The Associations Between Children's and Adolescents’ Suicidal and Self-Harming Behaviors, and Related Behaviors Within Their Social Networks: A Systematic Review

Social influences—including the suicidal and self-harming behaviors of others—have been highlighted as a risk factor for suicidal and self-harming behavior in young people, but synthesis of the evidence is lacking. A systematic review of 86 relevant papers was conducted. Considerable published evidence was obtained for positive associations between young people's suicidal and self-harming behavior and that of people they know, with those reporting knowing people who had engaged in suicidal or self-harming behaviors more likely to report engaging in similar behaviors themselves. Findings are discussed in relation to a number of methodological and measurement issues—including the role of normative perceptions—and implications for the prevention of suicidal and self-harming behavior are considered.

A wide range of terminology has been used to attempt to define suicidal and selfharming behavior (SSHB), both in research and in practice, and given that individuals reportedly engage in SSHB for myriad reasons (Laye-Gindhu & Schonert-Reichl, 2005;Nock, 2009;Scoliers et al., 2008), no one definition is universally accepted as all-encompassing. Furthermore, the utility of focusing on intention and separating out non-suicidal from suicidal self-harm has been debated (e.g., Kapur, Cooper, O'Connor, & Hawton, 2013;O'Carroll et al., 1996;Silverman, Berman, Sanddal, O'Carroll, & Joiner, 2007), particularly given the apparent uncertainty and=or ambivalence surrounding intention and motivation for some individuals (Dorpat & Boswell, 1963;Henriques, Wenzel, Brown, & Beck, 2005). For the purposes of the current paper, any act of self-injury-regardless of intention or motivation-is included under the umbrella term SSHB, in an attempt to capture all relevant behaviors.
SSHB is a major public health concern, representing the tenth leading cause of death Archives of Suicide Research, 21:185-236, 2017 Copyright # International Academy for Suicide Research ISSN: 1381-1118 print=1543-6136 online DOI: 10.1080DOI: 10. /13811118.2016 worldwide, and constituting about 1.5% of the international disease burden ). As such, the implementation of evidence-based prevention, intervention, and postvention strategies has become a priority, internationally (Hadlaczky, Wasserman, Hoven, Mandell, & Wasserman, 2011). Various social factors have been implicated in increasing risk of engagement in SSHB, including socioeconomic deprivation (Hawton, Harriss, Hodder, Simkin, & Gunnell, 2001), unemployment (Kposowa, 2001), and social isolation (Bearman & Moody, 2004), while others appear to reduce risk; such as social support (Greening & Stoppelbein, 2002). In particular, the impact of the SSHB of other people on an individual's own behavior, and the co-occurrence of such behaviors across groups of individuals has attracted much interest. Evidence has repeatedly been found for clusters of SSHB in time and space (e.g., Haw, Hawton, Niedzwiedz, & Platt, 2013), an increase in suicide attempts has frequently been recorded following widespread reporting of high-profile suicides (e.g., Niederkrotenthaler et al., 2012), and a contagion-like spread of such behaviors within shared environments has often been observed (e.g., Brent et al., 1989).
A number of theoretical models of suicidal behavior and their supporting literature have highlighted the importance of social factors in the development and trajectory of suicidal behavior. For example, in Williams' (2001) Cry of Pain model, social support represents a rescue factor which may prevent feelings of defeat and entrapment from developing into suicidal behavior. In addition, in Joiner's (2005) Interpersonal Theory, feelings of thwarted belongingness and perceived burdensomeness-both arguably concepts with social bases-interact with acquired capability to prompt suicidal behavior. More recently, O'Connor's (2011) Integrated Motivational-Volitional (IMV) model posits that social factors may be important at every stage of the pathway from suicidal thought to behavior. An individual's biopsychosocial context determines their premotivational phase; poor social problem solving may represent a Threat-to-Self Moderator, and thwarted belongingness, burdensomeness, and a lack of social support may represent Motivational Moderators. The presence of each of these may result in suicidal ideation or intention, but it is only when Volitional Moderators are also present, such as social learning or perceived social norms, that behavioral enactment will take place. Empirical evidence of the importance of social factors in differentiating between thought and action comes from a study by O'Connor, Rasmussen, and Hawton (2012), who showed that self-harm enactors differed from ideators and controls on perceived descriptive norms and reported exposure to self-harm in family or friends.
SSHB is particularly prevalent in young people (De O'Loughlin & Sherwood, 2005;Owens, Horrocks, & House, 2002), and perhaps due to their susceptibility to social influence in general (Brechwald & Prinstein, 2011), evidence suggests that the effects of social factors on the SSHB of young people may be particularly powerful (e.g., Haw et al., 2013;King & Merchant, 2008;Phillips & Cartensen, 1986). This may be compounded by the fact that those individuals who are most prone to social influence may be at an already heightened risk of engaging in damaging behaviors (Allen, Porter, & McFarland, 2006), and it has been repeatedly shown that young people's engagement in risky or health-damaging behaviors in particular may be vulnerable to social influence (e.g., Eisenberg, Neumark-Sztainer, Story, & Perry, 2005;Perkins, Perkins, & Craig, 2010;Van Der Vorst, Engels, Meeus, & Deković, 2006). This may be as a result of the improved self-status afforded by engaging in certain Suicidal Behavior in Children and Their Social Networks risky behaviors (Brechwald & Prinstein, 2011), and evidence has shown that selfharming behaviors are perceived as high-status behaviors in early adolescence .
The evidence for the co-occurrence of SSHB across groups and the apparent transmission of SSHB between individuals, together with the increased risk of SSHB in young people and the heightened susceptibility of young people to social influence (particularly with regard to status-gaining behaviors), argues for the importance of gaining a better understanding of the relationship between young people's SSHB and that of other people. To date, no comprehensive synthesis of the research in this area has been conducted, and findings appear somewhat inconsistent. Moreover, it is not always clear whether individuals involved in such research are explicitly aware of the behaviors of others, or whether knowledge is assumed based on presence in a particular geographic location or attendance at a particular school, for example. Such considerations may be important according to research from the field of social norms, which consistently indicates that individuals' perceptions of the social norms surrounding a particular behavior-regardless of the accuracy of those perceptions-are more predictive of their own engagement in that behavior than are actual norms (e.g., Perkins, 2007). Consideration of whether or not the relevant others are personally known to the individual (and if so, in what capacity) is also often omitted from articles, making it difficult to gauge whether accurate knowledge is likely, or to determine whether perceptions of unknown others' behavior is sufficient to influence one's own. A systematic search and comprehensive review of the available literature, with specific inclusion criteria which would enable synthesis of relevant findings to address these inconsistencies, was therefore conducted.

