Introduction: We investigated the longitudinal course of self-, other-, and dual-harm in adolescents, focusing on the infliction of physical injury on oneself, another person, or both parties, respectively. We examined the within-person transitions between these types of harm and whether relationships with peers and teachers predict individual harm trajectories. Methods: We used community-representative longitudinal data (N = 1,482; 52% male; 50% both parents born abroad). The participants self-reported self- and other-harm at 13, 15, 17, and 20 years. We assigned them to groups with self-, other-, dual- or no harm at specific assessments. Bullying victimization and relationship quality with classmates and teachers were assessed at 13 and 17. We estimated transition probabilities between the harm groups using latent Markov chain models. Results: At age 13, 3% of the sample engaged in dual-harm, 10% in self-harm only, and 7% in other-harm only. These percentages decreased in late adolescence. Initial dual-harm was often followed by sex-specific single-harm: most of the female participants transitioned to self-harm, and male participants to other-harm. Those in the initial dual-harm group were less likely to stop harming than those in the initial single-harm groups (p < 0.05). Adverse relationship experiences generally predicted harm. A positive teacher-student bond was associated with the cessation of single-harm. Conclusion: Single- and dual-harm in the form of physical injury typically emerge by mid-adolescence. After this point, adolescents commonly maintain harm, especially those who have presented with dual-harm. Helping adolescents cope with adverse relationship experiences and creating opportunities for positive relationship experiences could address these harmful behaviors.

Self-harm (e.g., injuring oneself by self-cutting with or without suicidal intent) and other-harm (e.g., physically injuring others) typically emerge during adolescence [1, 2]. These two sets of behaviors have important differences. For example, self-harm often occurs in private settings, whereas other-harm necessarily involves others [3, 4]. Furthermore, self- and other-harm have some distinct risk factors (e.g., internalizing psychopathology is a common risk factor for self-harm and anti-social behaviors for other-harm) [3, 5]. Female adolescents have a higher risk for self-harm than male adolescents, who, in turn, engage in other-harm more often than female adolescents [5, 6]. Yet, self- and other-harm also share risk factors such as childhood adversity and victimization [3, 7]. In addition, they are both often preceded by severe interpersonal stress or emotional dysregulation [8]. Although self- and other harm are rarely studied together, recent research has reported that 5–7% of the youth in the community engage in both types of harm (i.e., dual-harm) either concurrently or sequentially [3, 7].

Adolescents who engage in dual-harm are a unique at-risk group. For instance, they have higher levels of psychological, physical, and social impairments than those engaging in either self- or other-harm (i.e., single-harm) [3, 7‒9]. However, there is limited information on the developmental trajectory of single- versus dual-harm because previous research has relied on cross-sectional, retrospective, or lifetime cumulative data [3, 7, 10]. Studies on either severe self- or other-harm have suggested that early adolescence may be the onset period of these behaviors and adulthood may be a period of decline [1, 11‒14]. Nevertheless, it remains unclear when adolescents typically engage in single- versus dual-harm and how the different types of harm unfold over the individual life-span. To inform effective intervention mechanisms, it is important to pinpoint the ages at which single- and dual-harm are most prevalent and to identify which individuals are likely to keep engaging in harmful behaviors and which will disengage from them at some point.

The paucity of research on single- versus dual-harm extends to sex-based differences as well as the factors that promote risk or resilience. In previous papers based on our sample, the cumulative prevalence of dual-harm during adolescence was less sex-differentiated than that of single-harm [3], and sex differences in self-harm emerged after early adolescence [13]. These findings were not informative, however, about the developmental course of single- versus dual-harm in females and males.

As for factors promoting risk and resilience, childhood social adversity is a distal risk factor for subsequent self-, other-, and, in particular, dual-harm [3, 7, 15]. However, little is known about how current social relationships in adolescence shape within-person longitudinal trajectories of harm. During adolescence, peer and school contexts gain significance in the areas of individual development and mental health [16‒18]. Notably, stressful experiences in these contexts, such as victimization due to bullying and stress at school, become proximal risk factors for harm [16, 19]. On the resilience side, a positive teacher-student relationship can safeguard against the emergence of internalizing and externalizing symptoms [20], which are well-established correlates of self- and other-harm [3, 21]. Regardless, prior research has not illuminated the role of specific adolescent relationship experiences in single- and dual-harm trajectories. For example, it is unknown whether an improvement in the quality of relationships with classmates or teachers can help adolescents cease harming.

In this study, we investigated the development of single- and dual-harm across adolescence and the associated role of social relationships. Specifically, we examined (1) the age- and sex-related prevalence of harm, (2) how single- and dual-harm unfold over the individual lifespan in male and female adolescents, and (3) the possible association between individual harm trajectories and bullying victimization and supportive bonds with classmates and teachers. We hypothesized that adverse social experiences and a lack of positive ones are associated with an increased risk of engaging in harm. Likewise, we speculated that an improvement in relationship quality over time may be associated with an increased likelihood of stopping harmful behaviors. We also explored whether these associations differed between groups with initial incidents of either single- or dual-harm.

