Understanding the potential factors affecting carers' mental health during end-of-life home care: a meta synthesis of the research literature

Search and selection strategy

Studies for the project were identified through an electronic search of the literature from 1 January 2009 to 24 November 2019 in the following databases:

• MEDLINE (National Library of Medicine, Bethesda, MD, USA)

• Cumulative Index to Nursing and Allied Health Literature (CINAHL) Plus (EBSCO Information Services, Ipswich, MA, USA)

• PsycINFO^® (American Psychological Association, Washington, DC, USA)

• Social Sciences Citation Index (Clarivate Plc, London, UK)

• Excerpta Medica (EMBASE) (Elsevier, Amsterdam, the Netherlands)

• Cochrane Central Register of Controlled Trials (CENTRAL) (John Wiley & Sons, Inc., Hoboken, NJ, USA)

• Database of Abstracts of Reviews of Effects (DARE) (University of York Centre for Reviews and Dissemination, York, UK).

Dissertations and other grey literature were not searched owing to time constraints and the large number of research papers that the original search strategy returned. Policy literature and professional literature were also excluded as the focus was on peer-reviewed, empirical research published in academic journals. See Appendix 1 for the full search strategy.

Box 2 describes the shared inclusion and exclusion criteria for all qualitative, observational and intervention studies in the project.

In addition, studies included for the intervention synthesis had to (1) have carer mental health as an intervention outcome; (2) report that the intervention led to a significant change in a factor identified in the observational or qualitative syntheses as associated with carer mental health (see Box 118,19); (3) be a randomised controlled trial (RCT), non-randomised trial, controlled before–after study or interrupted time series; and (4) have a clear comparator in the form of usual care, enhanced usual care, ‘no intervention’ or waiting list controls.

For further information on project searches and inclusion/exclusion criteria see Bayliss et al. 18 and Shield et al. 19

Ten per cent of both titles/abstracts and full texts were screened independently by two reviewers. Over 90% agreement was established in each case, indicating that no further modifications to the inclusion and exclusion criteria were required. Subsequent papers were screened by one reviewer both at title/abstract and full-text screening stages.

Data extraction and quality appraisal process

Analysis

Individual factors were synthesised thematically into subthemes using Box scores21 (see Appendices 2 and 3). The association with mental health in each study for each subtheme was categorised as better/worse/no change in mental health. Here, direct statistical assumptions for the direction of effect on mental health were not assessed, thereby indicating only the general direction of the association. Subthemes were then mapped onto the overarching themes identified from our reviews of qualitative research and observational quantitative research (see Box 1 and Table 2). The themes had been developed with input from the carer RAP. The format of table presentation (in Appendices 2 and 3) was also informed by the carer RAP and was seen by them as a useful way of presenting the evidence.

A meta-analysis was not feasible for this review for the following reasons. First, the review aimed to assess whether or not a factor affected by the intervention had an impact on carer mental health, rather than assessing if the intervention per se had an impact on mental health, and there was limited information available from studies regarding relationships between factors and outcomes for such a meta-analysis. Second, there were very few instances in which two studies considered both the same factor and the same mental health outcome, such as depression (see Appendix 3). Finally, there was a large variation in the way factors were measured, making a meta-analysis less meaningful.

Critical Appraisal Skills Programme Quality Appraisal

When there was more than one study for a theme, the mean CASP QA scores for each theme was quite similar: patient condition had a mean score of 6.8 [standard deviation (SD) 2.5] out of 11, impact of caring responsibilities had a mean score of 6.0 (SD 3.0) and carer internal processes had a mean score of 6.9 (SD 2.5). The relationships and support themes had only one study each, but with relatively high mean scores of 9 and 8, respectively. These numbers hide considerable variety in the quality of studies, ranging from 3 to 11.

If we consider where studies did not meet CASP20 criteria or where there was insufficient information given to assess this, blinding was the main area in which criteria were not met. It would be difficult to blind recipients to the fact that they received the intervention, given the types of intervention tested. However, only 4 of the 13 studies reported any form of single blinding of assessors. Regarding the criterion of size of treatment effect, only Badr et al. 22 showed large effects [standardised mean difference (SMD) 0.8], one study showed medium effects (SMD 0.5)34,35 and three showed small effects (SMD 0.2),23,31,33 with the remaining studies showing no significant effects. Only six studies had the requisite precision of estimate of the treatment effect, in terms of providing both p-values and confidence intervals, or enough information for these to be reliably calculated. Criteria met by a majority of studies (eight each) were as follows: it was clear that all participants who entered the trial were accounted for at its conclusion; that groups entering the trial were similar at baseline; all important outcomes were considered, for example in terms of completeness of outcome data; and that participants had been randomly allocated to trial arms. For randomisation, three further studies indicated that this had taken place, but that there was insufficient reporting on how this was done to satisfy the CASP criteria. For eight studies it was clear that the benefits outweighed harms and costs, whereas for the remainder it was simply difficult to tell based on the information available, rather than any indication of harm. All studies but one were able to treat groups equally apart from delivering the intervention.

