Background

Euthanasia and assisted suicide (EAS) are becoming legal in an increasing number of countries (including Australia, Belgium, Switzerland, Colombia, and Canada) and US states [1,2,3], with many other countries debating the issue of legalisation. The UK government is considering whether to allow assisted dying for terminally ill people [4], reflecting a high level of public support but divergent views of politicians, faith groups, and clinicians [5, 6]. Within countries or states where EAS is legal, between 0.3 and 4.6% of deaths are attributed to EAS [3], and the number of EAS requests and deaths attributable to EAS is rising [7, 8]. Most countries operate under strict restrictions, allowing EAS for those with either a terminal illness or serious physical illness causing intolerable suffering that cannot be relieved, such as amyotrophic lateral sclerosis (ALS), or terminal cancer [9, 10]. This is strictly monitored, aiming to achieve the goal that all patients receive the best end-of-life care aligned with their preferences. In some countries (such as Belgium and the Netherlands) EAS is extended to people based on a psychiatric disorder deemed to be causing suffering that is ‘unbearable and without prospect of improvement’, such as treatment-resistant depression [11]; often termed psychiatric EAS [12]. Such cases are often characterised by loneliness or social isolation, and disagreement between clinicians [13], and not all such patients have received evidence-based treatments [14]. The procedures for granting EAS vary substantially between countries and contexts, and decision-making should always take account of modifiable risk factors for suicidal thoughts. There are no reliable figures on how many people travel abroad to access EAS, or the proportion of suicides in which an individual had tried to access EAS. There is also no clear evidence that rates of (non-assisted) suicide change significantly in countries after EAS is introduced [15].

Terminology classifying EAS is contentious, reflecting the strong views of patients, clinicians, and the public regarding the ethics of these practices [16,17,18]. The term euthanasia describes the act of deliberately ending a person’s life to relieve suffering, and can be defined as passive (i.e., by omitting treatment such as life support or life-saving drugs with patient consent) or active (i.e., through deliberately and knowingly administering a sedative or relaxant drug to a patient who does not otherwise need it, at a dose that has the aim of ending their life, with the patient’s consent). The term physician-assisted suicide (PAS) and the related term physician-assisted dying (PAD) apply to cases in which a physician provides a patient with a drug or intervention in the knowledge that the patient will use this to end their life [14, 19, 20]. Assisted suicide describes where active euthanasia is enacted by a relative or carer, who has obtained those drugs with the intention of supplying them to an individual, who takes them willingly with the intention of dying. Whilst there is a considerable degree of overlap between the terms euthanasia and assisted suicide, euthanasia is defined as a doctor ending someone’s life with their consent, whereas physician assisted suicide and assisted suicide define the act of supporting someone in ending their life by providing them with the means to do so [21]. Due to these conceptual overlaps, this review uses the unifying label Euthanasia and Assisted Suicide (EAS) to describe euthanasia, PAS, PAD, and assisted suicide.

The term desire for hastened death or desire to hasten death (DHD) is used among people with a terminal illness to describe suicidal thoughts or requests for assisted suicide or euthanasia, regardless of whether they are legal in the patient’s country [22]. DHD is highly correlated with depression and suicidal ideation and regarded as a marker of profound psychological distress [23], signalling a need to identify and address modifiable risk factors for suicidality. Among people with a terminal illness these include clinical factors, such as pain and depression, and social factors, such as loneliness, poor social support, and financial concerns [24,25,26]. Reasons for requesting EAS include perceptions of unacceptable quality of life (commonly due to loss of independence, mobility, communication, or participation in meaningful activity) and fear of future suffering or disability [27]. It is therefore important to understand the factors driving people to consider EAS (via DHD), highlighting whether any risk factors can be further modified. This might also inform the development of approaches developed to support people in making an informed decision as to whether to end their life.

