Introduction

The World Health Organisation (WHO) reports that 800,000 people die by suicide annually worldwide, and that for each of those deaths approximately 20 other people attempt suicide [1]. Amongst long-established risk factors such as mental illness and alcohol misuse, the WHO also hypothesises that a sense of isolation may be a risk factor for suicidality [1]. Whilst social isolation is defined objectively as a lack of social contacts, the concept of loneliness has been defined as the unpleasant experience that occurs when a person’s network of social relationships is deficient either quantitatively or qualitatively [2]; the subjective, unwelcome feeling of a lack or loss of companionship [2]; and as an inner, subjective sense of not being sufficiently socially connected [3]. Loneliness and social isolation are both distinct from the concept of social support; the emotional, tangible (practical), informational, and companionship support arising from either formal (professional) or informal (friends and relatives) sources [4]. Both social isolation and loneliness have recently become the focus of health and social policy research, and the adverse physical and mental health effects of loneliness are becoming increasingly well recognised [5], including the association between loneliness or social isolation and suicidality [6]. This is a particular concern amongst young people, in whom loneliness is prevalent and also stigmatising [5]. The hypothesised links between loneliness and suicidality are that lonely individuals feel dissatisfied with life [7], and that their perceptions of limited social connections contribute to a desire to die [8]. Other mechanisms suggested include the contribution of loneliness to chronic stress, the link with cognitive factors such as low self-esteem or an external locus of control, and the mediating role of depression [6]. Theoretical models of suicidality consider the role of thwarted belongingness; a construct describing what arises when the fundamental need to form and maintain strong, stable interpersonal relationships is unmet, resulting in feelings of disconnection [8, 9]. The Interpersonal Theory of Suicide considers thwarted belongingness as comprising (a) loneliness and (b) the absence of reciprocal caring relationships [8], and confirmatory factor analysis shows convergent associations of thwarted belongingness with loneliness [10]. The Integrated Motivational–Volitional (IMV) model of suicide proposes that factors such as thwarted belongingness and perceived burdensomeness (the view that one’s existence is a burden on friends, family and/or society) govern the transition from suicidal ideation to suicide attempt [11]. Studies conducted to date have found loneliness to be associated with suicidal behaviour in Canadian [12] and English [6] general population samples, even when adjusting for common mental disorder [6], and in specific populations limited to war veterans [13, 14], adolescents [15], and sexual minority groups [16]. However, no studies have investigated this in a bereaved population.

The loneliness arising from sudden bereavement is a topic worthy of further investigation because sudden bereavement is itself a risk factor for suicide [17], particularly where the death was due to suicide [18, 19]. Although the mediators of suicide risk after sudden bereavement remain unidentified, factors such as loneliness, social isolation, and poor social support are implicated [20]. Loneliness after sudden loss may arise from the abrupt loss of a confiding relationship or companionship. The taboo around sudden or violent losses can also engender social isolation due to others’ awkwardness and avoidance [21]. The perceived stigma of sudden bereavement, defined as the subjective awareness of others’ negative attitudes towards the loss [22], is likely to be a major contributor to loneliness after loss. This stigma arises not only from others’ avoidance but also from a perception of them gossiping about the death or casting blame [23]. Although stigma is perceived after all sudden deaths, most commonly due to embarrassment or a fear of appearing socially incompetent, it is particularly pronounced after deaths by suicide and other unnatural causes [24, 25]. This is likely to be due to the long history of religious and cultural sanctions against suicide, stigmatising the act and those mourning it [26], and others’ views about irresponsibility in the case of accidental death [21]. The perceived stigma of sudden bereavement is associated with risk of suicidal thoughts and suicide attempts [22], and there is also evidence that perceived stigma contributes to suicide attempt risk after suicide bereavement [25].

