Suicide is one of the biggest public health concerns worldwide (Bachmann, 2018). Globally, 800,000 people die from suicide every year, and 1.4% of all premature deaths are suicides (World Health Organization, 2014). This trend has remained consistent over the past 50 years (Franklin et al., 2017), continuously calling upon researchers, practitioners, and policymakers to address preventing suicide and creatively intervening with those who are at risk for suicide. Suicidal thoughts and behaviors (STBs) produce wide and long-lasting effects, constantly increasing stress to mental health systems. The effects of STBs go beyond an individual, with negative psychological, social, and economic cascade effects propagating to surrounding social groups and larger communities and countries. Thus, significant prevention, intervention, and postvention efforts are being developed and tested constantly (e.g., World Health Organization, 2014).

There are many variations within suicidal thoughts and behaviors, including suicide-related thoughts, urges, preparatory acts, or attempts (Klonsky et al., 2016). O'Connor et al. (2013) outline differences within STBs. Suicide-related thoughts (ideation) include having either passive or active thoughts of ending life without preparatory behaviors. Suicide intent (urges) refers to the urge to want to die. Suicide-related thoughts and suicide intent may not always go hand in hand. For example, someone might experience thoughts of ending his/her life but may not want to die. Preparatory behaviors refer to activities involved in planning toward suicide. Last, a suicide attempt is a non-fatal self-directed injury with an intent to die, and suicide is a fatal self-directed injury with the intent to die (Crosby et al., 2011). Each of these STBs reflects different levels of severity (and correspondingly increasingly intensive intervention care) and depending on whether transient or chronic mental health conditions co-occur, the level of intervention also differs. However, a common basis of STBs is the desire to escape from psychological and physical pain.

A closely related, yet distinct, behavior from STBs is the non-suicidal self-injury (NSSI). The NSSI refers to engaging in self-destructive behaviors (e.g., cutting, burning, and punching) without the intention to die (Nock et al., 2006). Approximately 20% of young adults have a history of NSSIs (Swannell et al., 2014), and 70% of adolescents in clinical settings have reported both a history of NSSI and a suicide attempt (Nock et al., 2006). The reasons behind engaging in NSSI vary, but it is often when individuals desire to escape from negative emotional states (Bentley et al., 2014). Generally, no suicidal intent is present when engaging in NSSI, although research has shown that NSSI increases the possibility of suicidal behavior (Andover et al., 2012) such that future suicide attempts were associated with past NSSI (Bryan et al., 2015).

One mechanism by which repeated NSSI increases the risk for suicide attempts might be through a behavioral reinforcement process. That is, repeated NSSI can become a negative reinforcement for individuals by achieving the desired consequences. For example, individuals can avoid negative emotions by paying attention to their physical pain and, thus, use NSSI as a strategy to cope with negative emotions (Joiner, 2005; Hasking et al., 2017). Progressively repeated NSSI will habituate individuals to not fear physical pain, ultimately lowering psychological resistance to engage in lethal self-harm (Joiner, 2005), thereby increasing the likelihood of suicide attempts. Additionally, certain mental health conditions such as depression seem to function as a transdiagnostic risk and maintenance factor of both NSSI and suicide (Klonsky et al., 2013), corroborating that STBs and NSSI are closely related. Attending to these shared characteristics of STBs and NSSI, the Centers for Disease Control and Prevention (CDC) recently proposed that both can be classified as self-directed violence (Crosby et al., 2011). Thus, it is essential to explore whether any common underlying factors exist that are associated with both STBs and NSSI.

Much prior research has focused on accurately identifying risk and protective factors associated with STBs and NSSI (see Franklin et al., 2017 for a review of STBs and Fox et al., 2015 for a review of NSSI). Internalizing psychopathology (e.g., anxiety, depression, hopelessness, and emotion dysregulation) and demographic variables (e.g., age, education, and employment) were two of the strongest risk factors for STBs, followed by externalizing psychopathology (e.g., aggressive behaviors and impulsivity), prior STBs, and social factors (e.g., abuse, family problems, isolation, peer problems, and stressful life events). Similarly, characteristics such as impulsivity, unpredictability, and hopelessness were strong risk factors for NSSI (Fox et al., 2015).