The lack of blinding of participants, and to some extent the researchers, was therefore a main weakness of this literature. As all the outcomes and many of the factors were based on self-report measures, participants may have been biased towards giving responses that they thought met with study expectations, and whatever effects were observed could have been due to this. However, it would be difficult to blind participants to the fact that they are receiving an intervention and to obtain measures through other means than self-report in these cases. Therefore, we could not fully remove this bias, but had to take it into account in the interpretation of results. However, few studies showed significant impact on mental health in any case, and only two showed medium to large effects. Just over half of studies reported p-values and confidence intervals. This may raise some concerns about the quality of interventions and the reporting in some of this literature, and therefore this literature’s ability to form a robust basis for progress in the field. We were probably less likely to see bias due to systematic differences in characteristics of the intervention and control groups (given randomisation, similarity at baseline and data completeness) or in the way they were treated, apart from the intervention itself. Regardless, the main issue for our synthesis remained the lack of investigation of factors as mediators or moderators.

Patient condition

A dyadic intervention (i.e. involving both patient and carer) that improved the patients’ and carers’ social quality of life (QoL) (relating to support from family and friends) also improved the carer’s emotional QoL (but not the patient’s). 31 However, it is difficult to ascertain if carers’ improved mental QoL here is due to patient improvement or their own improved sense of social support. The improvement in the patients’ overall QoL in another study was not associated with changes in carers’ distress or their QoL. 30

An intervention that reduced patient depression also improved two aspects of carers’ mental health (depression and anxiety). 22 In contrast, a study that reduced patients’ distress due to breathlessness had no impact on carer depression, anxiety or distress,24 and carers’ perception that the patients’ pain was better controlled was not associated with changes in carer anxiety and QoL. 32

Impact of caring responsibilities

One study found that reduction in carer burden was related to two aspects of carers’ mental health (depression and anxiety). 22 In contrast, Sulmasy et al. 33 found that carers with increased carer burden also had lower carer distress. The focus of the latter intervention was, however, to support carers as ‘surrogate’ decision-makers, and it may be that this places greater burden on carers, although such involvement may leave them less distressed. Ngyen et al. 30 found that reducing carers’ perceptions that care responsibilities were overly demanding was not related to carer mental health (as shown through QoL and distress levels).

Relationships

An intervention that increased levels of relatedness (i.e. quality of the carer–patient relationship) also improved two aspects of carers’ mental health (depression and anxiety). 22

Carer internal processes

Changing carers’ belief that pain is inevitable (fatalism) was not related to any change in carer mental health (QoL or anxiety). 32

An intervention that decreased the use of avoidant coping strategies (e.g. denial) and increased healthy behaviours (e.g. exercise) improved carer QoL. 31 Increased dyadic coping and stress communication by carers within the patient–carer relationship seemed to be associated with improved mental health (depression and anxiety). 34,35 However, the intervention only increased dyadic coping and stress communication for partner carers, and it in fact decreased stress communication in parents cared for by an adult child. Furthermore, path analyses did not show that common coping and stress communication mediated the effects on carers’ anxiety or depression. Finally, Mosher et al. 29 found that reducing carers’ perceived constraint in discussing patients’ illness with them was not related to changes in carers’ depression or anxiety levels.

Improving carers’ control over the care situation (through improved support for mutual decision-making or satisfaction with involvement in decision-making) was associated with lower carer distress. 33 A greater sense of autonomy (internal motivation/willingness to provide care) among carers was also associated with both lower depression and lower anxiety levels. 22

Regarding self-efficacy, an intervention that improved carers’ and patients’ confidence in their ability to manage the illness and its related caregiving also improved carer QoL. 31 Furthermore, an improved confidence in ability to manage a range of caregiving components was related to lower depression and anxiety. 22 However, improving carers’ confidence in managing their own emotions showed no relationship with either depression or anxiety. 29

Improving carers’ sense of mastery (both perceived and actual ability to perform caregiving) was associated with maintaining QoL over time. 23 In contrast, several studies found that improving carers’ perceived adequacy of performance or feelings of competence for caregiving was unrelated to changes in their mental health (depression and anxiety25,26 and distress27). Similarly, improving carers’ preparedness for caregiving showed no association with their mental health (QoL,30 depression and anxiety,25,26 and distress27,30).

Support

Improving carers’ satisfaction with care received (for both patient and family) was not associated with changes in their QoL. 28 Otherwise, no intervention studies investigated whether improving carers’ support or perceived support improved their mental health.

Summary

This intervention literature can at best indicate associations between modifiable identified factors and carer mental health, as only one study directly investigated whether or not factors may be mediators of carer mental health. 34,35 Overall, this literature may show some support for a link between identified factors (see Box 1) and mental health, but studies were few and the findings were mixed. Potential biases in this literature are likely to stem from a lack of blinding, rather than systematic differences between intervention and control groups.

Carers’ internal processes received most attention. There was an indication that improvement in carers’ sense of control, autonomy, self-efficacy and mastery in their management of caregiving were associated with better mental health. However, in contrast, several studies found no indication that improved sense of competence or preparedness for caregiving were related to carer mental health. These apparently contradictory results indicate a need for a better understanding of the underlying concepts that interventions aimed to target, to help resolve contradictions and develop more effective interventions. Regarding coping, there was some evidence that reduction in (dyadic) avoidance coping and increased healthy behaviours were related to improved mental health. 31 In contrast, Heymann-Horan et al. 34,35 found no evidence from path analysis that dyadic coping and stress communication were mediators in improvement of carer mental health. However, they noted that their intervention may only indirectly target coping, and highlighted the need to assess more directly targeted mechanisms (factors) and the range of coping strategies employed, which again indicates the importance of gaining a better understanding of underlying concepts.

