Assessing the relationship between agency and peer violence among adolescents aged 10 to 14 years in Kinshasa, Democratic Republic of Congo and Blantyre, Malawi: A cross-sectional study.

BACKGROUND: Interpersonal violence has physical, emotional, educational, social, and economic implications. Although there is interest in empowering young people to challenge harmful norms, there is scant research on how individual agency, and, specifically, the "power to" resist or bring about an outcome relates to peer violence perpetration and victimization in early adolescence. This manuscript explores the relationship between individual agency and peer violence perpetration and victimization among very young adolescents (VYAs) living in two urban poor settings in sub-Saharan Africa (Kinshasa, Democratic Republic of Congo (DRC) and Blantyre, Malawi). METHODS AND
FINDINGS: The study draws on two cross-sectional surveys including 2,540 adolescents 10 to 14 years from Kinshasa in 2017 (girls = 49.8% and boys = 50.2%) and 1,213 from Blantyre in 2020 (girls = 50.7% and boys = 49.3%). The sample was school based in Malawi but included in-school and out-of-school participants in Kinshasa due to higher levels of early school dropout. Peer violence in the last 6 months (dependent variable) was defined as a four categorical variable: (1) no victimization or perpetration; (2) victimization only; (3) perpetration only; and (4) both victimization and perpetration. Agency was operationalized using 3 scales: freedom of movement, voice, and decision-making, which were further divided into tertiles. Univariate analysis and multivariable multinomial logistic regressions were conducted to evaluate the relationships between each agency indicator and peer violence. The multivariable regression adjusted for individual, family, peer, and community level covariates. All analyses were stratified by gender and site. In both sites, adolescents had greater voice and decision-making power than freedom of movement, and boys had greater freedom of movement than girls. Boys in both settings were more likely to report peer violence in the last six months than girls (40% to 50% versus 32% to 40%, p < 0.001), mostly due to higher rates of a perpetration-victimization overlap (18% to 23% versus 10% to 15%, p < 0.001). Adolescents reporting the greatest freedom of movement (Tertile 3) (with the exception of girls in Kinshasa) had a greater relative risk ratio (RRR) of reporting a perpetrator-victim overlap (boys Kinshasa: RRR = 1.9 (1.2 to 2.8, p = 0.003); boys Blantyre: RRR = 3.8 (1.7 to 8.3, p = 0.001); and girls Blantyre: RRR = 2.4 (1.1 to 5.1, p = 0.03)). Adolescents with the highest decision-making power in Kinshasa also had greater RRR of reporting a perpetrator-victim overlap (boys: RRR = 3.0 (1.8 to 4.8, p < 0.001). Additionally, girls and boys in Kinshasa with intermediate decision-making power (tertile 2 versus 1) had a lower RRR of being victimized (Girls: RRR = 1.7 (1.02 to 2.7, p = 0.04); Boys: RRR = 0.6 (0.4 to 0.9, p = 0.01)). Higher voice among boys in Kinshasa (Tertile 2: RRR = 1.9 (1.2 to 2.9, p = 0.003) and Tertile 3: 1.8 (1.2 to 2.8, p = 0.009)) and girls in Blantyre (Tertile 2: 2.0 (1.01 to 3.9, p = 0.048)) was associated with a perpetrator-victim overlap, and girls with more voice in Blantyre had a greater RRR of being victimized (Tertile 2: RRR = 1.9 (1.1 to 3.1, p = 0.02)). Generally, associations were stronger for boys than girls, and associations often differed when victimization and perpetration occurred in isolation of each other. A main limitation of this study is that the cross-sectional nature of the data does not allow a causal interpretation of the findings, which need further longitudinal exploration to establish temporality.
CONCLUSIONS: In this study, we observed that peer violence is a gendered experience that is related to young people's agency. This stresses the importance of addressing interpersonal violence in empowerment programs and of including boys who experience the greatest perpetration-victimization overlap.