Participants and Procedures

We leveraged data from the community-representative longitudinal Zurich Project on Social Development from Childhood to Adulthood (z-proso; [22]). In 2004, 1,675 children from 56 primary schools were randomly selected from 90 public schools in Zurich, Switzerland’s largest city, to form the target sample. The sample stratification accounted for the sizes of the schools and the socioeconomic backgrounds (SESs) of the school districts. The sample was largely representative of first graders who attended public schools in Zurich. The participants were followed until they were 20 years old, which was in 2018. We used data from those who participated at least once from the ages of 13 to 20, which is when the sample was assessed for self- and other-harm (n = 1,482).

Information on participants’ biological sex was recorded at the onset of the study according to specifications in the school records. These records make use of data from the local population authority, which in turn contain information from each participant’s respective maternity clinic or, if they were born abroad, their passport. According to the legal requirements in Switzerland, clinics must designate the sex of an individual as either male or female. Hence, nonbinary individuals must be assigned to one of these categories. While further assessments will collect additional self-reported information on the gender identities of the participants, this information is currently unavailable. Of the study sample, 52% were categorized as male and 48% as female. Consistent with Zurich’s diverse population, the parents of the participants had been born in more than 80 different countries, most frequently in the former Yugoslavia, Portugal, Sri Lanka, Turkey, Germany, and Italy (see [13]). Overall, 50% of the participants had two parents born abroad, and 91% had been born in Switzerland. In 26% of the families, at least one parent held a university degree.

Data were collected in classroom-based settings with paper-and-pencil questionnaires for the participants up to age 17 and in a computer laboratory with computer-administered questionnaires at age 20. It took the participants roughly 90 min to complete each questionnaire. The participants received a cash incentive that increased from roughly USD 30 at age 13 to USD 75 at age 20. This study was approved by the ethics committee of the Faculty of Arts and Social Sciences, University of Zurich. The participants provided their written informed consent at each wave, and parents of children younger than 15 could decline their children’s participation in the study.

Variables

Indicators of Harm

The present study focused on harm involving physical injury. The participants reported incidents of self-harm at the ages of 13, 15, 17, and 20 using one item that indicated how often they had injured themselves on purpose during the previous month [13]. The example behaviors provided were “cut my arm,” “tore open wounds,” “hit my head,” and “tore out my hair.” The participants recorded their answers on a five-point scale (1 = never, 2 = rarely, 3 = sometimes, 4 = often, 5 = very often), and we dichotomized the assessments (0 = no self-harm and 1 = any self-harm). The assessment did not distinguish between suicidal and nonsuicidal self-injury. However, the examples primarily involved nonsuicidal self-injury. In addition, the participants reported incidents of other-harm at the ages of 13, 15, 17, and 20 using an indicator of assault taken from a broader delinquency scale developed by the z-proso team. The participants indicated whether they had “purposely hit, kicked, or cut someone and injured him or her in the process” and, if so, how often they had done this in the previous year. Potential victims included familiar and unfamiliar peers and adults. We dichotomized the assessments (0 = no other-harm and 1 = at least one incidence of other-harm).

We coded dual-harm in two ways. First, to analyze the average longitudinal course of single- versus dual-harm, we assigned the participants who reported both self- and other-harm at the same assessment to the dual-harm group. Likewise, we assigned the participants who reported self-harm but did not report other-harm to a self-harm-only group. Conversely, we assigned the participants who reported other-harm but not self-harm to an other-harm-only group. We assigned all the other participants to the no-harm group. Second, to analyze the within-person trajectories of harm, we combined assessments taken at the ages of 13 and 15 to represent harm in early and mid-adolescence (for readability, we henceforth refer to this period as “mid-adolescence”), which is the typical onset period of severe self- and other-harm [2, 23]. We also combined the assessments at the ages of 17 and 20 to represent late-onset harm as well as periods of harm maintenance versus cessation in late adolescence. We assigned the participants who reported self- and other-harm, either concurrently or sequentially, from the ages of 13 to 15 to the initial (i.e., mid-adolescent) dual-harm group and the participants who reported these forms of harm from the ages of 17 to 20 to the later (i.e., late adolescent) dual-harm group. We created initial and later single- and no-harm groups in a similar fashion.