Otherwise, intervention studies provided some evidence that improvement in patients’ condition (through improved QoL and reduced depression) may relate to improved carer mental health. However, improvement in patients’ physical symptoms showed no relationship. Interventions to reduce the impact of caring responsibilities through the reduction in carer burden or demands showed mixed results. Only one study considered relationships, indicating that improving patient–carer relationships may be associated with better carer mental health. Further, only one study considered support, in terms of carers’ satisfaction with care, and found no effect.

The correspondence between findings from the intervention studies and the wider literature will be considered in Synthesis of combined findings from qualitative, observational and intervention reviews.

Challenges for drawing firm conclusions from the data

Several issues affected our ability to draw firm conclusions from the intervention review given the remit of our synthesis. With the exception of Heymann-Horan et al. ,34,35 interventions were not focused on investigating factors per se and their impact on mental health. First, although all sought to improve mental health, for some this was not the primary outcome or aim. For instance, the main outcome for Sulmasy et al. 33 was improved decision-making, and for Parker Oliver et al. 32 it was carers’ perception of pain. For other studies mental health was often a part of a range of outcome measures, such as for Nguyen et al. 30 Second, the focus was on testing the impact of the intervention, not the mechanisms through which it worked or relationships between mediating factors and mental health. Our factors of interest may not be what the intervention sought to change (its target) to improve mental health, and may simply be included as another outcome to be measured alongside mental health outcomes (e.g. Parker Oliver et al. 32 and Sulmasy et al. 33). Furthermore, even when the intervention did seek to change a factor of interest, the factor and outcome may have been measured only once at the same time after baseline (e.g. Badr et al. ,22 Henriksson et al. 25 and Sulmasy et al. 33) or, if repeated measurements were available, these were normally not analysed to consider how an early change in a factor may lead to a later change in outcome. Therefore, we could only establish that the putative ‘factor’ and ‘outcome’ both changed together; we could not establish the sequence of change. The information gained from intervention studies was therefore in fact often similar to that gained from cross-sectional studies, and in some ways less informative to our investigation because the correlation between factor and outcome may not have been directly tested.

Interventions may furthermore be unsuccessful in changing a putative ‘factor’ sufficiently to have an effect, but it might still be an actual factor affecting carer mental health. Although we only included interventions that had a significant impact on a factor, the change in the factor may not have been large enough for it to influence mental health (statistically significant differences may not have made a clinically meaningful difference). Changes in the factors mainly appeared small even if significant (e.g. Mosher et al. 29 and Nguyen et al. 30), and effect sizes, when reported, were often small or at best moderate (e.g. Henriksson et al. ,25 Holm et al. 26 and Hudson et al. 27).

The delivery of dyadic interventions (involving both patient and carer) is fairly common in carer intervention research,22,29,31,33–35 but further complicates our understanding of mediating factors. These interventions may maximise effect by influencing patients, carers and their interactions. However, they may also dilute the effect on the carer if the patient takes precedence. Plausibly, whether or not dyadic interventions maximise or detract from the impact on carers may depend on their target factor, e.g. they may maximise for communication and shared decision-making, but detract for factors relating more to carers’ own needs as opposed to patients’.

A strength of this review was its novel approach in reviewing the trials literature to uncover further information about factors that potentially affect carers’ mental health, guided by comprehensive reviews from the qualitative and observational carer literature. Although this review provided little added information, an important contribution of the review was to highlight how the current interventions literature fails to reach its potential in improving our understanding of the factors and underlying mechanisms affecting carers’ mental health.

Search, selection and data extraction

Synthesis of intervention studies, Method, Box 2 and Appendix 1 provide the search strategy and inclusion/exclusion criteria for the project as a whole, and for the intervention review. Added inclusion criteria for the qualitative review were that studies had to have as their aim to investigate carers’ mental health from the perspectives of EOL carers themselves, and for the observational review that studies had to report on the relationship between a factor and carer mental health outcome. For full details of the search, selection, data extraction and QA for the whole project, see Bayliss et al. 18 for the qualitative review, Shield et al. 19 for the observational review and Synthesis of intervention studies, Method, in this report for the intervention review. A summary is provided below.

For all studies, 10% of both titles/abstracts and full texts were screened independently for inclusion by two reviewers. Over 90% agreement was established in each case, and subsequent studies were screened on title/abstract and full texts by one reviewer. Owing to project time pressures, and on advice of the project’s external Study Steering Committee [and notification to the National Institute for Health and Care Research (NIHR) Health and Social Care Delivery Research (HSDR) programme], this process represents some tightening of the search strategy and inclusion/exclusion criteria (i.e. most recent decade, fewer databases, Organisation for Economic Co-operation and Development (OECD) country and English or Scandinavian publications, and peer reviewed publications only), and simplification of the screening process.

