Dating violence (DV) is a highly prevalent issue that can have serious negative consequences on its victims, including depression (Kamimura et al., 2016), and predicts involvement in relationships with intimate partner violence (IPV) in adulthood (Temple et al., 2016; Taquette & Monteiro, 2019). Abuse, control, and harassment perpetrated by a (ex)dating partner can occur face-to-face (physical, verbal-emotional, sexual violence, threats; Fernández-González et al., 2012) or via digital technologies such as texting on mobile phones and social media (cyber-DV; Fernet et al., 2019). Studies estimate that between 12 to 90% of adolescents are victims of cyber-DV. This wide disparity in rates is due to significant differences in terminology use, construct definitions and methodological characteristics (e.g., sample characteristics, measures used; Caridade et al., 2019; Stonard et al., 2017). Evidence suggests that adolescent girls, as compared to boys, report more cyber-DV victimization (particularly sexual victimization; Cava et al., 2020; Howell, 2016; Reed et al., 2017; Zweig et al., 2013, 2014) and greater levels of distress as a result (Reed et al., 2017; Smith et al., 2018). Given its high prevalence rates and serious repercussions on the well-being of adolescent girls, at a vulnerable stage in their development, the need for effective preventive interventions is evident. While there is increasing interest in the topic, a holistic examination of risk and protective factors has yet to be achieved. The aim of the present study was to identify individual, interpersonal, and community risk and protective factors associated with the cyber-DV victimization of adolescent girls.

Cyber-DV Behaviors and Conceptualization

Cyber-DV can involve many abusive behaviors through technologies such as the control and stalking of a dating partner, sending and posting offensive comments about one’s partner, and sending text messages or emails containing threats (Bennett et al., 2011; Burke et al., 2011; Zweig et al., 2013). While various typologies of cyber-DV have been offered (e.g., Gàmez-Guadix & Mateos-Peréz, 2019; Fernet et al., 2019), the conceptualization by Watkins et al. (2018) is used in the present study and includes: 1) cyber-psychological aggression (e.g., using texting and social media platforms to humiliate, harass, or threaten one’s partner), 2) cyber-harassment (e.g., excessively calling or texting one’s partner to monitor their whereabouts and 3) cyber-sexual aggression (e.g., sending sexual images without a partners’ consent).

Cyber-Dating Violence as a Distinct Form of Dating Violence

Due to its online nature, cyber-DV is arguably qualitatively different than offline forms of DV and may have more negative impacts. For example, since cyber-DV does not happen face-to-face, it may be easier for perpetrators to abuse and harass (or continue harassing) their (ex)dating partners as they may be less aware of the consequences of their actions (Heirman & Walrave, 2008; Hellevik, 2019). Moreover, perpetrators can contact their victims at any time or place, making it difficult for them to escape cyber-DV behaviors (Zweig et al., 2014). Indeed, perpetrators of cyber-DV can easily access their victims, overcome geographical boundaries, keep track of their whereabouts, and send hurtful and denigrating information about them to a wide audience with a single click of a button. Such characteristics of cyber-DV can exacerbate victims’ feelings of helplessness, which can have deleterious impacts on their well-being (Borrajo et al., 2015; Paat & Markham, 2021; Peskin et al., 2017; Zweig et al., 2014). The permanent and constant nature of cyber-DV has also been suggested to cause more prominent suffering on its victims than offline forms of DV, and to increase their risk of revictimization. Moreover, some youth suffer both offline and cyber-DV, which may increase these negative impacts (Hellevik, 2019; Stonard et al., 2014). Although offline DV and cyber-DV are found to co-occur, the distinct characteristics of cyber-DV suggest that this form of DV may be associated with a unique set of risk and protective factors, which highlights the need to identify such factors separately.

Ecological Model Adapted to Interpersonal Violence Prevention

From a theoretical lens, socioecological models offer a comprehensive understanding of risk and protective factors by considering such factors across contexts (individual, interpersonal, community and societal) that may interact with one another to predict violence (re)victimization (Bronfenbrenner, 1977; Dahlberg & Krug, 2002). For the present study, the ecological model proposed by Dahlberg and Krug (2002) adapted to interpersonal violence prevention was used. It considers how factors arranged from the most proximal to the most distal from the individual may influence one another to predict (re)victimization. The individual level considers personal characteristics such as sociodemographic factors (e.g., age) and mental health issues (e.g., depression). The interpersonal level examines the contribution of past interpersonal victimization experiences on one’s risk of (re)victimization (e.g., offline forms DV, child sexual abuse). The community level documents neighborhood social processes (e.g., community support) and structural characteristics (e.g., neighborhood disadvantage) that influence (re)victimization. Lastly, the societal level refers to norms that create acceptance or intolerance for violence (e.g., culture blaming the victim). In addition to the identification of these factors, the ecological model suggests that to achieve and to sustain population-wide violence prevention, it is necessary to assess the influence of these factors simultaneously.