Theoretical models of suicide and suicidal behaviour, although they do not apply directly to EAS, propose that suicide arises from a complex combination of biological, social, psychological, and environmental factors, including social connectedness, interacting with adverse life events [28]. Whether a person acts on their suicidal ideation, including whether someone with terminal illness seeks EAS, is likely to be determined by the interplay of these influences. Thwarted belongingness is a key social connectedness construct in theoretical models of suicide that is also likely to apply to decision-making around EAS. It describes what arises when the fundamental need to form and maintain strong, stable interpersonal relationships is unmet, resulting in feelings of disconnection. This is postulated to influence the emergence of suicidal thoughts in someone who feels defeated and entrapped. The Interpersonal Theory of Suicide considers thwarted belongingness as comprising a) loneliness and b) the absence of reciprocal caring relationships [29]. Loneliness is defined as a subjective unpleasant feeling arising from a mismatch between a person’s desired and perceived level of meaningful social relationships [30]. Studies using the approach of confirmatory factor analysis have shown convergent associations of thwarted belongingness with loneliness [31] and there are also correlations of thwarted belongingness with other measures of social connectedness [32]. Loneliness is distinct from more objective measures of social connectedness, such as social isolation, social capital, and social network. Social isolation is an objective measure of the number of social connections, quantified in terms of social network size and number of meaningful ties [33]. It is therefore possible to have a large number of social contacts but still experience feelings of loneliness, or vice versa [34].

Loneliness is associated with a range of adverse health outcomes [35,36,37], including increased all-cause mortality [38] and suicide [39]. The association between loneliness and suicidal thoughts and attempts [36, 37, 40] is at least partially mediated by depression [40]. Social isolation is also associated with poor physical and mental health [41], and with suicidal thoughts, suicide attempts, and suicide [39, 42, 43]. Conversely, social support protects against suicidal ideation [44]. Given the high prevalence of loneliness and social isolation in people with terminal illness [45], and among people who request EAS for psychiatric indications [13], and its stigmatising nature [45], it is important to understand its contribution to decision-making around EAS. One previous systematic review of factors predicting request for or attitudes to EAS in older adults found no association with loneliness or with satisfaction with family relationships [46]. However, no systematic review to date has summarised associations between social connectedness and EAS across all age groups. In this review we aimed to describe the association between social connectedness and a) requested/actual EAS, b) attitudes to EAS, and c) DHD. We used the umbrella term social connectedness to describe a person’s perceptions of their relationships with others, including constructs such as loneliness, social isolation, social support, and other terms used to describe the quantity and/or quality of relationships (e.g., living alone; satisfaction with relationships). This met our objective to expand our scope to include measures beyond loneliness and social isolation, to capture all studies investigating associations with social disconnectedness.

Methods

We conducted the review in accordance with Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [47] (see Appendix 1a and 1b) and guidance for conducting narrative synthesis in healthcare [48]. We pre-registered the review protocol on PROSPERO (CRD42019160508).

Search strategy

We searched five psychological and medical electronic databases (PsycINFO, MEDLINE, EMBASE, Scopus, and Web of Science) from inception to November 2022. We conducted the initial search on 16th December 2019, repeated this in July 2021 as an interim update, and then finally on 17th November 2022 to capture all papers published until the end of October 2022. We used the following search terms: lonel* OR social isolation* OR social network$ OR social support* OR social contact* OR social relation* OR social capital* OR confiding relation* AND euthan* OR assist* suicide* OR assist* death*. Search terms were agreed by the interdisciplinary team, which included experts in loneliness research and a psychiatrist with experience of psychiatry in palliative care settings. We also reviewed the first 200 hits from Google Scholar for each combination of search terms. Finally, we also used reference searching of potentially eligible articles to identify additional citations.

Inclusion and exclusion criteria

Our inclusion criteria were any cross-sectional or longitudinal epidemiological study investigating an association between social connectedness (including loneliness, social isolation, social support, social capital, confiding relationships, and thwarted belongingness) and three outcomes a) EAS (requested or actual), b) attitudes towards EAS (including using hypothetical scenarios), and c) DHD. The first of these categories related to our main research question, but we included studies providing information on attitudes to EAS and on DHD to provide a broader perspective.