In addition to suicidality, unexpected bereavement is also associated with a range of incident psychiatric disorders [27], and it is possible that these might lead to or compound loneliness. Not all individuals will be adversely affected, and grief is a non-pathological process. However, risk of poor outcomes is likely to be governed by factors such as personal vulnerability [28], the quality of attachment to the person who died [29], and the degree to which social support buffers the negative impact of a traumatic life event [30]. Further work is needed to describe the relationships between loneliness, social isolation and suicidality after a sudden loss to further our understanding of the roles that social isolation and loneliness might play in the pathway to suicidality. No previous studies have measured the association between loneliness after the sudden loss of a close friend or relative and suicide-related outcomes. We aimed to investigate whether greater loneliness was associated with a higher probability of suicide attempt and suicidal ideation after sudden bereavement in a population-based sample of adults in the United Kingdom. We also aimed to test the hypothesis that associations differ by mode of bereavement (suicide bereavement versus non-suicide bereavement), such that the magnitude of the association would be greater in the suicide-bereaved, and that the main association might partly be accounted for by the perceived stigma of the bereavement.

Methods and materials

Study design

We analysed cross-sectional data collected in the UCL Bereavement Study: a 2010 survey of all staff and students aged 18–40 at 37 UK higher educational institutions (HEIs) who had experienced the sudden bereavement of a close friend or relative since reaching the age of 10. Full details of sampling for this closed online survey have been described elsewhere [22], including the survey instrument [25]. The focus was on young adults given policy concerns about their risk of suicide [31] and their tendency not to engage with services when in distress [32,33,34]. Sampling via global email lists avoided the biases associated with recruiting a help-seeking sample, and was felt to be the most efficient, comprehensive and pragmatic means of recruiting a hard-to-reach population of young adults [35].

Of 5085 respondents to the survey, we included those who: consented to participate; completed the loneliness item of the Social Functioning Questionnaire (SFQ) and all seven items within a scale measuring perceptions of social support received from family and friends; and specified their mode of bereavement (n = 3193). This sample included respondents who had experienced sudden bereavement due to sudden natural causes (n = 1952), sudden unnatural causes (n = 666), or suicide (n = 575).

Ethical approval

Ethical approval for analysing data from the UCL Bereavement Study was obtained from the UCL Research Ethics Committee in 2010 (reference number: 1975/002).

Measures

Exposure: loneliness

Our exposure was loneliness using eight items chosen to represent the construct of loneliness, as used in the seven-yearly Adult Psychiatric Morbidity Surveys (APMS), a regular representative population survey in England [36]. One item was taken from the SFQ [37]; an eight-item self-report scale used to measure current social function. Seven items were taken from a scale measuring perceptions of social support received from family and friends, used in the 1987 Health and Lifestyle survey in England [38]. These capture respondents’ sense of whether relatives and friends can: be relied upon; make them feel loved, valued and accepted; and support and encourage them. The items therefore convey the degree to which one’s relationships create a perception of social belonging and acceptance as well as of emotional connectedness, and companionship. By reverse scoring them in this loneliness measure, higher scores captured the perception of lacking belongingness, acceptance, connectedness and companionship in one’s social relationships.

The development of this continuous measure, scored from 0 to 17, is described in “Appendix”.

Outcome: suicide attempt

Our outcomes were: self-reported suicide attempt (“Have you ever made an attempt to take your life, by taking an overdose of tablets or in some other way?”) [39] and self-reported suicidal ideation (“Have you ever thought of taking your life, even though you would not actually do it?”) [40] post-bereavement. These standardised measures were again taken from the APMS [36], qualified by whether each was before or after the sudden bereavement, or both, to derive an incident measure.