Predictors and Control Variables

The relationship experiences of the participants were measured at the ages of 13 and 17. These are the respective onset ages of the assessment periods that we distinguished in our harm trajectory analyses (i.e., mid- and late adolescence). Bullying victimization was assessed with four items from the Zurich Brief Bullying Scale [24] at ages 13 (Cronbach’s α = 0.77) and 17 (α = 0.69). For example, the participants reported how often other people ignored or excluded them, laughed at them, mocked them, or insulted them in the previous 12 months (1 = never to 6 = almost every day; M = 1.70 [SD = 0.77] at age 13, M = 1.45 [SD = 0.57] at age 17). The relationship with classmates and teacher-student relationship were assessed with two three-item scales created by the z-proso team (relationship with classmates: e.g., “We have a really good sense of community within the class,” α = 0.77 at age 13, α = 0.75 at age 17; relationship with the main teacher: e.g., “My teacher treats me fairly,” “I get along with my teacher,” α = 0.77 at age 13, α = 0.78 at age 17). The participants provided their answers on a four-point Likert scale (1 = fully untrue to 4 = fully true). We recoded the scales so that higher values indicated less positive or more negative experiences to ease interpretation of the results (M [classmates] = 1.71 [SD = 0.61] at age 13, M = 1.67 [SD = 0.56] at age 17, M [teacher] = 1.85 [SD = 0.65] at age 13, M = 1.85 [SD = 0.59] at age 17). We kept 17-year-olds who did not attend educational training after completing their compulsory schooling (7%) in the main models (see below for information on how we handled missing data). We also replicated the results with a subsample limited to those participants who currently attended school (i.e., high school [29%], vocational training [54%], or other training, such as apprenticeships [10%]).

We adjusted our models for sex (0 = female, 1 = male) and household SES. The latter was assessed using the International Socio-Economic Index of Occupational Status (ISEI) [25], which is based on an occupation-specific income and required educational level. The scores ranged from 16 (e.g., unskilled worker) to 90 (e.g., judge; M = 45.74 [SD = 19.24]). We also tested whether academic achievement confounded any associations between the teacher-student bond and harm. We considered the grade point average at age 13, calculated as the mean of teacher-reported grades in math and language, and the type of school (i.e., high school, vocational, or other) at age 17.

Analytic Strategy

We examined the prevalence of single- and dual-harm from the ages of 13 to 20 using descriptive statistics and binary logistic regression models (to estimate sex differences) in SPSS. These analyses of the longitudinal course of single- and dual-harm are suitable for identifying periods in adolescence during which the average risk of harm is high. To examine how harm unfolds over the individual life span, we analyzed the within-person transitions between the four harm groups in mid-adolescence (i.e., 13–15 years) and thereafter (i.e., 17–20 years) using latent Markov chain models [26]. We specified these models in MPlus, using a maximum likelihood estimator with robust standard errors. We handled missing data with full information maximum likelihood [27]. The cell sizes for some transition patterns were small and thus rendered model estimation with covariates impossible in some cases. Therefore, we combined the three groups with harm in late adolescence (i.e., all participants with harm were assigned to a group with any harm) to analyze the predictors of transitions from self-, other-, and dual-harm in mid-adolescence to any versus no harm in late adolescence. In these models, relationship experiences were predictors of harm group membership in mid-adolescence (Part A) and the maintenance of any harm in late adolescence, depending on initial harm group membership (Part B; online suppl. Fig. S1 provides an illustration of the model specification; for all online suppl. material, see www.karger.com/doi/10.1159/000525296). Time-invariant covariates (i.e., sex and SES) were predictors of initial group membership and within-group differences in the transition probabilities [28]. We treated the indicators of relationship experiences as time-varying covariates. To assess the impact of changing relationships (e.g., a change in bullying victimization), we conditioned the effects of the experiences at age 17 on respective experiences at age 13.

Average Longitudinal Course of Single- and Dual-Harm

The prevalence of dual-harm decreased from the ages of 13 (3%) to 20 (<1%; Fig. 1). There were few or no changes in the prevalence of self-harm only (∼10%) and other-harm only (∼7%) from the ages of 13 to 15. Both forms of harm declined thereafter (8% and 3%, respectively, at age 20). Sex-related differences in single-harm emerged at each assessment: the female participants were more likely to engage in self-harm only, whereas the male participants were more likely to engage in other-harm only (Fig. 1). The male participants had a higher risk of engaging in dual-harm at age 13 but not thereafter.

Fig. 1.

Prevalence of single- and dual-harm from the ages of 13 to 20 in the overall sample and sex differences (results from the logistic regression analyses). Sex differences in dual-harm at age 20 could not be tested due to the low prevalence. For each comparison, the group (female/male) with the lower prevalence was chosen as the reference category to ease the interpretation of the odds ratios.

Fig. 1.

Prevalence of single- and dual-harm from the ages of 13 to 20 in the overall sample and sex differences (results from the logistic regression analyses). Sex differences in dual-harm at age 20 could not be tested due to the low prevalence. For each comparison, the group (female/male) with the lower prevalence was chosen as the reference category to ease the interpretation of the odds ratios.