Studies were included if quantitative data related to routine records of EAS (actual/requested), patients’ ratings (but not solely their caregivers as a proxy measure), or the accounts of involved healthcare professionals or carers. We included studies regardless of whether they used validated measures of the above constructs, and were interested in any effect measures reported.

We excluded studies that were not written in English, not empirical research, and not published in a peer-reviewed journal; excluding books, book chapters, reviews, abstracts, posters, dissertations and editorials. We also excluded studies considering passive euthanasia only, assuming that in such cases a patient would be too incapacitated to reflect on measures of exposure/outcome.

Screening

One researcher (EC) conducted an initial screen of all titles and abstracts, and a second researcher (MB) conducted an independent screening of a random sample of 20% of citations. Any disagreements were discussed with the wider team to resolve inclusion/exclusion. One researcher (EC) completed a full text screen of all papers, and a second researcher (MB) conducted an independent screen of a randomly selected 20% of full texts. Two other researchers (AP; RB) then conducted independent checks of all included full texts. As with the initial screen, disagreements were discussed within the review team to reach consensus.

Data extraction

One researcher (EC) extracted from each study details of sample demographics, details of key measures, analytic approach, findings and interpretations, tabulating these separately for each of the three outcomes. Two researchers (AP; RB) reviewed the data extracted for 100% of the studies. Disagreements were discussed within the review team to reach consensus.

Quality assessment

We conducted quality appraisal of all included studies using an adapted version of the Quality Assessment Tool for Systematic Reviews of Observational Studies (QATSO) [49]. As per a previous systematic review on the topic of loneliness [40] we expanded the original scoring of 0–5 to 0–9, with higher scores reflecting higher quality (Appendix 2). We rated included studies on the quality of design, response rate, whether validated exposure/outcome measures were used, and whether the analysis controlled adequately for potential confounders. Where studies did not state their threshold for statistical significance, we used a threshold of p < 0.05. One researcher (EC) rated quality for all papers. The quality of each paper was independently rated by at least two researchers (EC; MB; AP; RB). Any disagreements over study quality were discussed within the team to reach consensus. No studies were excluded from the review based on quality, but all findings were presented in the context of study quality.

The certainty of the evidence available was assessed using the Grading of Recommendations, Assessment, Development, and Evaluations (GRADE) scoring system [50] with GRADE criteria adapted for a narrative synthesis approach, according to Murad and colleagues [51] (see Appendix 3 for GRADE scoring criteria). Two researchers (AP; NA) independently assessed the certainty of evidence for each research question and met to resolve any inconsistencies.

Synthesis of findings

We tabulated details extracted from eligible studies, regardless of quality rating, by category of social connectedness measure. We created a table each for our three outcomes: requested/actual EAS; attitudes to EAS; DHD.

We conducted a narrative synthesis of the findings of those studies rated at a low or medium risk of bias, categorised by our three outcomes. We also noted how these findings related to legality of EAS in country/state sampled.

Results

Our initial database searches identified 2,757 citations, reduced to 2,253 following de-duplication. After title and abstract screening, 112 citations were judged potentially relevant. Following full text screening, we identified 30 studies eligible for inclusion. Google Scholar searches identified a further five eligible studies, resulting in a total of 35 papers. Our July 2021 search identified one further eligible paper, which was included. Our November 2022 search identified four potentially relevant studies, but all were excluded on full text screening. Throughout these searches one relevant study was identified through reference-searching and was included. A total of 37 eligible studies were included in the review (Fig. 1), accounting for 17,359 participants (range 6 to 7,534).