Covariates

We also measured the following covariates:

  • socio-economic status, using the UK Office for National Statistics Standard Occupational Classification [41];

  • depression, using the Composite International Diagnostic Interview (CIDI) screen for lifetime depression [42], qualified by whether the onset of these core symptoms of low mood and anhedonia occurred before or after the sudden bereavement, to derive an incident measure

  • possible personality disorder, using the 8-item self-report Standardised Assessment of Personality-Abbreviated Scale (SAPAS), a standardised instrument with demonstrated reliability and validity in screening for likely personality disorder in psychiatric out-patient samples [43, 44]

  • perceived stigma of the sudden loss, measured using the stigmatisation sub-scale of the validated Grief Experience Questionnaire (GEQ) [26]

  • network size, using a continuous measure of primary group size derived from the Interview Measure for Social Relationships (IMSR); a standardised instrument with demonstrated reliability [45], as used in the APMS [36] with waves of population norms [46]. This captures the numbers of friends and relatives (aged 16 years and over) respondents felt close to, sub-categorised into numbers of close relatives, extended family members, and friends.

We selected nine potential confounding variables on the basis of existing literature and clinical judgement: age, gender, socio-economic status, pre-bereavement depression, pre-bereavement (suicidal and non-suicidal) self-harm; and primary group size.

Missing data for model covariates and outcomes varied from < 1% (for age, gender, group size, pre-bereavement depression, pre-bereavement self-harm, pre-bereavement suicide attempt, and pre-bereavement suicidal thoughts) to 3% (for socio-economic status).

Statistical analysis

We described median values (and interquartile ranges) or means (and standard deviations) for our exposure variable and each descriptive variable, and reported p values for tests of univariate associations with each using linear regression. Our threshold for significance was set at p < 0.05.

We investigated the relationship between loneliness and outcomes using multilevel regression models with HEI as random effect, to take into account the clustering effect at the HEI level. We used multivariable logistic regression to investigate the relationship between loneliness (continuous measure) and two binary outcomes (post-bereavement suicide attempt, post-bereavement suicidal ideation), adjusting for socio-demographic factors, pre-bereavement psychopathology, and primary group size, as described above.

To test whether the effect of loneliness on outcomes varied by mode of bereavement (suicide versus non-suicide loss), we added this binary variable as an interaction term to adjusted models, using a less stringent p value threshold (p < 0.1) to reflect the limited statistical power of interaction tests. We also used multivariable logistic regression to investigate whether suicide bereavement is associated with post-bereavement loneliness, adjusting for the same potential confounders as in the list above.

Finally, we tested whether adding the perceived stigma of bereavement to our main adjusted models attenuated the association between loneliness and suicide-related outcomes, to provide evidence supporting a partially mediating effect. To set this in context, we also described the association between loneliness and perceived stigma.

Models were fitted using complete case analysis. We ran a priori defined sensitivity analyses to assess the robustness of our main findings when taking into account biases introduced by 3% missing data on socio-economic status, using best-case and worst-case scenarios to impute missing values.

All analyses were conducted using Stata version 13 (StataCorp. 2013. Stata Statistical Software: Release 13. College Station, TX: StataCorp LP).

Results

Participant characteristics

The majority of our sample of 3193 adults were female (81%), of white ethnicity (90%), bereaved by sudden natural causes (61%), and reporting the death of a relative (71%) (Table 1). The sample was relatively evenly split between participants aged 18–21 years (41%) and those aged 22–40 years (59%). Mean age of respondents was 25·1 years (standard deviation [SD] 6·3, whilst mean age at bereavement was 20.1 years (SD 6.04; median 19; interquartile range [IQR] 16–23). Mean time elapsed since bereavement was 5 years (SD 5·3; range 1 day–30 years). The age of the deceased varied from 0 (for miscarriage or stillbirth) to 100 years, with a median age of 47 years (IQR 23–63). In the total sample mean loneliness scores were 2.49 (SD 2.70; median 1; IQR 1–3). Median network size in this population was 11 (IQR 7–16; mean 12.9; SD 9.1). The proportion scoring positive for possible personality disorder was 37%, ranging from 22% reporting “difficulty making and keeping friends” and 22% describing themselves as a “loner”, to 63% reporting being “a perfectionist” and 72% as “a worrier”.

