Elsevier

Clinical Psychology Review

Volume 42, December 2015, Pages 156-167
Clinical Psychology Review

Meta-analysis of risk factors for nonsuicidal self-injury

https://doi.org/10.1016/j.cpr.2015.09.002Get rights and content

Highlights

  • We conducted a weighted random-effect meta-analysis of NSSI risk factor studies.

  • Results suggested significant, but weak, NSSI risk factor magnitude.

  • A prior history of NSSI was the strongest risk factor (odds ratio around 6).

  • Remaining risk factor magnitudes were low, suggesting limited clinical utility.

  • Continuous NSSI measurement resulted in stronger NSSI risk factor magnitude.

ABSTRACT

Nonsuicidal self-injury (NSSI) is a prevalent and dangerous phenomenon associated with many negative outcomes, including future suicidal behaviors. Research on these behaviors has primarily focused on correlates; however, an emerging body of research has focused on NSSI risk factors. To provide a summary of current knowledge about NSSI risk factors, we conducted a meta-analysis of published, prospective studies longitudinally predicting NSSI. This included 20 published reports across 5078 unique participants. Results from a random-effects model demonstrated significant, albeit weak, overall prediction of NSSI (OR = 1.59; 95% CI: 1.50 to 1.69). Among specific NSSI risk factors, prior history of NSSI, cluster b, and hopelessness yielded the strongest effects (ORs > 3.0); all remaining risk factor categories produced ORs near or below 2.0. NSSI measurement, sample type, sample age, and prediction case measurement type (i.e., binary versus continuous) moderated these effects. Additionally, results highlighted several limitations of the existing literature, including idiosyncratic NSSI measurement and few studies among samples with NSSI histories. These findings indicate that few strong NSSI risk factors have been identified, and suggest a need for examination of novel risk factors, standardized NSSI measurement, and study samples with a history of NSSI.

Introduction

Nonsuicidal self-injury (NSSI) is defined as direct and deliberate self-harm enacted without the desire to die (most often self-cutting; Nock, 2010). Lifetime prevalence rates of these behaviors range from 5.5–17% in community samples (among teens and adults respectively; Swannell, Martin, Page, Hasking, & St. John, 2014) and 50% in clinical samples (DiClemente et al., 1991, Penn et al., 2003). In addition to being dangerous in its own right, NSSI may be a risk factor for future suicidal behaviors (e.g., Asarnow et al., 2011, Bryan et al., 2014, ⁎Cox et al., 2012, Goldstein et al., 2012, Whitlock et al., 2013, ⁎Wilkinson et al., 2011, Guan et al., 2012). Given the dangerousness and prevalence of these behaviors, it is concerning that no intervention has been consistently shown to reduce NSSI compared to an active control group (see Brausch and Girresch, 2012, Glenn et al., 2015, Gonzales and Bergstrom, 2013, Nock, 2010, Washburn et al., 2012). These findings indicate that existing treatments do not target the processes that drive NSSI. The primary purpose of the present meta-analysis was to evaluate risk factors for NSSI, with the aim of providing a foundation for advancing the understanding and treatment of NSSI.

Before exploring these risk factors in more detail, it is necessary to differentiate risk factors from correlates (Kraemer et al., 1997). Correlates are associated with a given outcome, but the specific nature of this association is ambiguous. For example, if emotion dysregulation co-occurred with NSSI, emotion dysregulation would be a correlate of NSSI and it would be unclear how or why they were related. Risk factors, in contrast, temporally precede the outcome of interest and divide individuals into high and low risk groups (Kraemer et al., 1997). If emotion dysregulation preceded NSSI and distinguished those who would engage in future NSSI from those who would not, emotion dysregulation would also be a risk factor for NSSI. Causal risk factors are a specific type of risk factor that can be especially useful for prediction, theory development, and establishing treatment targets. Causal risk factors can be manipulated to change the probability that an outcome will occur. If emotion dysregulation were a causal risk factor, increases or decreases in emotion dysregulation would lead to subsequent increases or decreases in the likelihood of future NSSI. The majority of research on NSSI has focused on correlates (i.e., cross-sectional associations with NSSI), but in recent years there has been a proliferation of NSSI risk factor studies (i.e., longitudinal prediction of NSSI). Very few studies have examined causal risk factors for NSSI, so the present meta-analysis will focus more specifically on NSSI risk factors.