Fig. 1
figure 1

Flow chart of included studies

Full size image

Characteristics of studies

We categorised the 37 studies by our three outcomes: requested/actual EAS (Table 1; n = 9), attitudes to EAS (Table 2; n = 16), and DHD (Table 3; n = 14), with limited overlap between categories due to use of more than one category of outcome measure. Within each of these tables we sub-classified studies by type of social connectedness exposure measure. We did not conduct a meta-analysis due to heterogeneity of the range of measures and methodologies used in identified studies, as described below.

Table 1 Summary of study characteristics and findings for requested/actual EAS (n = 9) 
Full size table
Table 2 Summary of study characteristics and findings for attitudes to EAS (n = 16) 
Full size table
Table 3 Summary of study characteristics and findings for DHD (n = 14)
Full size table

The majority of studies were cross-sectional (76%; 28/37). Of the 24% (9/37) that used longitudinal approaches, one involved repeated cross-sectional surveys of three consecutive representative samples of older adults, exploring societal changes in attitudes [63]. Across all included studies, the gender distribution was 50% female, and the mean age was 63 years. Included studies sampled populations in Australia (n = 2), Austria (n = 1), Canada (n = 4), France (n = 1), Germany (n = 2), the Netherlands (n = 3), Switzerland (n = 1), New Zealand (n = 1), the United Kingdom (UK) (n = 1), and selected states in the United States of American (USA) (n = 24). Figures do not add to 37 as some studies were conducted across multiple locations.

We ascertained whether EAS was legal in the country/state of sampling for 13/27 (89%) of studies, whilst acknowledging that people do travel abroad to enact EAS. Those sampling in countries where EAS was legal were set in Germany, the Netherlands, Switzerland, Canada, and selected US states (Oregon, Washington). The remaining four studies were conducted across multiple locations in which EAS legislation differed [66, 67, 69, 75].

The majority of studies surveyed patients with terminal illnesses or who were already in the process of accessing EAS (n = 31). Other studies, primarily assessing attitudes to EAS, sampled people not defined by a physical health condition or by a request for EAS (n = 6). Recruitment methods included: newsletter advertisements; direct recruitment from medical centres, euthanasia organisations, palliative care units or research registers; and referrals from healthcare professionals. Data were collected using survey methods (relatives; clinical staff; patients, with or without carer support), by interview or questionnaire; as well as using data from EAS application forms; and routine data from clinical or audit records within services. Sample sizes ranged from n = 6 to n = 7,534.

Exposure and outcome measurement

Overall, only 16% (6/37) of studies used validated measures for both exposure and outcomes, and 24% (9/37) used unvalidated measures for both (Tables 1– 3). The most common pattern was for use of a validated exposure measure with an unvalidated outcome measure (38%; 14/37), and 14% (5/37) used an unvalidated exposure measure and a validated outcomes measure. Three studies (8%) used a mixture of validated and unvalidated measures for either exposure or outcome. These methodological issues influenced judgements over study quality (Table 4).

Table 4 Quality appraisal of included studies (n = 37)
Full size table

Exposure measurement

To measure exposure, 24/37 studies (65%) used validated measures of social connectedness. The majority of studies (68%, 25/37) measured social support, whilst 11% (4/37) measured loneliness, and 19% (7/37) used other measures (i.e., social isolation, thwarted belongingness, satisfaction with relationships). Measures used to capture social support included the Social Support Questionnaire (SSQ) [86], the Social Networks in Adult Life Survey [87], the Duke University of North Carolina Functional Social Support Questionnaire (FSSQ) [88], the Medical Outcomes Study – Social Support Survey (MOS-SSS) [89], the Family Relationships Index (FRI) [90], the Coping Orientations to Problems Experienced Scale (COPE) [91], and the Multidimensional Scale of Perceived Social Support (MSPSS) [92]. Measures used to capture loneliness included the De Jong Gierveld Loneliness Scale [93] and the UCLA Loneliness Scale [94]. Some studies used instruments capturing social connectedness that had been specifically validated in palliative care populations, such as the State-of-Suffering V (SOS-V) [54], whilst others used specific items from validated measures e.g., the Quality of Death and Dying Questionnaire (QODD) [95].

