A systematic review of cancer caregiver interventions: Appraising the potential for implementation of evidence into practice

Abstract Objective nformal caregivers provide substantial support for people living with cancer. Previous systematic reviews report on the efficacy of cancer caregiver interventions but not their potential to be implemented. The aim of this systematic review was to explore the potential for cancer caregiver interventions to be implemented into practice. Methods We searched three electronic databases to identify cancer caregiver interventions on 5 January 2018. We operationalised six implementation outcomes (acceptability, adoption, appropriateness, feasibility, fidelity, and costs) into a tool to guide data extraction. Results The search yielded 33 papers (27 papers from electronic databases and six papers from other sources) reporting on 26 studies that met review criteria. Fewer than half the studies (46%) contained evidence about the acceptability of interventions from caregivers' perspectives; only two studies (8%) included interventions developed with input from caregivers. Two studies (8%) addressed potential adoption of interventions, and no studies discussed intentions, agreement, or action to implement interventions into practice. All studies reported on intervention appropriateness by providing a rationale for the interventions. For feasibility, on average less than one‐third of caregivers who were eligible to be involved consented to participate. On fidelity, whether interventions were conducted as intended was reported in 62% of studies. Cost data were reported in terms of intervention delivery, requiring a median time commitment of staff of 180 minutes to be delivered. Conclusions Caregiver intervention studies lack components of study design and reporting that could bridge the gap between research and practice. There is enormous potential for improvements in cancer caregiver intervention study design to plan for future implementation.

third of caregivers who were eligible to be involved consented to participate. On fidelity, whether interventions were conducted as intended was reported in 62% of studies. Cost data were reported in terms of intervention delivery, requiring a median time commitment of staff of 180 minutes to be delivered.
Conclusions: Caregiver intervention studies lack components of study design and reporting that could bridge the gap between research and practice. There is enormous potential for improvements in cancer caregiver intervention study design to plan for future implementation. KEYWORDS cancer, caregiver, carer, dissemination, framework, implementation, intervention, oncology, outcomes 1 | BACKGROUND Informal caregivers provide substantial practical and emotional support for people living with cancer, and in doing so, many receive minimal support themselves. Previous studies have outlined the negative impacts associated with being a caregiver, including depression, 1 burden, 2 social isolation, 3 loss of self-identity, 4 sleep deprivation, 5 financial burden, 6 and significant changes to their lives. 2 The role they take on in caring for the person with cancer is extensive, demanding, and often without training or resources. 7 Many research papers focus on the development and evaluation of interventions aimed at improving the experience of caregivers, including several reviews of caregiver interventions. [8][9][10][11][12][13][14][15][16] Of these, Northouse et al presented a meta-analysis of the efficacy of caregiver intervention studies categorising interventions as psychoeducational, skills training, and therapeutic counselling. They concluded that interventions had beneficial small to medium effects on burden, coping, self-efficacy, and quality of life. 15 More recently, Ferrell and Wittenberg 11 performed an updated review, identifying an increase in trials and the growing need to translate evidence into practice. Similarly, a review article drawing upon the literature and stakeholder perspectives from an in person meeting attended by more than 75 invited researchers, clinicians, advocates, and representatives recommended the implementation of successful interventions. 17 Previous reviews have focused on the efficacy of caregiver interventions but not their potential to be implemented into practice. Implementation science frameworks contribute to reducing the evidence to practice gap 18 by applying a theory to identify factors that may evaluate implementation success. 19 Proctor et al 26 developed a framework of implementation outcomes, defined as the "effects of deliberate and purposive actions to implement new treatments, practices and services" (p65). This framework has eight implementation outcomes: acceptability, adoption, appropriateness, (implementation) costs, feasibility, fidelity, penetration, and sustainability. Of these, the first six are relevant to the earlier stages of implementation, whereas penetration is relevant mid-implementation and sustainability applies to longer-term implementation. This framework has been applied to inform a variety of research topics including standardised multidisciplinary team meeting templates, 20 shared decision-making, 21 cervical cancer prevention, 22 and uptake of human papillomavirus (HPV) vaccines. 23 Caregiver interventions show promise for potential implementation into practice. 11,15 However, we know little about whether interventions are designed and reported in a way that supports implementation. 24,25 There is a need to explore the implementation potential of existing cancer caregiver intervention studies to guide the development of future caregiver research. The aim of this review is to describe and appraise the cancer caregiving literature to explore the potential for caregiver interventions to be implemented into practice. To identify studies for inclusion in this review, three electronic databases were searched, Cumulative Index to Nursing and Allied Health Literature (CINAHL) Complete, MEDLINE Complete, and PsycINFO Complete, representing the fields of nursing, medicine, and psychology. The terms used in the search were caregivers (as a subject heading) and cancer. No limitations were applied for language or publication date. The search was performed on 5 January 2018.
The reference lists of papers meeting the inclusion criteria were scanned for additional papers for possible inclusion in the review.
We also searched reference lists of eight recent systematic reviews on caregivers of people with cancer. 8,10-16