Introduction

Interpersonal violence, the intentional use of force or power on a family member or unrelated individuals [1], is particularly salient in adolescence due to the significance of violent experiences on developmental trajectories, with considerable physical, emotional, educational, social, and economic implications [2-4]. Bullying is particularly prevalent in this stage of life, rising in early adolescence, peaking in late adolescence, and decreasing in young adulthood [2,5,6]. Data from 144 countries participating in the Global School-based Health Survey show that 10% to 65% of adolescents 11 to 15 years report having been bullied in the past year [4]. These global figures mask significant variation by geography and gender [7], with high prevalence noted in the African region (43.5%, 43.0% to 44.3%) [8] and greater violence exposure among boys compared to girls [8,9].

Interpersonal violence, including peer violence perpetration and/or victimization, is informed by the interplay of personal and contextual factors [2,6]. At the individual level, age, gender, history of adverse childhood experiences (ACEs), medical/physical conditions that affect the nervous system, and psychological characteristic, i.e., impulsivity, attention deficit, and substance abuse are associated with increased risk of violence, including bullying and general aggression. At the family and peer levels, experiencing and/or witnessing harsh parenting, domestic violence, and antisocial peer behaviors increase the risk of adolescent violent behaviors. At the community level, community violence, lack of social cohesion, and cultural norms inform peer violence [6]. These socioecological factors intersect with structural factors, including poverty, racial/ethnic discrimination, and gender discrimination to inform violent behaviors and outcomes [10,11]. While these factors are traditionally examined in relation to perpetration and victimization separately, there is considerable similarity in the personal and social factors associated with perpetration and victimization, leading to a growing interest into the perpetrator–victim overlap [12].

Preventing all forms of violence is a fundamental human right enshrined in the Convention on the Rights of the Child and in the Sustainable Development Goal 16.2 [13]. Interventions to prevent violence generally address risks and protective factors in specific populations, such as adolescents, and/or structural determinants of inequities and discrimination [14]. At the intersection of these approaches, women and girl’s empowerment is a particular mechanism invoked to reduce violence against women, by increasing women’s agency, while challenging hegemonic forms of masculinity that promote male dominance and aggression through greater bargaining power [15,16]. A number of studies support raising girls’ voice and agency to prevent gender-based violence [17], but a few studies also alert to the potential backlash associated with challenging the gender order by raising women’s autonomy and reducing male dominance, which has, in some cases, resulted in increased gender-based violence [18-20].

While there is growing interest in empowering young people to challenge harmful norms and become the agents of their own lives, there is little understanding of how agency relates to peer violence perpetration and victimization in this stage of the life course, when gendered expectations intensify, encouraging young girls to stay home and to rely on others to meet their needs, while promoting boys’ independence and emancipation [21]. Agency is defined as the ability to set and achieve one’s goals by mobilizing resources [22,23]. Agency is multidimensional including “the power to” or the ability to act according to one’s goals, “the power over” or the ability to influence others, and the “power with” or the ability to exercise collective action [24]. It touches on all spheres of life and expands significantly in this critical stage of the life course [25-27]. While current efforts focus on collective empowerment to prevent peer violence, less is known about the potential of individual agency (“power to”) to refrain from or engage in peer violence in early adolescents and how these associations differ by gender and by social context.

The Global Early Adolescent Study (GEAS), a multisite investigation of gender socialization and adolescent health and well-being, offers a unique opportunity to explore these associations, using validated multidimensional measures of agency exploring freedom of movement, voice, and decision-making power among very young adolescents (VYAs). We used these measures to assess the relationship between agency and peer violence perpetration and victimization among poor urban VYA in Kinshasa, Democratic Republic of Congo (DRC) and Blantyre, Malawi. We hypothesize that boys and girls with more freedom of movement would have greater exposure to violence (victimization and perpetration) and that greater voice and decision-making among boys would be associated with violence perpetration. We anticipate a more complex relationships for girls as greater agency could increase their ability to resist aggression but could also increase victimization if they do not conform to the gender order, as suggested in the systematic review of empowerment programs for adolescent girls [17].

Methods

This study is reported as per the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline (S1 STROBE Checklist).

