Effectiveness of palliative care interventions offering social support to people with life‐limiting illness—A systematic review

Individuals managing the challenges of life‐limiting illness require adequate social support to maintain quality of life. Qualitative research reports that patients value highly the social support obtained in palliative care interventions such as day care and group therapies. This systematic review aims to summarise existing quantitative evidence on palliative care interventions that facilitate social support. Research literature was systematically searched using electronic databases and key journals. Searches returned a total of 6,247 unique titles of which sixteen were eligible for inclusion. Interventions include group therapies, group practical interventions and palliative day care. Outcome measures and study designs were heterogeneous. Only one study used a validated outcome measure of social support. Benefits were influenced by participant characteristics such as baseline distress. Partial economic evaluation was attempted by two studies. Methodological challenges include attrition and use of outcome measures that were insensitive to change. Statistically significant results were reported in psychological and physical domains. Evidence is limited due to methodological issues and a scarcity of quantitative research, particularly regarding long‐term benefits and cost‐effectiveness. Interventions may be more beneficial to some groups than others.

Palliative care aims to improve quality of life for patients with life-limiting illness and their families (Hui et al., 2013), providing support for patients to live as actively as possible (World Health Organisation, 2002). The social world of an individual has potential to contribute to, or alleviate, suffering in life-limiting illness (Garcia-Rueda, Valcarcel, Saracibar-Razquin, & Solabarrieta, 2016). Pain, fatigue and other symptoms can limit opportunities to engage with others, so that declining physical function is paralleled by increasing social restriction (Lloyd, Kendall, Starr, & Murray, 2016). Friends and family members may struggle to accept the diagnosis and feel unable to relate to patient experiences (Wilson & Luker, 2006). Social relationships may also be limited by stigma around illness and death (Garcia-Rueda et al., 2016).
However, social support is associated with better outcomes in advanced cancer (Applebaum et al., 2014) and other incurable diseases (Tomaka, Thompson, & Palacios, 2006). Obtaining emotional support from others, a common coping mechanism in advanced cancer, is associated with better quality of life and reduced anxiety and depression (Nipp et al., 2016). People coping with the existential challenges of approaching death can find interpersonal relationships to be a key component of experiencing meaning in life (Haug, Danbolt, Kvigne, & DeMartinis, 2016), and social comparisons with people in a similar situation may be helpful when trying to establish a new sense of normality in the presence of advanced illness (Lobb et al., 2013). Therefore, psychological stressors that accompany life-limiting illness might be alleviated through social support (Crunkilton & Rubins, 2009).
Patients and referring clinicians agree that additional social support is helpful (Bradley, Frizelle, & Johnson, 2010; thus, informal relationships cultivated in palliative care are valuable to stakeholders, but appear underacknowledged by research (Wilson & Luker, 2006).
A broad range of palliative services can facilitate social supportincluding home visits or remote support delivered via telephone or Internet. However, opportunities to get out of the house and engage with others in a dedicated environment are thought to be beneficial for well-being, by relieving both physical and psychosocial isolation (Bradley et al., 2010;Stevens, Martin, & White, 2011). Palliative day care offers psychosocial support alongside clinical services-patients in the UK are motivated to attend palliative day care because of potential gains in social support (Goodwin, Higginson, Myers, Douglas, & Normand, 2002;Kernohan, Hasson, Hutchinson, & Cochrane, 2006). Peer relationships developed in face-to-face group therapies or support groups are also highly valued by patients (Taylor-Ford, 2014). Thus, this review focuses on solely on interventions taking place outside of the home.
Reviewers have noted a scarcity of quantitative evidence in the evaluation of palliative care (Aoun & Nekolaichuk, 2014), particularly so for interventions with social goals that can be challenging to define and measure (Bradley et al., 2010;Stevens et al., 2011).
The evidence base for economic evaluation is poorly developed (Gardiner, Ingleton, Ryan, Ward, & Gott, 2016), yet it is vital to demonstrate effectiveness and make pragmatic decisions about resource allocation in the context of an ageing population and increasing healthcare demands (Woodthorpe & Foster, 2016). There is moderate evidence that complex interventions can improve quality of life in palliative care, but it is unclear which components of these interventions confer benefit (Catania et al., 2015). The research above suggests that provision of social support may be an influential component of psychosocial interventions in palliative care.
This systematic review aims to summarise available evidence on the effectiveness and cost-effectiveness of palliative care interventions that facilitate social support, by including studies using at least one quantifiable, validated patient outcome measure.