See Synthesis of intervention studies, Method, for data extraction and QA information for the intervention studies. For qualitative studies, first order themes were extracted for 10% of studies by two researchers and carer RAP members, and the remaining data extraction was carried out by one reviewer. Second order themes were created by one researcher and reviewed by a second researcher, and presented to the carer RAP for any comments. For the observational review, the data extraction template was tested independently by two reviewers on 10% of included studies, any differences were discussed and the data extraction template was clarified, and the remaining data extraction was carried out by one reviewer with a random sample of 10% checked by another.

Quality appraisal of qualitative studies used the CASP Qualitative Studies checklist. 38 The QA of observational studies used an adapted version of the Newcastle–Ottawa Scale (NOS) for cohort and case-control studies,39 modified to encompass cross-sectional studies based on an adjusted NOS scale. 40 QA was carried out independently by two reviewers on 10% of the studies. Over 90% agreement was achieved, and subsequent studies were quality assessed by one reviewer.

Synthesis of findings from the qualitative, observational and intervention reviews

The starting point for the synthesis was the thematic framework template developed from the qualitative review in collaboration with carer RAP members (see Box 1 for overarching themes and Bayliss et al. 18 for the full thematic framework). Findings from the observational review were synthesised thematically into subthemes using Box scores21 and mapped onto the relevant overarching themes within the qualitative framework template. Materials were sent to RAP members, discussed between researchers and RAP members in an online meeting, and amended to improve the clarity and meaningfulness of presentation to carers (see Shield et al. 19 for the observational results). The same process was then repeated for the intervention review (see Analysis and Appendices 2 and 3).

Next, findings within subthemes from all three reviews were combined into one document, which indicated (1) which findings were from the quantitative reviews and which were from the qualitative review, (2) where there was a direct match between the qualitative subthemes and quantitative subthemes and (3) where there was no evidence of a relationship from the quantitative findings (see Appendix 4). This draft document was sent to the RAP and discussed in a further meeting for feedback. Members were asked to consider the following:

• How clear is the information presented?

• Do these themes/subthemes make sense? Are they logical? Do they reflect reality?

• Do the themes/subthemes need to be changed in any way?

The document was considered a useful way of summarising all the findings, but again there were some suggested changes to make it clearer and more meaningful to carers. See Appendix 4 for the final presentation.

Finally, findings from the individual reviews and combined syntheses were presented to additional carers and practitioners, commissioners and policy-makers for their assessment of the validity, importance and relevance of findings to their respective stakeholder groups.

Patient condition

Patient condition was the theme investigated by the largest number of observational studies: 31 studies reporting on 95 bivariate investigations into relationships between patient condition factors and mental health. A further six studies reported multivariate analyses only, which confirmed findings from bivariate investigations. Five intervention studies also considered factors within this theme, the second most investigated within intervention research. Within qualitative research, carers mentioned issues related to this theme as important to their mental health in 19 out of 33 studies.

Within the theme, the quantitative research appeared mainly to focus on factors relating to patient information recorded in clinical records (e.g. diagnosis, treatment) or measured and monitored in clinical practice (e.g. QoL, functional status, symptoms). The quantitative evidence was most consistent in linking severity of patients’ psychological symptoms to worse carer health, and patient QoL to better carer health, whereas evidence for physical and functional decline was more mixed. The qualitative carer narratives confirmed the importance of patient condition to carer mental health, but seemed to reflect more the emotional connotations of the physical and cognitive decline or expected decline in the person cared for.

Impact of caring responsibilities

This theme received modest consideration by observational research: 14 studies and 34 bivariate investigations. A further three studies reported multivariate analyses only, confirming themes from the bivariate analyses. Only three intervention studies considered this theme. Within qualitative research, carers mentioned issues related to this theme as important to their mental health in 18 studies.

A considerable amount of qualitative evidence contributed to this theme. Carers considered how their mental health was affected by caregiving workload, conflicting or added responsibilities; the exhaustion, physical impact felt and crises experienced; the lack of rest, time for self-care and respite, and the impact of employment. Carers also highlighted the effects of isolation, loneliness and the inability to socialise. The observational evidence complemented the qualitative findings, showing consistent relationships between carers’ mental health and lifestyle adjustments (e.g. negative life changes) or demands of caregiving (e.g. difficulty of tasks, time spent on caregiving). Similarly, standard measures of carer burden were consistently related to mental health (covered under the heading relating to ‘Workload/physical burden/carer workload’ and that relating to the impact of ‘Caring for patient’ in Appendix 4). Intervention findings were few and mixed, showing negative, positive or no impact of carer burden.

Relationships

Relationships received little consideration in observational research: eight studies and 16 bivariate investigations. A further two studies reported only multivariate analyses, which confirmed themes from the bivariate analyses. Only one intervention study considered relationships. Within qualitative research, carers mentioned issues related to this theme as important to their mental health in 13 studies.