| Selection criteria
Studies were included in this review if they met the following criteria: (i) Participants were informal (unpaid) adult (18+ y) caregivers who had an active role in the provision of care for an adult with cancer; (ii) interventions were programmes, supports, sessions, or resources provided to, and directed towards supporting, caregivers to improve their own functioning or assist them in providing support for the patient (eg, programmes focusing on upskilling caregivers); (iii) study designs included at least two conditions (eg, randomised controlled trials and quasi-experimental studies), one of which must have been a control condition (eg, active controls, waiting list controls, and treatment as usual [TAU] controls); and (iv) study outcomes focused on the caregiver. Pilot and feasibility studies were eligible for inclusion.
Studies were excluded if they met the following criteria: (i) They focused on spouses or other family without establishing that they had caregiving roles; (ii) 25% or more of patients had conditions other than cancer; (iii) the interventions focused on both patients and caregivers (interventions where minimal content was delivered to patients were eligible for inclusion, however); and (iv) the study design included two or more experimental conditions without a control condition. These exclusion criteria were established to ensure a focus on cancer caregiver interventions. Review papers were excluded from selection.

| Study selection
Two authors (A.U. and C.J.G.) performed the eligibility assessment independently in an unbiased standardised manner. C.J.G. undertook an initial screening of papers, on the basis of title and then abstract.
Both A.U. and C.J.G. then assessed papers on the basis of a full-text review. Disagreements between reviewers were resolved through consensus. Deferring to a third reviewer was not necessary.

| Data extraction
From each study meeting the selection criteria, data were extracted on study characteristics and the implementation outcomes of the interventions. Data extracted on study characteristics included (i) country of origin, (ii) aim, (iii) caregiver demographic characteristics (sample size, sex, and age), (iv) patient diagnosis, (v) relationship between caregiver and patient, (vi) study design, (vii) intervention details (format, content, setting, and who delivered the intervention), (viii) theory underpinning intervention (explicit statement required), (ix) evidence of prior pilot testing of intervention, (x) comparison condition, (xi) outcome measures, (xii) key findings, and (xiii) whether the conclusions were supported.
Operationalisation of the Proctor et al 26 taxonomy of implementation outcomes (acceptability, adoption, appropriateness, feasibility, fidelity, and cost) guided the extraction of data on intervention implementation outcomes (see Table 1). This framework was selected because of the alignment between the implementation outcomes and the aims of the review. We selected this framework in preference to others in the implementation science discipline as it draws on a conceptual framework that addresses a range of outcomes. The outcomes are defined in a comprehensive manner that facilitates measurement for the purposes of a systematic review.
The framework was operationalised into a data collection tool by  Complete, n = 2120), of which 2682 were duplicates (see Figure 1).
Of the remaining 4501 entries, 103 were retained following screening the titles of papers. After reviewing the abstracts, 61 papers did not meet the selection criteria and were excluded. The full texts of the remaining 42 papers were reviewed, of which 27 papers were finally included in the review.
An additional six papers meeting the selection criteria came from other sources. Inspection of the reference lists of previous systematic reviews in the area 8,11-16 enabled identification of a further five papers that met the selection criteria. One further paper was identified from a preliminary search that abandoned because the search was too narrow. Being relevant, this paper was included in the review. Checking the reference lists of the included papers resulted in no further papers being identified for inclusion. In total, 33 papers were included in the review, representing 26 studies (Table 2).

| Study characteristics
An overview of study characteristics is presented in Table 2. Intervention characteristics and implementation outcomes tables are attached as supporting information (Table S1).