| ME THOD
The review was conducted according to PRISMA guidelines (Moher, Liberati, Tetzlaff, & Altman, 2009). A review protocol was not registered. This review is part of an ongoing project funded by the Economic & Social Research Council. The primary reviewer (NB) conducted searches, data extraction and quality appraisal. A representative sample of 10% of identified titles and papers was independently checked by a second reviewer (MLW) to reduce risk of bias.
Any discrepancies during the process were discussed with the entire team (NB, MLW & CFD).

| Inclusion criteria
Palliative care interventions were considered to offer opportunities for social support if they facilitated face-to-face interactions with other people, outside of the individual's home. Participants are adult outpatients (at least 18 years old and currently living in the community) with a diagnosis of life-limiting (incurable) illness, including but not limited to cancer. Group activities, structured group interventions and settings such as palliative day care were included. Publications were included that reported at least one validated, quantifiable patient outcome measure, in any domain; including perceived social support, quality of life, psychological distress and symptoms. Mixed methods papers were included.
Only English language papers were included due to resource limitations. There was no other restriction on comparators or study design.

| Exclusion criteria
Interventions for children, long-term care residents or palliative inpatients were not included in this review, as issues of social wellbeing are expected to differ in these populations. Interventions consisting entirely of clinical appointments, individual tasks or home-based activity were excluded. Publications were excluded that were purely descriptive or used only qualitative evaluation.

| Quality appraisal
The quality of included papers was appraised using a structured checklist approach designed for disparate data (Hawker, Payne, Kerr, Hardey, & Powell, 2002). The flexibility of this checklist in assessing diverse study designs was considered advantageous for this review.
Quality assessment considered nine domains: abstract, introduction, method, sampling, analysis, ethics and bias, results, transferability and implications. Each domain was scored out of 4, with higher scores indicating better quality, giving a maximum score of 36.

| Data extraction and synthesis
Data were extracted summarising: study design, sample characteristics, outcome measures, results and methodological difficulties reported. This included results of subsequent analysis and additional publications from the same study. A meta-analysis was not appropriate due to very high heterogeneity of included studies; thus, results are presented as a narrative summary. POMS, pain and suffering scale Improvements in mood disturbance (p = .02) and pain experience (p = .04) after one year Participants with higher baseline distress benefited more from intervention

| RE SULTS
Cost-minimisation analysis using resource use and patient outcomes, no significant result (Lemieux et al., 2006) Quality of life analysis reported separately (Bordeleau et al., 2003) See also: report on recruitment and enrolment (Goodwin et al., 2000) MQOL.
No statistically significant result.

POS.
Better pain control at "baseline" (p = .065) Better symptom control at 6-8 weeks (p = .053), lost by 12-15 weeks of 6,149 papers were excluded based on title and abstract, and the vast majority of these were not relevant to this topic. A total of 116 papers were read in full and assessed for eligibility. Sixteen studies met the criteria and are included in this review (PRISMA flow chart- Figure 1).

