The co-occurrence of aggression and self-harm: Systematic literature review

Abstract

Background

Epidemiological research supports an association between aggression and self-harm through data on the frequency with which individuals exhibit both behaviours. Unbiased evidence, however, is needed to draw conclusions about the nature and extent of co-occurrence.

Method

Systematic review of published studies was undertaken to evaluate whether or not the frequency with which aggression and self-harm co-occur is beyond that which would be expected by chance. Outcome measures included: (a) between-group differences on a standardised aggression/self-harm measure – the groups defined by scores on a measure of the other behaviour; (b) correlations between the two behaviours; (c) co-occurrence rates in populations defined by the presence of either behaviour; (d) co-occurrence rates in populations not defined by either behaviour. Odds ratios were calculated for studies presenting complete frequency data.

Results

123 studies, some yielding more than one type of result, met the inclusion criteria. Most case-control studies found elevated levels of aggression in self-harming populations (or self-harm in aggressive populations) compared to controls. The majority of correlational, co-occurrence rate, and odds ratio data found aggression and self-harm to be associated.

Limitations

Results were subject to descriptive synthesis only and thus, unable to report an overall effect size.

Conclusions

Evidence suggests that aggression and self-harm frequently co-occur. Such evidence necessitates more theoretical discussion and associated research on the source and nature of co-occurrence. Nonetheless, individuals who present with one behaviour may be considered an ‘at-risk’ group in terms of exhibiting the other. Such evidence holds implications for practice (e.g. risk assessment).

Introduction

Aggression and self-harm are important public health issues. While violent crime rates have fallen in recent years (ONS, 2013a), it remains a significant problem – incurring substantial costs to individuals and society (see Dubourg and Hamed, 2005 on economic and social costs). The Crime Survey for England and Wales (CSEW) estimates that around 1.9 million violent incidents towards adults occurred during 2012/2013 (ONS, 2013a). Police-recorded statistics for ‘violence against the person’ are significantly lower (approximately .6 million) – the discrepancy due to factors such as victims’ unwillingness to report offences and recent Home Office re-classification of crimes historically categorised as violent (ONS, 2013a). Annual rates of CSEW-reported violent incidents and police-recorded ‘violence against the person’ have therefore been estimated, respectively, at 42 per 1000 and 11 per 1000 (ONS, 2013a). While incidences of homicide in England and Wales in 2012/2013 were comparatively rare (n=552: ONS, 2013a), minor assaults remain frequent – especially when no physical injury was involved (e.g. ‘assault without injury’, ‘aggressive behaviour’); CSEW (ONS, 2013a) reporting 103,000 such incidences against children aged 10–15 and 830,000 against adults aged 16 and over.

Epidemiological data on incidence of self-harm is harder to obtain for a number of reasons. No national statistics exist and regional statistics are often narrow in scope, with focus on, for example, only certain types of self-harm (e.g. self-poisoning) or certain age-groups (e.g. youths). While there are other possible sources for rates of self-harm, the most commonly cited is hospital attendance. The NHS recorded 110,960 hospital admissions following episodes of self-harm for the year ending August 2012 – nearly 90% being cases of self-poisoning (HSCIC, 2013). Hospital presentations that do not lead to admittance are more frequent, with regional estimates suggesting a figure of 170,000 episodes of self-poisoning and a further 30,000 episodes of self-injury per annum (Horrocks et al., 2003; Kapur, 2009). Given that many episodes of self-harm do not lead to healthcare presentation, these figures are likely to be underestimates (Kapur et al., 1998, Hawton et al., 2002, Rodham et al., 2005, Taylor and Cameron, 1998). Suicide statistics, however, are more reliable, with the ONS reporting that 6045 people died by suicide in 2011 – the estimated rate of deaths from suicide estimated at 11.8 per 100,000 (ONS: Office for National Statistics, 2013b). The World Health Organization (2009) estimates that for every person who dies by suicide, there are at least 20 suicide attempts, with Kapur (2009) suggesting that over half of those who die by suicide will have a history of self-harm. Epidemiology aside, self-harm as an important public health issue is perhaps most aptly emphasised under consideration of the distressing consequences to patients, victims, and their families, and the eventual costs to society (see Currier et al., 2006, Kapur et al., 2001, Kapur et al., 2002, Kennelly, 2007, Knapp et al., 2011 on costs and consequences).

While seemingly different, even opposing behaviours, it has long been theorised that aggression and self-harm are linked. From a psychodynamic perspective, Freud (1905/1953, 1917/1953) viewed suicide as aggression turned inward and proposed that aggression underlies both suicidal and violent behaviours. More recently, Plutchik and colleagues (e.g. Plutchik, 1994, Plutchik et al., 1989a, Plutchik and van Praag, 1990a) regard evidence of their association as a marker for a shared aggressive impulse to act violently towards the self and others. Even the use of legal and psychological terminology over the years such as ‘self-murder’ (e.g. see Gates, 1980 on ‘felo-de-se’), ‘self and other-directed violence’ (e.g. Myers and Dunner, 1984), ‘outward and inward-directed aggressiveness’ (e.g. Plutchik, 1994), or ‘self-directed and other-directed aggression’ (e.g. Hillbrand, 2001) has often implied a conceptual link. While the two behaviours may have different clinical and forensic implications, it is therefore surprising that aggression and self-harm research emerged historically as two separate fields (Lubell and Vetter, 2006). This empirical separation did not last long though. Over the years, many researchers began to find evidence that supported the theoretically proposed link between aggression and self-harm.