| Country of origin
Almost half of the studies (n = 12, 46%) were conducted in the United States. 29

| Implementation outcomes
The implementation outcomes across studies are presented in Table   S2 (acceptability, adoption, and appropriateness), Table S3 (feasibility), and

| Acceptability
In almost half the studies (n = 12, 46%), there was no evidence to indicate the acceptability of interventions from caregivers' perspectives.
For 11 studies (42%), acceptability data were reported, which were collected via surveys (n = 5), 30,36,52,53,59 focus groups (n = 2), 37,38,58 interviews (n = 2), 46,54 and engagement with the intervention or debriefing (n = 2). 31,43,55 In one further study, acceptability data were collected on certain aspects of an intervention (ie, feedback was gathered on some aspects of the intervention but not others). 35 For the two remaining studies, acceptability data may have been collected, but insufficient information was provided to enable firm judgements to be made. 45,50 Most studies (n = 21, 81%) did not report on the acceptability of interventions from the perspectives of other stakeholders (ie, stakeholders other than caregivers). These data were available for three studies and were collected via focus groups (n = 2) 37,38,58 and surveys (n = 1). 30 Stakeholder acceptability data were collected on some components of an intervention in one further study 50 and may have been obtained in another study. 45 Caregivers appeared to have limited input into intervention development. In most studies (n = 17, 65%), there was no evidence of caregiver involvement in the development of interventions. Caregivers were directly involved in the development of the intervention in one study (4%), 58 and in eight studies (31%), caregivers were involved in separate studies, such as focus groups, that informed the interventions. 39

| Adoption
No studies reported on intentions, agreement, or action to implement interventions into practice. However, two studies reported issues about the potential adoption of the interventions. In one, researchers reported that health care providers held reservations about possible implementation, 28 and in the other paper, researchers noted that the intervention may not be sustainable in the longer term because of the costs involved in delivery. 46

| Appropriateness
In all studies, interventions were considered a good fit, with solid arguments presented as to why the interventions were appropriate for the cancer caregivers.
Few interventions were targeted towards specific population groups who may have high support needs or may benefit from interventions, such as caregivers experiencing high levels of distress or those from minority cultural backgrounds (n = 2, 8%). 31,59 One intervention targeted caregivers with high distress levels, 59 and another one focused on caregivers who had experienced sleep difficulties for at least 1 month. 31

| Feasibility
Most caregivers screened were eligible for inclusion in the studies (median = 84%, IQR = 52% to 90%, data available from 65% of studies). On average, less than one-third of eligible caregivers consented to participate (median = 28%, IQR = 17% to 55%, from 69% of studies). Most caregivers who consented to be involved commenced the interventions.

| Fidelity
In the majority of studies, interventions appeared to be conducted as intended (n = 19, 62%). In one study, researchers reported that caregivers did not engage with one aspect of the intervention (an online forum). 45 For the remaining studies, no information about intervention fidelity was reported (n = 9, 35%).
On average, caregivers completed all aspects of the interventions, such as attending all sessions provided (median = 100%, IQR = 84% to 100%, from 54% of studies).
No changes to interventions during the studies were reported, and no changes to the dose, delivery, or strategies during the studies were reported.

| Costs
The time commitment data were available for n = 19 (62%) of studies.
For most studies (n = 22, 85%), aside from the occupations (and, in some cases, experience) of those who delivered the interventions, no information was provided on the training and expertise required to deliver the interventions. In two studies (8%), staff training was provided, 33,58 and in a further two studies (8%), the training and experience required was unspecified. 27,52