Close modal

Within-Person Trajectories of Single- and Dual-Harm

The transition models included the participants who had completed both assessments in mid- or late-adolescence (n = 1,407; Fig. 2). In the following, we provide rounded estimates of group sizes (n) and transition probabilities based on the estimated models. The results revealed that the mid-adolescent dual-harm group (n = 78) had a higher risk of maintaining any harm in late adolescence than all other groups (comparison to the other-harm group: p = 0.002; self-harm group: p = 0.013; no-harm group: p < 0.001). The most common transition was from the dual-harm group to the self-harm group. The mid-adolescent self-harm (n = 208) and other-harm groups (n = 171) had similar risks of maintaining any harm later on (p = 0.384). However, the stability of the type of harm differed between the two groups. One in three members of the initial self-harm group remained in the self-harm group, whereas one in six members of the initial other-harm group remained in the other-harm group. One in eight members of the initial no-harm group (n = 950) transitioned to either the single- or dual-harm group in late adolescence, possibly indicating a later onset of harm.

Fig. 2.

Latent transition probabilities between harm groups in mid- and late adolescence based on the estimated model; the rank orders of the transitions are given for the overall sample (in parentheses) and female and male subsamples, with 1 indicating the most likely and 4 indicating the least likely transition (n= 1,407, n[male] = 720, n[female] = 687; including respondents with valid assessments at 13 and 15 or 17 and 20). For sex-specific transitions, we do not provide exact numbers because of the small cell sizes (n< 5) in some cases to avoid reporting potentially identifying information. We give select probabilities of relatively frequent transitions in the text.

Fig. 2.

Latent transition probabilities between harm groups in mid- and late adolescence based on the estimated model; the rank orders of the transitions are given for the overall sample (in parentheses) and female and male subsamples, with 1 indicating the most likely and 4 indicating the least likely transition (n= 1,407, n[male] = 720, n[female] = 687; including respondents with valid assessments at 13 and 15 or 17 and 20). For sex-specific transitions, we do not provide exact numbers because of the small cell sizes (n< 5) in some cases to avoid reporting potentially identifying information. We give select probabilities of relatively frequent transitions in the text.

Close modal

Our inspection of sex-specific transitions revealed that the female participants in the initial dual-harm group (n = 27) mostly transitioned to the self-harm group (58%), whereas the male participants in the dual-harm group (n = 51) mostly transitioned to either the no-harm (53%) or other-harm group (26%; Fig. 2). Self-harm was relatively stable in both the male (22% of n = 74) and female participants (38% of n = 135). Other-harm was often stable in the male participants (21% of n = 121). Most female members of the initial other-harm group who maintained harm transitioned to the self-harm group (29% of n = 50). While the members of the initial no-harm group (females: n = 476; males: n = 474) tended to stay in this group, the male participants who left predominantly transitioned to the other-harm group (9%), whereas female participants who left tended to fall into the self-harm group (10%).

Predictors of Initial Harm Group Membership and Subsequent Maintenance of Any Harm

The results of Part A of our models with covariates revealed that recent bullying victimization and a lack of positive relationship experiences with classmates and the main teacher were associated with an increased probability of membership in the mid-adolescent harm groups rather than the no-harm group (Table 1). Bullying victimization was also associated with an increased probability of membership in the initial dual-harm group compared to the other-harm group and, marginally, the self-harm group (online suppl. Table S1). The relationships with classmates varied marginally between the dual- and other-harm groups. Likewise, the relationship with the main teacher varied marginally between the dual- and single-harm groups.

Table 1.

Predictors of initial group (i.e., latent class) membership and subsequent transitions to any versus no harm (odds ratios with 95% confidence intervals in parentheses, p values in second row)

 Predictors of initial group (i.e., latent class) membership and subsequent transitions to any versus no harm (odds ratios with 95% confidence intervals in parentheses, p values in second row)
 Predictors of initial group (i.e., latent class) membership and subsequent transitions to any versus no harm (odds ratios with 95% confidence intervals in parentheses, p values in second row)

The results of Part B of the models defined the predictors of any harm (vs. no harm) in late adolescence and took initial harm group membership into account (Table 1). The female participants in either the mid-adolescent dual-harm or self-harm groups had a higher risk of subsequent harm maintenance than the male participants from these groups. For members in the initial dual-harm group, relationship experiences did not predict harm maintenance. In the single-harm groups, worsening of the teacher-student bond was associated with an almost three times (self-harm) and more than four times (other-harm) higher risk of harm maintenance rather than cessation. In other words, a supportive teacher-student relationship was associated with a higher likelihood of harm cessation. In contrast, changes in bullying victimization were not associated with harm maintenance versus cessation. Online supplementary Tables S2 and S3 show the replication of these results when educational achievement is taken into account. The worsening of relationships with classmates was moderately associated with an increased likelihood of transitioning from the initial no-harm group to the any harm group; bullying victimization had a marginal effect in the same direction.