Risk and Protective Factors Associated with Offline DV and Cyber-DV

Individual level factors (e.g., sociodemographic factors, mental health variables, behaviors) are found to be responsible for a large portion of the explained variance of DV victimization (Jain et al., 2010), and potentially cyber-DV victimization. Regarding sociodemographic factors, conflicting results about age were identified. For example, in sample of 626 youth, Sánchez et al. (2015) found that age was positively associated with cyber-DV victimization with the size of the association depending on the abusive typology measured. However, Smith et al. (2018), concluded that age was not significantly associated with cyber-DV victimization. Regarding mental health factors, emotion dysregulation (Foshee et al., 2015) and psychological distress (Caridade et al., 2019) were found to be positively associated with DV. As for behaviors, adolescents who spent more time using technological devices (Zweig et al., 2014) and who engaged in sexual activity (Van Ouytsel et al., 2016) were also more likely to be victims of cyber-DV.

A number of studies have shown that previous exposure to interpersonal violence is related to DV victimization (Copp et al., 2015; Ellis & Wolfe, 2015), and studies are beginning to identify interpersonal violence experiences (e.g., adverse childhood experiences) associated with cyber-DV victimization (Lachapelle et al., 2022). Hébert et al. (2019) meta-analysis examining the influence of peer and family factors on DV victimization showed that adolescents with a history of child sexual abuse (CSA) were at an increased risk of DV. Suffering offline forms of DV victimization was also found to be associated with an increased risk for experiencing cyber-DV victimization (Fernet et al., 2019; Zweig et al., 2013). For example, Cava et al. (2020) identified offline physical DV victimization as a predictor of cyber-aggression victimization. Lachapelle et al. (2022) found that adolescents and young adults who reported experiencing parental neglect and witnessed interparental violence were at a heightened risk of cyber-DV victimization (Lachapelle et al., 2022). In the same study, reporting experiencing CSA was not associated with cyber-DV victimization.

Community social processes (i.e., social cohesion) and structural characteristics (i.e., neighborhood disadvantage) may also be associated with the likelihood of cyber-DV victimization (Caridade & Braga, 2020; Foshee et al., 2015). For example, Garthe et al. (2018) found that social cohesion was positively associated with offline DV. Moreover, in a meta-analysis by Wincentak et al. (2017), adolescents living in neighborhoods with higher levels of socio-economic disadvantage showed increased rates of physical DV victimization. Yet, Foshee et al. (2015) found that living in communities with greater levels of neighborhood disadvantage was not related to DV offline victimization, suggesting that all youth are vulnerable to this form of DV.

The identification of individual factors associated with a reduced vulnerability to cyber-DV is essential for the design of efficient preventive programs. For example, the achievement of good academic results was linked to a lower likelihood of experiencing offline DV victimization (Cleveland et al., 2003). To our knowledge, researchers have yet to identify individual protective factors associated with cyber-DV victimization. However, Hinduja and Patchin (2017), noted that youth reporting higher levels of individual resilience (i.e., positive recovery or adaptation following adversity) were less likely to be cyber-bullied. For those who were cyber-bullied, resilience seemed to act as a buffer, significantly reducing youth’s likelihood of experiencing negative school outcomes. As expected, lower frequency of technology use also acted as a protective factor from cyber-bullying victimization (Kowalski et al., 2014).

The examination of interpersonal protective factors is another important avenue for helping youth to foster healthy relationships in adolescence. For example, in a recent longitudinal study examining cyber-DV victimization, parental monitoring was protective of cyber-DV engagement in three domains (electronic monitoring, coercion, and harassment) in younger (6th to 9th grade) but not older youth (9th to 12th grade; Thulin et al., 2022). Parental support (Hébert et al., 2019), and parental monitoring (Livingston et al., 2018) were associated with a reduced vulnerability to DV victimization.