Outcome measurement

Studies measured outcomes in three main ways:

Requested/actual EAS; measured using routine clinical records, such as GP, palliative care or physician records, or registers of those contacting palliative care services or voluntary sector organisations (representing those with terminal illness or advocating for EAS) seeking information on how to make a request. Death by EAS was measured objectively using records in hospices, medical centres, and voluntary sector organisations registering deaths by EAS. Four studies captured request for EAS [53, 55, 56, 60]. A further four studies included not only requests but also cases where a request for EAS had ended in death by EAS [11, 52, 58]. One study involved patients ranking lack of support in relation to other reasons for having requested PAS [59].

Attitudes towards EAS; captured using a range of unvalidated measures, including using standardised questions taken from national surveys or from previous research studies, justified based on face validity. Agreement with these measures was elicited from patients using questionnaires or interviews. In some cases, attitudes to EAS were elicited from caregivers to capture their views on EAS, whether they would support or oppose the patient’s decision to seek EAS, the perceived likelihood that the patient would request EAS and whether the patient had discussed EAS with them or hoarded drugs. Seventeen studies captured attitudes towards EAS [13, 59, 61, 63,64,65,66,67,68,69,70,71,72,73,74,75,76].

Desire for hastened death (DHD); measured using the Schedule of Attitudes Towards Hastened Death (SAHD) [83], or the Desire for Death Rating Scale (DDRS) [79], and in some studies using unvalidated measures of DHD. The SAHD is a self-report inventory and has been validated formally [83, 96]. The DDRS is a clinician-rated measure and constitutes a set of questions created for a study published in 1995 [79], since used widely as a set of standardised questions, but not formally validated. Nevertheless, the DDRS is highly correlated with the SAHD [83]. Fifteen studies captured DHD [22, 23, 57, 59, 62, 74, 77,78,79,80,81,82,83,84,85].

Methodological quality

We judged 19% (7/37) of studies to be at low risk of bias, 46% (17/37) at medium risk, and 35% (13/37) at high risk (Table 4). The main limitations identified included use of purposive or opportunistic sampling and/or unvalidated measures, lack of adjustment for key confounders, and uncorrected multiple testing. Only ten studies (27%) used representative sampling, and 21 (57%) used multivariable regression modelling. The majority (29/37; 78%) reported their response rate. This varied substantially depending on the methodology used (median response rate = 50%, IQR = 38%-79%).

Association between social connectedness measures requested/actual EAS (n = 9)

Overall, nine studies investigated associations between social connectedness and requested/actual EAS, of which four were rated as having a medium risk of bias and five at a high risk of bias. Focussing on those rated at medium risk of bias, one Dutch study of applications to an end-of-life clinic reported an association between loneliness and a request for EAS being rejected, although loneliness was one of at least 20 sociodemographic and clinical variables tested (Snijdewind et al., 2015) [11]. All other studies reported the absence of a significant association. A study in a Dutch sample of patients with terminal cancer found no associations between loneliness, unsatisfactory social contacts, insufficient support or subjective isolation and a request for EAS (Ruijs et al., 2014) [53]). A US study of patients with terminal illness recruited from end-of-life advocacy organisations and hospices in Oregon State found no association between social support and a request for EAS (Smith et al., 2015) [60]). A US study of the relatives of patients in Oregon State with terminal illness recruited from health services, amyotrophic lateral sclerosis organisations, and euthanasia advocacy organisations found no association between three social connectedness measures and requested/actual EAS (Smith et al. (2011) [52]. Note that EAS was legal in all the above settings at the time of sampling. Together these studies do not support an association between social connectedness and requested or actual EAS due to the potential for Type I error in the only study reporting an association with loneliness, as well as other methodological issues.

Our GRADE rating denoted low certainty of this evidence for the outcome of requested/actual EAS.

