Non-suicidal self-injury (NSSI) is defined as deliberate self-inflicted harm to one's body tissue without the intent to die and for purposes that are not socially sanctioned (Turecki and Brent, 2016)—a prevalent and debilitating public health concern across demographic groups and clinical populations (Heath et al., 2009; Klonsky, 2011). Burning, cutting, scratching, and self-hitting are common behaviors in NSSI (Cipriano et al., 2017). NSSI is currently a diagnostic criterion for borderline personality disorder, although recent conceptualizations suggest that NSSI occurs across multiple clinical disorders and may occur independent of any psychiatric comorbidities (Zetterqvist, 2015). Given the prevalence and trans-diagnostic nature of NSSI, a workgroup published a proposal in the Diagnostic and Statistical Manual of Mental Disorders 5 (DSM-5) to include NSSI disorder under conditions for further study (American Psychiatric Association, 2013).

Although the majority of NSSI research has been conducted on adolescent populations, up to 21% of adult psychiatric inpatients engage in NSSI (Briere and Gil, 1998). NSSI is particularly prevalent among psychiatric inpatients with a history of suicide attempts (SA) (Langbehn and Pfohl, 1993; Andover and Gibb, 2010). The relationship between NSSI and SA is complex. Although frequently considered distinct entities on the basis of the presence/absence of suicidal intention, studies consistently demonstrate high levels of comorbidity between non-suicidal and suicidal self-injury (Zetterqvist, 2015). Furthermore, research shows that common heritable factors are shared by NSSI and suicidal ideation/SA (Maciejewski et al., 2014). Indeed, a substantial proportion of individuals who attempt suicide report NSSI (Franklin and Nock, 2016), and a history of NSSI increases risk of suicide following or during psychiatric inpatient treatment (King et al., 2001; Hunt et al., 2007). This association may be expected as NSSI is commonly understood as a maladaptive affect regulation strategy (Klonsky, 2009; Nock, 2009) that may increase risk of suicide through an acquired capability for lethal self-harm (Joiner et al., 2012; Chu et al., 2018).

Evidence suggests that additional risk factors, both environmental and individual, may play a role in the etiology of NSSI. Disturbances in attachment have been linked to both NSSI (Adam, 1994; Critchfield et al., 2008) and suicidal behaviors (Adam et al., 1996; Lessard and Moretti, 1998; Stepp et al., 2008). Attachment theory suggests that experiences with caretakers beginning in infancy are internalized as working models of self and other (Griffin and Bartholomew, 1994) that profoundly influence an individual's ability to regulate affect (Schore, 2015) and manage relationships (Levy and Blatt, 1999; Crowell et al., 2008). Four primary adult attachment styles have been described in the literature: secure, preoccupied (also referred to as anxious), dismissing, and fearful, the latter three of which are considered variants of insecure attachment (Bartholomew and Horowitz, 1991). Characteristics associated with preoccupied attachment appear to be particularly related to NSSI. Increased levels of self-blame and self-criticism have been noted in individuals with preoccupied attachment (Bartholomew and Horowitz, 1991). Recent studies support the role of negative associations with the self in motivating NSSI through the belief that one deserves punishment (Franklin and Nock, 2016). Moreover, preoccupied or anxiously attached individuals tend to bolster their unstable sense of self-worth through seeking excessive reassurance of intimacy in personal relationships, leaving them vulnerable to extreme distress when their intimacy needs are not met (Griffin and Bartholomew, 1994). Thus, individuals with preoccupied attachment disturbances may be more likely to engage in NSSI in the context of acute affective dysregulation, specifically in response to interpersonal stress. Accordingly, several studies have found a direct effect of insecure or anxious (i.e., preoccupied) attachment style on NSSI (van der Kolk et al., 1991; Gratz et al., 2002; Stepp et al., 2008; Martin et al., 2011).

Much of affect regulation is learned through the formative attachment experiences in infant-caregiver dyads (Mikulincer et al., 2003; Mikulincer and Shaver, 2008). Acute cognitive and emotional states associated with imminent suicide risk may play a role in the development and maintenance of NSSI (Cuenca, 2013). In particular, relief of emotional pain has been shown to serve a powerful reinforcing function for NSSI (Brown et al., 2002; Nock and Prinstein, 2004). Likewise, recent work has highlighted the importance of emotional pain and entrapment in the etiology of SA (Shneidman, 1993, 1999; Orbach et al., 2003; Rasmussen et al., 2009; Yaseen et al., 2012, 2016; Ducasse et al., 2017; Galynker et al., 2017). Thus, the same acute emotional states associated with SA may also be related to NSSI.

Current research suggests that regulation of emotional states may mediate the relation between attachment insecurity and NSSI in young adults (Kimball and Diddams, 2007). Some researchers posit that individuals with insecure attachment styles may be less able to regulate painful affective states (Schaffer, 1993) and are more prone to experiencing intolerable emotional pain, leading to affect regulation in the form of NSSI (Cuenca, 2013). Preoccupied attachment may therefore increase susceptibility to experiences of emotional pain, which would thus mediate its relationship with NSSI.