Like other forms of interpersonal violence, DV protective mechanisms are embedded in the broader community. For example, school safety and a positive school climate (Guo, 2016; Kowalski et al., 2014) were predictive of lower rates of cyber-bullying victimization.

To summarize, while scholars have identified risk and protective factors across the socioecology associated with offline forms of DV among youth, limited research is available on the risk and protective factors associated with cyber-DV in adolescent girls specifically. Yet, theory and empirical findings highlight the relevance of these factors in the risk of cyber-DV victimization. Of the studies available, most focus on the identification of these factors at a specific ecological level (e.g., individual) without consideration of the factors at other ecological levels (e.g., interpersonal, community). Moreover, most studies on cyber-DV focus on risk in detriment of protective factors (Caridade et al., 2019). Yet, scholars found that protective factors were critical for effective prevention efforts targeting cyber-DV (Peskin et al., 2017; Smith-Darden et al., 2017). As such, a comprehensive examination of risk and protective factors associated with the cyber-DV victimization of adolescent girls is necessary.

Purpose of the Present Study

Considering the high rates of cyber-DV victimization and the severity of its consequences on youth, this study aimed to offer a holistic examination of risk and protective factors associated with cyber-DV victimization of adolescent girls, across three contexts, simultaneously: 1) individual sociodemographics (age, belonging to a ethno-racial minority background) and mental health variables (dissociative symptoms, emotion dysregulation, post-traumatic stress symptoms, resilience), 2) interpersonal victimization experiences (offline forms of DV; verbal-emotional DV, sexual DV, physical DV, threats as well as a history of child sexual abuse) and 3) community factors (community support, community resilience, neighborhood material disadvantage and neighborhood social disadvantage; Dahlberg & Krug, 2002). Regarding risk factors, we hypothesized that being older, belonging to ethno-racial minority backgrounds, reporting clinical levels of dissociative symptoms, of post-traumatic stress symptoms, having higher emotion dysregulation scores, experiencing offline forms of DV, child sexual abuse, and living in a socially and materially disadvantaged neighborhoods will be associated with an increased risk of cyber-DV victimization. As for protective factors, having higher scores of self-reported resilience as well as higher levels of perceived community support and community resilience will be associated with a lower risk of cyber-DV victimization. Given that patterns of violence in romantic and intimate relationships typically emerge during adolescence, the identification of risk and protective factors may inform the elaboration of public health programs which may contribute to reduce the prevalence of DV, including DV revictimization in adulthood.

Method

Participants and Procedure

The present study used data drawn from a larger sample of 708 adolescent girls aged 14 to 18 years old. The sample retained for the analyses consisted of 456 girls (M = 16.17 years old, SD = 1.28) who had been in a dating relationship in the 12 months preceding the study or who were in a dating relationship at the time of recruitment. A total of 91% of the participants were born to parents of Canadian origin. Detailed sociodemographic characteristics are outlined in the results section. Participants were solicited using paid Facebook ads between November and December 2022 to complete the study’s 25-min questionnaire on the online survey platform Qualtrics. The questionnaires were administered in French. In the recruitment poster and in the consent form, participants were informed that they could, if they wished, be entered into a draw to win 1 of 10 gift cards of 50$ to the Appstore, iTunes or Amazon. Informed consent was obtained from the participants prior to completing the questionnaire. The study was approved by the institutional Research Ethics Committee of Université du Québec à Montréal.

Measures

Demographics

Participants reported on their sociodemographic and romantic relationship information (age, ethnicity, the nature of their relationship with their (ex)dating partner and length of the relationship in months).

Cyber-Dating Violence

Experiences of cyber-DV in the last 12 months were evaluated using a 17-item adapted version of the Cyber Aggression in Relationships Scale (CARS; Watkins et al., 2018). This instrument measures three dimensions of cyber-DV: cyber-harassment (8 items, e.g., “My partner kept tabs on my whereabouts without my permission”), cyber-psychological aggression (5 items, e.g., “My partner used information posted on social media to put me down or insult me”) and cyber-sexual aggression (4 items, e.g., “My partner tried to make me talk about sex online when I did not want to”). Each item is rated on a 4-point Likert scale: (0) Never to (3) 6 times or more in the last 12 months. The scale’s internal consistency was high both in the original sample (a= 0.86; Watkins et al., 2018) and in the current sample (a = 0.88). The total CARS score was dichotomized as (0) never happened and (1) happened at least once.