To prevent the burden of self-, other-, and dual-harm among adolescents, we must improve our understanding of their longitudinal course and how everyday social relationship experiences may influence these behaviors. The results of our community-representative longitudinal study indicate that two out of three adolescents who initiate dual-harm by mid-adolescence will maintain harmful behaviors thereafter. This number exceeds the maintenance rates of groups with initial self-harm only (two out of five) and other-harm only (two out of six). In our results, adverse social relationship experiences were consistently associated with the emergence of harm, and a select few were also associated with the maintenance of harm.

As discussed in our previous paper [13], the prevalence of self-harm in our sample was generally within the range of the estimates reported in the international literature [1]. The rates of assault were slightly lower in our data compared to some previous evidence [14]. This could be due to our narrow definition of other-harm as causing physical injury. Adding to the previous evidence, our study found a considerable overlap between the self- and other-harm groups (i.e., a high prevalence of dual-harm), especially in early adolescence, which is a typical onset period of self-harm [11‒13] and other-harm [14]. Although the peak prevalence of dual-harm at age 13 is lower than the previously reported cumulative prevalence of dual-harm for the entire adolescent period (e.g., 7% in our own research [3]), the numbers jointly indicate that more than a third of the adolescents who switch back and forth between self- and other-harm do so within short periods (e.g., a year). Furthermore, our data revealed that the typical developmental trajectory is from dual- to single-harm. Thus, dual-harm may reflect the fact that early adolescents frequently experiment with different ways of coping with overwhelming stress.

Previous research has mostly supported the higher prevalence of self-harm in female populations compared to male populations but left room for further clarification [6]. In our sample, self-harm was not differentiated by sex before mid-adolescence [13]. Notably, the results presented here indicate that, starting in early adolescence, the female participants were more likely than the male participants to engage only in self-harm. Thus, future research on sex differences may elect to distinguish youths who only conduct self-harm from those with comorbid other-harm.

Furthermore, our analyses of the individual trajectories of harm showed that initial dual-harm is followed by the subsequent sex-specific channeling of harm into self-harm by female participants and other-harm by male participants. This dynamic may indicate that adolescents increasingly adhere to gender stereotypes as they age, given that other-harm is usually conceived of as a masculine behavior [29] and the example behaviors given in our self-harm assessment may over-represent female-typical behavior [30].

The low likelihood of transitioning to no harm after engaging in dual-harm highlights the strong need for effective interventions tailored especially to adolescents with dual-harm. These adolescents tend to have several unique psychological characteristics, such as low self-control [7], and more adverse childhood experiences than those with single-harm [3, 15]. Higher stress paired with poorer abilities to regulate emotions may, in and of themselves, explain the elevated risk of maintaining harmful behaviors in these adolescents. In addition, dual-harm is likely indicative of comorbid internalizing and externalizing psychopathology [3], which is a known risk factor for prolonged histories of psychological strain [31]. It is also possible that dual-harm represents a more generalized and frequent engagement in harmful behaviors (e.g., across different contexts). In turn, adolescents with dual-harm may have less room, time, and opportunity than those with single-harm to learn adaptive strategies for emotion regulation. To illuminate this further, future research is needed to identify differences and similarities in the form, function, severity, and frequency of self- and other-harm among adolescents with single- versus dual-harm.

Prior evidence, including other evidence based on our sample, has shown the strong impact of childhood victimization experiences and social disconnection on adolescent dual-harm in particular [3, 7]. Our study shows that similar associations exist between early adolescent relationship experiences and harm. These new results show that observations of students’ social relationships, for example in the school context, may help identify those who are currently at risk of harming and offer them support. In fact, the protective role of supportive relationships with classmates and teachers against single- and dual-harm could mirror the relevance of the school context to adolescents’ daily lives. In turn, the link between the recent worsening of relationships with classmates and the transition from no harm to any harm (i.e., the onset of harm) in late adolescence further supports the conclusion that adverse peer group dynamics at school can be precursors to the emergence of harm. Overall, our findings align with the conceptualization of harm as a behavioral response to overwhelming interpersonal stress [8]. One mechanism likely involved in these links is psychological strain, especially in the form of dysregulated negative emotions (e.g., sadness, anxiety, and anger) [32, 33]. However, our study design is not suitable for inferring causality, and the associations between harm and social relationship experiences can be of a reciprocal nature [34].