Sampling and recruitment

The GEAS is a prospective study currently conducted in 10 countries (12 sites) across five continents. In this study, we used baseline (cross-sectional) data from the 2 GEAS sites in sub-Saharan Africa: Kinshasa, DRC and Blantyre, Malawi. S1 IRB and S2 IRB outline the Institutional Review Board (IRB) submission and study protocol for both sites. Both sites used the same survey instruments and eligibility criteria (adolescents aged 10 to 14 years who provided adolescent assent and parental consent to participate and who were able to understand the questionnaire), but sampling and data collection procedures differed.

In Kinshasa, in-school and out-of-school adolescents were selected form two urban poor communes selected by Save the Children for an intervention (Growing Up Great!). Schools were purposively selected in the intervention and control zones after stratification by commune and school type (public school, religiously affiliated non-for-profit schools, and private religious schools). In the intervention arm, 25 students per school were selected from the Growing Up Great! school clubs. In the control arm, 25 adolescents were randomly selected after stratification by gender and age. Out-of-school adolescents were recruited from the same communes and randomly selected from a listing of households. A total of 2,840 adolescents (2,016 in school and 826 out of school) were selected and provided assent and parental consent to participate. All participants who were approached completed the survey.

In Blantyre, sampling was a two-stage process with the selection of schools and students for an intervention (Very Young Adolescence 2.0), which was ultimately canceled due to the Coronavirus Disease 2019 (COVID-19) pandemic. Eight public urban schools that offered sixth grade were listed and purposively allocated to “intervention” or “control” status in consultation with the city education authorities and head teachers. Class registers for sixth grade in each of the schools were used to select between 200 and 550 adolescents per school depending on number of students at each school. The sample for the present analysis includes 1,694 participants, out of the 2,140 students who had been approached before the COVID-19 shutdown.

In both sites, cases missing more than 15% of data were dropped (n = 10 (0.4%) in Kinshasa and n = 139 (8.2%) in Malawi). In addition, 27 (0.9%) respondents in Kinshasa and 53 (3.1%) respondents in Blantyre were excluded due to missing information on the main independent and dependent variables. The analytical samples included 3,753 adolescents (n = 2,540 in Kinshasa and n = 1,213 in Blantyre).

Data collection occurred in schools and/or community centers in November 2017 in Kinshasa and February to March 2020 in Blantyre. In Kinshasa, adolescents were interviewed face to face by trained interviewers due to low literacy rates, while in Blantyre, adolescents self-completed the survey using Audio Computer-Assisted Self-Interview features. All surveys were uploaded to the SurveyCTO server and compiled into deidentified datasets. The study received ethical approval from both sites (Kinshasa School of Public Health and College of Medicine, University of Malawi). The study was deemed exempt for secondary data collection by the Johns Hopkins School of Public Health’s IRB.

Measures

S1 Table outlines the questions and responses used from the GEAS questionnaire to answer the research question in this manuscript.

Our outcome measure of peer violence perpetration and victimization was based on responses to the following two questions: (i) During the last six months, have you slapped, hit, or otherwise physically hurt another boy or girl in a way that they did not want? (ii) During the last six months, have you been slapped, hit, or otherwise physically hurt by another boy or girl in a way that you did not want? We created a four-level categorical variable: “no peer violence,” “victimization alone,” “perpetration alone,” and “perpetration and victimization.” This classification was intended to identify the high level of overlap between perpetration and victimization while isolating differential influences of agency on each component. While the GEAS study collected information about the gender of the protagonist, we did not include this information in the indicator due to small sample size concerns. However, we report the distribution of violence by respondent and protagonist gender in the result section.

We used the GEAS cross-culture measures of agency that reflect three common dimensions of women’s empowerment (freedom of movement, voice, and decision-making power), which are salient in early adolescence. The dimensions were adapted to the developmental stage of early adolescence, when young people have limited autonomy. The measures were validated in formative research across 11 settings in 5 continents [28]:

freedom of movement (5 items) measuring young people’s ability to circulate in their community without adult supervision (Polychoric Alpha of 0.71 in Kinshasa and 0.87 in Blantyre);

voice (7 items) assessing young people’s ability to express their needs and opinions (Polychoric Alpha of 0.78 in Kinshasa and 0.85 in Blantyre); and

decision-making (4 items) evaluating young people’s influence over daily decisions (Polychoric Alpha of 0.75 in Kinshasa and −0.77 in Blantyre).