| Group therapeutic interventions
Group therapy had statistically significant effects on psychological outcomes including mood disturbance (Goodwin et al., 2001;Spiegel, Bloom, & Yalom, 1981), helplessness (Edmonds et al., 1999;Kissane et al., 2007), emotional impact of stressful events (Classen et al., 2001), emotional regulation (Classen et al., 2001), coping (Spiegel et al., 1981), intrusive thoughts and depression (Kissane et al., 2007). Four studies reported survival, with three studies finding no effect (Classen et al., 2001;Edmonds et al., 1999) and one study reporting significant survival effect ten years after the intervention (Spiegel et al., 1981). Two studies reported significant effects on perception of pain (Classen et al., 2001;Goodwin et al., 2001). All studies were randomised, which increases the strength of this evidence. Intervention effective at all time points for participants over 65 years  Relationship between caregiver age, caregiver QOL and patient QOL (Shahi et al., 2014) See also reports on role of exercise (Cheville et al., 2010) and social worker  N/A POMS, HLC, maladaptive coping.  (Classen et al., 2001;Goodwin et al., 2001)-more distressed participants benefitted more from the intervention. Edmonds et al. (1999) controlled for baseline psychological differences, intervention compliance and the use of self-help strategies, reporting a statistically significant result in helplessness only. However, the control group of this study received considerable support (CBT tasks, supportive telephone calls and relaxation exercises). The Duke-UNC Functional Social Support Questionnaire was used by Edmonds (Edmonds et al., 1999) and identified no significant change. However, the authors report that the questionnaire prioritised support received from the family unit and therefore appears insensitive to gains in social support occurring as a result of the intervention. Miller et al. (2005) measured baseline social support, stating that it was not expected to be influenced by the intervention (without further elaboration) and controlling for this variable; dropouts from this study had higher baseline social support, suggesting that the group format was not suitable or relevant for these individuals.  suggests social support should be monitored in future studies-warning that minimising participant burden can impede study design through the use of insufficient outcome measures. Poor sensitivity and floor effects of outcome measures were reported (Edmonds et al., 1999;Goodwin et al., 2001).

| Palliative day care
Day care was reported to have a statistically significant effect on symptoms Leppert et al., 2014) and hope (Higginson, Gao, Amesbury, & Normand, 2010). Studies with longer study designs found this was not apparent over time, possibly due to attrition and small sample size Higginson et al., 2010). The majority of participants in an observational study benefitted from 8 weeks of day care (Kilonzo, Lucey, & Twomey, 2015), but attrition makes it difficult to establish whether the intervention was beneficial for non-completers. Improved symptomology reported by Leppert et al (2014) could have been influenced by baseline differences in physical activity and symptom burden.
Methodological difficulties included large individual fluctuations over time (Sviden, Furst, von Koch, & Borell, 2009) and difficulty in obtaining a true baseline prior to day care attendance due to ethical concerns Sviden et al., 2009)
Social support was not used as an outcome measure. Rummans et al. (2006) identified improvements in social well-being through quality of life subscales. A pilot study with mixed methods (Roulston et al., 2012) appeared to have positive effects on mood and perceived health; all five participants cited social support as a useful component of the intervention. Tsianakas et al. (2017) reported that questionnaires used were not sufficiently sensitive, and draws from qualitative insights to suggest that a specific social support measure be used in future work. was not identified; however, the difference in resource use appeared to be larger for participants with higher distress at baseline.

| Economic evaluation
Douglas , reporting on Goodwin et al. (2003), presented evidence on intervention cost and highlighted challenges of obtaining accurate cost estimates for multidimensional interventions such as day care.
Participant health and social care resource use appeared to differ between the intervention group and comparison group-patients accessing palliative day care accessed fewer community services.
Conclusions were limited by group size and extent of missing data.

| Quality appraisal
The quality appraisal method used in this review (Hawker et al., 2002) was selected to allow for methodological heterogeneity, as evidence-based practice cannot rely solely on randomised controlled trials in areas of ethical sensitivity and fluctuating or deteriorating health. All studies were rated as fair for quality (more than 23 points out of a maximum of 36) or good (more than 31 points), although this frequently reflected the quality of reporting rather than the study itself. The lowest scoring domain was ethics and bias, reflecting that issues of confidentiality, consent and bias were not sufficiently articulated in the articles. There is some risk of reporting bias at study level, and it is possible that publication bias may have limited the likelihood of null results being identified. Consistent with other reviews (Singer et al., 2016), cancer diagnoses dominate the included studies.

