Conceptualizing bullying

Bullying, as first defined, entails direct and indirect aggression that is intentional, repeated, and involving a physical or social imbalance of power, typically in school [19, 20]. In 2014, the Centers for Disease Control and Prevention (CDC) expanded this definition to entail “unwanted aggressive behavior(s) by another youth or group of youths who are not siblings or dating partners that involves an observed or perceived power imbalance and is repeated multiple times or is highly likely to be repeated. Bullying may inflict harm or distress on the targeted youth including physical, psychological, social, or educational harm” [21]. Direct aggression includes physical and verbal forms—such as hitting, shoving, or name-calling—that occurs in the presence of the target [21]. Indirect aggression entails exclusion, rumors, or gossip typically to decrease social status and occurs through a third party or when the target is not present [21]. In addition to physical or verbal aggression and indirect or relational aggression, other types of bullying include damage to property [21] and cyberbullying, or bullying through electronic forms of contact [22].

Among school-aged children and adolescents, these definitions of bullying have been criticized because of their de-emphasis of more serious, injurious, or aggravated forms of aggression, the difficulty of identifying and measuring power imbalances, and restriction to aggression in schools [23–25]. Questions of definition in this age group also have focused on the extent to which bullying overlaps with other forms of violence or victimization and the distinctions between forms of bullying [26]. Among those who have been bullied, the experience of multiple (for example, physical and verbal) forms of bullying is associated with poorer outcomes [27]. These results corroborate the broader literature on poly-victimization, which consistently finds that exposure to a greater number of adverse events is associated with worse psycho-social and health outcomes [28]. Thus, definitions of bullying should distinguish this form of violence from others (like dating violence) and should capture multiple types of bullying to improve estimates of prevalence and to capture gender-differentiated forms of bullying perpetration and victimization [29, 30].

Despite efforts to develop standard definitions for bullying in adolescents, the global literature reflects a range of terms and definitions. A common global definition may be difficult in practice, as descriptors and social mechanisms of bullying may differ cross-culturally [31–37]. Groups, such as UNESCO and UN Women, focus more broadly on child violence in schools, and include gender-based violence and sexual violence under this umbrella term [2, 3]. For this study, we follow the CDC’s definition of bullying, allowing for victimization to be during school hours, on the way to and from school, during school-related events, or during in-school or out-of-school tutoring [21].

Academic and psychosocial effects of bullying

An emerging global body of evidence suggests a dose-response relationship between bullying victimization and academic and psychosocial outcomes. Experiences of bullying have been associated with lower scores for reading and math in 15 Latin American countries [16]. In a recent global systematic review and meta-analysis, the authors found that both traditional and cyber-bullying had strong relationships with a range of mental-health disorders, probable relationships with tobacco and illicit-drug use, and possible relationships with alcohol use, poor academic achievement, psychosomatic symptoms, social isolation, overweight and obesity [17]. In a meta-analysis of quasi-experimental studies in the Americas, United Kingdom, Ghana, Korea, and Italy, bullying victimization led to poorer short-term developmental outcomes, including higher internalizing and externalizing symptoms and reduced academic achievement [38]; however, these adverse effects appeared to diminish in the long-term, highlighting the potential for resilience in the victims of bullying. A smaller study among Virginia public-high-school students found that the perceived prevalence of teasing and bullying among ninth graders and teachers was positively associated with dropout rates four years later [39] and negatively associated with school-wide passing rates on state-mandated achievement tests [40, 41].

The role of schools in a social-ecological model of girls’ bullying victimization

While gender parity in schooling has been achieved at primary level in Nepal, girls continue to lag behind boys in terms of overall attainment [42]. Lower secondary school, which begins at grade six, represents a time when girls’ risk of school dropout begins to increase, peaking at grade 9 [43]. The elements of a secondary school environment designed for improving educational and psychosocial outcomes for girls is unstudied in LMICs, though there is emerging evidence of an association between school-level gender equity and girls’ pass rates [44].