NSSI risk factor research has focused primarily on the ability of internalizing symptoms (e.g., depression, hopelessness, anxiety), affect dysregulation, and prior self-injurious thoughts and behaviors (including both suicidal and nonsuicidal behaviors) to predict future NSSI. Research on these factors reflects many of the most popular current theories, which link NSSI to emotional problems (especially affect dysregulation and depressive symptoms) and other forms of self-injurious thoughts and behaviors. For example, some theories suggest that people may choose to engage in NSSI to cope with internalizing symptoms (Nixon, Cloutier, & Aggarwal, 2002) and to decrease feelings of numbness or emptiness (Peterson, Freedenthal, Sheldon, & Andersen, 2008). Consistent with these theories, internalizing symptoms have been linked to NSSI in numerous studies, with cross-sectional research demonstrating higher levels of depression and anxiety (e.g., Selby et al., 2012, Gollust et al., 2008, Nock et al., 2006) as well as disordered eating (e.g., Paul, Schroeter, Dahme, & Nutzinger, 2014) among individuals with a history of NSSI. Extending this research, numerous studies have examined the longitudinal association between NSSI and internalizing symptoms, with a specific focus on depression, anxiety, and eating disorders.

Regarding affect dysregulation, the majority of NSSI theories propose that emotion dysregulation is a central component in understanding why people engage in NSSI (e.g., Chapman et al., 2006, Brown et al., 2002, Selby and Joiner, 2009). According to these theories, people who engage in NSSI have particularly high levels of emotion dysregulation, and these feelings drive them to engage in NSSI as a way to improve their mood. The hypothesized mechanisms through which this affect regulation occurs varies (e.g., painful distraction redirecting attention, Chapman et al., 2006; disruption of ruminative processes, Selby & Joiner, 2009), but the conclusion is the same: people engage in NSSI (or other impulsive behaviors such as binge drinking) because they have labile emotions, and these behaviors then serve to regulate their emotions. Cross-sectional research examining this theory has been mixed. Regarding self-report studies, people who engage in NSSI demonstrate higher levels of self-reported emotion dysregulation (e.g., Glenn et al., 2011, Nock et al., 2008, Franklin et al., 2013) and negative affect (e.g., Bresin, 2014, Victor and Klonsky, 2014) than those who do not. However, experimental (e.g., Franklin et al., 2010, Kaess et al., 2012, Weinberg and Klonsky, 2012, Bresin and Gordon, 2013) and psychophysiological studies (e.g., Franklin et al., 2010, Glenn et al., 2011, Kaess et al., 2012) have often failed to reveal this pattern.

In contrast to these mixed findings, a large body of research has revealed that the majority of people who engage in NSSI report that doing so helps them to feel better, and this finding has been demonstrated across self-report, experimental, and psychophysiological measures (e.g., Nock and Prinstein, 2004, Brown et al., 2002, Bresin and Gordon, 2013, Franklin et al., 2010, Russ et al., 1992, Weinberg and Klonsky, 2012). Together, this work suggests that mood improvement may be a central function of NSSI engagement. Extending these findings, emerging research has examined longitudinal associations between NSSI and numerous types of self-reported affect dysregulation, such as emotional suppression, emotional reactivity, and negative affect.

Finally, previous behavior is often one of the strongest predictors of future behavior. Accordingly, many researchers have examined whether a history of NSSI is predictive of future NSSI. This is an especially important risk factor to examine in conjunction with other factors to help discern the unique importance of a given factor above and beyond a history of these behaviors. Moreover, a large body of research has demonstrated that NSSI and suicidal thoughts and behaviors are highly comorbid (e.g., Lloyd-Richardson et al., 2007, Tang et al., 2011, Brunner et al., 2007, MacLaren and Best, 2010). Extending upon research examining NSSI as a risk factor for future suicidal behaviors, researchers have also examined whether these thoughts and behaviors are predictive of future NSSI.

In addition to these more frequently studied variables, researchers have also examined many additional potential NSSI risk factors. These less frequently studied factors include borderline personality disorder, externalizing symptoms (e.g., aggression, conduct problems), impulsivity, patient prediction (i.e., self-reported likelihood of engaging in NSSI in the future), and gender. The principal goal of the present meta-analysis was to summarize this burgeoning NSSI risk factor literature. To accomplish this, we addressed three basic questions within the NSSI risk factor literature.