| D ISCUSS I ON
Baseline psychological distress influenced intervention effectiveness: more distressed participants reported more benefit. This is in accordance with a meta-analysis of 61 studies on psychosocial treatment across all cancer stages, which concluded pre-intervention distress moderates intervention effects, more so than intervention format, setting or dose (Schneider et al., 2010). Baseline social support might influence participant's experience of an intervention and its acceptability (Miller, Chibnall, Videen, & Duckro, 2005).
Participant gender  and age (Chock et al., 2013) may be relevant, but research is limited. The relevance of caregiver quality of life was highlighted as an important influence on patient quality of life (Clark et al., 2013;Rummans et al., 2006;Shahi et al., 2014).
Survival differences were reported by one study, but not confirmed by three other studies reporting this outcome. This discordance might reflect sociocultural differences between studies and populations, for example, it has become more acceptable for cancer to be discussed in public (Spiegel et al., 2007). It is also possible that the reported survival difference (Spiegel, Kraemer, Bloom, & Gottheil, 1989) is anomalous: the survival curves of the intervention and control groups did not diverge until after notable attrition had taken place, and the 12 participants remaining in the control group by this time differed markedly from regional survival norms for breast cancer (Fox, 1998). Despite ambiguity over survival, psychosocial interventions can influence other clinical outcomes (Temoshok & Wald, 2002). Sufficient dose may be necessary for effectiveness (Miller et al., 2005), supported by survival analysis of the Edmonds study (Cunningham et al., 1998) identifying differences in active engagement with the intervention to be significantly associated with survival. An adjusted meta-analysis reported that, with sufficient dose, psychosocial and behavioural interventions may prolong survival for at least some patients with cancer (Xia, Tong, & Feng, 2014); interestingly, a subsequent meta-analysis across cancer populations reported a survival benefit in group interventions, but not individualised interventions (Fu et al., 2016).  (Stevens et al., 2011). Distinguishing between types of attrition and reporting on reasons for withdrawal can be revealing (Higginson et al., 2013).
Selection of outcome measures and other elements of research design should be informed by the experience of other researchers (Gaertner et al., 2016).

| Limitations of review
This review has limitations associated with search strategy and eligibility criteria. We used eight electronic databases, supplemented with hand searching of key journals and checking all citations and references. However, only English language articles and grey literature were searched due to resource limitations, introducing a source of bias and limiting the comprehensiveness of this review.
Intervention descriptions were occasionally insufficient to determine whether or not social support was facilitated-we did not include individualised therapeutic tasks completed in a room alongside others without explicit reference to group communication (Imriea & Troop, 2012)-it is therefore possible that relevant interventions were not included. A number of interventions were retrospectively evaluated as having enabled social support by the researchers, but were designed for a different purpose such as self-management (Roulston et al., 2012). All interventions involved multiple components, so it is possible that benefits such as symptom control would be more appropriately attributed to other components such as clinical input or self-management skills. The use of mixed methods in research can help to untangle the relationship between components and outcome (Higginson et al., 2013).
The criteria for inclusion in this review were developed from scoping searches, using patient experience of psychosocial palliative care interventions. We focussed explicitly on interventions taking place outside of the home, informed by a qualitative metasynthesis concluding that a change in scenery and getting out of the house were necessary for alleviation of both physical and psychosocial isolation (Bradley et al., 2010). This meant that well-established examples of community-based support programmes, for example the Good Neighbourhood Partnership in Ireland  and the Neighbourhood Network in India (Sallnow, Kumar, & Numpeli, 2013), were excluded from the review. However, models of home-based palliative care (which may or may not include facilitation of social support) are better represented than other settings in the research literature (Brereton et al., 2017), and we are aware of an ongoing systematic review of community-led support interventions for adults living at home with palliative care needs (Mcloughlin, Furlong, et al., 2015). An additional consideration is the increasing availability of different formats, for example social networking interventions (Owen, Bantum, Pagano, & Stanton, 2017).
Given the high prevalence of limited mobility in this population, it would be appropriate for future reviews to compare social support facilitation across different settings and formats.

| CON CLUS ION
Responding to the needs of people with life-limiting illness requires consideration of social well-being, including their perception of support from others. Psychosocial palliative care services and interven-