Table 1 Univariate associations of loneliness scores (continuous variable) with participant characteristics (n = 3193)
Full size table

Univariate associations of loneliness with socio-demographic and clinical characteristics, and measures of social connectedness, are shown in Table 1. Loneliness scores were significantly higher in men, those who defined themselves as single or non-white, those classified as of lower socio-economic status, those reporting depression or self-harm prior to bereavement, those reporting post-bereavement suicidal thoughts or attempts, those screening positive for possible personality disorder, and those with greater perceived stigma scores.

Association between loneliness scores and outcomes

In an adjusted analysis (Table 2), higher loneliness scores were associated with a significantly higher probability of post-bereavement suicide attempts (AOR 1.19; 95% CI 1.14–1.25) and of post-bereavement suicidal ideation (AOR 1.24; 95% CI 1.20–1.28). Stepped adjustments are reported in Supplementary Table 3.

Table 2 Estimates of the associations between loneliness scores and outcomes in bereaved participants (n = 3193)
Full size table

There was no evidence to support an interaction of mode of bereavement with suicide attempt (p = 0.7211) or suicidal ideation (p = 0.6343). Associations with suicide attempt were similar whether someone had been bereaved by suicide (AOR 1.21; 95% CI 1.10–1.33; p = 0.005) or by non-suicide causes (AOR 1.19; 95% CI 1.12–1.25; p < 0.001). Similarly, associations with suicidal ideation were similar whether someone had been bereaved by suicide (AOR 1.26; 95% CI 1.16–1.37; p < 0.001) or non-suicide causes (AOR 1.23; 95% CI 1.19–1.28; p < 0.001). There was no significant association between suicide bereavement and loneliness (adjusted coefficient 0.22; p = 0.063; 95% CI − 0.12 to 0.45), when adjusted for the same nine covariates included in the main model.

There was a significant association between loneliness and perceived stigma of bereavement, in an unadjusted model [unadjusted coefficient 0.574 (95% CI 0.525–0.623; p ≤ 0.001) and when adjusting for the same nine covariates (adjusted coefficient 0.532 (95% CI 0.480–0.583; p ≤ 0.001)].

Adding scores for perceived stigma of sudden bereavement to final models (Table 2) slightly attenuated the association between loneliness and suicide attempt (AOR 1.13; 95% CI 1.08–1.19), and between loneliness and suicidal ideation (AOR 1.18; 95% CI 1.14–1.23), with both associations remaining significant.

We added a post hoc test for an interaction with stigma (dichotomised at the median into high versus low scores) and found no evidence for an interaction with suicide attempt (p = 0.6343) or suicidal ideation (p = 0.2053).

Sensitivity analysis

In sensitivity analyses simulating potential biases introduced by 3% missing data for socio-economic status the magnitude, direction and precision of adjusted odds ratios for the association between loneliness scores and outcomes were unchanged.

Discussion

Main findings

The findings of this analysis of British cross-sectional data support our hypothesis that people who feel lonely after the sudden death of a friend or relative are more likely to experience suicidal thoughts and to make a suicide attempt after the loss, even when adjusting for social network size and for the perceived stigma of the bereavement. The magnitude of our risk estimates were such that the odds of making a suicide attempt or of having suicidal thoughts increased by 19% and 24%, respectively with each point increase on our loneliness measure. Potential explanations for this association include loneliness contributing to substance misuse, to a sense that no-one would understand their problems, and to the intensity of fantasies about reunion with the deceased, otherwise considered a normal grief reaction. The associations remained significant after adding stigma to final models, and we found no interaction with high versus low stigma score. This is surprising because of the observed association between the stigma of sudden bereavement and suicidal ideation and attempt [22]. This suggests that a sense of loneliness is independently associated with probability of suicidality among people who experience sudden bereavement.