AIMS OF REVIEW
The primary aim of the current review was to investigate whether relationships exist between child=adolescent SSHB and the SSHB of people they know. Having examined this, the review also aimed to identify whether perceptions of others' SSHB-and their potential inaccuracies and biases-are considered in the literature with regard to associations with child=adolescent SSHB, or whether accurate knowledge of the behavior of others is routinely assumed. Finally, the review aimed to explore whether any specific literature exists around the perceived social norms of SSHB (and their relationship with child= adolescent behavior and attitudes).

Eligibility Criteria
Papers were included in the current review if: (i) they were original, published, peer-reviewed journal articles; (ii) they were written in English; (iii) they reported the investigation of associations between an individual's SSHB and that of (specific) people they know, 1 or any influence of others' SSHB on one's own SSHB; (iv) they focused on a child and=or adolescent (up to 19 years old) population 2 ; and (v) a reasonable standard of inferential statistical analyses was conducted or the paper reported on qualitative data. The study selection process is illustrated in Figure 1.

Study Characteristics
Eighty-six relevant papers were identified for inclusion in the current review (full details of reviewed studies can be found in Tables 1-4). The SSHBs examined (in terms of both children's=adolescents' behavior and that of the people they know) ranged from thoughts of self-harm, through self-harm, suicide plans, threats, and attempts, to death by suicide, as well as some cumulative scales of unspecified ''suicidal behavior.'' The reference group examined (i.e., the ''others'' to which studies referred) also varied widely, including such groups as friends, peers, parents, siblings, and other relatives. Research was conducted within a range of populations in terms of age (range ¼ 5-19 years) and location (23 different countries) and in a variety of settings (including schools, the community, inpatient and outpatient mental health services, emergency departments and other healthcare settings).
A similarly diverse range of methods and analyses were employed. Measures of both child=adolescent SSHB and that of their networks were obtained through child=adolescent reports of both (e.g., through questionnaires, standardized measures or interviews), third party reports, analysis of official records=national statistics, secondary analysis of previously collected data, psychological autopsy, observation, and a mixture of child=adolescent report and one or more other method(s). Again, this information is detailed within Tables 1-4. Analyses ranged from simple t-tests and odds ratios, through (mainly logistic) regression, to the generation of complicated statistical models. Such diversity in target behavior, reference group, setting, methods, measures, and analyses, rendered meta-analysis unfeasible. Instead, findings are presented in narrative form, according to the reference group with whose behavior associations were explored, with a separate section for qualitative studies.

Associations with Family SSHB
Twenty-three papers looked at the relationship between young people's SSHB and that of members of their family, with most focusing on family in general (firstand second-degree relatives, first-degree 1 Papers which failed to define the nature of the young person's relationship with the reference group were excluded (e.g., ''people you know''), as were those in which reference groups were not necessarily present in the young person's social network (e.g., celebrities or fictional characters), in order to minimize the likelihood that the data reflected guesswork or more general perception of overall rates. 2 In accordance with World Health Organization (2013) and UNICEF (2011) definitions of adolescence.
Suicidal Behavior in Children and Their Social Networks only, or unspecified), and a small number focusing on siblings, or one or both parents. Papers in this section looked mainly at suicidal ideation or attempts in both children=adolescents and their family members, with only a small minority incorporating suicide deaths or a cumulative scale of general suicidality, and one which looked specifically at self-poisoning (intent not specified). Full details of the family studies can be seen in Table 1.
Positive Findings. Twenty (87.0%) of these studies reported positive associations between children's=adolescents' SSHB and that of their family members, and those with a family history of SSHB were repeatedly found to be more likely than those without, to engage in SSHB themselves (An, Ahn, & Bhang, 2010;Bridge, Brent, Johnson, & Connolly, 1997;Cheng et al., 2014;Garfinkel, Froese, & Hood, 1982;Gartrell, Jarvis, & Derksen, 1993;Goldstein et al., 2005;Gould, Fisher, Parides, Flory, & Shaffer, 1996;Johnson, Brent, Bridge, & Connolly, 1998;Kerfoot, 1988;Marusic, Roskar, & Hughes, 2004;McKenry, Tishler, & Kelley, 1982;Myers, Burke, & McCauley, 1985;Pfeffer, 1984;Pfeffer, Conte, Plutchik, & Jerrett, 1980;Pfeffer, Normandin, & Kakuma, 1994;   Study 2-Own self-harm at time 0 was positively associated with higher levels of perceived self-harm in friends at 9 months, and perceptions were positively associated with own self-harm at 18 months. Again, effects were moderated by gender You et al. (2013)  Self-report questionnaires Suicidal ideation, ''behaviors'' Suicide death Friends No differences were found between those who were or were not exposed to suicide in friends, on suicidal ideation or behaviors Family or friends Those who first self-harmed between waves reported that their friends held more positive views of self-harm, than did non-self-harmers. Repeat self-harmers were more likely to have friends or family who self-harmed, and who were more positive about self-harm, compared to non-self-harmers CROSS-SECTIONAL  7,018 Icelandic 9-10th graders (schools) Thoughts of self-harm, self-harm with and without intent to die