In conjunction, recent research on STBs and NSSI has focused on identifying protective factors. Beyond demographic variables, the most notable protective factors of STBs and NSSI include a sense of social belongingness (Marraccini and Brier, 2017), social support (Kleiman and Liu, 2013), feelings of hope (Jiang et al., 2020), and emotional intelligence (Cha and Nock, 2009) to name a few. In exploring these protective factors, it was highlighted that the presence of protective factors is not merely the absence of risk factors. For instance, not feeling hopeless does not automatically translate into feeling hopeful about self and future. In this vein, Franklin et al. (2017) emphasized that protective factors that are not a simple reversal of risk factors (e.g., no psychopathology or no alcohol use) should be set a priori to be explored. Identifying protective factors allows the intentional fostering of the cultivation of skills by interventions to increase these protective factors where applicable.

A factor that may be negatively associated with both STBs and NSSI that is theoretically plausible for potential successful intervention is self-compassion. Self-compassion is operationalized as a healthy attitude toward oneself when experiencing life challenges and feelings of self-inadequacy (Neff, 2003, 2020). Researchers draw from philosophical and neuroscientific literature to conceptualize self-compassion. Neff (2003) referred to Buddhist teachings to draw connections between compassion for others and compassion for self, in which compassion refers to being touched by others' sufferings and pain and opening oneself to be connected to others' pain. Similarly, self-compassion was conceptualized as being connected with suffering and pain while sending warm acceptance to the experience of suffering of the self. Gilbert (2005) argued that being compassionate to others and self physiologically activates a soothing system (parasympathetic nervous system). In sum, self-compassion allows someone to be in touch with sufferings with kindness, providing a sense of soothing.

Self-compassion prepares a person to be less judgmental toward oneself, less isolated, and more balanced in perspectives so that he or she does not need be overwhelmed by negative emotions, mistakes, or failures (Neff et al., 2007). Because intense negative feelings and attitudes about the self (e.g., self-loathing or self-isolation) often accompany STBs and NSSI, self-compassion may counteract these negative emotional states. Indeed, self-compassion has been found to be negatively associated with STBs (Tielke, 2016; Rabon et al., 2018), and self-compassion has been found to moderate the effects of negative life events on STBs and NSSI (Jiang et al., 2016; Xavier et al., 2016; Chang et al., 2017; Hasking et al., 2019). With psychological and emotional well-being associated with STBs or NSSI, meta-analyses examining correlational studies have also found that self-compassion was inversely related to depression, anxiety, and stress (MacBeth and Gumley, 2012) and positively associated with psychological well-being (Zessin et al., 2015). Furthermore, considering that self-compassion could be cultivated through repeated practice, intervention studies have found that self-compassion-oriented interventions reduced psychological symptoms (e.g., anxiety, depression, eating behaviors, stress, rumination, and self-criticism) and increased life satisfaction and mindfulness (Ferrari et al., 2019; Wilson et al., 2019). In sum, empirical evidence points to exploring self-compassion as a viable factor that may be negatively associated with STBs and NSSI.

Two theoretical frameworks on STBs and NSSI, respectively, the interpersonal theory of suicide (IPTS; Joiner, 2005) and the cognitive–emotional Model of NSSI (CEM-NSSI; Hasking et al., 2017) can be extended to qualify the function of self-compassion. The IPTS argues that suicidal desires emerge when someone has a negative perception of oneself in relation to others (Joiner, 2005; Van Orden et al., 2010). Specifically, suicidal ideation is likely to emerge when an individual lacks a sense of belonging, accompanied by feelings of loneliness (thwarted belongingness) while perceiving him/herself to be a burden to others colored by self-hatred (perceived burdensomeness). Experiencing loneliness and feeling burdensome to others precisely reflects a lack of common humanity and being kind to oneself (Rabon et al., 2019). That is, self-compassion may reduce those self-defeating perceptions even in the face of mistakes and failures by understanding that feeling lonely is an inevitable aspect of human life, thereby allowing one to treat oneself with kindness.