Individual Level Factors

Dissociative Symptoms

Dissociative symptoms were assessed using an adapted version of the Adolescent Dissociative Experiences Scale (Armstrong et al., 1997) called the Adolescent Dissociative Experiences Taxon (ADES-T; Martinez-Taboas et al., 2004). The original ADES demonstrated good internal consistency (a = 0.93) in a sample of adolescents aged 12 to 18 years (Armstrong et al., 1997). Internal consistency was high (a=.82) in the current sample. The scale has 8-items assessing four subscales of dissociative symptoms: 1) amnesia (e.g., “I find writings, drawings or letters that I must have done but I can't remember doing”); 2) absorption (e.g. “I feel like I am in a fog or spaced out and things around me seem unreal”); 3) passive influence, and 4) depersonalisation that forms the total score of dissociative symptoms. Participants indicated if the statement applied to them using a Likert scale ranging from (0) Never to (10) Always. The ADES-T total score was dichotomized as (0) nonclinical levels (scores 0–3.99) and (1) clinical levels (scores 4–10) of dissociative symptoms, following Armstrong et al. (1997) clinical cut-off guidelines.

Emotion Dysregulation

Emotion dysregulation symptoms were measured with an adapted version of the Difficulties in Emotion Regulation Scale Short-Form (DERS-SF; Kaufman et al., 2016). This version includes 17 items that load onto six emotion dysregulation subscales: 1) nonacceptance of emotional responses (e.g., “When I'm upset, I feel guilty for feeling that way”); 2) difficulties in engaging in goal-directed behavior (e.g., “When I’m upset, I have difficulty getting work done”); 3) impulse control difficulties; 4) lack of emotional awareness; 5) limited access to emotional regulation; and 6) lack of emotional clarity. The Likert scale ranges from (0) False to (5) True. The internal consistency of the DERS-SF was α = 0.91 derived from data from three adolescent samples (Kaufman et al., 2016), and α = 0.86 in the current sample. The total DERS-SF score was kept as a continuous variable.

Post-traumatic Stress Symptoms

Post-traumatic stress symptoms were measured using an Abbreviated version of the University of California at Los Angeles Post-Traumatic Stress Disorder Reaction Index (UCLA PSTD-RI; Cohen et al., 2008. The Cronbach’s alpha of the UCLA-PTSD-RI in the current sample was α = 0.90. The Likert scale ranges from (0) Not at all to (4) Almost always in the past month (e.g., “I have nightmares, including dreams about what happened”). The total post-traumatic stress symptom score was dichotomized as follows: (0) non-clinical (scores 0–19) and (1) clinical levels (scores 20 and over; Cohen et al., 2008).

Resilience

Resilience was measured using the 6-item Brief Resilience Scale (BRS; Smith et al., 2008) designed to assess adolescent’s ability to bounce back or recover following adversities. The BRS ranges on a 5-point Likert scale: (0) No, not really to (5) Yes, absolutely (e.g., “It does not take me long to recover from a stressful event”). The internal consistency of the French version of the BRS was α = 0.84 (Jacobs & Horsch, 2019) and α = 0.79 in the current sample. The total continuous BRS score was used in the current study.

Interpersonal Level Factors

Offline forms of DV

Physical DV, verbal/emotional DV, and threats were measured using an adapted version of the Conflict in Dating Relationships Inventory (CADRI; Wolfe et al., 2001). The adapted version uses 2 items from the original CADRI (Wolfe et al., 2001) as well as 6 of the 7 items of the CADRI Short-Form (CADRI-SF; Wekerle et al., 2009). The 8-item version measures three subscales of offline DV: verbal/emotional DV (3 items, e.g., “How often has your boyfriend/girlfriend or partner or most recent ex (if you are single today)…[said] things to make you angry”), physical DV (3 items, e.g., “How often does your boyfriend/girlfriend or partner or most recent ex (if you are single today)…slap you in the face or pull your hair”) and threats (2 items, e.g., “How often does your boyfriend/girlfriend or partner or most recent ex (if you are single today)…threaten to hurt you or harm you"). The internal consistency of the original CADRI was 0.83 and 0.70 in the present study. The Likert scale ranges from (0) Never to (3) 6 times or more in the last year. Sexual DV was measured with the Sexual Experiences Scale (SES; Koss et al., 2007). The 9-item scale is measured on a Likert scale from (0) Never to (3) 6 times or more in the last year (e.g., “How often did your boyfriend or girlfriend or partner or most recent ex (if you are single today) …kiss, fondle, or touch when you didn't want to, using arguments or pressure”). The Cronbach’s alpha of the SES in the original scale was 0.70 and 0.81 in the current sample. The scores for each DV subscale were dichotomized following a (0) never happened and (1) happened at least once in the last year.