A striking evidentiary feature is the notable overlap in risk factors associated with each behaviour. Plutchik and colleagues (e.g. Plutchik, 1994, Plutchik et al., 1989a, Plutchik and van Praag, 1990a) have suggested that as many as 23 ‘psychosocial’ risk factors may be causally implicated in both suicide and violence (e.g. hopelessness, impulsivity). For neurobiologists, there appears to be so much overlap (e.g. shared serotonergic dysfunctions in the prefrontal cortex leading to cognitive vulnerabilities in functions such as impulse control and decision-making), that many neurobiological reviews prefer to integrate findings into one paper (e.g. Goethals et al., 2005, Golden et al., 1991, Henry and Demotes-Mainard, 2006, Jokinen et al., 2010a, Mann and Currier, 2009, Roggenbach et al., 2002). One particularly prominent risk factor that is garnering attention in both aggression and self-harm research is childhood abuse (physical and sexual) or neglect, and the proposed effect on subsequent emotional and cognitive development (e.g. Beaver, 2008, Boudewyn and Liem, 1995, Eron et al., 1991, Fergusson et al., 1996, Fergusson and Lynskey, 1997, Gratz et al., 2002, Herrera and McCloskey, 2003, Lipschitz et al., 1999, Low et al., 2000, O’Keefe, 1994, Raine et al., 2001, Swanston et al., 2003, van der Kolk et al., 1991, Wiederman et al., 1999, Yates and Carlson, 2003). However, while some researchers (e.g. Hillbrand, 2001) have presented such research as evidence of co-occurrence, overlap in risk factors is not the same as co-occurrence. Thus, while we can utilise such evidence to support a link between aggression and self-harm, we should not use it to support proposals that the risk of one behaviour increases as a function of the presence of the other. For the latter proposal, we must look to more direct evidence – specifically, to a body of studies finding that the behaviours frequently co-exist within an individual.

Notable evidence for the co-occurrence of aggression and self-harm began to emerge in the 1970s and 1980s, with the publication of a number of studies finding elevated prevalence of one behaviour in populations defined by the presence of the other. For example, patients with a history of suicidal behaviour often revealed a history of violence (e.g. Inamdar et al., 1982, Plutchik et al., 1989b), and prisoners incarcerated for violent offences showed evidence of prior self-harm, such as scars resulting from self-cutting (e.g. Bach-y-Rita, 1974). Since then, the evidence has continued to accumulate. Irrespective of the definitions employed, the measures and methods used, or the populations selected, researchers continue to find that a significant proportion of individuals with a history of aggression tend to have a history of self-harm (e.g. Cairns et al., 1988, Flannery et al., 2001, Lidberg et al., 2000, Virkkunen et al., 1989) and vice versa (e.g. Buri et al., 2009, Hunt et al., 2006, Olsson, 1999, Robinson and Duffy, 1989) – with prevalence exceeding 40% in some studies (e.g. Bach-y-Rita and Veno, 1974, Conner et al., 2000, Conner et al., 2002, Nijman and Campo, 2002, Rosenblatt and Greenland, 1974). Similarly, epidemiological data reveal elevated co-occurrence rates in a variety of populations such as psychiatric (e.g. Asnis et al., 1994, Fennig et al., 2005, Pfeffer et al., 1983a), clinical (e.g. Bergman and Brismar, 1994a, Bergman and Brismar, 1994b, Hasin et al., 1988), forensic (e.g. Maden et al., 2000, Stalenheim, 2001), and community (e.g. Gould et al., 1998, Suokas et al., 2010). Finally, cross-sectional studies lend additional support for co-occurrence, via the common finding that people who score highly on measures of aggression (or self-harm) score higher on measures of self-harm (or aggression) compared to controls (e.g. Brent et al., 2002, Dervic et al., 2006, Grosz et al., 1994, Korn et al., 1997, Mann et al., 2005, Michaelis et al., 2004, Oquendo et al., 2000, Renaud et al., 2008).

Despite evidence of co-occurrence, and the implications this may hold for risk assessment and monitoring of aggression and self-harm, current practice does not appear to reflect such research. While joint measurement of the risk of aggression and self-harm have been incorporated into research instruments such as the Overt Aggression Scale (Yudovsky et al., 1986) and the Suicide and Aggression Survey (Korn et al., 1992), most UK clinical and forensic risk assessments do not acknowledge one behaviour to be a significant predictor of the other. For example, standardised risk assessment of violence is not a component of recommended psychiatric assessment procedures following an episode of self-harm (e.g. NICE, 2004). One reason for this situation is that, despite suggestive literature, co-occurrence beyond chance has not been conclusively established. While we found one non-systematic review of ‘self-directed’ and ‘other-directed aggression’ (Hillbrand, 2001), no review was identified that methodically observes the extent (and limits) of the association between aggression and self-harm as it co-occurs within individuals. Clearly, such research is a pre-requisite for theorising further on the nature of aggression and self-harm as co-occurring and, indeed, as independent behaviours. While causal links are rarely established between co-morbid conditions, evidence of a reciprocal relationship regarding risk has important predictive value in clinical and forensic settings. Formulations of either type of behaviour, for the purposes of risk assessment and service planning and provision, may need modification upon objective assessment of the degree and nature of their association. We have therefore undertaken a systematic review of the literature to answer the following question: Is there good evidence that aggression and self-harm co-occur in individuals more commonly than chance would predict and if so, to what extent?