The CEM-NSSI outlines that both emotion regulation and cognitive factors interactively influence NSSI, grounded in existing emotion regulation theories (Chapman et al., 2006) and social cognitive theory (Bandura, 1989, 1997). Specifically, emotional reactivity (e.g., temperament), representations of the self (e.g., “I am not worthy”), and NSSI-related cognitions (e.g., “cutting will provide immediate relief”) work synergistically, creating a vicious feedback loop that increases the risk for NSSI (Hasking et al., 2017). Holding a positive self-attitude (e.g., “I will be kind to myself”) can be introjected to weaken this negative feedback loop reducing self-judgment. Given the multiple developmental trajectories and risk factors that are associated with NSSI, Abdelraheem et al. (2019) recently emphasized the importance of finding modifiable factors that could introject or undermine the pathways and the effects of risk factors. After synthesizing 25 studies, they found that self-compassion, along with interpersonal difficulties, self-esteem, and impulsivity, can be a potentially modifiable factor. Modifiable factors can be targets for treatment, and much evidence suggests that self-compassion is a quality that can be cultivated through practice (Neff and Germer, 2013). In sum, self-compassion has the potential to counteract propositions of central aspects of both theories leading to STBs and NSSI.

To the best of the current knowledge, only one systematic review has explored the associative patterns of self-compassion with suicidal ideation and NSSI in 16 studies, finding that self-compassion was inversely associated with suicidal ideation or NSSI (Cleare et al., 2019). Although this review was informative, there were two limitations. First, Cleare et al. (2019) only focused on how self-compassion relates to suicide ideation and not any additional variations within STBs. Second, no meta-analysis was conducted, limiting an understanding of the relative importance of self-compassion against other factors.

This current study started with the goal to specifically address these two limitations. As such, first, we explored the relations between self-compassion and all STBs, expanding from a sole focus on suicide ideation in Cleare et al. (2019). Notwithstanding the qualitative differences in characteristics and the prevalence of each suicide-related variable, exploring all variants of suicide (ideation, thoughts, attempts, and plans) allows for the more comprehensive assessment of how self-compassion is related to STBs. Second, by conducting a meta-analysis, the current study derived effect sizes of correlations between self-compassion and STBs and NSSI. Quantifying effect sizes can be informative for future research when comparing the magnitude of their findings.

This meta-analysis investigated whether and how self-compassion was associated with STBs and NSSI, examining studies published since 2003 when the construct of self-compassion was introduced and operationalized. A review of a total of 18 studies found that self-compassion was negatively associated with both STBs and NSSI, with moderate and small effect sizes, respectively. Furthermore, positively worded subscales of self-compassion were more strongly inversely associated with STBs, although a small sample size (k = 4) cautions against making definitive conclusions. These findings lend support to arguments positing that underactivation of a “positive pathway” (e.g., low self-compassion) should be considered in conjunction with the role of risk factors (e.g., high hopelessness) activating “negative pathway” toward suicide and NSSI (Chang et al., 2017). This dual-factor model of mental health (Wang et al., 2011) can account for the complex interactions of risk factors and (lack of) protective factors in understanding suicide and NSSI.