Our interaction tests found that feeling lonely had no greater effect on suicidality among people bereaved by suicide than in people bereaved by non-suicide deaths, despite findings from the same sample identifying the suicide-bereaved as having a significantly higher probability of post-bereavement suicide attempt [25]. This again is surprising, and suggests that for the suicide-bereaved other factors relating to suicide loss make additional contributions to their suicide risk. It is plausible that the stigma of suicide may condition the suicide-bereaved to normalise loneliness as an expected consequence of suicide loss. Instead, other factors influencing suicide risk are likely to include psychiatric disorder [18, 19], substance misuse [47], and a reluctance to seek help for mental health problems [48]. Risk of mood disorders and post-traumatic stress disorder are elevated in people bereaved by suicide [19]. People who have experienced suicide bereavement describe binge drinking or drug taking as a coping strategy in the immediate aftermath of the death [47]. They are also less likely to consult their general practitioner than bereaved controls, despite their greater probably of mental illness [19], which may be due to the stigma of suicide loss reducing motivation to seek help [49,50,51,52,53]. Self-stigma is certainly an important factor in dissuading people from seeking help for mental health problems, primarily due to reluctance to disclose a mental health condition [54]. Suicide bereavement is characterised not only by stigma, but also by a sense of responsibility, rejection, and shame when compared to bereaved controls [55]. A more detailed exploration of patterns of help-seeking for suicidality among the suicide-bereaved in relation to shame and stigma may be warranted, as it is possible that a reluctance to seek help accounts for their greater risk of suicide attempt. Finally, it is possible that carer burden prior to suicide loss might give rise to higher levels of pre-bereavement loneliness [56], although this requires specific testing.

In our sample there was a high overall proportion of people with post-bereavement suicidal ideation (47%), greatly exceeding those for the maximum lifetime prevalence of suicidal ideation (20.6%) in any corresponding age group within the APMS 2007 representative sample of the English population [36]. The proportion of those reporting a suicide attempt (7%) in our sample, during a period spanning a mean of 5 years since their loss, was similar to the maximum lifetime prevalence of suicide attempt (7.3%) in any corresponding age group within the same APMS 2007 representative sample [36]. Among a group in which suicidality is commonplace, screening for loneliness may be a useful means of identifying individuals at risk of suicide attempt and initiating a conversation about appropriate ways of alleviating loneliness in the context of that person’s unique circumstances.

Results in the context of other studies

Few other studies have used nationwide population-based samples to investigate the associations of loneliness with suicide-related outcomes. Previous work in Canada found a strong association between loneliness and suicidal ideation (OR 10.5; 95% CI 8.4–13.1) and self-harm (OR 13.5; 95% CI 9.3–19.6) in a representative household sample [12]. However, that study measured loneliness using one unvalidated item capturing responses to the question “How often do you feel alone?”, rated on a 5-point scale from 0 (never) to 4 (very often), and odds ratios were unadjusted [12]. An analysis of English household survey data found that those with higher levels of loneliness were significantly more likely to report past year suicidal ideation (AOR 11.09; 95% CI 6.91–17.79) and past year suicide attempts (AOR 17.37; 95% CI 5.51–54.72) [6]. This analysis adjusted for age, sex, educational qualifications, ethnicity, marital status, wealth, employment status, alcohol dependence, social support, physical health conditions, stressful life events and common mental disorder, and the magnitude of the effect sizes were large. However, confidence intervals were wide, and the analysis was based on the single item from the SFQ, whereas ours used the SFQ item together with seven complementary items capturing perceptions of belongingness, companionship, acceptance, and connectedness.