School-level factors are widely recognized to influence bullying victimization [45–47]; thus, schools may be an ideal entry point to address girls’ bullying victimization globally. While the elements of a favorable school climate continue to be debated, schools that foster the following practices may reduce peer victimization: bystander intervention; pro-social norms enforced by diligent monitoring; healthy, non-coercive relationships between teachers and students; and consistent communication of expectations between students, teachers, and families [48–57]. In randomized trials, school-based anti-violence programs have effectively reduced perpetration and victimization [58]; however, the effects of program components were not estimated, and rigorous, multi-armed intervention studies are lacking in low- and middle-income countries [59].

Study objectives

This study used baseline data from the evaluation of the Room-to-Read (RtR) Girls’ Education Program (GEP) in Nepal to assess the relationship of a conceptually grounded gender-equitable school (GES) index [44] with girls’ risk of bullying victimization, adjusted for potential confounders at the individual and school levels.

Ethics

The Emory University Institutional Review Board [Approval#00103052] and the Nepal Health Research Council [Approval#237/2018] approved the study. Written informed consent was collected for all adult participants, and written parental consent and written assent for all minor participants. In accordance with GAGE and Emory University policies, field staff received training on additional human subjects considerations when conducting research among children and adolescents.

Nepal setting

According to the 2016 Nepal Demographic and Health Survey, adolescents aged 10–19 years comprise 23% of the country’s population [42]. Girls in Nepal remain less likely than boys to complete secondary school, with increased risk of dropout beginning in early adolescence [43]. Barriers to girls’ adolescent capability development and associated social- and health-related challenges arise in part as a result of inequitable gender norms and entrenched social hierarchies that privilege men [62, 63]. Nepali girls and women still often are expected to be caretakers of the family and remain undervalued in society [64]. Practices like son preference, menstrual restrictions, dowry, child marriage, and polygamy persist and are symbolic of underlying gender-inequitable norms [62]. Girls and women also may experience unequal gender power relations, compromised voice and agency within their household and community, restrictions on their movement, and less access to formal schooling. Intimate partner violence (IPV) is an accepted norm among many men and women, and women’s exposure remains high [63].

Study sites

The present study was undertaken in Tanahun and Nuwakot districts in Nepal. Both districts are characterized, overall, as having considerable socioeconomic disadvantage and low human development.

Nuwakot, located in the central region of Nepal, falls below the national average with respect to adult literacy (60% versus 66%), schooling attainment (3.3 versus 3.9 grades), and per capita income ($1,086 versus $1,160 PPP) [65]. Of the 12-year-old girls in Nuwakot in 2011, about 65% had finished primary school, 33% had finished lower secondary school, and 0.5% had completed secondary school [65]. The median age at first marriage in the province is 19.7 years among women 20–49 years [42]. That said, structural change is evident, with an influx of new schools, hospitals and improved roads in recent years [62].

In Tanahun district, per capita income ($1,072 PPP) falls below the national average, though average levels of schooling, 4.2 grades, are slightly higher [65]. Of the 12-year-old girls in Tanahun in 2011, an estimated 63% had finished primary school (Grades 1–5), 36% had finished lower secondary (Grades 6–8), and only 0.3% had completed secondary (Grades 9–10) [65]. In 2016–2017, the median age at first marriage across the province was 18.6 among women 20–49 years [42].

Sample eligibility and response rates

The present analysis uses baseline data from the GAGE evaluation of the RtR GEP, which is underway in Nuwakot and Tanahun districts. We aimed to enroll approximately 580 grade six girls attending the 24 schools (12 per district) in which RtR was scheduled to roll out its GEP in 2018. GAGE provided the sampling frame for this cohort. We also aimed to establish a comparison cohort of 1,276 girls by taking a census of girls aged 12–14 years or in grade six in 40 randomly selected clusters (wards of <275 households or randomly selected segments of larger wards). Using the compact segment sampling method allowed us to produce a self-weighed sample with sufficiently high participation [66]. Eligible wards for the comparison cohort were adjacent to the wards in which the RtR GEP was being implemented and had a sufficient population of eligible girls. From this sampling frame, clusters were selected using probability proportional to size. The achieved sample at baseline included 571 girls in RtR GEP schools and a comparison sample of 1,127 girls. Baseline response rates were 99.8% of eligible RtR girls and 99.9% of eligible community girls, respectively.

