The association between social relationships and self-harm: a case–control study in Taiwan

Self-harm is a major public health issue which is substantially more common than completed suicide and a high-risk group for repetition [1‐3]. Despite its relatively low prevalence (i.e. 150-500/100,000 [4‐6]), self-harm with repetition had significantly raised risk for suicide [2]. Defined as an act of self-injury or self-poisoning with non-fatal outcomes regardless of the intention to die [7], self-harm is a construct differing from completed suicide in terms of the underlying intention of death, the combination of different methods and their lethality, and the prior degree of hopelessness, amongst other distinctions [8]. Suicide on the other hand is a fatal self-inflicted act before which a person may have carried out one or more self-harm acts [9]. Previous research has suggested that the pathways between social rejection or other stress and suicidal ideation may be moderated by social support; however, more understanding is needed of the role of these potentially protective factors [10‐13].

Social support, defined in terms of the quality of perceived support from social relationships, was found to confer protection through help-seeking augmentation [14], and was postulated as a rescue factor which moderates stress and suicidal behaviour [12]. However, it has received less research compared to other etiological factors in relation to self-harm. More comprehensive measures of social relationships have been suggested as important if for clarifying the role of social support and social isolation in the complex phenomenon of self-harm [12]. Social support as a broad entity has been defined as including functional (e.g. information, instrumental, practical, emotional support) and structural (e.g. social network, network outside the family, household size) components of support derived from social networks [15]. Measurement strategies seek to characterize both quality and quantity of social support through which social networks may promote health and well-being [15‐17]. Although suicide has been postulated as a result of a breakdown in social ties [17], little attention has been paid to social relationships among cases of self-harm. Adequate social support has been found to modify help-seeking behaviour in these groups [14], to regulate personal response to stressors and prevent anxiety or depression [18, 19], and to have both direct effects on mental health as well as modifying the negative impact of life events [20, 21]. In addition, social networks and interactions in social contacts may affect social functioning and thus the severity of mental illness [22, 23]. The degree of help-seeking as well as social relationships has been found to differ between men and women [24]. Thus, we hypothesized that the level of social support and networks would be lower among people with self-harm compared to a control group, and such association would differ by gender after controlling for physical and psychosocial confounders.

Using a sample of cases of self-harm in Taiwan, we sought to investigate the extent to which social relationships associate with self-harm under adjustment for factors recognised to be associated with suicidal behavior [11], including sociodemographic factors and physical and mental health.

The physical, psychosocial and demographic characteristics of the 124 cases and 124 age- and gender-matched controls were collected during a one year period and are described in Table 1. The mean age of the cases was 34.7 with a standard deviation of 12.8, and the male-to-female gender ratio was approximately 1:4 (80.6% females). The control sample was matched effectively on these factors. Among the case group, self-poisoning had been carried out by 73.7% and self-injury by 28.3%. The cases were characterized as having a higher proportion of participants with divorced, separated or widowed status of marriage, shorter duration of education, and higher prevalence of unemployment. In terms of social support and network levels, cases had significantly lower levels of self-reported confiding support, practical support, smaller networks beyond the household, and their level of isolation was higher than the controls.

In Table 2, associations between z-scored social relationship measures and case/control status were compared using conditional logistic regression models. After adjustment, a strong association persisted between case status and the standardized score of isolation (OR = 2.92; 95% CI: 1.44, 5.95), and a significant negative association was found between self-harm and larger household size (OR = 0.55; 95% CI: 0.32, 0.94). Social support subscale measures were not significantly different between cases and controls after adjustment.

Stratifying primary analyses by gender (Table 3), associations with higher levels of social isolation and smaller household size were only significant in female participants; however, most coefficients did not differ substantially and confidence intervals overlapped, suggesting insufficient evidence for effect modification.

In this sample of people who accessed emergency care after self-harm, social networks and household size were found to be more limited and smaller compared to a hospital control group matched on age and gender. Our findings supported the view that prior structural social support has more influence on self-harm than the functional social support provided by the nominated closest person to the participants [29‐31]. The results implied that the importance of networking from family and friends was more than that of perceived support from the closest person, in relation to self-harm within this East Asian society.