| DISCUSSION
With recent calls for a need to focus on implementation of interventions, 17,60,61 this review aimed to explore the implementation potential of cancer caregiver intervention studies. Although the reviewed studies focused on efficacy, there is a need to design, conduct, and report research that can be implemented into practice. 24 The main finding from this review was that studies were not designed or reported in a way to maximise the potential for interventions to be successfully implemented. We also gained insights about the challenges of operationalising implementation outcomes from an established framework. practice. There appears to be two key issues. Firstly, studies were not designed in ways that would maximise their potential to be successfully implemented. Secondly, in other instances, there is limited information reported relevant to implementation. Restrictions in reporting research in terms of journal requirements and required word counts may limit the opportunity to report evaluation data that includes outcomes of relevance to implementation.
There are other key findings from this review to highlight. The first is that consumer input into intervention development was notably low (acceptability outcome). In performing this review, we differentiated between studies that had active engagement with consumers as part of the project design, those studies that had developed interventions that were informed by the research team identifying a need, and those that had no consumer involvement. Consumer involvement into interventions is considered best practice, 62 and it was surprising to find a paucity of studies utilising caregiver input. Future research should engage caregivers as team members and promote active roles in the development and refinement of caregiver interventions.
A further finding was the tendency for studies to recruit broadly rather than targeting groups more in need of support. This is in the context of consent rates that, while varied, had a median of less than a third of those approached across the studies, meaning that while many caregivers were eligible, this did not translate to enrolment. A recent systematic review and meta-analysis exploring the efficacy of psychological interventions on anxiety in cancer patients found that low psychological distress at baseline was a key reason for low effectiveness, with authors advocating for screening and assessment of anxiety as an inclusion criterion before enrolment in psychological interventions. 63 Caregivers not experiencing a problem may have low motivation to spend time in an intervention study they see as not relevant to their situation. Others have noted the need to increased research for vulnerable caregiving populations and risk stratification to target those most in need of support. 17 Targeting groups in need of support is an important avenue for future caregiver intervention research.
This review also found that while most caregivers screened were eligible (feasibility outcome), this frequently was not well reported.
Future studies should clearly report about the participants who were assessed for eligibility in accordance with CONSORT criteria and flow charts. 64 There was also limited evidence available about intentions, agreements, or actions to implement interventions into practice (adoption outcome). There could be various reasons for this including that adoption is regarded as being outside the scope of conduct and reporting of studies, with adoption frequently reported at 6, 12, or 18 months after initial implementation. The lack of funding to test implementation processes has been acknowledged. 65 Information about adoption agreements with service providers or potential would be a useful addition to papers reporting trials of interventions, even when the focus is on efficacy.
This review has operationalised Proctor's implementation outcomes framework. While there are other potential frameworks, 19,66 this framework was selected as the six domains resonated with the scope of the review. In practice, the operationalisation and data extraction allowed for key information to be assessed and findings support this framework as being appropriate for this review. Frameworks can be used to plan and design studies to strengthen the potential for implementation, 24 and there may be potential to build on these results and use the Proctor framework in this context. This could strengthen the implementation potential of new studies. A recent literature review has outlined instruments to assess implementation outcomes, and addition of these measures could be considered in future trials. 67 We did not include two implementation outcomes: penetration (the integration of an intervention within its setting) and sustainability (extent to which an intervention is maintained) given these are longer-term outcomes. 26 This review focused on cancer caregiver interventions, and issues of implementation potential may not be unique to this content.

| Study limitations
This review has limitations to consider. Firstly, this review focused on implementation potential utilising a specific framework applied to the reporting of the original trial, but this may not mean that interventions were not implemented into practice. Studies may show limited implementation potential according this extracted data but may have been successfully implemented into practice. It appears that there are few published reports around implementation of cancer caregiver interventions; however, it was beyond the scope of our review to ascertain this. A further limitation is that it is important to acknowledge the diversity of cancer caregiving interventions in the literature. We screened abstracts broadly, and criteria focused on specific cancer caregiver interventions; for example, we omitted interventions directed at caregiver and patient dyads. This criterion was applied to ensure these interventions were specifically for caregivers. This review was conducted in the context of numerous caregiving reviews focusing on efficacy, and our aim was to complement these through exploring implementation potential.

| Clinical and research implications
There are numerous implications for future research. Exploring any relationship between implementation outcomes and efficacy of interventions was outside the scope of this review, but this could be relevant for future research to inform optimal delivery on implementation outcomes. Exploring the potential of the implementation outcomes framework to plan and design studies may lead to stronger potential for implementation for cancer interventions. Additionally, given the findings of this review, the development and conduct of high-quality cancer caregiver interventions that are able to be implemented into practice is essential.

| CONCLUSIONS
Interventions must be cost-effective and accessible; planning for implementation is important. 24 Our findings suggest that the reporting of cancer caregiver interventions demonstrates limited capacity to be translated into practice. This is of significant concern as it may indicate