All RtR GEP schools were visited and included in this analysis. Schools in the comparison clusters of Tanahun and Nuwakot districts were eligible for inclusion in this analysis if they were in the comparison area, were attended by a girl in the comparison sample, and were government schools, private schools, or government-supported public/community schools that had a grade six and above. For this analysis, we further restricted schools in the comparison clusters to those with greater than 40% of the reported female grade six enrollees sampled and excluded all schools with fewer than three girls sampled. The final school sample consisted of the 24 RtR GEP schools and 25 community schools, representing 750 girls in grade six. Twenty-one of the girls in this sample were missing outcome data and were removed, bringing the final sample to 729. Each girl attended one of 49 schools in the sample.

Data collection

The field team included 30 female researchers who received a 13-day training undertaken jointly by GAGE team members at Emory University and the Centre for Research on Environment, Health, and Population Activities (CREHPA). The training included didactic sessions on field procedures, the survey questionnaires, methods for conducting the interview, GAGE policies, and ethics in human-subjects research. The field team received didactic and experiential training in the conduct of face-to-face computer-assisted personal interviews (CAPIs) using tablets and REDCap version 8.3.2. REDCap is a web-based data systems application developed by Vanderbilt University with funding from the National Institutes of Health that allows for the secure collection, storage, and transfer of multi-relational data and user-defined access by multi-institutional partners at any location in the world [67]. The training also included experiential role-plays, a pre-test, and a full pilot. The pre-test and pilot informed final revisions to the questionnaires and REDCap application. Fieldwork was undertaken from July 4, 2018 to September 2, 2018. After data collection, one debriefing session was held to give interviewers a platform to share feedback.

Adolescents were interviewed at home, in a private area where parents could see them but not overhear their answers. Interviews with adolescents took an average of 61 minutes. The school questionnaire, completed by a school administrator, consisted of 12 modules covering staff, policies, facilities and materials, student population, accessibility, support for students, and outside organizations operating within the school. School administrator interviews took place at the school building and lasted an average of 42 minutes. All questionnaires were developed in English, reviewed by RtR and the Overseas Development Institute, translated into Nepali, and back-translated into English to identify and resolve differences in the meaning and interpretation of the questions before piloting.

Measures

To develop the questionnaire for this study, the study team relied on previously piloted and successfully implemented questionnaires from studies of adolescents in Nepal, studies of adolescents in the core Gender and Adolescent: Global Evidence consortium, and widely validated youth development scales (Table 1) [68]. Each question set that was included in the questionnaire was aligned with specific learning modules in the GEP program and was piloted and adapted for comprehension in the local setting and population.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Survey modules and items collected.

https://doi.org/10.1371/journal.pone.0253128.t001

The adolescent questionnaire included 22 modules that covered the following domains: educational experiences in pre-kindergarten and primary school; positive youth development—including self-confidence, self-efficacy, positive self-identity, social competence, emotional competence, critical thinking, and leadership competence [72–75]; academic performance—including tests of reading comprehension and math skills [76]; marriage, job, and educational aspirations; voice, agency and freedom of movement; livelihoods, financial literacy, and economic empowerment; a safe and enabling environment; participation and civic engagement; adolescents’ social networks, reference groups, and role models; gender attitudes and norms—based on questions about masculinity, femininity, and gender relations from published scales [64, 77–79] experiences of school-based aggression and victimization [80]; health and risky behavior; and program benefits.

Analysis

To assess the relationship between the gender-equitable school (GES) index, a contextual factor, and girls’ risk of peer victimization, we employed a multi-level approach [92]. First, we examined the distribution of all individual adolescent and school-level variables using frequencies, means, and standard deviations. Next, tetrachoric correlations among study variables were assessed separately by the level at which they were modeled (individual or school). As the outcome variable is a weighted count of victimization types that also is over-dispersed (scaled Pearson statistic different from 1), we used multilevel negative binomial regression for the estimation. We fit a series of models, the first of which was an unconditional random-intercept model (Model A) to estimate the amount of variability in victimization across schools. Subsequently, we fit a model containing the random intercept and the two GES components (Model B), followed by a model in which all school-level variables were also included (Model C). Finally, we estimated a full model (Model D), in which all individual- and school-level variables were modeled simultaneously. As the analytic focus was on the relationship of gender-sensitive school environment and victimization, we grand-mean-centered the individual-level variables so that the GES regression coefficients could be interpreted as the log count of victimization types, adjusted for potential confounding by individual variables [93]. Students per grade was grand-mean-centered for interpretability. We used PROC GLIMMIX in SAS (version 9.4, SAS Institute Inc., Cary, NC, USA) for the analysis.