Strengths and limitations

We analysed data from a large, UK-wide sample of 3193 bereaved adults using a newly-validated loneliness measure with good face and content validity, and internal consistency, albeit without data validating it against an established measures of loneliness. Our hypotheses were formulated on the basis of current theory and clinical experience, and our models were adjusted for pre-selected potential confounders, including pre-bereavement psychopathology. Results were robust to sensitivity analysis simulating the biases potentially introduced by missing data. However, our use of cross-sectional data limits interpretation of the chronology of the pathways between loneliness scores (measured currently) and outcomes (measured at any time after the bereavement), and it is possible that suicidality engenders further loneliness or that the relationships are bidirectional. Our logistic regression did not capture the time between bereavement and suicidality outcome measures, nor whether respondents related their suicidality to the bereavement. As bereavement was as distal as 30 years for some participants, and a mean of 5 years, it is possible that the experience of current loneliness may relate to more recent factors, including current mental illness. It is also possible that suicidality since bereavement was explained by other unmeasured factors. Further longitudinal work is necessary to understand the directionality, and potential reinforcing effect, of these relationships, and how the strength of associations varies over time. Our sample was predominantly white, female and highly-educated, and this limits generalisability of our findings to settings outside HEIs and in other countries.

Clinical and research implications

This study has identified loneliness after a sudden bereavement to be associated with suicidality, whether or not the loss was due to suicide. Our findings suggest a need for agencies that support bereaved people, including counsellors, general practitioners, voluntary sector organisations, and informal networks, to inquire about loneliness and consider ways of addressing this. However, the stigma of admitting to feeling lonely may preclude identification of the problem, particularly in those who wish to mask their sense of isolation. As yet the evidence base for interventions that address loneliness and prevent adverse mental health outcomes is in its infancy. However, systematic reviews of studies evaluating the effectiveness of interventions to address loneliness in people with existing mental health problems [57] and in the general population [58]. These tentatively favour interventions addressing maladaptive cognitions, such as biases in cognition towards negative aspects of the social context. Pending trial evidence in bereaved populations, referral to a peer bereavement support group may be an acceptable means of increasing connectedness among bereaved people, particularly those who feel alienated from their non-bereaved peers. Such groups may provide the only access to people who can understand them, or the only setting in which expressing grief is acceptable [59]. They may also be a setting in which selective attention towards others’ failure to offer support after loss, and other maladaptive cognitions, might be challenged. Even for those who have family and friends around them, peer support groups offer the potential for connectedness with people who have a shared understanding of the experience of bereavement, as well as the opportunity to help others [59].

Longitudinal work to explore the nature of the relationships between loneliness, stigma, social support, and suicidality, would help determine the comparative contribution of loneliness, stigma, and poor informal support as putative mediators of the association between sudden bereavement and suicide-related outcomes. Understanding these relationships would help determine whether there is a need to develop and trial individual-level or community-level anti-stigma interventions among those have been bereaved and who feel lonely and stigmatised. Investigating the relationship between loneliness and hopelessness is also warranted. Theoretical models of the interrelationship between loneliness and suicidal ideation suggest that the effects of loneliness on future suicidality operate through its influence on hopelessness [60]. However, when tested, findings support an alternative model in which hopelessness predicts both loneliness and suicidality, with no direct relationship between loneliness and suicidality beyond hopelessness [60]. Our study did not measure hopelessness, but testing these hypotheses in a bereaved population would contribute further to our understanding of these mechanisms. Finally, further investigation of the relationship between loneliness and entrapment is warranted using longitudinal approaches, as entrapment is a key factor in theoretical models of suicide [61].

Conclusions

Our results confirm an association between feeling lonely after the sudden death of a friend or relative and an increased risk of suicidal thoughts and suicide attempt, which is not explained by a lack of friends or family or by feeling stigmatised by the loss. Our most striking finding is that suicide bereavement is not associated with feeling lonely, and that loneliness had no greater effect in people bereaved by suicide, despite their greater risk of suicide attempt. Further work using longitudinal data is needed to understand the potentially complex associations between loneliness, stigma, mental illness, social support and suicidality in people who experience bereavement. This will help identify how we might mitigate the negative health effects of sudden bereavement, as part of suicide prevention efforts.