Self-harm
Family or friends Self-harm in friends or family differentiated between various groups (e.g., those with and without suicidal thoughts; those with self-harm with intent to die and those with thoughts). In males, self-harm of peers differentiated those with self-harm without intent to die and those with thoughts. There were also strong associations between self-harm groups and self-harm in others 380 9th-12th graders in the US (schools) Self-report questionnaires Suicidal ideation, attempt ''Suicidal behavior'' (not specified) Family or peers Those with ideation or attempts reported more suicidal behavior in their family than those without (Continued ) Self-harm with and without suicidal intent Self-harm in friends, mother and father, suicide attempt in parents Friends or parents Self-harm in friends and mothers was strongly associated with own suicidal self-harm, less so with non-suicidal self-harm. Self-harm in fathers and parents' suicide attempts were associated with own suicidal self-harm, but not non-suicidal McMahon et al. Self-report questionnaires Suicidal ideation Suicidal ideation Parents or peers More suicidal ideation was reported in those whose mothers (but not fathers) or peers had suicidal ideation, than those whose mothers or peers did not. Peer (Continued ) Associations between child=adolescent SSHB and their mothers' SSHB appeared to be of particular importance, with such associations often found to be stronger than those with fathers' or other relatives' SSHB (Kerfoot, 1988;Pfeffer, 1984;Pfeffer et al., 1994;Pfeffer et al., 1998;Pfeffer et al., 1984;Tischler & McKenry, 1982). Each of these findings was obtained through a range of methods, including self-report (e.g., Gartrell et al., 1993), therapist ratings (Pfeffer et al., 1980), clinical records (e.g., Johnson et al., 1998), family report (e.g., Bridge et al., 1997) or a combination of multiple methods (e.g., McKenry et al., 1982). The increased risk of suicide attempt or death in those with a family history of either was found in two studies to be beyond that contributed by shared psychopathological variables (Gould et al., 1996;Johnson et al., 1998).
Only one (5.0%) of the papers reporting positive findings used a longitudinal design (Tucker & Wiesen-Martin, 2015). Using a sample of 1,055, this paper indicated that siblings tended to be similar to each other in their suicidal ideation, and that older siblings' suicidal ideation predicted younger siblings' later suicidal ideation.

CROSS-SECTIONAL STUDIES.
Nineteen (95.0%) of the papers reporting positive findings were cross-sectional in their design. Although most papers identified widespread associations between children's=adolescents' SSHB and that of their family members, one paper indicated that mothers' suicide death is only associated with that of female adolescents, while fathers' suicide death is associated only with that of males (Cheng et al., 2014).
Three studies (13.0%) failed to identify any associations between child=adolescent SSHB and that of their family members, and all employed school=community samples (Cerel, Fristad, Weller, & Weller, 1999;Kebede & Ketsela, 1993;Marcenko, Fishman, & Friedman, 1999). Notably, all three looked at the actual death of a family member by suicide-one specifically at the suicide death of a parent (Cerel et al., 1999) and the others at suicide deaths in the family in general. LONGITUDINAL STUDIES.
One (33.3%) of the papers reporting negative findings used a longitudinal design (Cerel et al., 1999). In 358 participants, no difference was observed in suicidality between those whose parents had died by suicide and those whose parents had died through other causes.

CROSS-SECTIONAL STUDIES.
Two studies (66.7%) found no associations between a family history of suicide and adolescents' own suicide attempts or ideation using cross-sectional methods.

Associations With Friends'=Peers' SSHB
Sixteen papers explored associations between child=adolescent SSHB and that of their friends or peers. Papers covered self-harm through to suicide attempt in adolescents, and self-harm through to suicide death in friends=peers. Authors varied in their use of terminology, with reference groups referred to as friends, close friends, peers, and acquaintances, and definitions were rarely given. 3 Notably, all studies referred to adolescents, with no relevant research identified within child samples. Full details of these studies can be seen in Table 2.
Positive Findings. As with family studies, the majority of studies in this section (eleven-68.8%) reported positive associations between the SSHB of adolescents and that of their friends, with an increased likelihood of those whose friends engaged in SSHB doing so themselves, and vice versa (Alfonso & Kaur, 2012;Brent et al., 1993;Cerel, Roberts, & Nilsen, 2005 Four (36.4%) of the studies with positive findings used a longitudinal design, revealing that having friends who self-harm=attempt suicide predicts one's own later self-harm= suicide attempt (respectively). It was suggested however, that depression (Liu, 2006), adverse life events (e.g., parental discord, getting in trouble with the police), and previous thoughts of self-harm (Hasking et al., 2013) may have moderated the relationship. Importantly in terms of the relevance of perceptions to associations, one study found a reciprocal relationship between perceptions and behavior . Positive relationships were found both between adolescent self-harm at baseline and their perceptions of their friends' self-harm at 9-month follow-up, and between their perceptions of their friends' self-harm at the 9-month follow-up and their own self-harm at 18-month follow-up (although this was moderated by gender).

CROSS-SECTIONAL STUDIES.
Seven (63.6%) of the positive studies were cross-sectional. One study suggested that associations between suicidal behaviors appeared to be particularly strong between close friends, compared to more distant acquaintances (Ho et al., 2000).
Two (40.0%) of the negative findings came from longitudinal studies. Brent et al. (1996) failed to identify any increase in suicidal behavior at follow-up in those with friends who had died by suicide, despite higher baseline rates of psychopathology. Additionally, Giletta et al. (2010) found no increase in self-harm following friends' self-harm, although friends' depressive symptoms predicted an increase in adolescents' self-harm.

CROSS-SECTIONAL STUDIES.
Three (60.0%) of the papers with negative findings used cross-sectional designs. Brent et al. (1993) found that higher levels of suicidal ideation in those whose friends had died by suicide were almost entirely accounted for by depression, and despite identifying no associations with adolescents' suicidal behavior, Brent et al. (1992) found much higher rates of depression in those whose friends had died by suicide.