We examined descriptive characteristics of NSSI risk factor studies to shed light on the types of studies that have been conducted and to investigate potential strengths and gaps in the literature. Specifically, we examined the number of NSSI risk factor studies, variation in NSSI measures, follow-up lengths, and sample characteristics (i.e., sample age, history of psychopathology, and NSSI frequency over the follow-up period).

To summarize the findings across NSSI risk factor studies, we estimated the magnitude of the overall combined effect of all risk factors and the magnitudes of each individual risk factor category. We employed meta-analytic methods for this estimation because risk factor magnitudes vary substantially across studies. For example, *Prinstein and colleagues (2010) found that depressive symptoms strongly and significantly predicted future engagement in NSSI among adolescents; however, *Hankin and Abela (2011) did not find a significant association. Considering findings in isolation makes it difficult to determine the true magnitude of a risk factor. Meta-analytic methods overcome this limitation by combining results across studies using dynamic weighting procedures.

Clinicians are often asked to assess whether their clients are at heightened risk for engaging in future self-harming behaviors. For a variety of reasons (e.g., stigma, fear, parental consequences, longer hospital stays), some clients may be unwilling to disclose that they want or plan to engage in NSSI in the future. However, it remains important for clinicians to accurately identify clients at high risk of engaging in these behaviors to better tailor treatment and prevention efforts for those clients (e.g., asking parents to help get rid of razors and scissors around the home; ensuring clients are closely monitored when receiving inpatient care). Similarly, screening measures that can be administered on a large-scale in school or other settings could be especially helpful at identifying individuals at risk and then funneling treatment and prevention resources to those who most need them.

Accordingly, in addition to looking at the magnitude of these risk factors, we also considered their clinical utility. We defined clinical utility as the degree to which a given factor increases the absolute odds of engaging in NSSI. The prevalence rate for engaging in NSSI over a one-year period is approximately 0.9% among adults (Klonsky, 2011). Accordingly, the absolute odds of an adult engaging in NSSI any given year is .009, meaning approximately one in every 100 adults will engage in NSSI in a one-year period. If a risk factor has a weighted odds ratio of two, this factor would double the odds of next-year NSSI engagement to two in every 100 adults. In contrast, if that factor had a weighted odds ratio of 10, it would increase the odds ten-fold, resulting in next-year NSSI engagement in approximately nine of every 100 adults. To our knowledge, there is no cross-national study of past year prevalence rates of NSSI among child and adolescent populations, but rates in these populations are likely 2–3 times higher than in adult populations (Swannell et al., 2014). As such, the same risk factor magnitude may imply higher clinical utility in an adolescent sample compared to an adult sample.

The effect of a risk factor may change in important ways under different conditions. In the present meta-analysis, we examined four potential moderators of NSSI risk factor magnitude. The first moderator was NSSI measurement type. Measures of NSSI are highly variable across studies, with NSSI assessments ranging from single-item open-ended questions to extensive questionnaires, checklists, and interviews. Some checklists include indirect methods of self-harm (e.g., self-poisoning, substance ingestion) and normative behaviors (e.g., picking at a wound; Lloyd-Richardson et al., 2007) whereas others exclude these types of behaviors. Still other researchers include only new instances of NSSI, excluding those individuals who engaged in NSSI at baseline. This high variability in the content assessed across NSSI measures raises concerns about the validity and reliability of results and compromises the ability to make inferences across studies. In the present meta-analysis, this heterogeneity precluded tests of moderation by specific measures due to the very small number of studies employing any one measurement tool. Instead, we examined moderation across binary (i.e., grouping NSSI engagement into “yes” versus “no” categories) or continuous (i.e., assessing NSSI frequency using interval or ratio scales) measures of NSSI. We expected that binary measurement of NSSI would produce weaker prediction, as it may not sufficiently assess important features of NSSI behavior (e.g., behavior frequency, severity) that may improve predictive power.