In qualitative studies carers highlighted the impacts of relationship changes, strains or conflicts caused by the patient’s illness or by caregiving. This included the patient sometimes not going along with treatment (‘non-compliance’), which could cause distress if the carer feels the patient is not doing as well as they possibly could do and they feel unable to influence this. Quantitative research, both observational and intervention, seemed to be more focused on the general quality of the patient–carer relationship and communication. What little evidence there is suggests that higher quality was related to better carer mental health. Wider family relationships were not highlighted in the qualitative research and considered only in observational studies. These indicate that families’ ability to cope with stressors, cohesion, supportiveness or levels of conflict were related to mental health.

Finances

This received the least consideration in quantitative research: only six observational studies through eight bivariate investigations, and no intervention studies. Within qualitative research, carers mentioned issues related to this theme as important to their mental health in 14 studies.

Overall, finances have received less consideration than other themes. However, the impact of costs, concerns about finances, access to benefits and impact on work appear to have been emphasised more in the qualitative literature by carers themselves than in the quantitative research. Still, observational study findings indicate a relationship between finances and carer mental health when sufficiency or insufficiency of resources are considered, as opposed to the level of income per se, and also highlight impact on work as a factor.

Carer internal processes

This received modest investigation by observational research: 13 studies and 36 bivariate investigations. Furthermore, four multivariate analysis-only studies found associations for preparedness, but mixed results for coping, which corresponds with bivariate analysis findings. This theme, however, received the largest focus by intervention research, with 11 out of 13 studies including a factor within this theme. Carers mentioned issues related to this theme as important to their mental health in 22 qualitative studies (15 studies noted factors detrimental to mental health and 13 noted strategies to improve it).

In the qualitative literature, carers highlighted how a loss of self-determination and autonomy and a lack of control affected their mental health, as well as a lack of experience of acting as a carer. Lack of control included uncertainty about future events and progression. The impact of transitions, and coming to terms with these, also featured strongly in carers’ narratives. The above seems to highlight the dynamic and uncertain nature of caregiving from carers’ perspectives.

Carers also stressed ways of coping as important, which moves us on to potential strategies to support mental health. Having time for respite, and using strategies to enable such time, was seen as positive for mental health. Carers also mentioned the use of positive self-talk for coming to terms with the situation and retaining positivity, and of spirituality for acceptance and lessening of isolation. They also highlighted the suppression of their own emotions and needs, which may be detrimental longer term but can be seen as a rational strategy for managing day-to-day caregiving from a carer’s perspective.

Observational study findings on coping strategies were very mixed. There were some indications in this literature that lack of acceptance, avoidance or substance abuse are related to worse carer mental health, whereas being optimistic and having time for oneself are positive. Intervention research indicated that a decrease in avoidant coping and an increase of healthy behaviours were associated with improved mental health. However, overall there was little clarity within this literature, and, to make progress, quantitative research may need more conceptual clarity in terms of how coping strategies are defined and are expected to work, and better methods for studying them (e.g. moving from generic coping measures of hypothetical scenarios to more investigation into real-life situations).

Quantitative research, in particular interventions, focused considerably on self-efficacy, mastery and preparedness. Carers themselves appeared generally to emphasise this less, although they mentioned a lack of carer experience, autonomy and control as factors. Observational studies indicated that self-efficacy and preparedness overall were associated with better mental health. Intervention studies also indicated that improved self-efficacy, confidence, mastery, autonomy, control and communication related to improved carer mental health, but showed no such relationships for preparedness or sense of competence.

Support

This was the second largest category for observational research, considered by 18 studies, with 42 bivariate investigations and one multivariate study confirming that good support was related to better mental health. Support was considered in only one intervention study. However, carers were most likely to mention issues related to this theme as important to their mental health: it was mentioned in 29 out of 33 qualitative studies (22 studies noted factors detrimental to mental health and 23 noted strategies to improve it).

Carers highlighted features of the formal care system as having a negative impact on their mental health, such as shortfalls in the availability and quality of care; disjointed care; and a lack of information, practitioner skill and adequate pain management. They also noted features of interaction with practitioners, such as a lack of empathy, poor communication, not listening to wishes of patients and carers, not recognising carers’ expertise and lack of collaboration. Some also experienced cultural barriers related to language. In terms of strategies to support mental health, carers highlighted the importance of formal support and access to information. Observational study findings within this theme showed little consistency of focus and clarity. However, unmet needs in general were associated with worse mental health, and carer satisfaction with available support was associated with better health (conversely, the one intervention study found no relationship regarding satisfaction). Observational studies of communication showed little relationship with mental health, but focused on components within the communication (e.g. dialogue pace), not its perceived quality, accessibility or adequacy. Accessible information was positively related to good mental health in one study. Findings regarding receipt of services per se were mixed and difficult to interpret without further information on what was provided, its perceived usefulness and how well it matched with carers’ needs.

Regarding informal support, qualitative studies found a lack of informal support to be detrimental. Conversely, strategies to support mental health included support from friends and family and a sense of community and shared responsibility. Support and information from others in the same situation was also important. Some observational studies found informal support to be positively related to mental health whereas others showed no relationship, but again it would be important to ascertain what support was provided, its usefulness and how well it matched with the carer’s needs.

Contextual factors

Contextual factors were considered in observational research only, in which they represented the third largest body of evidence: 16 studies with 104 bivariate investigations, and a further seven studies with multivariate analysis only that mainly confirmed bivariate results.