Associations With Multiple Sources' SSHB
Forty-three papers did not look at individual reference groups, instead exploring the relationship between SSHB in multiple others (e.g., friends, family, romantic partners), and children's=adolescents' own SSHB.
The full range of behaviors of interest was covered by papers in this section, both in terms of child=adolescent behavior, and the behavior of people they know. Full information for these papers can be seen in Table 3.
Four (9.8%) of these studies included clinical samples; the remainder were school= community samples.
Nine papers with positive findings (22.0%) employed longitudinal designs. Adolescents' SSHB was predicted by the earlier SSHB of others, and in fact friends' or family members' suicide attempts were found to be among the strongest predictors of adolescents' future suicide attempts (e.g., Borowsky et al., 1999;Nanayakkara et al., 2013). One study suggested that boys may be more susceptible to the influence of friends, while girls are susceptible to that of both family and friends (Abrutyn & Mueller, 2014).

CROSS-SECTIONAL STUDIES.
Thirty-two (78.0%) of the studies with positive findings were cross-sectional. Those who attempted suicide or self-harmed were more likely to report knowing people who also did so (e.g., Corder et al., 1974;Deliberto & Nock, 2008). In terms of the aims of the current review relating to social norms and perceptions, one paper reported relevant findings . In this paper, group norms for self-harm (defined by the authors as ''the beliefs, attitudes and behavior of friends and peers'') were associated with self-harm, but only in boys.
Two papers (4.9%) also reported explicit influence of others' SSHB over adolescents' own. As well as finding that those who self-injured were more likely to report a family history of suicidal ideation than those who did not self-injure, 38.3% of Deliberto and Nock's (2008) self-injuring participants explicitly reported that they first got the idea to do so from their peers (and 13.3% from the media). Additionally, as well as statistical associations between adolescent self-harm and that of their friends and family, O'Connor et al. (2014) report that 13.3% of their adolescent participants explicitly stated that family members' self-harm or suicide attempts influenced their own self-harm, and 23.2% reported that the same was true of their friends' self-harm or suicide attempts.
Negative Findings. Only two papers (4.7%) found no associations between children's= adolescents' SSHB and that of others (Razin et al., 1991;Tomori & Zalar, 2000), and both were cross-sectional and sampled from non-clinical populations. Neither found any differences in reported suicide attempts by friends or family between those who had attempted suicide and those who had not.

Qualitative Studies
Four qualitative papers were selected for inclusion in the current review on the basis that while exploring general risk factors or characteristics of SSHB, each found some reported influence of those behaviors in others on the child's=adolescent's own. Full details of these papers can be seen in Table 4. Beekrum, Valjee, and Collings (2011) reported that a family history of attempted suicide or suicide death was indicated as a potential influence over the non-fatal suicidal behavior of respondents, with many respondents explicitly describing instances in which they had witnessed the suicidal behavior of a family member or friend result in some desired outcome. This observation may well have encouraged their own suicidal behavior, with the expectation that it might aid them in achieving some goal in the same way. Indeed, some participants reported instances where their own suicidal behavior had improved their situation. Herrera, Dahlblom, Dahlgren, and Kullgren (2006) found that suicide among friends sometimes acted as a trigger for respondents' own suicide attempts. Aside from these overt reports, many of the other triggers identified in this paper featured themes of loss or abandonment. One could arguably view the suicidal actions of a friend or relative as their afflicting both loss and abandonment upon an individual, so although these accounts do not explicitly refer to the suicide of friend, the resulting outcomes may be related. This is not, however, explored in this paper, and death of a relative generally (i.e., not by suicide), was also alluded to as important, by several participants. Orbach, Gross, and Glaubman (1981) reported that one of the common characteristics of most of the children they studied-all of whom had threatened or attempted suicide-was a suicidal parent in their family (usually their mother). In some cases, parents had openly spoken about their own or the child's potential suicide in front of the child, even offering a choice of weapons with which the child might take their life, so it might be argued that to those children, suicide became a particularly ''real'' concept and a possible addition to their behavioral repertoire.
In the final qualitative study, Tingey et al. (2014) reported a number of instances in which participants described imitating others' suicide attempts, as well as concerns that others might have been aware that they copied their behavior. One participant also compared their suicide attempt to a previous attempt by a cousin, describing their disappointment that their family's reaction had not been as supportive towards them in the aftermath as it had been towards their cousin; perhaps suggesting that elicitation of a similar reaction may have been part of the motivation for their own attempt.

Summary of Findings in Relation to Aims
The current systematic review aimed to investigate whether relationships exist between child=adolescent SSHB and the SSHB of people they know; to identify whether perceptions of others' SSHB-and their potential inaccuracies and biases-are considered in the literature or whether accurate knowledge is assumed; and to explore whether any literature exists around the perceived social norms of SSHB or normative influence.
Overall, the vast majority of the literature suggests that there are positive associations between children's=adolescents' SSHB and that of people they know. Such findings were obtained both through cross-sectional and longitudinal inquiry. The literature is wide-ranging in terms of where associations appear to lie, with some studies reporting the strongest relationships with family members' behavior (e.g., Rotheram-Borus et al., 1996) and others reporting that friends' behavior is particularly predictive of that of the child=adolescent (e.g., Lewinsohn et al., 1994). Different explanations for these findings have been proposed, each with their own merits. For example, family associations may be particularly strong as a result of the shared time spent with one's family, experiencing shared outcomes of events (e.g., ; or peer associations may be stronger because young people may look to their friends for behavioral guidance (e.g., Brechwald & Prinstein, 2011).
It was found that very little distinction was made in the literature between children's=adolescents' perceptions of the behaviors of others, and their actual knowledge of those behaviors; the two were typically assumed to be synonymous. Although other methods were used, self-report was by far the most common method of obtaining data, and the implications of relying on adolescents' self-reports will be discussed below. In terms of practical applications of the research to practice, as discussed by Brechwald and Prinstein (2011), if inaccurate perceptions are related to certain behaviors, employing interventions which correct those misperceptions may be effective in reducing related behaviors, as has been the case in other behavioral domains (e.g., Berkowitz, 2004;McAlaney, Bewick, & Hughes, 2010).
Only one paper reviewed touched upon normative influence . O'Connor et al. suggest that as well as others' behavior (descriptive norms), others' positive attitudes towards those behaviors (injunctive norms) may be associated with individuals' own behaviors. If overestimations are present for either of these norms, particularly given the ''invisible'' nature of such concepts as suicidal ideation (which is obviously more difficult to observe than are suicide attempts or deaths), individuals' behavior may be increased-as has been observed for other damaging behaviors (e.g., Borsari & Carey, 2003;Clemens, Thombs, Olds, & Gordon, 2008;Labrie, Grossbard, & Hummer, 2009;Lewis & Neighbors, 2004).