History of Child Sexual Abuse

Self-reported child sexual abuse was measured with 1-item “Has anyone ever touched you sexually when you didn't want to, or manipulated or forced you to have sex?” (Finkelhor et al., 1990). Child sexual abuse was dichotomized as (0) never happened and (1) happened.

Community Level Factors

Neighborhood Disadvantage

Neighborhood levels of material (education, employment, and income levels in the neighborhood) and social (proportion of individuals living alone, being a single parent, and being separated, divorced, or widowed) disadvantage were evaluated using the Material and Social Deprivation Index (MSDI; Gamache et al., 2019) derived from census data based on participants’ postal codes. The local versions of the index were used (Gamache et al., 2019). The index divides the level of deprivation into quintiles ranging from (1) Most privileged to (5) Most deprived. The quintiles for both dimensions of disadvantage were used as ordinal independent variables in this study.

Community Support

Perceived levels of community support were assessed with an adapted version of the Community Support Scale (CSS; Hamby et al., 2015). The 6-item instrument is designed to evaluate “the degree to which members in the community get along, help one another and support neighborhood youth” (Hamby et al., 2015, p. 2). The Likert scale ranges from (1) Not true to (4) Very true (e.g., “People in my neighborhood offer to help one another”. The original CSS scale’s Cronbach’s alpha was α = 0.80 which was replicated in the current sample α = 0.79. The total CSS score was used in the current study.

Community Resilience

Perceived levels of community resilience were measured using the Transcultural Community Resilience Scale (TCRS; Cénat et al., 2021). The 28-item TCRS measures the adolescent’s perception of the capacity of their communities to share their resources, as well as the support from and interactions with the rest of the community to foster the resilience of its members (Patel et al., 2017). The Likert scale of the TCRS ranges from (1) Totally disagree to (5) Totally agree. In the original scale and in our sample, the Cronbach’s alpha was excellent (α = 0.96). The TCRS has three subscales 1) community strengths and supports (e.g., “I get involved in my community’s activities”), 2) community trust and faith, and 3) community values. The total TCRS score was used in the current study.

Statistical Analyses

Descriptive analyses and a hierarchical logistical regression were conducted for this study using SPSS version 27. A three-step logistic regression was used to identify whether any of the independent variables were linked to cyber-DV victimization. The logistic regression was carried out using the forced entry method. The results are summarized in Table 1. The variables were entered according to the ecological level that they apply to. Variables entered in the first step (individual level) included the participants’ age, ethnicity, post-traumatic stress, dissociation, emotion dysregulation and resilience. In the second step (interpersonal level), child sexual abuse, physical DV, verbal/emotional DV, sexual DV and threats were entered into the model. In the third step (community level), community support, community resilience, neighborhood material disadvantage and social disadvantage were included into the model. None of the independent variables correlated at more than r = 0.70. The data met the assumption of multicollinearity as the collinearity statistics (i.e. tolerance and variance inflation factor, VIF) were within accepted limits (Coakes, 2005; Hair et al., 1998).

Table 1 Means, standard deviations of variables and correlations between variables
Full size table

Results

Descriptive Statistics and Prevalence Rates of Cyber-DV

When self-reporting their ethnic or cultural groups, participants could select more than one option. Most participants (91%) reported that their parents were in the Québec or Canadian ethnic or cultural group. The remainder of the participants reported that their parents were Western European (7.5%), First Nations (e.g., Inuit, Metis; 3.7%), Latino-American (3.7%), Asian (2.9%), Eastern European (2.2%), African American/Black African (2.2%), North African/Middle Eastern (1.8%), Caribbean/Antilles (1.5%), and Other (.2%). Regarding the participants’ age at the time of recruitment, 13.4% of the sample reported being 14 years of age, 15 years of age (18%), 16 years of age (23%), 17 years of age (29.4%), and 18 years of age (16.2%). Over half (60.3%) of the sample reported experiencing at least one episode of cyber-DV victimization in the past year. More specifically, 44.5% experienced cyber-harassment, 43.9% cyber-psychological aggression, and 17.3% cyber-sexual DV. In addition, approximately half (51.5%) of the sample reported a history of child sexual abuse. Moreover, 63.3% of the participants reported experiencing offline DV victimization. More specifically, 56.6% of adolescent girls experienced verbal/emotional DV, 30% sexual DV, 16% physical DV, and 7.7% threats. In terms of neighborhood material disadvantage, approximately a fourth of the participants (19.8%) were living in disadvantaged neighborhoods whereas 16.8% of participants were living in very disadvantaged neighborhoods. Regarding neighborhood social disadvantage, 20.8% were living in disadvantaged neighborhoods whereas 19.3% of participants were living in very disadvantaged neighborhoods.