Characteristics of adolescent girls

Girls in the sample ranged in age from 9 to 17 years, with a mean age of 11.8 (Table 2). Approximately 40% had attended preschool. The mean age at starting first grade was 5.8 years, and 70% of girls reported having received financial assistance in primary school. On average, girls’ caregivers (mostly mothers) had received schooling through third grade. Mean household SES scores were approximately zero, with the lowest tertile ranging from negative 10 to negative 2 and the highest ranging from 1 to 8. On a scale of 1 to 4, average reported peer leadership was low, at 1.9, whereas average reported peer gender equity was high, at 3.6. On average, girls in our sample reported 3.12 victimizations in the prior week. Of these, 2.84 were direct victimizations and 0.27 were relational victimizations. The most frequently reported item was teasing, whereas the least commonly reported item was making up rumors. Girls from RtR schools were significantly less likely to have attended preschool and reported higher disability scores than their counterparts at community schools. They also reported higher levels of physical victimization in the prior week.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Characteristics of adolescent girls (N = 729), Nuwakot and Tanahun districts, Nepal 2018.

https://doi.org/10.1371/journal.pone.0253128.t002

Characteristics of schools

The overwhelming majority of sampled schools were government or community schools, with fewer than five percent private (Table 3). About two thirds were in urban municipalities, and located in Tanahun or adjacent areas of Gorhka and Lamjung districts. A large majority of schools did not report using corporal punishment as a means of discipline. On average, schools enrolled 37 students per grade, although this number ranged from about 8 to almost 150. SEC ranged from -3.89 to 2.31, with mean of zero. Mean GES scores also were approximately zero, ranging from -3.03 to 4.97 for component 1 and -3.42 and 2.78 for component 2. RtR schools were significantly more likely to be in Nuwakot and public. In addition, RtR schools were more likely to have a higher economic condition, had more students per grade, and had higher average reported disability scores.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Characteristics of schools, community sample (N = 25) and RtR school sample (N = 24), Nuwakot and Tanaun districts, Nepal 2018.

https://doi.org/10.1371/journal.pone.0253128.t003

School environment and bullying victimization

In multilevel models, no school-level covariates were significantly associated with weekly rate of either total bullying victimization or direct bullying victimization, although caregiver education was negatively associated with both direct and total bullying victimization, and higher levels of disability were positively associated with total victimization (Table 4). The generalized GES component showed a significant negative relationship with weekly relational bullying victimization in Models C (with all school-level covariates) and D (with all school- and individual-level covariates). In the full model, a one-point increase in the generalized GES was associated with a 26% decrease in relational bullying victimization in the prior week. In full models, attending a government or community school also was positively associated with the weekly victimization rate of relational bullying, accounting for a 16-fold increase versus attending a private school.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Multilevel models of gender equitable school-level exposures and total victimization, N = 729 adolescent girls in Nuwakot and Tanahun districts in Nepal, 2018.

https://doi.org/10.1371/journal.pone.0253128.t004

Strengths and limitations

The scope of this study, which considered individual, household, peer-network, and school-level factors, allowed the research team to better understand how the nested environments in which adolescent girls operate may impact their risk of peer victimization. While direct victimization in the prior week was more commonly reported than relational victimization, the scale measure used included more items to measure direct aggression. Almost 90% of girls in our sample reported having some access to cell phones, and more than a third of these could access the internet. As cell phone and internet access expand, increasing opportunities for cyberbullying, accurately measuring relational aggression may require new items to be added to existing, validated scales. Additionally, we did not ask girls for the gender or other characteristics of perpetrators, limiting our ability to draw conclusions about the relative prevalence of same-gender versus cross-gender bullying and the relative prevalence of different forms of aggression perpetrated by girls versus by boys. Finally, under GAGE’s ethics policies, we were not able to collect individual-level data on experiences of violence perpetrated by adults. Measures of corporal punishment based on administrator report may be underestimates due to the illegality of the behavior [103].