Older carer age generally seemed to be associated with better carer mental health, and being female with worse mental health. The remaining findings consisted of the occasional significant relationship set against a much larger set of non-significant relationships for the same factors. Therefore, it is difficult to draw further conclusions from this research at present. A notable gap in the literature is the limited research on caregiving experiences and outcomes relating to differences in ethnicity, race or culture.

Summary

In general, the findings of the qualitative, observational and intervention literature fell within the same themes and arrived at similar or complementary results, adding validity to the findings overall.

Patient condition was the most researched theme, possibly owing to the availability of clinical data, but also indicating how important patients’ well-being is to carers’ mental health, in particular patients’ psychological symptoms and QoL. However, we must not assume that by simply focusing on improving patient well-being, carers’ mental health will also be fully addressed. Rather, it is important to ensure that the patients’ basic well-being is covered, so that carers do not have to expend added energy and frustration on getting adequate formal support to meet patients’ clinical needs. We need to be aware that patients’ psychological symptoms may be particularly distressing to carers and may require particular focus to address both these symptoms and carers coming to terms with these symptoms. Finally, within EOL care elements of patients’ decline is natural and inevitable, but carers may need emotional support to cope with distress related to this.

The impact of caring responsibilities emerged as an important theme for carers themselves and was the focus of a fair amount of observational research. The evidence seems quite consistent that the life changes and increased workload associated with caregiving were related to worse carer mental health. General measures of carer burden also showed consistent relationships with mental health. However, to help identify where to focus interventions, it is important to consider further what is measured within different carer burden measures to clarify which factors here are the likely predictors of carers’ mental health (and whether they are likely precursors to mental health, rather than being influenced by mental health themselves). Potential interventions or remedies within this theme are wide ranging, including changes to employment law, help with negotiating change, support with care tasks, and respite to prevent exhaustion and to promote self-care.

Relationship with the patient seems to be important to carer mental health but is under-researched and merits further study. In terms of implications for supporting carers, it is important to be aware of the role relationships may play and potentially of how to facilitate open and constructive communication during adjustment to terminal illness. This includes situations in which the patient does not understand the impacts of the carer’s responsibilities for caring, and, for example, may resist carers bringing in added help. Dyadic interventions may be of particularly high value in addressing factors that can be exacerbated or eased by relationship dynamics, compared with interventions focused on carers only. Particularly ‘dysfunctional’ relationships or families may still require more specialist help.

Finances was the least researched theme, but appears important to carer mental health and clearly warrants further research. However, the potential sensitivity of this topic to families and its likely political implications may be elements limiting research so far, and would need to be considered and tackled for future studies. In terms of current implications for practice, findings highlight the importance of ensuring that carers are able to access the benefits that they are entitled to and are aware of any entitlements for flexible working and carer leave under their terms and conditions of employment. This often requires that carers recognise themselves as ‘carers’, and they may sometimes also require support to come to terms with this label. Future initiatives at policy level to support carers should consider employment law and benefits provision.

Internal carer processes formed the main focus for intervention research, but also received substantial attention within qualitative and observational research. Quantitative findings relating to coping strategies were very mixed, which may reflect several issues: a failure to meaningfully measure strategies, the fluidity and changeability of such strategies, difficulty of establishing direction of causality between strategies and health, and that a ‘maladaptive’ strategy may ‘do the job’ of getting the carer through the situation at a given time. Quantitative research here needs more conceptual clarity and better methods to progress our understanding of coping strategies. In addition, quantitative research, in particular interventions, focused considerably on self-efficacy, mastery and preparedness. Both observational and intervention studies indicate that improved self-efficacy is beneficial, whereas results for preparedness and related concepts, such as mastery or competence, were mixed. Although promising, contradictions in findings again highlight that we need a better understanding of these constructs, and how best to improve and measure them, to progress further.

There appeared to be a discrepancy in emphasis placed on self-efficacy, mastery and preparedness between observational/intervention research and carers themselves. Carers mentioned lack of caregiving experience, but seemed more likely to emphasise loss of self-determination, autonomy and control, which may instead reflect the impact of the caregiving situation on them. Researchers may focus on self-efficacy, mastery and preparedness because they see them as more amenable to intervention than other factors. However, this focus may also reflect a general ‘self-management’ perspective in health and social care delivery. Interventions to boost self-management-related factors may indeed improve carer mental health if sensitively designed and delivered. However, such initiatives could also add burden by placing added responsibility and onus on carers and detracting from factors that carers themselves see as important in preserving their mental health. In strategies to support mental health, carers themselves focused more on allowing for respite and emotional acceptance, which are compatible with a ‘self-management’ perspective but give more emphasis to respite and interventions to enable acceptance than the more ‘active’ self-management often championed. Although interventions to foster self-efficacy show considerable promise in improving carer mental health, it is therefore important to ascertain carers’ own perspective on such interventions and incorporate elements that work for them.