Self-compassion was negatively associated with STBs and NSSI. The medium and small effect sizes, respectively, are comparable to the effect sizes of other factors that are negatively associated with STBs and NSSI. For instance, school belongingness was negatively associated with suicidal thoughts and behaviors (OR = 0.54) (Marraccini and Brier, 2017), and engaging in physical activities was also negatively associated with suicidal ideation, such that physically active individuals were less likely to report suicidal ideation compared with those who were physically inactive (OR = 0.87) (Vancampfort et al., 2018). Furthermore, self-compassion was negatively associated with depression and shame (Johnson and O'Brien, 2013) and hopelessness (Zhou et al., 2013), all of which are risk factors for suicide (Davidson et al., 2011) and NSSI (Selby et al., 2012), with a large effect size of −0.54 reported in the association between self-compassion and psychopathology (MacBeth and Gumley, 2012). In sum, the role and effect of self-compassion appear comparable with other identified factors, and due to self-compassion's association with psychological distress that often co-occurs with STBs and NSSI, it is not surprising that self-compassion emerged as a significant factor negatively associated with STBs and NSSI.

While self-compassion is conceptualized as a trait with stable individual differences (rank-order consistency), the argument has been made that it is a skill that can be cultivated. Multiple randomized controlled trial results implementing self-compassion interventions across a wide array of individuals (individuals with depression, college students, mental health professionals, and adolescents) consistently found support for its effectiveness (Wilson et al., 2019; see for a review, Ferrari et al., 2019), with moderate effect sizes. For example, self-compassion-related therapies and interventions were effective in increasing self-compassion (g = 0.52; g = 0.75) and reducing anxiety (g = 0.46; g = 0.57) and depressive symptoms (g = 0.40; g = 0.66) (Ferrari et al., 2019; Wilson et al., 2019). Thus, implementing self-compassion interventions, which increase self-compassion, may be explored in the future as a potential factor to lower the likelihood of occurrence of and reduce the effects on STBs and NSSI.

Second, there were effect size differences between STBs and NSSI on the associations with self-compassion, such that the effect size of self-compassion was larger for STBs than NSSI. There may be several plausible explanations for this difference. One explanation might be that it is due to the different nature of STBs and NSSI. To elaborate, emotion dysregulation undergird the maintenance and exacerbation of both STBs and NSSI (Kranzler et al., 2016), but NSSI is additionally distinctively characterized by an uncontrollable urge for the action of self-injury. This characteristic makes NSSI essentially a behavioral problem (Nock et al., 2006), with emotional distress instigating this self-destructive behavior. Thus, with NSSI, attending to emotion dysregulation is important, and learning behavioral techniques to distract the action of NSSI is equally as important. Because self-compassion is associated with internal thoughts and emotions, as well as behaviors of NSSI, diffusion in the size of effect sizes might have led to an overall smaller effect size with NSSI compared with STBs. Indeed, other studies that report the association between self-compassion with behavioral indicators also show small effect sizes. For instance, in a meta-analysis that examined the association between self-compassion and health-promoting behaviors, a small effect size of r = 0.25 was shown (Sirois et al., 2015). On the contrary, studies that have examined the association between self-compassion with maladaptive thoughts and emotions showed large effect sizes (MacBeth and Gumley, 2012). In sum, NSSI having a behavioral component may be one plausible explanation for smaller effect size compared with STBs.

When comparing against each self-compassion subscale, positively worded (or “pure” self-compassion) subscales of self-compassion showed larger effect sizes than negatively worded subscales in the associations with STBs and NSSI. Multiple plausible explanations can exist. Our findings may reflect ongoing debates on the psychometric properties (and perhaps conceptual clarity) on what constitutes self-compassion and how it should be measured and calculated. If both factors equivalently reflected the general factor of self-compassion, it is unclear why such different effect sizes would emerge. Utilizing a large data of 11,685 participants, Neff (2020) recently found support for a single two-factor model (utilizing higher-order one factor of self-compassion) or a correlated six-scale model (utilizing all six subscale scores independently) to specify the SCS. The authors recommended against using correlated two factors of compassionate self-responding and non-compassionate self-responding, without a hierarchical one factor of self-compassion. However, our findings seem to lend support to distinguishing two higher-order factors with “non-compassionate self-responding” and “compassionate self-responding” with clearly distinctive associative patterns (direction of the association and the magnitude of association) between these two factors. As Neff (2020) mentioned, conceptualization and measurement of self-compassion ultimately is an empirical question that will become further clarified with accumulated evidence. Our findings lend support to distinguishing these two factors.