Methods and Measurement
The prevailing use of cross-sectional design and quantitative data contributes to a certain lack of clarity regarding whether children=adolescents are influenced by the behavior of people they know, whether they choose to associate with people they believe engage in similar behaviors to themselves, or whether engaging in those behaviors themselves renders them more likely to be aware of= over-estimate the prevalence of similar behavior in others. The literature employing longitudinal methods-as well as the qualitative evidence-supports the findings from cross-sectional studies, and suggests that exposure to SSHB in others increases children's=adolescents' engagement in those behaviors, but there are also a minority of longitudinal studies (n ¼ 3) which failed to find any influence. The findings of Prinstein et al. (2010) suggest that there may in fact be a reciprocal relationship between perceptions of others' SSHB and one's own; these authors found that adolescents' self-injury at baseline was related to perception of friends' self-injury 9 months later, but that perceptions of friends' self-injury was also related to own self-injury 9 months later. Other researchers however, failed to find such effects (Giletta et al., 2013).
An abundance of research findings gathered in school settings should be considered with further caution. Due to the process of recruiting from school populations, it is possible that many of the young people who might have been of particular interest in terms of the research aims were excluded. In some institutions in Scotland for example, researchers are required to obtain parental consent for anyone under 16, which means that only those chil-dren=adolescents whose parents wish them to participate will be allowed to do so. If a child=adolescent or their family has experience of SSHB or there are other particular issues in the family which might make SSHB more likely, parents may decide that the research would be too distressing for their child, and decline to participate. Similarly, those pupils who the literature would suggest are most at risk of SSHB (e.g., those with psychological problems, those from dysfunctional homes, or those with problems at school or with friends; see Webb, 2002) may be particularly likely to miss school as a result (e.g., through ill-health, truancy), and their potentially interesting data is therefore lost. Those participants who dropped out between waves 1 and 2 of Hasking et al.'s (2013) study scored higher on the Self-Harm Behavior Questionnaire than those who completed follow-up, indicating that it is at times those participants who are most at risk, who fail to participate. These issues may even be demonstrated at the organizational level-with some authors reporting that schools that declined to participate in their research had experienced more recent exposure to deaths by suicide than had those schools that participated (Ho et al., 2000). Furthermore, although all of the reviewed studies that employed clinical samples (n ¼ 12) reported positive findings, the relatively low number of such studies and the fact that most were conducted in the United States means that it is unclear whether patterns displayed in clinical settings would be as consistent as in community settings on a larger scale.
A heavy reliance on self-report methodology further complicates the picture, as self-report by definition enables the reporter to provide only that information to which they are privy, or indeed that which they choose to provide, and the potential bias that this affords may be particularly pertinent with a topic as sensitive in nature as SSHB. Data of this nature might be vulnerable to recall bias, social desirability, shame=embarrassment etc., such that a dataset relying heavily on self-report data may be somewhat less accurate than researchers might hope. O'Connor et al. (2014) suggest that the lower than expected self-harm rate they observed in their Northern Irish sample may reflect a society-wide reluctance to disclose personal information as a result of ''The Troubles'' and associated sectarianism, as opposed to a genuinely low rate of self-harm. The practice employed by many researchers, of informing participants that those deemed at high risk of suicidality will be referred to support services or reported to their parents (e.g., Marcenko et al., 1999;Watkins & Gutierrez, 2003) may further discourage participants from admitting to suicidal thoughts or behavior. In support of this, Marcenko et al. (1999) claim that research into SSHB better reflects participants' willingness to disclose their SSHB, than their actual SSHB. There is also the potential for inaccuracy with self-reported data, though it could be argued that individuals' perceptions of events-regardless of accuracy-are more important than the actual events themselves, in terms of the resultant impact on that individual. This has been shown to be the case in the alcohol literature, where perceived norms of peer alcohol use have been found to better predict personal use than peers' actual alcohol use (Perkins, Haines, & Rice, 2005). As such, self-report might be the ideal method for obtaining information regardless of accuracy, and the (in)accuracy itself, and its relation to the individual's own behavior, is of most interest. Moreover, a number of researchers made concerted attempts to avoid these types of biases-for example by using multiple data sources (e.g., Corder et al., 1974;Johnson et al., 1998) or collecting data about others' behavior directly from those individuals (e.g., Bearman & Moody, 2004;Feigelman & Gorman, 2008;Thompson et al., 2009)-and findings were nevertheless comparable to those studies which did not employ such measures.
A clear methodological limitation of research in this field is the lack of feasibility of experimental manipulation, which ordinarily assists researchers in determining whether apparent effects are the result of variables of interest, or whether other factors are responsible for outcomes. Needless to say, it would be impossible for example, to randomly expose a proportion of participants to SSHB in people they know and then compare how their own behavior develops in relation to an unexposed group. However, a small amount of experimental research has been conducted in this area, and similar findings have been found to those of the studies reviewed here. Using a self-aggression paradigm, Berman and Walley (2003) found that participants tended to engage in similarly self-aggressive behaviors as their (fictitious) opponents, in a reaction time task for which the ''loser'' was required to self-administer electric shocks. Those participants whose opponent engaged in high self-aggression on losing trials also tended to self-administer an increasing severity of shock, whereas those whose opponents engaged in low self-aggression also tended to self-administer less severe shocks. Sloan, Berman, Zeigler-Hill, and Bullock (2009) later replicated these findings. While these studies are interesting and provide us with an approximation of information that we would be unlikely to be able to obtain in such a controlled manner directly, they are lab-based, highly contrived and thus lacking in mundane realism, which limits the extent to which the results can be generalized to SSHB in the real world. As such, more naturalistic, ecologically valid research, controlling for as many other variables as is appropriate and feasible, may be the most rigorous method researchers currently have at their disposal for exploring these issues.