Second, we examined study population as a moderator. NSSI risk factors studies have included general samples (i.e., participants were not selected for psychopathology or NSSI history), clinical samples (i.e., participants were selected based on a history of psychopathology), and NSSI samples (i.e., participants were selected based on a history of NSSI). We hypothesized that general sample studies would produce the strongest NSSI prediction. This is because when self-injurers are compared to other self-injurers, there are relatively few differences between the two groups other than the potential risk factor. As a result, any observed effects would be specific to the risk factor under investigation. However, when self-injurers are compared to non-injurers (especially from a general sample), there are many differences between the groups besides the potential risk factor. In those cases, psychopathology, self-injury history, and other confounding factors may combine with the risk factor under investigation to produce larger observed effects.

Third, we explored the effects of sample age. Based on current literature, it was unclear whether prediction would be stronger for adult or adolescent samples. As such, these analyses were exploratory.

Fourth, we examined the impact of the type of measure (i.e., binary versus continuous) used to predict NSSI. Importantly, odds ratios are linked to measurement scales and should be interpreted as such. Specifically, an odds ratio reflects the change in odds per one unit of measurement. Binary measures only have one unit (i.e., yes versus no), whereas continuous measures have a wide range. Accordingly, odds ratios from binary measures tend to be larger. We hypothesized that binary prediction cases would result in larger odds ratio magnitudes than continuous prediction cases.

Section snippets

Study retrieval and selection

For the purposes of the present meta-analysis, NSSI was defined as any intentional act of self-harm enacted without the desire to die. To be included, we required that papers include longitudinal prediction of NSSI in any population, country, and year, using any predictor variable prior to January 1st, 2015. We identified studies by searching on PubMed, PsycInfo, and Google Scholar. To provide the most comprehensive meta-analysis possible, we included a wide range of search terms. This is

Number of published reports across time

A total of 20 published reports and 16 unique study samples were included in the present meta-analysis. The earliest published report was *Van der Kolk, Christopher, and Perry (1991); the next qualifying study was not published until 17 years later (*Zanarini et al., 2008).

Prediction cases and trends across time

These 20 published reports produced a total of 247 prediction cases. Of these, we excluded 79 cases. Sixty-six of these prediction cases were excluded because they were used across multiple time points within one study and 13

Discussion

Each year, millions of people purposely hurt themselves without wanting to die (Klonsky, 2011, Nock and Prinstein, 2004). Historically, the majority of research on NSSI has been cross-sectional; however, prospective risk factor research is growing. Risk factor research is critical for advancing the conceptualization, prediction, and treatment of NSSI. The primary goal of the present meta-analysis was to synthesize the NSSI risk factor literature. Results highlighted several statistically

Conclusions

The present meta-analysis synthesized data from nearly a decade of research examining NSSI risk factors. Results suggested significant, but weak, NSSI prediction and highlighted variables that might represent risk factors for NSSI. More importantly, however, these results emphasized that we currently lack strong risk factors for NSSI. Additionally, the present meta-analysis highlighted extreme heterogeneity across NSSI measurement, limiting our ability to accurately identify NSSI risk factors.

Role of funding sources

This work was supported in part by funding from the Military Suicide Research Consortium (MSRC), an effort supported by the Office of the Assistant Secretary of Defense for Health Affairs (Award No. W81XWH-10-2-0181). Opinions, interpretations, conclusions and recommendations are those of the authors and are not necessarily endorsed by the MSRC or the Department of Defense.

Contributors

KRF and JCF designed the study, conducted literature searches, and provided summaries of previous research studies. All authors assisted in coding included studies for this meta-analysis. KRF conducted the statistical analysis. KRF wrote the first draft of the manuscript and all authors contributed to and have approved the final manuscript.

Conflict of interest

All authors declare that they have no conflicts of interest.

Acknowledgment

Dr. Ribeiro was supported in part by a training grant (T32MH18921) from the National Institute of Mental Health.

References1 (73)

  • M.K. Nock et al.

    Non-suicidal self-injury among adolescents: Diagnostic correlates and relation to suicide attempts

    Psychiatry Research

    (2006)
  • M.K. Nock et al.

    The emotion reactivity scale: development, evaluation, and relation to self-injurious thoughts and behaviors

    Behavior Therapy

    (2008)
  • J.V. Penn et al.

    Suicide attempts and self-mutilative behavior in a juvenile correctional facility

    Journal of the American Academy of Child and Adolescent Psychiatry

    (2003)
  • M.J. Russ et al.

    Pain perception in self-injurious patients with borderline personality disorder

    Biological Psychiatry

    (1992)
  • J. Whitlock et al.