Support, and the quality of support provided, appeared to be important to carer health. Aspects of formal support formed a large part of carers’ narratives in qualitative studies. The relationship between formal support and carer mental health was also considered by a substantial number of observational studies, but notably was nearly absent from intervention research, although interventions to improve aspects of formal support should be possible. However, for all its investigations, observational research appeared to fail to measure the features of formal support that were important to carers (e.g. co-ordination of care, sufficiency of information, empathy), and there should be greater future research emphasis on features of service provision likely to affect carer mental health. Informal support received less attention from both carers and quantitative research, but still appears important for carer mental health. The findings from the qualitative research can provide good pointers for improvements in service delivery that matter to carers and, similarly, important elements to foster within social networks. When carers have limited networks of family and friends, it would be important to consider alternative social network options, such as bringing in peer support.

Contextual factors were considered only in observational research. Findings indicate that younger carers and females are more likely to report adverse mental health outcomes. However, results were too mixed and lacking in significant findings to provide helpful guidance. It is important for future research to investigate why younger carers and females may be vulnerable, to guide interventions. Research also needs clearer rationales for choosing other contextual factors for investigation. Service providers need to be aware that younger carers and females may be more at risk of mental health impacts, despite a common assumption that women may be more ‘naturally suited’ to the carer role than men.

Comparison with previous and recent literature

The present review identified similar themes to previous quantitative and qualitative reviews of the carer literature in the period 1998–2008,16,41 again validating our findings. However, our review also added quantitative findings on relationships, finances and formal support to the earlier reviews. 18,19 Findings from more recent literature (2020–21) also fit within our present themes; in terms of patient condition, worse carer mental health was associated with severity of patient symptoms, particularly psychological,42–44 and worse patient functioning. 43,44 Comorbidities, duration of illness and frequency of hospital admissions may also matter. 43 Regarding the impact of caring responsibilities, more caregiving hours and less relaxation44 and greater carer burden43 also related to worse mental health. More active coping45 and less negative religious coping46 were associated with better health, and being female was associated with worse health. 44 Formal care support also appeared to matter. 44

What each method contributes

The qualitative literature is invaluable in providing carers’ inside perspective, including their understanding of what improves or worsens their mental health and solutions that would help. Interventions and support initiatives that are out of touch with carers’ own realities are unlikely to be effective. The qualitative research was therefore used to provide the framework of themes for the overall synthesis. Limitations of this literature include that it is difficult for carers to consider the carer population as a whole and patterns within it. For instance, individuals will be able to see that a factor (e.g. patient depression) made them depressed, but cannot easily know if other carers experienced the same factor without becoming depressed because other factors protected them from depression (e.g. supportive social networks). Understanding the carer population as a whole and the mechanisms through which impact is mediated requires a group perspective in addition to an individual perspective. Furthermore, although qualitative studies provided information on carers’ own solutions, identification of solutions was normally not the focus of this qualitative literature. Therefore, the qualitative studies may provide an incomplete, rather than a comprehensive, picture of carers’ own solutions.

The observational literature can statistically test for significant relationships between factors and carer mental health at a group level, and can both confirm individual perspectives and uncover additional patterns less obvious to individuals. As it often has larger samples and does not seek to manipulate associated factors, in contrast to interventions, it can more easily encompass a broader range of variables and allows for a wider investigation of factors. However, this may also be a weakness, as current literature sometimes gives the impression that factors are included because measures are easily available and easy to add (e.g. some patient condition, formal support and context variables in particular), but without further consideration as to why they should have an impact on carer mental health. A parallel issue is that this literature may converge on similar factors because researchers want studies to be comparable, which can be commendable in progressing the field. However, if they are not the most important factors, just the most commonly measured and easily accessible, this may mainly add ‘noise’ and potentially lead to confounding bias. Finally, this observational literature may tell us little about causality, particularly as it constitutes predominantly cross-sectional, rather than longitudinal, research and it gives us little direct information on what solutions or interventions may help.

Intervention studies are designed to help establish causality and could therefore be a crucial addition to our understanding of factors affecting carer mental health. However, a few issues limited their usefulness for our review. First, these studies naturally limited themselves to factors that can feasibly be manipulated in a trial (and whose equipoise ideally can be maintained), but this does not mean that other factors are not important. For instance, no intervention studies in this review considered finance. General improvement in carers’ financial situation would require larger system changes, and manipulating finances for individuals would be beyond the remit of intervention research. Similar considerations would also apply to many factors under the main theme of impact of caring responsibilities. Contextual factors, including carer age and sex, would not be possible to manipulate experimentally. There is less of a rationale for not including factors associated with formal support delivery within intervention research, but trials of service models or components can be complex because they sit within larger health-care systems (and when service models have been tested, their focus is less likely to be on carer mental health). Second, interventions that involved modifiable factors may still not have been successful in changing the factor sufficiently to demonstrate an impact on outcome, even if the factor itself genuinely affects mental health. A lack of change in mental health may not be due to the factor, but to an ineffective intervention. Finally, when factors were found to be associated with carer mental health, the aim was not normally to investigate the factors’ causal relationship with mental health. Therefore, owing to intervention study design or analysis this literature provided little added information on causality, and only indicated association, similar to what could be gained from observational cross-sectional research. Nevertheless, the combination of the three types of literature provided insights into factors associated with carers’ mental health and the state of the field that we could not have gained by viewing one strand on its own.