Terminology=Definitions
An issue which makes synthesis of findings challenging, and conclusions drawn somewhat tentative, is the breadth of terminology used. There is debate across the field regarding the similarity or relatedness of self-harm and suicide attempt, and the utility of differentiating between suicidal and non-suicidal self-harm (e.g., Kapur et al., 2013;O'Carroll et al., 1996;Silverman et al., 2007). Some papers in the current review refer to non-fatal suicide attempts and self-harm synonymously, paying little attention to suicidal intent (e.g.,  while others refer to and measure self-harm with and without suicidal intent separately (e.g., Mars et al., 2014). Behaviors termed as self-harm also vary across the reviewed literature from relatively less severe behaviors such as pinching, preventing wounds from healing (e.g., Alfonso & Kaur, 2012) or self-biting (e.g., You et al., 2013), to more dangerous and potentially lethal acts such as self-poisoning (e.g., Kerfoot, 1988) or jumping from a height (e.g., Hawton et al., 2002). The meaning derived from terms such as ''self-harm,'' ''self-injury,'' ''suicide attempt'' etc. is likely to not only differ across research teams, but also across participants. The same may also be true of reference group terms such as ''family member,'' ''friend,'' or peer,'' which also varied across studies.
Further, the use of the general term ''suicidal behaviors'' by many researchers (e.g., Myers et al., 1985) may conceal useful information around specific behaviors, and result in the incorrect generalization of findings across different behaviors within a spectrum of SSHB. Harkavy-Friedman et al. (1987) compared subgroups of those with different SSHB on experience with the behavior of different groups, and found that adolescents who ideate or who attempt suicide have more experience with family suicidal behavior than those who neither ideate nor attempt suicide. They were not however, different from each other in experience with family suicidal behavior. Conversely, those who made suicide attempts had more experience with peer suicidal behavior than those who only ideated, who in turn had more experience than those without any SSHB. Further, Mars et al. (2014) noted different associations with friends' and family's self-harm and suicide attempts between adolescents' who engaged in SSHB with and without suicidal intent. These findings have implications for the interpretation of the results of studies which group together reference groups (e.g., those which ask generally about ''people you know''; hence their exclusion from the current review), and those which group together behaviors (e.g., into one ''suicidal behavior'' variable).

Samples
Samples were frequently large and overall a wide range of ages, ethnicities, and social situations were represented. Despite this, however, the relative rarity with which people actually tend to engage in SSHB means that often, samples of those individuals will actually be quite small in real terms, potentially making associations difficult to detect. Perhaps as a result of this, there are gaps in the literature in terms of specific behaviors (e.g., there are no family-focused papers which address self-harm specifically). Some researchers explicitly report being unable to explore potentially interesting aspects of the data due to the limited number of individuals engaging in target behaviors (e.g., Nanayakkara et al., 2013), and other researchers may have clumped together groups of data for the same reason.
There may also be some limitation to the representativeness of findings from some of the reviewed studies due to the employment of somewhat restrictive inclusion criteria. For example, a large population-based survey of Korean adolescents  only included data from households in which all members agreed to take part and did not include single-parent households, which potentially increased the likelihood of excluding individuals who might be particularly vulnerable. Other studies that have included parental presence in the home as a variable suggest that those from single-parent households might be at particularly high risk of these kinds of behaviors (e.g., Garfinkel et al., 1982). Additionally, studies such as that of Kerfoot (1988) and Tischler and McKenry (1982) made use of very specific samples (i.e., children and adolescents referred to psychiatric services following an episode or self-poisoning, and adolescents treated in an emergency room for suicide attempt, respectively), such that findings may illustrate a specific vulnerability of that specific group of psychiatric inpatients, or those who seek=require emergency medical help, as opposed to something characteristic of those engaging in SSHB generally, or those who are never referred to health services. Nevertheless, comparable findings were obtained from a range of other samples and from studies with less restrictive inclusion criteria, so these concerns may be minimal.
A final important observation regarding the samples studied in the reviewed papers is that although samples were taken from all over the world, all studies were published in English (potentially resulting in some inclusion bias) and the majority of studies were in fact undertaken in the western world, particularly in the US and the UK. The World Health Organization (2014) reports that the majority (75%) of suicides take place in low and middle income countries, so there are issues with trying to generalize the findings of a predominantly wealthy, western sample, to suicidal behavior worldwide. Although findings were relatively uniform across the samples examined, different patterns might be observed in non-English language studies or in studies of the relatively under-sampled developing world. More research into issues surrounding SSHB internationally is desperately needed.

Unexamined Potential Confounds
A number of factors which may have affected individuals' behavior or the way in which they coped with exposure to trauma, and which therefore might have an impact on the research findings reviewed here, were largely ignored in the literature. Few papers considered for example, the length of time which had elapsed since exposure, and many only asked participants to report on recent exposure (e.g., within the past year). The number of exposures experienced by an individual was similarly overlooked, so it is unclear whether numerous exposures are more likely to result in habituation or cumulative distress. The closeness of the relationship between the child=adolescent and the other(s) to whose SSHB they were exposed might also have determined the impact of that exposure, and how profoundly it was felt or experienced, but exploration of this was limited. Whether or not individuals sought or received any support in dealing with their exposure to others' SSHB might also have altered outcomes for them, but this was also generally omitted from the literature. Finally, a number of the longitudinal studies reviewed failed to provide any information on baseline levels of SSHB or exposure, rendering it difficult to determine whether exposure to the SSHB of others actually resulted in children's=adolescents' increased SSHB, or whether those young people were either already engaging in SSHB themselves, or had previously experienced the SSHB of others.