Finally, our study adds a new dimension to the existing knowledge of harm by providing insights into the role of social relationships in the within-person trajectory of harm after its first initiation. In this study, peer relationships (i.e., relationships with classmates and bullying victimization) were mainly involved in the emergence of harm. In contrast, a supportive teacher-student bond was associated with single-harm cessation (not the onset) in late adolescence. This emphasizes the relevance of adolescent-adult relationships in overcoming mental health problems, such as self- or other-harm or other potentially underlying psychopathology (e.g., depressive symptoms and personality disorders [21]). In particular, positive teacher-student bonds may encourage adolescents to disclose their problems and encourage teachers to facilitate access to professional support systems [35]. Positive teacher-student bonds may also increase adolescents’ sense of self-worth, self-efficacy, and belonging in the school context. In turn, this could decrease the urge to self-harm or inflict harm on others. Consequently, programs to improve the bonds between teachers and their students, which typically focus on school-related outcomes in the first place [36], could also be fruitful in counteracting prolonged histories of adolescent self-harm or other-harm. Plus, reducing time and performance pressure in the school context could provide more room for relationship-building between teachers and their students. In addition, improving mental health awareness among teachers and adolescents could be a promising avenue [37], enabling both agents to make even better use of the opportunities inherent to their relationship (e.g., by increasing their abilities to talk about mental health-related topics, if needed). Nevertheless, we cannot ignore the possibility of reverse causation. For example, when students cease other-harm behaviors, teachers may like these students more or be less afraid of them in turn, thus improving the teacher-student bond. Furthermore, associations between changes in social relationships and changes in harm could also result from changes in unobserved confounders, such as self-regulation skills.

The lack of an association between the teacher-student bond and harm maintenance in adolescents in the initial dual-harm group could indicate that adolescents with this severe pattern of harm have specific needs. However, the small group sizes in the transition analyses may have resulted in a lack of power to detect within-group differences. Therefore, future research with larger dual-harm groups is needed to identify the precursors of harm maintenance versus cessation among these adolescents.

Limitations

We assessed self- and other-harm with single-item self-reports, which may be prone to social desirability. Furthermore, we used a narrow definition of self-harm (i.e., direct self-injury) and severe other-harm (i.e., injury inflicted on another person), with the intent of combining a mirroring pair of actions (i.e., both involving physical injury) [3]. Using a wider definition for each of the two behaviors (e.g., including any violent offenses intended to damage another person) may have resulted in higher prevalence rates and different correlates for each group. Moreover, the 1-month time frame of the self-harm assessment may have resulted in an underestimation of the sizes of some groups, including the initial dual-harm group. Three-year time intervals between some assessments may have caused an underestimation of the rates of harm maintenance and the strength of the associations with the covariates measured at the beginning of each assessment period. In addition, we relied on official records of the participants’ biological sex, limited to a binary categorization, and were unable to examine the role of (nonbinary) gender in the development of self-, other-, and dual-harm. Finally, our models focused exclusively on the associations between social relationships and trajectories of single- and dual-harm. Future research is needed to assess the role of psychological strain beyond these associations.

Overall, our findings demonstrate that the majority of adolescents who report dual-harm as well as many of those who report single-harm are on a trajectory toward prolonged maintenance of harm. Poor relationships in peer and school contexts indicate a high risk of the emergence and maintenance of harm. Notably, our study shows the importance of strengthening adolescents’ capacities to cope with the challenges inherent to everyday social relationships and counteracting bullying victimization before harm becomes chronic. Interventions that create opportunities for certain positive social experiences, such as those with teachers, may be able to prevent and even interrupt individual trajectories of harm.

We are grateful to all the participants who provided data for the z-proso study and to the research assistants involved in data collection.

All procedures complied with the Helsinki Declaration of 1975, as revised in 2008. The study was approved by the Ethics Committee of the Faculty of Arts and Social Sciences, University of Zurich (2018.2.12). The participants provided their written informed consent at each wave, and the parents of children younger than 15 years of age could choose not to have their children participate in the study.

During the previous 3 years, Denis Ribeaud, Manuel Eisner, and Lilly Shanahan received funding for independent research from the Swiss National Science Foundation.

Funding for independent fundamental research was received from the Swiss National Science Foundation (100014_132124, 100014_149979, 10FI14_170409, 10FI14_198052/1, 10531C_189008) and the Jacobs Foundation (2010-888, 2013-1081-1). The content presented here is solely the responsibility of the authors and does not necessarily represent the views of these foundations.

Annekatrin Steinhoff and Lilly Shanahan conceived of the paper. Manuel Eisner and Denis Ribeaud designed the z-proso study and collected the data. Annekatrin Steinhoff analyzed the data. Annekatrin Steinhoff conceptualized and drafted the manuscript, with critical input from Lilly Shanahan. Annekatrin Steinhoff, Denis Ribeaud, Manuel Eisner, and Lilly Shanahan critically revised the manuscript for important intellectual content.

The datasets generated for this study will not be made publicly available due to the presence of sensitive personal information. Scientists can direct requests for reanalysis to the first author.