Hierarchical Logistic Regression

Results of the logistic regression analysis are summarized in Table 2. Step 1 was statistically significant (χ2 = 31.63; p = < 0.001) and accounted for 10.2% of the variance of the risk of being a victim of cyber-DV victimization in the past year with higher dissociation and emotion dysregulation identified as significant predictors. Step 2 was significant (χ2 = 102.84, p = < 0.001); with the interpersonal variables added, the model explained 38.4% of the variance in cyber-DV victimization, an increase of 28.2% of the explained variance. In the second step, only sexual DV and verbal-emotional DV were significant in increasing the odds of cyber-victimization. The 3rd and last step of the logistical regression was also significant (χ2 = 141.62; p = < 0.001), and the final model explained 40.1% of the variance of cyber-DV victimization (ΔR2 = 1.86%). The overall classification accuracy was 66.5%. Once other variables were controlled for, none of the individual level variables were significantly associated with cyber-DV victimization. At the interpersonal level, girls who experienced sexual DV were 2.55 times more likely to experience cyber-DV in the last year (OR = 2.554; 95% CI = [1.352–4.828]). Moreover, results indicate that adolescent girls who experienced verbal-emotional DV were 5.66 times more likely to experience cyber-DV victimization in the last year (OR = 5.661; 95% CI = [3.434–9.333]). Adolescent girls who experienced threats in the context of their romantic relationship were 9.53 times more likely to experience cyber-DV victimization in the last year (OR = 9.528; 95% CI = 1.066–85.131). A history of child sexual abuse and physical DV was not associated with cyber-DV. At the community level, with every unit decrease of neighborhood social disadvantage, adolescent girls were 1.41 times more likely to experience cyber-DV (OR = 1.408; 95% CI = [0.661–3.000]). Community support, community resilience and local material disadvantage were not associated with cyber-DV victimization in the final model once other variables were controlled for.

Table 2 Hierarchical Stepwise Logistic Regression Cyber-Dating Violence
Full size table

Discussion

Cyber-DV victimization of adolescent girls is a prevalent problem in North America with adverse consequences on the well-being of victims. The aim of the present study was to offer a holistic examination of risk and protective factors associated with cyber-DV victimization in a sample of adolescent girls from Quebec, Canada, across three ecological contexts (Dahlberg & Krug, 2002). In our sample, 63.3% of adolescent girls reported experiencing at least one form of cyber-DV in the last year which is substantially higher than the 24% prevalence rate found in a previous study (Hinduja & Patchin, 2020). However, a potential explanation for the variability in prevalence rates is that Hinduja and Patchin (2020) used only five items to assess cyber-DV victimization behaviors, whereas our study used eighteen items, capturing a broader range of cyber-DV manifestations. In addition, participants of our study completed our questionnaire between November 2021 and December 2022, during the COVID-19 pandemic. Therefore, these elevated rates could be attributed to the fact that most social interactions, including those in the context of dating, moved online, with youth spending more time on their technological devices. In fact, an American National Intimate Partner Violence Hotline found that reports of cyber-DV increased by 101% from 2019 to 2020, with many of these reports made by youth (the Hotline, 2022). Although cyber-DV victimization prevalence rates may decline post-pandemic given that social interaction restrictions have decreased, pre-COVID-19 rates of this form of violence were already elevated. Thus, a better understanding of risk and protective factors associated with cyber-DV remain necessary to effectively prevent this deleterious form of violence.