Strengths and limitations of the project

A considerable strength of the project was the comprehensiveness of a robust synthesis approach, combining reviews of qualitative, observational and intervention literature. This enabled us to gain a full overview of current knowledge, benefit from strengths of each approach and gain reassurance regarding validity of findings from triangulation. Using the qualitative review to provide a framework for the other reviews helped to provide an overall structure and, importantly, ensured that this structure was rooted in carers’ own considerations.

Another major strength of the project was its stakeholder engagement and consultation. We worked with a carer RAP throughout and conducted wider stakeholder consultation at the end of the project with additional carers, patients, practitioners, commissioners and policy-makers (this work will be reported elsewhere). Considerable time and resources were spent on involving these audiences and maintaining meaningful engagement during the COVID-19 pandemic. This was actively used as a means of assessing and ensuring the validity, importance and relevance of findings to key stakeholders. Furthermore, learning how to adapt communication of findings so they were meaningful to these audiences was crucial, to help ensure that findings would reach those who may be able to utilise them to influence change and improve the daily lives of carers.

The review was limited to studies in OECD countries, publications in English or Scandinavian languages and studies on adult carers of adult patients. Therefore, findings may be less relevant outside these contexts. Furthermore, we only considered mental health during caregiving but there is additional important work required to ensure that carers are supported after the patient’s death. The search was also limited to the time period 2009–19, but comparison with overview reviews of the carer literature in the period 1998–200816,41 indicates that carer issues remain quite consistent over time.

A challenge for the synthesis was the ‘sprawling’ nature of the literature regarding study focus and objectives, methods, variables and measures, which was greater than expected. However, this is a finding in its own right, and highlights the need for a clearer focus and direction of enquiry to make meaningful progress in this field.

Implications for research

The combination of the three types of literature enabled us to identify contradictions, weaknesses and gaps in the field overall. Consequently, this project provides several implications for future research:

• There needs to be more research on finance and relationships and their impact on carer mental health.

• There are gaps in our knowledge of how carer experience is affected by ethnicity, race and culture.

• Observational research needs to ensure that it focuses on factors that are important to carers (e.g. more meaningful formal support delivery variables) and variables that can illuminate what affects carers’ mental health, rather than what can be easily measured.

• Intervention studies should consider more investigation into formal support delivery components that matter to carers and focus on ‘self-management’ interventions that work for carers.

• Intervention research needs to employ designs and analysis that utilise its full power to investigate causal relationships between factors and carer mental health.

• Both observational and intervention research would benefit from more use of longitudinal design with repeated measures and path analysis, and from clearer conceptualisation of key variables, for example carer burden, coping strategies and mastery.

• There should be consideration of when dyadic interventions or carer-only interventions may be most effective.

• Many factors important to carer mental health (e.g. finance, life impact factors, context) are not feasible to study using RCTs, and more use of quasi-experimental design may be considered.

• This research field as a whole would benefit from greater use of conceptual frameworks and theories to guide further research enquiry and ensure more coherence and, possibly, agreement on core sets of factors that should be included in observational and intervention study measurement.

Implications for practice

The factors likely to affect carers’ mental health during EOL caregiving cover a wide spectrum, including the patient’s condition, the impact of caregiving on carers’ everyday lives, the sufficiency of their finances, the quality of their relationships, their internal psychological processes, the formal and informal support provided, and demographic context variables. This means that the solutions for sustaining or improving carers’ mental health also need to be multipronged, and that different stakeholders are likely to be able to influence different factors. Some factors are amenable to change, whereas others may serve as ‘red flags’ that a carer’s mental health may be adversely affected and he/she may require closer monitoring to identify added need for support. Many factors are likely to be interlinked, for example relationships, informal support and impact of caring responsibilities on carers’ lives, and we need to be aware of such potential cross-factor influences.

The synthesis indicates that the following are important in supporting carers:

• Maintaining patient well-being, and monitoring and responding to the impact on carers of patients’ psychological symptoms and general decline.

• Improving employment terms and conditions for carers and providing help with negotiating transitions and support with care tasks and respite to prevent exhaustion and promote self-care.

• Bolstering the relationship between patient and carer and within the wider family, and facilitating open and constructive communication.

• Ensuring carer access to benefits and awareness of entitlements to flexible working and carer leave (which may include helping carers to recognise themselves as ‘carers’).

• Improving benefits and employment conditions for carers longer term.

• Improving carers’ sense of self-efficacy in their carer role, enabling carers to take breaks and help maintain a positive outlook and acceptance.

• Improving formal service delivery, to ensure it is, for example, co-ordinated, responsive, listening and provides good information and communication.

• Assessing carers’ social networks and providing peer support.

• Being aware of and supporting carer groups at risk of worsening mental health, such as younger or female carers.

Patient condition

Impact of caring responsibilities

Relationships

No factors identified.

Finances

No factors identified.

Carer internal processes

Support

Patient condition

Impact of caring responsibilities

Relationships

Finances

No factors identified.

Carer internal processes

Support

No factors identified.

Patient condition

Impact of caring responsibilities

Relationships

Finances

No factors identified.

Carer internal processes

Support

No factors identified

Patient condition

Impact of caring responsibilities

Relationships

No factors identified.

Finances

No factors identified.

Carer internal processes

Support

No factors identified.