1.
Plener
PL
,
Schumacher
TS
,
Munz
LM
,
Groschwitz
RC
.
The longitudinal course of non-suicidal self-injury and deliberate self-harm: a systematic review of the literature
.
Borderline Personal Disord Emot Dysregul
.
2015
;
2
:
2
.
2.
Loeber
R
,
Hay
D
.
Key issues in the development of aggression and violence from childhood to early adulthood
.
Annu Rev Psychol
.
1997
;
48
:
371
410
.
3.
Steinhoff
A
,
Bechtiger
L
,
Ribeaud
D
,
Eisner
M
,
Shanahan
L
.
Self-, other-, and dual-harm during adolescence: a prospective-longitudinal study of childhood risk factors and early adult correlates
.
Psychol Med
.
2022
:
1
9
.
Epub ahead of print
. http://dx.doi.org/10.1017/S0033291722000666.
4.
Hooley
JM
,
Franklin
JC
.
Why do people hurt themselves? A new conceptual model of nonsuicidal self-injury
.
Clin Psychol Sci
.
2017
;
6
(
3
):
428
51
.
5.
Farrington
DP
.
Origins of violent behavior over the life span
. In:
Vazsonyi
AT
,
Flannery
DJ
,
DeLisi
M
, editors.
The Cambridge handbook of violent behavior and aggression
.
Cambridge
:
Cambridge University Press
;
2018
. p.
3
30
.
6.
Bresin
K
,
Schoenleber
M
.
Gender differences in the prevalence of nonsuicidal self-injury: a meta-analysis
.
Clin Psychol Rev
.
2015
;
38
:
55
64
.
7.
Richmond-Rakerd
LS
,
Caspi
A
,
Arseneault
L
,
Baldwin
JR
,
Danese
A
,
Houts
RM
,
.
Adolescents who self-harm and commit violent crime: testing early-life predictors of dual harm in a longitudinal cohort study
.
Am J Psychiatry
.
2019
;
176
(
3
):
186
95
.
8.
Shafti
M
,
Taylor
PJ
,
Forrester
A
,
Pratt
D
.
The co-occurrence of self-harm and aggression: a cognitive-emotional model of dual-harm
.
Front Psychol
.
2021
;
12
:
586135
.
9.
Steeg
S
,
Webb
RT
,
Mok
PLH
,
Pedersen
CB
,
Antonsen
S
,
Kapur
N
,
.
Risk of dying unnaturally among people aged 15–35 years who have harmed themselves and inflicted violence on others: a national nested case-control study
.
Lancet Publ Health
.
2019
;
4
(
5
):
e220
-
8
.
10.
O’Donnell
O
,
House
A
,
Waterman
M
.
The co-occurrence of aggression and self-harm: systematic literature review
.
J Affect Disord
.
2015
;
175
:
325
50
.
11.
Brown
RC
,
Plener
PL
.
Non-suicidal self-injury in adolescence
.
Curr Psychiatry Rep
.
2017
;
19
(
3
):
20
.
12.
Muehlenkamp
JJ
,
Xhunga
N
,
Brausch
AM
.
Self-injury age of onset: a risk factor for NSSI severity and suicidal behavior
.
Arch Suicide Res
.
2019
;
23
(
4
):
551
63
.
13.
Steinhoff
A
,
Ribeaud
D
,
Kupferschmid
S
,
Raible-Destan
N
,
Quednow
BB
,
Hepp
U
,
.
Self-injury from early adolescence to early adulthood: age-related course, recurrence, and services use in males and females from the community
.
Eur Child Adolesc Psychiatry
.
2021
;
30
(
6
):
937
51
.
14.
Shulman
EP
,
Steinberg
LD
,
Piquero
AR
.
The age-crime curve in adolescence and early adulthood is not due to age differences in economic status
.
J Youth Adolesc
.
2013
;
42
(
6
):
848
60
.
15.
Carr
MJ
,
Steeg
S
,
Mok
PLH
,
Pedersen
CB
,
Antonsen
S
,
Kapur
N
,
.
Adverse childhood experiences and risk of subsequently engaging in self-harm and violence towards other people-“dual harm”
.
Int J Environ Res Public Health
.
2020
;
17
(
24
):
9409
.
16.
Steinhoff
A
,
Bechtiger
L
,
Ribeaud
D
,
Eisner
M
,
Shanahan
L
.
Stressful life events in different social contexts are associated with self-injury from early adolescence to early adulthood
.
Front Psychiatry
.
2020
;
11
:
487200
.
17.
Prinstein
MJ
,
Giletta
M
.
Peer relations and developmental psychopathology
. In:
Cicchetti
D
, editor.
Developmental psychopathology: theory and method
.
Hoboken, NJ
:
Wiley
;
2016
. p.
1
53
.
18.
Nelson
EE
,
Leibenluft
E
,
McClure
EB
,
Pine
DS
.
The social re-orientation of adolescence: a neuroscience perspective on the process and its relation to psychopathology
.
Psychol Med
.
2005
;
35
:
163
74
.
19.