Association between social connectedness measures and attitudes to EAS (n = 16)

Overall, 16 studies investigated associations between social connectedness and attitudes to EAS, of which one was rated as having a low risk of bias, ten at a medium risk of bias, and five at a high risk of bias.

Focussing on the only study rated as low risk of bias, which sampled Austrian care-dependent adults, this reported an association between a crude measure of social isolation (living alone) and reported approval of availability of euthanasia as well as of hypothetical utilisation of euthanasia but it was not associated with approval of availability of assisted suicide or hypothetical utilisation of assisted suicide (Stolz et al., 2017) [65]. Thwarted belongingness was not associated with any of these four outcomes (Stolz et al., 2017) [65]. Whilst this study was rated as good quality methodologically, there was a clear issue of generalisability in that those sampled were community-dwelling care-dependent adults over 50 years of age, defined by having a physical or mental disability that was expected to last at least 6 months. In not being defined by a life-limiting illness, their attitudes to EAS were elicited in a situation in which many would not be considering EAS. Additionally, EAS was illegal in Austria at that time.

Focussing on the ten studies in the studies rated at medium risk of bias, in a US study of patients with multiple sclerosis in the states of Oregon (where EAS was legal) and Michigan (where EAS was illegal), poor social support was negatively significantly associated with having ever thought about assisted suicide but there was no evidence to support an association with social isolation (Berkman et al., 1999) [67]. However, in a US-wide study of patients with multiple sclerosis, there was no association between perceived social support and attitudes to EAS Marrie et al. (2017) [75]. In a study of patients with amyotrophic lateral sclerosis in Canada, the US, and the UK (in the context of EAS being legal in Canada and two US states) there was a significant negative correlation between availability of social support and whether patients were contemplating EAS (Achille & Ogloff, 2004) [68]). However, a German study of patients with amyotrophic lateral sclerosis found no association between perceived social support and attitudes to EAS (Lulé et al., 2014) [74], and a US study sampling patients with amyotrophic lateral sclerosis in Oregon and Washington found no significant differences in social support between patients who would and would not consider EAS (Ganzini et al., 1998) [72]. In both studies EAS was legal for those sampled. In a US-wide study of patients with terminal illness no association was found between social support and whether patients had discussed EAS or hoarded drugs (Emanuel et al. (2000) [13]. In a US study of patients in New York state with HIV, there was a significant association between two dimensions of social support and attitude to EAS (Breitbart et al. (1996) [71]. However, this had issues of generalisability because attitudes to death among HIV-infected patients sampled in the mid-1990s would have been conditioned by the public understanding of HIV prognosis at that time.

A number of studies in the medium risk of bias category sampled people who were not defined by life-limiting illness. This means that findings (all found no evidence to support an association) would not be generalisable to those with terminal illness. One Dutch study of older adults (not defined by life-limiting illness) found no association between loneliness and attitudes to EAS (Buiting et al., 2012) [63]. A US study sampling older adults in Indiana state (also not defined by life-limiting illness) found no association between loneliness or perceived social support and attitudes to EAS (Cicirelli, 1997) [64]. Similarly, a US study sampling older adults (with no apparent life-limiting illness) in New Jersey, New York, and Pennsylvania states found no association found between social support and attitudes to EAS (Himchak, 1997) [73].

Together these studies did not provide good evidence to support an association between social connectedness and attitudes to EAS given issues of generalisability, methodology, and the general issue of the validity of attitude measurement as a proxy for EAS probability.

Our GRADE rating denoted very low certainty of this evidence for the outcome of attitudes to EAS.

Association between social connectedness measures and desire for hastened death (n = 14)

Overall, 14 studies investigated associations between social connectedness and DHD, of which six were rated as having a low risk of bias, four at a medium risk of bias, and four at a high risk of bias.

