remained in the study over 28 months. The authors reflected that the high baseline nurse-reported empathy and sense of coherence, and low burnout, may indicate a ceiling effect to the selected outcome measures.75 Empathy was measured using the Empathy Construct Rating Scale,85 also used by La Monica et al. 64 in the hospital study reported above. The finding in both studies that there may be a ceiling effect to measuring empathy in this way suggests a limited capacity to measure improvements in empathy over time.
Of the six studies,64,66,70,71,75,82 three reported it was not possible to mask nursing staff to experimental allocation.64,66,75 Two studies reported concealing experimental allocation from the research team.66,71 Two studies deployed measures to control contamination of control conditions66,71 and one study identified pathways through which contamination may have occurred.82
Only Chenoweth et al. 66 formally measured fidelity to the intervention, and they reported variation in implementation between clusters, suggesting that the time frame of the study limited implementation (post-intervention assessment was at 6 and 14 months following planned start of intervention implementation). In one of the two hospital studies, La Monica et al. 64 reported that all eligible nurses attended the training but did not report if the training was delivered as planned. They speculated that the follow-up assessment period may have been timed too early (4-week assessment period following 4-week training period), and did not allow the results of the training to embed into practice. In the other hospital study, attendance at training or any aspect of fidelity went unreported.82 Searcy82 reported that staff attending the training fed back that 2 hours of training was insufficient. Pålsson et al. 75 and McGilton et al. 71 speculated that the intervention period may have been too short to effect change (10 months and 7 months, respectively). Pålsson et al. 75 did not report if clinical supervision was delivered as planned but did report number and length of sessions, and attendance at sessions (74%). McGilton et al. 71 reported that the protocol was adhered to in intervention delivery and 70% of care providers and 50% of supervisors attended the whole programme. Finnema et al. 70 did not report assessing if the staff training and supervision was delivered as planned, but did use findings from a staff self-assessment of nursing skills to infer that the observed increase in emotion-oriented skills meant that emotion-oriented care was applied more often than before. Five studies64,66,71,75,82 speculated that the impact of the intervention may have been affected by other factors, such as leadership support of the intervention and by other initiatives taking place at the same time as the intervention.
In summary, previous experimental and quasi-experimental work in this field has raised a number of potential feasibility issues that can inform future study design and implementation. Hospital settings presented a number of validity issues including a possible reluctance of patients to rate nurses negatively while still in receipt of care and/or a ceiling effect to empathy rating instruments deployed to date. One of the hospital studies did not include patients who were ‘confused’64 and the other relied on a post-discharge written survey,65 and so neither study illuminated how best to include people with a cognitive impairment in care evaluations. Response rates were markedly higher for the study surveying people while still an inpatient than for the study surveying people post-discharge. In care-home studies more inclusive of people with cognitive impairment, feasibility issues were identified with written questionnaires, and methods involving observations were not associated with feasibility issues. Most studies did not pay attention to fidelity of actual delivery of the planned intervention. Some studies reported that the intervention may have been delivered in too small a dose, and/or that follow-up assessment may have occurred too rapidly. Five64,66,71,75,82 studies speculated on the importance of a variety of contextual factors in affecting the implementation process but data were not gathered to enable in-depth exploration of this potential influence.
Discussion
This chapter aimed to provide an overview of the evidence base on the evaluation of interventions for compassionate nursing care. It began by reporting on the qualitative work in this field, work that has focused on relationship-centred approaches to improving care, interventions that we would classify as a combination of care model and staff support intervention types. These studies often provide detailed information about the intervention and its inherent mechanisms for change, and often include analyses of the factors enabling and inhibiting sustainable change. However, they do not examine in depth the process of implementation itself and commonly pay only superficial attention to the influence of actors, networks and resources on the impact and sustainability of these interventions. In addition, as would be expected with a qualitative approach, there is only weak objective evidence of effectiveness of the interventions deployed in these studies in relation to impact on patient outcomes.
As identified in our systematic review, there is a wide range of intervention studies in which compassion has been addressed in a variety of ways, including through staff training, staff support or introducing a new care model. However, the overall strength of work in the field limits the conclusions that can be drawn to inform health-care policy and practice. No study reported in the systematic review included sufficient detail of its intervention to enable replication and further evaluation. This state of play limits the capacity of nurses and others to include effective strategies in their own practice, but also limits the construction of a coherent evidence base to guide managers and practitioners in improving services.86–88
In relation to the nature and strength of the existing evidence base, overall quality of the evidence was low and it appears that the few higher-quality studies are less likely to report positive results. No intervention was tested more than once and the majority of studies used before-and-after designs that are intrinsically weak. Patient-based outcomes were not routinely included, especially in relation to the evaluation of training interventions and nurse support interventions.
Our analysis did, however, highlight the feasibility issues associated with this field of research that have formed the backdrop to the design and implementation of the study reported in subsequent chapters.
Conclusions
Although there have been many published studies that appear to offer potential solutions to deficits in compassionate care, this is a body of literature that does not offer a definitive way forward for policy and practice in this important area of health care. This is especially challenging in a context in which the need for more compassion in health care is professed from national government to frontline practitioners. Greater conceptual clarity, better designed and reported interventions, including a focus on implementation processes, and evaluations using stronger research designs are urgently required.
To date, no evaluations of initiatives of this kind have enabled a robust assessment of the effectiveness of interventions on patient care, linked with the use of theory-based interventions reported with sufficient clarity to support optimal implementation, impact and sustainability. The study reported in the following chapters is the first stage in designing a rigorous mixed-methods evaluation incorporating experimental design to understand what works best in improving care.