Karanikola
MNK
,
Lyberg
A
,
Holm
AL
,
Severinsson
E
.
The association between deliberate self-harm and school bullying victimization and the mediating effect of depressive symptoms and self-stigma: a systematic review
.
Biomed Res Int
.
2018
;
2018
:
4745791
.
20.
Murray
AL
,
Obsuth
I
,
Speyer
L
,
Murray
G
,
McKenzie
K
,
Eisner
M
,
.
Developmental cascades from aggression to internalizing problems via peer and teacher relationships from early to middle adolescence
.
J Youth Adolesc
.
2021
;
50
(
4
):
663
73
.
21.
Nock
MK
,
Joiner
TE
Jr
,
Gordon
KH
,
Lloyd-Richardson
E
,
Prinstein
MJ
.
Non-suicidal self-injury among adolescents: diagnostic correlates and relation to suicide attempts
.
Psychiatry Res
.
2006
;
144
(
1
):
65
72
.
22.
Ribeaud
D
,
Murray
AL
,
Shanahan
L
,
Shanahan
M
,
Eisner
M
.
Cohort profile: the Zurich project on social development from childhood to adulthood (z-proso)
.
J Dev Life Course Criminol
.
2022
;
8
(
1
):
151
71
.
23.
Plener
PL
,
Kaess
M
,
Schmahl
C
,
Pollak
S
,
Fegert
JM
,
Brown
RC
.
Nonsuicidal self-injury in adolescents
.
Dtsch Arztebl Int
.
2018
;
115
:
23
30
.
24.
Murray
AL
,
Eisner
M
,
Ribeaud
D
,
Kaiser
D
,
McKenzie
K
,
Murray
G
.
Validation of a brief self-report measure of adolescent bullying perpetration and victimization
.
Assessment
.
2021 Jan
;
28
(
1
):
128
40
.
25.
Ganzeboom
HBG
,
De Graaf
PM
,
Treiman
DJ
,
De Leeuw
J
.
A standard international socio-economic index of occupational status
.
Social Sci Res
.
1992
;
21
(
1
):
1
56
.
26.
Kaplan
D
.
An overview of Markov chain methods for the study of stage-sequential developmental processes
.
Dev Psychol
.
2008
;
44
(
2
):
457
67
.
27.
Schafer
JL
,
Graham
JW
.
Missing data: our view of the state of the art
.
Psychol Methods
.
2002
;
7
(
2
):
147
77
.
28.
Muthen
BO
,
Asparouhov
T
.
LTA in Mplus: transition probabilities influenced by covariates
.
2011
. Available from: https://www.statmodel.com/examples/LTAwebnote.pdf.
29.
Fleming
PJ
,
Agnew-Brune
C
.
Current trends in the study of gender norms and health behaviors
.
Curr Opin Psychol
.
2015
;
5
:
72
7
.
30.
Sornberger
MJ
,
Heath
NL
,
Toste
JR
,
McLouth
R
.
Nonsuicidal self-injury and gender: patterns of prevalence, methods, and locations among adolescents
.
Suicide Life Threat Behav
.
2012
;
42
(
3
):
266
78
.
31.
Sourander
A
,
Jensen
P
,
Davies
M
,
Niemelä
S
,
Elonheimo
H
,
Ristkari
T
,
.
Who is at greatest risk of adverse long-term outcomes? The Finnish from a boy to a man study
.
J Am Acad Child Adolesc Psychiatry
.
2007
;
46
(
9
):
1148
61
.
32.
Kretschmer
T
.
What explains correlates of peer victimization? A systematic review of mediating factors
.
Adolesc Res Rev
.
2016
;
1
(
4
):
341
56
.
33.
Steinhoff
A
,
Bechtiger
L
,
Ribeaud
D
,
Murray
AL
,
Hepp
U
,
Eisner
M
,
.
Self-injury and domestic violence in young adults during the COVID-19 pandemic: trajectories, precursors, and correlates
.
J Res Adolesc
.
2021
;
31
(
3
):
560
75
.
34.
Burke
TA
,
Hamilton
JL
,
Abramson
LY
,
Alloy
LB
.
Non-suicidal self-injury prospectively predicts interpersonal stressful life events and depressive symptoms among adolescent girls
.
Psychiatry Res
.
2015
;
228
(
3
):
416
24
.
35.
Halladay
J
,
Bennett
K
,
Weist
M
,
Boyle
M
,
Manion
I
,
Campo
M
,
.
Teacher-student relationships and mental health help seeking behaviors among elementary and secondary students in Ontario Canada
.
J Sch Psychol
.
2020
;
81
:
1
10
.
36.
Duong
MT
,
Pullmann
MD
,
Buntain-Ricklefs
J
,
Lee
K
,
Benjamin
KS
,
Nguyen
L
,
.
Brief teacher training improves student behavior and student-teacher relationships in middle school
.
Sch Psychol
.
2019
;
34
(
2
):
212
21
.
37.
Fortune
S
,
Sinclair
J
,
Hawton
K
.
Adolescents’ views on preventing self-harm. A large community study
.
Soc Psychiatry Psychiatr Epidemiol
.
2008
;
43
(
2
):
96
104
.