This differential associative pattern with suicide may be reflective of two different underlying systems that activate compassionate self-responding and non-compassionate self-responding (Muris and Otgaar, 2020). For instance, Gilbert and Kirby (2019) argued that (self) compassion physiologically activates the soothing system (parasympathetic nervous system), whereas non-compassionate responding activates the threat-defense system (sympathetic nervous system). Linking assessment of self-compassion to physiological measures, in addition to self-report assessment, to examine the activation and deactivation of the soothing system and the threat-defense system, could be one way to clarify the nature of self-compassion. At a minimum, three dimensions reflective of compassionate self-responding (mindfulness, common humanity, and self-kindness) appear to assess self-compassion reliably, and their associations with correlates such as STBs can also be reliably interpreted. This could be further supported through the comparable effect sizes of total SCS and positively worded self-compassion subscales with STBs.

There was high heterogeneity among studies on both STBs and NSSI. Exploration of plausible moderating variables (gender, types of SCS scales, original long vs. short version, types of suicide/NSSI scales, and study quality) did not significantly explain the heterogeneity. However, subgroup analysis results found that sample type was a significant moderator explaining heterogeneity in effect sizes. This is consistent with previous studies that show different severity levels of STBs and NSSI per sample characteristics. Those who were diagnosed with a psychiatric disorder and with past suicide attempts were more likely to report higher levels of STBs (Franklin et al., 2017). Those with psychiatric disorders were more likely to report NSSI (Fox et al., 2015). The STBs and NSSI are not prototypically expected responses to stressful experiences, and how individuals respond to stress depends on their current level of overall functioning. This explains why there was a moderating effect of sample type.

Regarding other tested moderating variables, the results on gender were surprising, especially given the rich literature on gender differences on STBs and NSSI. For STBs, previous studies show mixed results in the association between gender and suicide, such that men are at higher risk of suicide (Hedegaard et al., 2018), while other studies have found comparable risks between genders (Kaplan et al., 2012). Similarly, for the gender difference on NSSI, some studies have found that women are more likely to engage in NSSI (see for a review, Bresin and Schoenleber, 2015), while other researchers have found no differences in engagement between genders (e.g., Garisch and Wilson, 2015). With self-compassion alone, on the other hand, Yarnell et al. (2015) found that men showed higher self-compassion scores than women (d = 0.18) after a meta-analysis of 71 studies. Thus, we speculate that the effect of gender was offset due to mixed directionality on self-compassion, STBs, and NSSI by gender. For SCS scale type, when used in total score, the full SCS and short-SCS seem to yield comparable results, thereby showing a non-significant moderating effect. Except for a few studies, most studies used a validated measure of STBs (k = 10) and NSSI (k = 5), and most studies (14 of 18 studies) were rated as good quality, which may explain the nonsignificant moderating effect.

Multiple factors may explain the heterogeneity, although they were not testable within the current meta-analysis. One plausible explanation is that there was significant variation in social environments and conditions of survey administration (Imrey, 2020). For instance, there were significant geographical regional differences (e.g., different states within the United States, multiple countries worldwide, such as the United States, China, and Portugal). However, no theoretical and empirical support exists to hypothesize and test whether a certain regional and social condition would be predictive of a stronger association with STBs or NSSI than others. Furthermore, categorizing regions also seemed rather arbitrary, with difficulties determining whether state-level categorization (given that most studies were from the United States) or country-level categorization is appropriate. Another reason for high heterogeneity might be due to the nature of the correlate. High heterogeneity is common in meta-analyses that test the incidence or prevalence of a phenomenon (Imrey, 2020). Other meta-analyses on suicide and NSSI also reported high heterogeneity (Huang et al., 2017).