Results of our study indicated that once all the potential risk and protective factors were accounted for, our model explained an important proportion of the variance of cyber-DV victimization (40.1%), emphasizing the relevance of these multilevel factors in understanding the contexts in which this form of victimization unfold and for whom. However, only a handful of factors remained significantly and independently associated with cyber-DV victimization in the final step of our multivariate model. In terms of risk factors and in partial support of our hypothesis, verbal/emotional DV, sexual DV, threats by an (ex)dating partner, and neighborhood social disadvantage remained associated with cyber-DV victimization. Contrary to our hypothesis, none of the protective factors remained significantly associated with cyber-DV in the final step once all other variables were controlled for.

Our results are consistent with past studies showing that offline forms of DV victimization are strongly associated with an increased risk of cyber-DV victimization (Fernet et al., 2019; Zweig et al., 2013). They showed positive associations between three forms of offline DV victimization (verbal-emotional, sexual, threats) and cyber-DV victimization, with exposure to threats showing the largest effect size (OR = 9.53). Similarly, to our findings, Cava and Buelga (2018) found positive links between verbal/emotional DV and forms of cyber-DV (cyber-control and cyber-aggression). Moreover, Zweig et al. (2014), identified a positive association between verbal-emotional DV and cyber-sexual aggression, whereas our study found a significant association between sexual DV and cyber-DV. Contrary to our hypothesis, past experiences of physical DV were not associated with cyber-DV victimization which could be explained by its low prevalence (16%) in our sample. Given the cross-sectional nature of our study, the temporality between cyber-DV and offline DV in our sample cannot be established. However, past longitudinal findings tend to indicate offline DV may precede cyber-DV experiences in the following year (Temple et al., 2016). Future studies are needed to better disentangle the associations between offline DV and cyber-DV as the way they relate to one another remains unclear. For example, some researchers postulate that digital technologies offer perpetrators of offline DV an additional mean to abuse and exert control over their (ex)dating partners (Korchmaros et al., 2013; Lara, 2020). Others suggest that the online context creates an environment in which individuals who would not engage in offline forms of dating violence may be less inhibited to do so online, as their actions are more removed or potentially anonymous, and because they are not exposed to the direct consequences of their actions on their victim (Cheung et al., 2021; Hellevik, 2019). More longitudinal studies and studies with perpetrators are needed to (dis)confirm these hypotheses.

Contrary to our first hypothesis, adolescent girls living in communities with lower levels of census tract social disadvantage were more likely to be victims of cyber-DV compared to girls living in more disadvantaged communities. This finding could be partially explained by the routine activities’ theory by Cohen and Felson (1979) which posits that youth’ online routine activities may expose them to cyber-DV victimization. This theory also suggests that cyber-DV behaviors are more likely to manifest themselves in the absence of a capable guardian who would otherwise protect youth from perpetrators of cyber-DV by monitoring or controlling their online activities (Cohen & Felson, 1979). Adolescent girls in communities with greater levels of social disadvantage may have to share technological devices with other members of the family (e.g., having a shared computer in a communal space in the home). Thus, the lack of online privacy may reduce their usage and improve parental supervision. Similarly, girls in more socially disadvantaged neighborhoods may be less likely to have an intelligent phone with roaming or have less data availability on their monthly plan, reducing their usage of cellular functions that require data (e.g., social media). Despite this hypothesis, we did not evaluate the level of parental monitoring of participants technological devices and did not question participants about the availability of devices with internet and their usage, which could be unmeasured confounding factors. Future study should examine these factors to help make sense of this counterintuitive finding.

Ecological Model

The hierarchical model allowed us to examine which factors remained significantly linked with the cyber-DV victimization of adolescent girls, while considering such factors across multiple contexts, simultaneously. Individual level factors (emotion dysregulation and dissociative symptoms) were no longer significantly associated with cyber-DV victimization once interpersonal variables were accounted for. Interestingly, the added variation was quite low (1.36%) when neighborhood variables were entered into the model, suggesting that most of the variation in cyber-DV victimization was found at the individual (10.1%) and interpersonal levels (28.2%). This finding is consistent with other research suggesting that more proximal, individual or interpersonal risk factors (e.g., offline DV) may be more predictive of DV victimization than community level risk factors (e.g., neighborhood disadvantage; Dahlberg & Krug, 2002). In addition, experiencing multiple forms of DV seems to have an additive effect on the adolescent girls’ vulnerability to other forms of DV, including cyber-DV. The high percentage of explained variance in cyber-DV (40.1%) that was obtained suggests that cyber-DV victimization, like other forms of interpersonal victimization, occurs through an interplay of many variables, acting across multiple contexts (Dahlberg & Krug, 2002). Our results provide additional evidence that support the practically and applicability of socioecological models as a theoretical lens for interpersonal violence prevention research.