The items were scored on a 4-point Likert scale and summarized into three continuous measures. The measures were not normally distributed, and, therefore, we created score tertiles, based on the distribution of the indicator by site. Correlations between the three agency scores ranged from 0.1 to 0.6.

Following Blum and Heise’s ecological models [10,29], we considered individual, family, peer, school, and neighborhood covariates that were related to interpersonal violence [11]. At the individual level, we explored age, educational attainment (appropriate grade for age), perceptions of gender stereotypical traits, and history of ACEs (none, 1, 2, 3, and 4 or more). Gender stereotypical traits were captured in a 7-item scale assessing adolescent’s perceptions of male toughness and female vulnerability [30]. Family factors included family structure (no parent, single parent, and both parent household), parent connectedness (assessed by the question “do you feel close to your main caregiver? By close, we mean you talk to that person and tell them about personal and important things,” dichotomized as “a lot” or “less than a lot”), parental awareness/monitoring (assessed by caregiver’s knowledge of adolescents’ friends, school performance, and general whereabouts and dichotomized “yes” or “no”), and wealth assets (quintiles); peer factors considered peer composition (dichotomized as “same gender friends” or “opposite gender friends”) and time spent with friends (dichotomized as “every day” or “less than every day”). Finally, community level factors included neighborhood social cohesion (assessed by “how likely is it that an adult in your neighborhood would do something like intervene if children or teenagers were (a) damaging property; (b) spraying paint on walls (graffiti); (c) bullying or threatening another person; and (d) fighting with another person”). An additive index was created and dichotomized as “low” and “high” based on the median in each site.

Analysis

We first described the distribution of the individual, family, peer, and community characteristics of the study population by gender and site and tested for gender differences using chi-squared tests or unpaired t tests (in the case of gender norm perceptions and agency scores). We then estimated the percentage of adolescents reporting violence perpetration and victimization in the last six months by gender and site and tested for gender differences using chi-squared tests. Next, we tested for differences in the distribution of the four categorical measure of peer violence by tertile of each agency measure, using chi-squared tests. Next, we ran multivariate multinomial logistic regression models to evaluate the relationships between agency indicators and violence victimization or perpetration or both, using no violence as the reference category and adjusting for all other covariates (individual, family, peer, and community factors) in each site. The models produce adjusted relative risk ratios (RRRs), which indicate how the risk of the outcome falling in one if the comparison groups (victimization, perpetration, or victimization/perpetration overlap) compared to the risk of being in the reference group (no violence) changes with one unit change of the independent variable (agency tertiles), after adjusting for other covariates [31]. All analyses were stratified by gender to investigate gender differences in the association between agency and peer violence experiences. Likewise, we stratified the analysis by site to account for cultural and study implementation differences. We tested for differences in the associations between agency and peer violence by school status in Kinshasa but found not overall interactions and therefore combined these groups into one. Finally, we pooled our data across sites and ran a multivariate multinomial logistic regression model among boys and girls separately to see if the associations between agency measures and violence differed by site. We tested the overall interactions for boys and girls using the command “testparm.” All analysis were conducted using Stata 14.2 [32].

Results

The characteristics of the study samples are presented in Table 1. A total of 2,540 adolescents were included in Kinshasa, evenly distributed between girls (49.8% (n = 1,266)) and boys (50.2% (n = 1,274)). A total of 1,213 adolescents responded in Blantyre with 50.7% of girls (n = 615) and 49.3% of boys (n = 598). A majority of adolescents, 64.1% in Kinshasa and 60.4% in Blantyre, were between 10 and 12 years old. In Kinshasa, 60% of adolescents were living with both parents, while this percentage dropped to 31% in Blantyre. Most adolescents had experienced ACEs, with 19% to 28% reporting 4 ACEs or more. Adolescents were more likely to fall below grade level in Blantyre, while 26% of girls and 31% of boys in Kinshasa were out of school. Adolescents generally perceived a high level of social cohesion in their community (70.5% to 74.3%). A majority of adolescents in both sites perceived unequal stereotypical gender traits (mean: 4.2 to 4.5 on a scale from 1 to 5), which considered boys as tough and girls as vulnerable.