This meta-analysis has several limitations that may inform future studies on this topic. First, the sample size of this meta-analysis was small. Although the small sample size reflects the maturity of the self-compassion and STBs and NSSI literature, the results of this study suggest that more pointed future examinations could be conducted. For instance, examining the role of self-compassion in samples that are at high risk for STBs/NSSI (e.g., trauma clients and individuals with relapse of major depressive disorder) could be fruitful, given that self-compassion may be an especially difficult quality to embody among those who are psychologically persistently unhealthy. Relatedly, interpretations about the SCS subscale relations to STBs should be interpreted with caution because the results were based on four studies. Given that the psychometric properties of the SCS with six subscales or three positively worded subscales are an ongoing debate (Neff et al., 2019; Muris and Otgaar, 2020), future studies should specifically focus on exploring the nature of relations between SCS subscales and STBs/NSSI. Similarly, the moderating effect of sample type was significant, but because each level (four types) contained a small number of studies (e.g., clinical patients, k = 1), conclusions should be cautiously made. At a minimum, consistent with previous studies (Franklin et al., 2017), this study found that the psychological functioning of individuals is important to consider.

Second, all but two studies included in this meta-analysis were cross-sectional studies, limiting conclusions on whether self-compassion is a protective precursor to reduced STBs and NSSI. Conceptually, it is plausible that self-compassion is a causal protective predictor, but future studies should infuse longitudinal design and elucidate the nature of associations more in detail. Third, the delineation of different subtypes of suicide was unavailable in this meta-analysis because there were only two or three studies for each of the subtypes of STBs (e.g., suicide ideation, suicide thoughts, and suicide attempt). It was deemed inappropriate to summarize results from such few studies. Further accumulation of empirical studies on this topic would be necessary to draw generalizable claims. Fourth, all eligible studies could not be analyzed as the final sample because correlations were not reported and available in some studies. Contacting the authors for correlation tables was unsuccessful with a few studies, which prevents a more comprehensive convergence of scholarly evidence. Future studies could ensure the inclusion of basic descriptive statistics such as means, standard deviations, and correlations to advance the literature on self-compassion and STBs/NSSI at large.

Last, examining other variables in the association between self-compassion and STBs/NSSI was not conducted in this study, and it focused on establishing self-compassion as an important factor in STBs/NSSI. Future studies could examine mechanisms through which self-compassion affects STBs and NSSI. This would be an especially welcomed addition to the literature that potentially bolsters whether underactivation of “positive pathway” according to the dual-factor model is relevant to STBs and NSSI. An examination of other protective factors (e.g., sense of belongingness, hope, and optimism) in conjunction with self-compassion and comparing their relative strength of association with STBs and NSSI would be a fruitful area of future research.

Despite these limitations, this meta-analysis provides a useful starting point for future research to continue to explicate the nuanced associations between self-compassion and STBs and NSSI. Clinically, this study points to the possibility of infusing self-compassion interventions to increase self-compassion so that a protective pathway for STBs and NSSI could be considered, although more evidence should accumulate through longitudinal and interventions studies. Several group programs, such as the Mindful Self-Compassion (MSC) program (Neff and Germer, 2013) and C ompassion-F ocused T herapy (CFT; Gilbert, 2014), are carried out in clinical settings targeted to increase self-compassion and compassion. However, the degree of change per person has not been identified, nor whether these programs are similarly effective across varying degrees of psychopathology. Thus, more converging evidence of rigorous randomized controlled trials with at-risk populations for STBs and NSSI should be accumulated to recommend with confidence before self-compassion interventions can be more widely applied across different settings.

The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author/s.

JJ designed and executed the review, performed data analyses, and wrote the paper. HS collaborated with analyzing data, wrote the paper, and edited the final manuscript. All authors contributed to the study.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.