Clinical Implications

Despite the serious consequences of cyber-DV on adolescents’ mental and physical health, there continues to be a lack of prevention and intervention efforts. One of the main practical implications of our study is that due to the potential co-occurrence between offline DV and cyber-DV, and its high prevalence among adolescents, universal school programs (i.e., aiming to educated the general population of youth) should target offline DV, for example, teaching youth the fundamentals of healthy relationships and conflict negotiation, but also specifically address the online form with specific modules and concrete activities (Galende et al., 2020). As such, universal school programs could target risk and protective factors common to both forms of DV, simultaneously, achieving greater preventative power. As mentioned previously, cyber-DV involves distinct features as it can happen 24/7, it is difficult to escape from, hurtful information can be spread to a wide audience, and it can cause greater feelings of helplessness (Cava & Buelga, 2018). In addition, it is more difficult for youth to identify it, as cyber-DV behaviors are more subtle, and may not be interpreted as violence (Galende et al., 2020). Therefore, universal school programs could be helpful to teach youth and bystanders (such as peers, teachers, and school counselors) how to recognize the warning signs and manifestations of cyber-DV, but also equip adolescents with the knowledge and tools to protect themselves from it (Galende et al., 2020). The results of our study are also important for practitioners to assist them in identifying consequences associated with cyber-DV as well as educate youth who engage in this form of violence on its expressions and impacts. Lastly, educating the broader community on cyber-DV and its harms is essential.

Strengths, Limitations, and Directions for Future Research

Our results are limited by the use of self-report and retrospective measures completed by adolescent girls from a limited geographical region. Therefore, our results may not be generalizable to all adolescent girls in Canada or North America. Researchers should include non-female identifying participants and participants with various sexual orientations in their sample to be more representative of the general population. A sample inclusive of all genders and sexual orientations, if in sufficient numbers, would offer additional variables to explore as potential contributors to cyber-DV violence victimization and perpetration. Self-report measures may be biased by the perceptions of adolescents or social desirability. Moreover, given the cross-sectional nature of our study, it is impossible to establish temporality between our study variables. Longitudinal designs are needed to establish the temporal relations between offline and online forms of DV (Lu et al., 2018; Thulin et al., 2022). It is also important to mention that our study did not examine the associations between specific forms of offline DV and specific forms of cyber-DV (e.g., sexual DV and cyber-sexual aggression) highlighting the need for future studies to examine these effects separately. Such analyses may offer more nuances that could be helpful in designing preventive interventions against cyber-DV. In addition, future research should be conducted to disentangle the associations between neighborhood structural characteristics and neighborhood social processes in their contribution to cyber-DV victimization, while accounting for possible confounding factors (e.g., parental monitoring, usage). The application of census tract boundaries also may not reflect true levels of disadvantage nor residents’ perceptions or their neighborhood, suggesting that researchers should move towards a blend of objective and subjective measures of neighborhood processes to gain a better comprehension of how such factors are related to cyber-DV (Jain et al., 2010; Johnson et al., 2015). To obtain a more comprehensive understanding of cyber-DV, researchers should examine risk and protective factors associated with both experiencing and perpetrating this form of DV.

Regarding the strengths of our study, many modifiable risk factors as well as neglected protective factors associated with cyber-DV victimization were analysed across multiple ecological levels at once, allowing us to examine which factors at which levels of the ecology were the most significant. Our sample was large and diverse in terms of victimization histories, providing key data on cyber-DV victimization in adolescent girls.

Conclusion

Despite its limitations, this study identifies factors associated with cyber-DV victimization using a holistic perspective. An important conclusion of our study is that experiencing certain forms of offline DV is associated with a substantially higher risk of experiencing cyber-DV victimization. Results accentuate the need for additional research to clarify the directionality of the associations between offline DV and cyber-DV as well as the need for more studies to focus on the examination of mediating factors in their links to cyber-DV victimization. Fundamentally, this study identifies potential targets for preventive interventions and inspires future research in the field. It is hoped that evidence-based interventions will ultimately reduce youths’ risk of suffering, and engaging in cyber-DV victimization and its associated negative impacts, at a vulnerable stage in their development.