Suggested Explanations for Findings
A number of possible explanations exist for the findings of this review. The first issue worth consideration is the causal direction between individuals' own behavior and their reports of that of others. It is possible that individuals who engage in SSHB erroneously report that they know others who also do so, on account of their believing that others probably behave in similar ways to them (as is the case with the false consensus effect; Prinstein & Wang, 2005), or that individuals tend to associate with individuals who behave in similar ways to them (e.g., Joiner, 2003). Evidence exists that although peer-selection effects may play a role, socialization effects are almost certainly present You et al., 2013), and the associations found between family members with whom one does not choose to associate and the contagion effects in forced settings such as hospitals (e.g., Gould, Petrie, Kleinman, & Wallenstein, 1994) or police custody (e.g., Cox & Skegg, 1993) argue in favor of socialization effects as an explanation. Rosen and Walsh (1989) suggest that a need to belong to groups may partly contribute to the clusters of self-harm which they observed in adolescent inpatient settings, so conformity to perceived norms may play an important role in the transmission of these behaviors. A related mechanism through which such behaviors are transmitted is proposed by Taiminen (1992), who suggests that out of empathy for a fellow human being who has suffered, individuals may project their best qualities onto people who engage in suicidal behavior, thus increases the extent to which they can relate to those individuals, inadvertently resulting in an increased capacity to relate to the suicidal behavior itself. By this logic, if individuals believe suicidal behavior to be widespread or normative among people they know, their ability to relate to it may be increased, and their risk of engaging therein thus increased also.
The nature of SSHBs specifically may make them particularly prone to social influence. Allen et al. (2006) found that those participants who are more susceptible to social influence are also more prone to psychological problems such as depressive symptoms. Given that depression is relatively common in those who engage in SSHB (and vice versa), associations observed between individuals' behavior and that of people they know may be the result of a cumulative effect of both depression and a greater propensity for social conformity. Indeed, Mittendorfer-Rutz, Rasmussen, and Wasserman (2008) claim that the associations they found between family suicidal behavior and individuals' own may be the result of a combination of both imitation or social modelling, and a genetic predisposition to psychiatric disorder. Watkins and Gutierrez (2003) propose a diathesis-stress model of the effects of exposure to others' suicidal behavior. They suggest that simply having knowledge of an individual ending their life would not in itself trigger another individual to do the same, but that if subsequent events occur for that individual which cause them distress with which they struggle to cope, they might recall that someone they knew ''solved their problems'' by ending their life, and see suicide as a feasible option to solve their own problems. In support of this notion are the findings of Swanson and Colman (2013), who found that exposure to the suicidal death of someone known personally predicted adolescent suicidal ideation and attempts 2 years later, but only in the presence of previous stressful life events. It is possible that these proposed effects hold for perceived normative SSHB as well as for specific instances of exposure; that is, the belief that other people engage in SSHB may act as a prompt for one's own, given a particular threshold of distress has been reached. This notion is in keeping with O'Connor's (2011) IMV model of suicidal behavior, in which suicidal ideation is proposed only to convert into action given certain additional motivational and volitional triggers; e.g., the belief that others in one's social network engage in SSHB. The findings of O'Connor et al. (2012) support this; reports of friends and family engaging in SSHB were identified as a volitional factor that differentiated adolescent self-harm ideators from enactors.

A Possible Protective Effect of Exposure to Suicide Death
A common (although not absolute) finding throughout the current review is that exposure to a suicide death was less often associated with children's=adolescents' own behavior, than was exposure to other, non-fatal behaviors. This may indicate that experiencing the death of someone else by suicide may have a qualitatively different impact on an individual than does witnessing a non-fatal attempt, or non-fatal self-harm. One study that looked at both family suicide attempts and deaths found that adolescent suicidal behavior was related to first-degree relatives' suicide attempts, but not deaths (Pfeffer et al., 1994), and the same pattern has been found with regard to friends (Ho et al., 2000). Notably, in all three of the family studies, and three out of the four friends=peers studies that found no associations with children's=adolescents' behavior, fatal behaviors were focused upon. Anecdotal evidence has also suggested that exposure to suicide deaths may in fact work to inhibit the suicidal behavior of an individual; as a result of witnessing the damage and misery it can cause (Brent et al., 1996).
Further support for this ''protective'' notion can be taken from the consistently reported increase in suicidal behavior following mass-media reporting of celebrity or high-profile suicides, internationally (e.g., Cheng, Hawton, Lee, & Chen, 2007;Etzersdorfer, Voracek, & Sonneck, 2004;Pirkis, Burgess, Francis, Blood, & Jolley, 2006). In this kind of ''remote'' or impersonal situation, individuals may be exposed to details of the suicide (which they can use to imitate it) and characteristics of the deceased (to which they might relate), but are never exposed to the pain suffered by the bereaved (which may have acted as a deterrent). Indeed, another paper in the current review  found that media reporting of suicide had a greater influence on suicidal behavior than did the suicidal behavior of people known to the individual. Some authors even argue that this apparent inhibitory effect may not be specific to fatal behaviors only. Hasking et al. (2013) argue that the protective effect they found against engaging in self-harm in those who knew others who did so, may be due to their having experience of the impact that self-harming has on those around the individual. These ideas are of course speculative, and require further investigation. A handful of other studies reviewed which also looked at suicide deaths found positive associations with adolescents' behavior (e.g., Bridge et al., 1997;Cheng et al., 2014;Garfinkel et al., 1982), so the notion of a protective effect cannot provide an adequate explanation in all circumstances. Brent et al. (1992) provide a potential alternative explanation for the lack of associations found between adolescents' behavior and their reports of knowing someone who has died by suicide. Far from experiencing a protective effect, their participants who were exposed to friends' suicide death had higher lifetime exposure to suicidality prior to the ''target'' death, than those who were not (currently) exposed, such that previous exposure had had a habituating effect resulting in less distress following subsequent exposure. Alternatively, the previous exposures may have resulted in those individuals being at an already optimal level of distress, with an increased (compared to those without exposure) but stable risk of suicidality that subsequent exposures did not affect. The data support this latter suggestion, with those with exposure exhibiting higher levels of past, current, and new-onset psychiatric disorder than those without, suggesting that the exposed individuals are indeed operating at an increased level of psychological distress. As the majority of papers fail to take into account past exposure, an already established optimum impact of exposure previous to the one currently studied cannot be ruled out, and an apparent lack of association may simply be an artefact of this effect.