    Nonsuicidal self-injury as a gateway to suicide in young adults

    Journal of Adolescent Health

    (2013)
  • T. *Andrews et al.

    Predictors of onset for non-suicidal self-injury within a school-based sample of adolescents

    Prevention Science

    (2013)
  • M.D. *Anestis et al.

    Predicting nonsuicidal self-injury episodes over a discrete period of time in a sample of women diagnosed with bulimia nervosa: An analysis of self-reported trait and ecological momentary assessment based affective lability and previous suicide attempts

    International Journal of Eating Disorders

    (2012)
  • J.O. *Roaldset et al.

    Predictive validity of the MINI suicidal scale for self-harm in acute psychiatry: A prospective study of the first year after discharge

    Archives of Suicide Research

    (2012)
  • B.A. *van der Kolk et al.

    Childhood origins of self-destructive behavior

    American Journal of Psychiatry

    (1991)
  • S.E. Victor et al.

    Daily emotion in non-suicidal self-injury

    Journal of Clinical Psychology

    (2014)
  • M.C. *Zanarini et al.

    The 10-year course of physically self-destructive acts reported by borderline patients and axis II comparison subjects

    Acta Psychiatrica Scandinavica

    (2008)
  • J. Bjärehed et al.

    Deliberate Self‐Harm in 14‐Year‐Old Adolescents: How Frequent Is It, and How Is It Associated with Psychopathology, Relationship Variables, and Styles of Emotional Regulation?

    Cognitive Behaviour Therapy

    (2008)
  • A.M. Brausch et al.

    A review of empirical treatment studies for adolescent nonsuicidal self-injury

    Journal of Cognitive Psychotherapy

    (2012)
  • K. Bresin et al.

    Changes in negative affect following pain (vs. nonpainful) stimulation in individuals with and without a history of nonsuicidal self-injury

    Personality Disorders: Theory, Research, and Treatment

    (2013)
  • M.Z. Brown et al.

    Reasons for suicide attempts and nonsuicidal self-injury in women with borderline personality disorder

    Journal of Abnormal Psychology

    (2002)
  • R. Brunner et al.

    Prevalence and psychological correlates of occasional and repetitive deliberate self-harm in adolescents

    Archives of Pediatrics and Adolescent Medicine

    (2007)
  • L.J. *Cox et al.

    A longitudinal study of nonsuicidal self-injury in offspring at high risk for mood disorder

    The Journal of Clinical Psychiatry

    (2012)
  • C. Delmo et al.

    Kiddie-SADS present and lifetime version (K-SADS-PL)

    Auflage der deutschen Forschungsversion. Frankfurt am Main: Klinik für Psychiatrie und Psychotherapie des Kindes-und Jugendalters der Universität Frankfurt

    (2000)
  • J.C. Franklin et al.

    The functions of nonsuicidal self-injury: Support for cognitive-affective regulation and opponent processes from a novel psychophysiological paradigm

    Journal of Abnormal Psychology

    (2010)
  • J.C. Franklin et al.

    Recent and frequent nonsuicidal self-injury is associated with diminished implicit and explicit aversion toward self-cutting stimuli

    Clinical Psychological Science

    (2013)
  • J.C. *Franklin et al.

    Low implicit and explicit aversion toward self-cutting stimuli longitudinally predict nonsuicidal self-injury

    Journal of Abnormal Psychology

    (2014)
  • J.C. Franklin et al.

    Risk factors for suicidal thoughts and behaviors: A meta-analysis of 50 years of research

    (2015)
  • Glenn, C. R., Blumenthal, T. D., Klonsky, E. D., & Hajcak, G. (2011). Emotional reactivity in nonsuicidal self-injury:...
  • C.R. Glenn et al.

    Evidence-based psychosocial treatments for self-injurious thoughts and behaviors in youth

    Journal of Clinical Child and Adolescent Psychology

    (2015)
  • T.R. Goldstein et al.

    Predictors of prospectively examined suicide attempts among youth with bipolar disorder

    Archives of General Psychiatry

    (2012)
  • S.E. Gollust et al.

    Prevalence and correlates of self-injury among university students

    Journal of American College Health

    (2008)
  • Cited by (324)

    View all citing articles on Scopus
    1

    References marked with an asterisk indicate studies included in the meta-analysis.

    View full text