Reasons for social isolation include sociodemographic changes such as ageing and unemployment, loss of a partner, isolated living environment, and societal position [32], and isolation may also be caused by physical and mental disorders or arise after major life events [33]. In this analysis we controlled for a range of potential confounders including physical and mental health. It might be that, given the resource flow and social support derived from the social networks, it was the structural conditions and quantity of support that determined individual adaptive behaviour as suggested by Berkman and colleagues [16]. In other words, the quantity of social relationships might be more influential in self-harm aetiology than the quality of support received through specific networks. It might also be that, within the Taiwan context, alternative social networks may dilute the “no escape” type of thought in the process when one formulates suicidal ideation or self-harm plans [12] or may form a pathway towards support in need for people experiencing suicidal crisis [34]. The results add to the previous model in suggesting that social isolation or rejection could play a major role in the pathway to suicidal risk, with a lesser influence of perceived/ functional social support [12]. However, further research is required to confirm this relationship.

Findings from randomized controlled trials have suggested that increasing personal access to peer support may result in some reduction of self-harm repetition or suicide ideation [35, 36]. For example, short-term periodic telephone contacts and networking from trained family or friends nominated by young women with self-harm and premorbid mood-related symptoms may decrease suicide ideation and attempts [36]. Our observational findings supported this and suggest a need for further evaluation of interventions to improve networking for people at risk of self-harm.

Strengths of our study include the relatively large case and control groups compared to previous studies [37‐39] and a relatively comprehensive social support and network measurement with a multidimensional approach. However, case control studies are recognized to be methodologically challenging and inferences should be drawn with caution. The fact that both case and control participants were sampled from a single general hospital in an urban district of Taipei may limit generalizability and findings require replication in other settings. Hospital controls might also lead to under-estimated associations of interest because of excess morbidity and physical disability associated with smaller social networks. A larger sample size would have increased statistical power to detect differences and negative findings should be interpreted with caution. Further, more specific matching of the control group could have taken place (for example with respect to employment, education, marital status and residence) as a means of adjustment; however, over-matching creates potential problems for case control studies, in addition to reducing the feasibility of a study where there are too many groups requiring individual matching; therefore we took a relatively parsimonious approach and adjusted for other potential confounding factors in regression models. Reverse causality, a related issue, also cannot be excluded in a case control design. For ethical and logistical reasons, cases could not be interviewed immediately after the index self-harm episode and it is possible that a subsequent change in social relationships followed on from the self-harm which was being rated when the CPQ was carried out (despite instructions to the contrary).

The study has implications for mental health practice and policy makers. In an open referral system such as that operating in Taiwan’s health service, social network plays an important potential protective role through the availability of others to support, listen, and respond to people at risk. Network members in the lay or healthcare system may also act to encourage prompt medical review and thus reduce self-harm risk through indirect means. Although interventions have yet to be developed with demonstrable efficacy in relation to preventing self-harm acts [40], recent studies have suggested the need for psychiatric and social approaches to reduce isolation, especially in middle-age [41]. Future studies could explore the needs of isolated groups and pilot strategies to improve living circumstances, as well as investigate the benefits of targeting high risk groups for enhanced social networks. Replication of the findings in this and similar cultural context is also required, as is prospective longitudinal research.

More limited social networks were associated with self-harm after adjustment for potential confounders. Enhanced social structure and raising awareness of networking people with self-harm to community resources may be important for self-harm management in Asian societies and elsewhere.

Written informed consent was obtained from the patient for publication of this report and any accompanying images.

CW is currently a nursing lecturer and epidemiological researcher who has published papers on topics related to suicide prevention and help-seeking behaviour among people with self-harm. SL is a visiting psychiatrist who has broad publications regarding people with major psychiatric disorders including alcohol use disorders, suicidal behaviour and self-harm, depressive and anxiety disorders. HH is the leader of research assistants in the research team and has been involved in psychiatric field studies for more than a decade. CC is an epidemiologist and senior researcher at the Institute of Psychiatry who excels in study design and statistical analysis relating to the field of oncology and psychiatry. RS is Professor of Psychiatric Epidemiology and Clinical Informatics at King’s College London Institute of Psychiatry with interests in late-life mental health and clinical data resources, in addition to psychosomatics including self-harm and suicide.