The high rate of NSSI among adult psychiatric patients, coupled with the strong association between NSSI and suicide, makes this population particularly vulnerable (Pompili et al., 2015). While several studies have examined the relationship between insecure attachment and affective state factors on the presence of NSSI, to our knowledge no investigations have analyzed the interrelation of the two factors in an adult suicide risk sample. Thus, we sought to address this gap in the literature by examining the association between attachment characteristics, acute cognitive and emotional states previously associated with short-term suicide risk, and NSSI in a sample of adult patients hospitalized for acute suicide risk.

Based on the existing literature, we first hypothesized that preoccupied attachment would be associated with lifetime NSSI among attachment styles (secure, preoccupied, dismissing, fearful). Second, we hypothesized that emotional pain would be associated with lifetime NSSI among the acute cognitive and emotional states previously associated with acute suicide risk (entrapment, panic/dissociation, ruminative flooding, fear of dying, emotional pain). Finally, we hypothesized that emotional pain would mediate the relationship between preoccupied attachment and NSSI. We tested these hypotheses cross-sectionally, using presence of lifetime NSSI as the outcome.

In this study of adult psychiatric inpatients hospitalized for suicide risk, half of the sample (48.5%) endorsed lifetime NSSI, highlighting the prevalence of this behavior among a high risk population. Moreover, contrary to popular conceptions, NSSI was equally prevalent among men and women in this sample. Few studies have examined risk factors for NSSI in adult samples (Kapur et al., 2013). To our knowledge, this is the first paper to analyze the relationship between attachment styles, acute cognitive and emotional states, and NSSI among high suicide risk adult psychiatric inpatients.

In support of our first hypothesis, we found that preoccupied attachment style was the sole independent predictor of lifetime NSSI among attachment disturbances. This finding parallels prior research suggesting a link between preoccupied attachment and NSSI, particularly in adolescent populations (Wright et al., 2005; Stepp et al., 2008), but also in adults (Critchfield et al., 2008). Individuals with preoccupied attachment, characterized by a sense of unworthiness or unlovability of the self, and a positive working model of the other, may favor coping behaviors, such as NSSI that are hurtful to self but elicit rescue from caregivers (Griffin and Bartholomew, 1994; Bolen et al., 2013) or that instantiate negative beliefs about oneself (Franklin and Nock, 2016).

Of the acute cognitive and emotional states examined, emotional pain at admission was the sole independent predictor of lifetime NSSI. Hence, the key role of emotional pain is not only traceable in suicidal ideation and SA (Ducasse et al., 2017) but in NSSI as well. The observed data highlight the importance of identifying the automatic negative reinforcement function of NSSI that serves to relieve negative affective states, such as emotional pain (Nock and Prinstein, 2004). Although entrapment and ruminative flooding were each individually predictive of NSSI in univariate analyses, they did not remain significant in the binary logistic regression. Hence, in our sample intolerable emotional pain appeared to be more closely associated with lifetime NSSI in comparison to other acute cognitive and affective states. Although the SCI subscales are intercorrelated, the low VIF values suggest that this did not unduly affect the reliability of the multivariable logistic regression model.

We also found preoccupied attachment style to be correlated with emotional pain. This result is in agreement with literature documenting the interplay between preoccupied attachment and negative emotional reactivity (Levy et al., 2005).

In support of our third hypothesis, emotional pain at admission partially mediated the relationship between preoccupied attachment style and lifetime NSSI. NSSI may serve as a potent coping behavior for regulation of negative affective states, such as unbearable emotional pain, for individuals with relationship histories characterized by instability (Linehan, 1993; Chapman et al., 2006; Kleindienst et al., 2008).

From a neurobiological perspective, the endogenous opioid system plays a role in pain threshold and perception and is a good candidate for involvement in NSSI, a behavior often associated with the need to feel pain and/or to relieve emotional tension. Patients with a history of NSSI were found to have significantly lower levels of cerebrospinal fluid (CSF) beta-endorphin and met-enkephalin when compared with non-NSSI patients (Stanley et al., 2010). Moreover, the opioid system seems to be implicated in social pain as well (Lutz et al., 2018). A history of chronic stress may lead to a dulled endogenous opioid response to acute stress, possibly increasing vulnerability to emotional pain. NSSI may be an effort at increasing the endogenous opioids to re-establish homeostasis (Stanley et al., 2010). Future studies should explore possible treatment approaches that can modulate opioid system function in patients with NSSI (Turner et al., 2014).

Although beyond the scope of this report, many other factors contribute to suicide and NSSI risk, both additively and in interaction with the factors examined here, and would deserve attention. For example, sleep disturbances, and in particular insomnia (Pompili et al., 2013), may both mark distress and further reduce distress tolerance and impulse inhibition, leading to increased risk.

AM and C-YC performed the first literature search, the first statistical analyses, and wrote a first draft of the manuscript. ZH revised the paper and performed an updated literature search. JB contributed to the writing of the paper. RC performed further statistical analyses and contributed to the writing of the last version of the manuscript. ZY was responsible of data collection, proposed the topic, closely supervised AM and C-YC in the manuscript preparation and revised the manuscript. IG and PR were responsible of data collection and supervised the manuscript writing.