Most of the six studies rated at low risk of bias related to patients with advanced cancer. In an Australian study of patients with terminal cancer, the three dimensions of social support investigated (family cohesion; number of social supports, satisfaction with social supports) were associated with DHD (Kelly et al., 2003) [80]. However, in three US studies of patients with terminal cancer in New York state there was no evidence to support a significant association between perceived social support and DHD (Breitbart et al., 2000) [22] and no association found between social support and DHD (Rosenfeld et al., 2000) [83] (Rosenfeld et al., 2014) [85]. Similarly, in a Canadian study of patients with metastatic cancer, social support was not associated with DHD (Rodin et al. (2007) [82]). Finally, in a US study sampling patients with advanced AIDS in New York State, social support was not associated with DHD (Rosenfeld et al. (2006) [84]. Note that of these studies, EAS was legal only in Canada.

Of the four studies rated at medium risk of bias, two were US analyses of the same sample of patients with terminal cancer in New York State, which presented weak evidence to support an association between poor social support and DHD (O’Mahony et al., 2005) [23] (O’Mahony et al., 2010) [81], and weak evidence to support an association between living alone and DHD (O’Mahony et al., 2005) [23]. The other two studies in this category reported no significant association. A German study of patients with amyotrophic lateral sclerosis found no association between perceived social support and DHD (Lulé et al. (2014) [74]. In a Canadian study of patients with terminal illness in hospital palliative care units there was no association between perceived social support and DHD (Chochinov et al., 1995) [79]. Of these studies, EAS was legal only in Germany and Canada.

Together these studies, did not provide good evidence to support an association between social connectedness and DHD given contradictory findings and issues of methodology. DHD is also a poor proxy for EAS probability, with issues over comparability of measurement in areas where EAS is and is not legal.

Our GRADE rating denoted moderate certainty of this evidence for the outcome of DHD.

Discussion

Main findings

Our review identified a body of literature that investigated the associations between a broad range of social connectedness measures and those capturing actual or requested EAS (our main research question), as well as those more distally related to it (attitudes to EAS; DHD). Generally, there was no evidence to support an association between different constructs of social connectedness and requested or actual EAS, and very little evidence to support an association with attitudes to EAS or an association with DHD. The quality of studies was poor, with 62% rated as medium/high risk of bias, with GRADE ratings indicating that the certainty of recommendations varied from very low to moderate. No studies related specifically to psychiatric euthanasia, which is striking given the high prevalence of loneliness in such cases [13]. No studies related to EAS in the context of intellectual disability and/or autism, in which the prevalence of loneliness is also high [97]. It is possible that the nature of any associations between social connectedness and EAS in the context of severe psychiatric disorder, intellectual disability and/or autism will differ from those in the context of terminal physical illness.

Comparing our findings to a previous systematic review focussed on older adults [46], both their review and ours concluded that there was no or little evidence to support an association between loneliness or low satisfaction with family relationships with request for or attitudes to EAS. In our review, hopelessness and depression were risk factors supported by more consistent evidence for an association with DHD and attitudes and EAS, even though this was not an inclusion criterion. The seven studies rated at low risk of bias in this review (albeit all cross-sectional) identified hopelessness and/or depression to be associated with DHD [22, 80, 82,83,84], or with attitudes to EAS [75], apart from one, which found no association of depressed affect with positive attitudes to EAS [65].

Strengths and limitations

To our knowledge, this is the first systematic review of published studies describing the association between social connectedness and measures relating to EAS across all age groups, using a broad range of concepts of social connectedness and EAS-related outcomes. Data extraction was conducted and reviewed independently by four authors, achieving a high level of rigor in distilling key findings and characteristics. Use of the GRADE framework to evaluate the strength of evidence for each outcome added to the robustness of our conclusions. Clear sub-categorisation was used to address substantial heterogeneity across exposure/outcome measures. We used an accepted quality appraisal tool, the (QATSO) [49], to delineate the limitations of the current evidence base. Our synthesis of findings focussed on higher quality studies, so that conclusions were based on those with lower risk of bias. We acknowledged that participants who had requested EAS (rather than enacted EAS) may not be representative of those who die by EAS, and that DHD may not correlate well with enacted EAS, which is why these outcomes were grouped separately.