Limitations of the Review Process
The current review was susceptible to many of the limitations common to other reviews, particularly on account of the strict inclusion criteria employed. For example, the inclusion of only peer-reviewed journal articles necessarily excludes the grey literature, which it has been claimed is likely to result in exaggerated reports of effects (e.g., McAuley, Pham, Tugwell, & Moher, 2000). It is possible therefore that the findings of the current review overstate associations as a result of publication bias of positive findings. However, the presence of several papers in the current review reporting negative findings may somewhat minimize concerns in this regard.
Another potential limitation intrinsic to this and many reviews is the use of the same data set by authors of multiple papers. Specifically, many of the papers reviewed here use data from the National Longitudinal Survey of Adolescent Health (Add Health), which despite consisting of high-quality, seemingly generalizable data, renders the overall data set under review somewhat smaller than it at first appears, and the multiple studies which use that data, susceptible to similar limitations. Independent findings, however, repeatedly support those of the Add Health survey, so this concern may also be nominal.
Finally, due to the diversity of methods, samples, analyses and definitions employed by study authors, meta-analyses were unfortunately not feasible, such that the current review is limited to providing a descriptive summary of findings. Given that meta-analysis is increasingly considered the gold-standard of research synthesis, it is regrettable that it was not possible in this case. This is perhaps indicative of the disparate terminology and divergence of theoretical approach across the SSHB literature generally. A more consistent, uniform approach across the field would arguably aid in clarifying some of the issues that remain uncertain.

Future Directions and Practical Implications
Although associations between children's=adolescents' SSHB and that of people they know are apparent, the current review identified a number of conflicting findings, so firstly, systematic research around the factors which affect associations (e.g., nature of relationship to others, behavior in question, psychopathology and environmental characteristics) is necessary to determine exactly where associations lie, in order that they might be addressed through intervention.
Furthermore, research is necessary to determine the exact mechanism(s) by which associations between child= adolescent SSHB and that of people they know occur. Research to date has provided a mixture of findings, and a more comprehensive understanding, using more systematic approaches, may assist in the development of effective interventions. For instance, if the SSHB of other people impacts upon that of a child=adolescent through socialization processes, service providers might aim to introduce assessment of exposure to such behaviors when assessing risk. This may help to identify those at high risk as a result of exposure, and in particular those for whom risk may be especially high as a result of exposure combined with other, more classical risk factors (e.g., depression, impulsivity). Alternatively, if SSHB develops (or is maintained) as a result of shared group identity or reward processes, interventions should be designed which address the social constructs behind these identities, and aim to provide alternatives.
Research is also needed to determine the extent to which normative perceptions impact upon the associations evident in the research to date. If it is merely the perception of others' SSHB and attitudes towards that behavior which is associated with a child's=adolescent's own, rather than the behaviors or attitudes themselves, more information about those perceptions would be useful. If heightened perceptions of SSHB in others or perceptions of more positive views of those behaviors in others are sufficient to increase one's own engagement, interventions should be designed which aim to address these perceptions and promote healthier norms, thereby potentially reducing any related increase in behavior. These types of interventions have proven effective in reducing engagement in a wide array of other health-damaging behaviors, and may be similarly effective in reducing SSHB. Indeed, Wang et al. (2011) note the importance of designing school-based programs which focus on increasing appropriate peer norms and improving attitudes towards life and help-seeking.
As it stands, the current findings highlight the potential impact of other people's actual or perceived SSHB in the development of young people's own SSHB. As such, schools, families and professionals working with young people should aim to familiarize themselves with the social environments in which their young people operate, risk assess and monitor the well-being of their young people, and aim to educate their young people on self-care and available sources of support and advice. Where feasible, monitoring of actual or perceived SSHB in friends and family, and responding with the appropriate support, might prove a valuable addition to existing practices designed to protect young people from harm.

CONCLUSION
Overall, the current review identified a vast array of published evidence for positive associations between children's=adolescents' SSHB and that of people in their social networks. Methodological inconsistencies make direct comparison and synthesis of findings across the literature difficult, but despite variation in methods, samples and settings, the identification of associations is highly consistent (perhaps with the exception of the suicide death of others, which is slightly less consistently associated with an individual's own SSHB). The findings of this review suggest that associations exist internationally, and the existence of such widespread associations warrants further investigation. In particular, findings highlight the potential utility of considering the impact of social networks during intervention development.
One factor that potentially underlies many of the studies reviewed, and which may result in negative consequences, is the potential for a discrepancy between the extent to which individuals believe others are engaging in SSHB, and the extent to which they actually are. Findings from social norms research in other behavioral domains indicate that perception of others' behavior does not always correspond with what those others report themselves, and heightened perceived norms are consistently related to an increase in one's own behavior (e.g., Borsari & Carey, 2003;Clemens et al., 2008;Labrie et al., 2009;Lewis & Neighbors, 2004). The literature reviewed here relies heavily on self-reports of others' behavior, so it is possible that these reports are overestimated, and that individuals' own SSHB is increased as a result. The extent to which young people's perceptions of others' SSHB are discrepant from reality, and whether or not those perceptions influence young people's own SSHB is an important, yet entirely under-researched consideration. Future research should focus on assessing the impact of normative perceptions on young people's SSHB and explore the mechanisms through which influence is exerted, with a view to developing preventative interventions.

AUTHOR NOTES
Jody Quigley, Division of Psychology, Faculty of Natural Sciences, University of Stirling, Stirling, UK.