The main limitation of our review is that the majority of studies included in the review had methodological limitations, mainly due to unvalidated exposure/outcome measures, cross-sectional design, unadjusted analyses, and multiple testing. Some studies sampled populations without a life-limiting illness, or in countries where EAS was illegal, or included relatives’ views. None of the longitudinal studies had sufficient statistical power to analyse fatal outcomes, and any fatal outcomes were included within combined measures of requests (enacted and non-enacted) for EAS. However, conducting larger longitudinal studies may not add much to knowledge given the difficulties of following up patients who had made EAS requests longitudinally. Finally, our strategy did not include identification of grey literature, and there also remains the possibility of publication bias in under-representing unpublished studies reporting no statistically significant associations [98].

Clinical, policy and research implications

The lack of evidence to support an association between social connectedness and requested/actual EAS and related measures suggests that other factors may dominate end-of-life decision-making processes. Patients who seek EAS experience intolerable suffering that cannot be relieved. Addressing potentially modifiable factors such as loneliness and poor perceived social support would improve quality of life, as would addressing depression and hopelessness (which are clinical factors likely to make a more substantial contribution to requests for EAS). Given the contribution of loneliness to depression in older people [99] and the bidirectional association between social disconnectedness and poor mental health [100, 101], loneliness and depression are likely to be closely interlinked. Whilst depression (and hopelessness, as a symptom of depression) are modifiable using evidence-based treatments, loneliness is also potentially modifiable, whether in general population samples [102] or among those with pre-existing mental illness [103]. This is a growing research field, with great interest in the development and evaluation of interventions to address loneliness [102,103,104]. Intervening in this way might be one means of preventing the onset of depression that might otherwise lead to DHD and/or EAS.

The reasons given by patients for requesting EAS, including perceived poor quality of life and fear of future suffering, might be allayed by Advance Care Planning; a process for discussing their preferences and priorities for their future care, including situations in which they might refuse treatment (or request EAS, where legal). In the context of serious physical illness Advance Care Planning should start by checking mental capacity and follow local clinical guidelines [105]. In the context of psychiatric disorder, Advance Care Planning can include discussions about psychiatric admission, use of electroconvulsive therapy, and refusal of treatment, but there is a comparative lack of clinical guidelines to inform this process [106]. Family involvement is very important in this process [107], but for patients who feel less socially connected this may be problematic.

Given limited support for an association between social connectedness and EAS, the rationale for inquiring about social connectedness in patient populations who might consider EAS is therefore that intervening is likely to improve quality of life. Clinical guidance on the assessment and management of patients with terminal illness, severe physical disability and severe or enduring mental distress, including for those who request EAS, should therefore include inquiring about loneliness, isolation and social contacts alongside standard questions about pain and untreated mental or physical health problems. Information on social connectedness would provide valuable opportunities for clinical and voluntary services serving such populations to consider what support might be indicated to reduce distress.

In view of the methodological limitations of many included studies, we recommend that future research studies employ representative sampling and use of validated measures and appropriate statistical models. We also acknowledge the limited applicability of existing theoretical models of suicide to the phenomenon of EAS and note the need for a conceptual review to inform appropriate iterations of such models.

Conclusions

This review found no evidence to support an association between social connectedness and requested/actual EAS, and weak evidence to support an association between social connectedness and related measures (attitudes to EAS; desire for hastened death). The strength of the evidence was generally low, mainly due to high risk of bias, and there is a need for further high-quality research investigating these associations. Nevertheless, modifying any distressing aspects of poor social connectedness could improve quality of life in those who experience intolerable suffering due to physical or mental illness that cannot be relieved. The evidence base for interventions to address loneliness and social isolation is growing, and more work is needed to develop and evaluate effective interventions to target these in a range of settings, including end-of-life care.