Table 1

Sociodemographic, family, peer, and community characteristics by gender and by site.

		Kinshasa, DRC			Blantyre, Malawi
		Girls n = 1,266	Boys n = 1,274	p-value	Girls n = 615	Boys n = 598	p-value
Age	10 to 12 years	66.1% (837)	62.0% (790)	0.03	63.7% (392)	57.0% (341)	0.02
	13 to 14 years	33.9% (429)	38.0% (484)		36.3% (223)	43.0% (257)
Educational attainment	Out of school	26.1% (330)	30.9% (394)		Data not availablea
	Lower than age expected grade	14.3% (181)	14.2% (181)	0.02	62.3% (383)	73.8% (441)	<0.001
	Age expected grade or higher	59.6% (755)	54.9% (699)		37.7% (232)	26.3% (157)
ACEs	0 ACEs	25.4% (322)	18.7% (238)		31.9% (196)	21.2% (127)
	1 ACEs	21.7% (275)	23.0% (293)		21.0% (129)	21.6% (129)
	2 ACEs	18.8% (238)	21.0% (267)	0.002	16.4% (101)	16.1% (96)	<0.001
	3 ACEs	12.5% (158)	14.5% (185)		11.7% (72)	12.7% (76)
	4 + ACEs	21.6% (273)	22.8% (291)		19.0% (117)	28.4% (170)
Household structure	Both parents	59.2% (749)	56.0% (714)		30.9% (190)	38.5% (230)
	One parent	26.9% (340)	29.5% (376)	0.25	62.0% (381)	53.5% (320)	0.01
	Grandparents and other	14.0% (177)	14.4% (184)		7.2% (44)	8.0% (48)
High parental closeness b		60.9% (771)	62.2% (792)	0.51	78.5% (483)	74.4% (445)	0.09
High parental monitoring and awareness about adolescent’s friends, school performance, and whereabouts b		50.2% (635)	46.1% (587)	0.04	98.4% (605)	99.0% (592)	0.34
Peer composition b	Any opposite gender friends	38.6% (489)	45.8% (583)	<0.001	72.5% (446)	80.3% (480)	0.002
Time spent with friends b	Nearly every day	45.4% (575)	59.0% (752)	<0.001	Data not availablea
Social cohesion	High	70.5% (893)	74.3% (946)	0.04	71.5% (440)	70.6% (422)	0.71
Gender stereotypical trait c		4.5 (0.6)	4.4 (0.7)	<0.001	4.2 (0.8)	4.3 (0.7)	<0.001
Agency (mean score on a scale 1 to 4, SD)	Freedom of movement overall	1.4 (0.6)	1.8 (0.8)	<0.001	1.8 (0.7)	2.3 (0.8)	<0.001
	Tertile 1	1.0 (0.01)	1.0 (0.01)	0.90	1.2 (0.2)	1.2 (0.2)	0.50
	Tertile 2	1.4 (0.2)	1.5 (0.2)	<0.001	1.9 (0.2)	1.9 (0.2)	0.15
	Tertile 3	2.3 (0.5)	2.5 (0.6)	<0.001	2.9 (0.5)	3.1 (0.5)	0.06
	Voice overall	2.4 (0.7)	2.5 (0.7)	<0.001	2.9 (0.7)	3.0 (0.7)	0.009
	Tertile 1	1.7 (0.3)	1.7 (0.3)	0.09	2.1 (0.5)	2.2 (0.5)	0.21
	Tertile 2	2.5 (0.2)	2.5 (0.2)	0.61	3.1 (0.2)	3.1 (0.2)	0.12
	Tertile 3	3.2 (0.3)	3.2 (0.3)	0.37	3.7 (0.2)	3.7 (0.2)	0.10
	Decision-making overall	2.7 (0.9)	2.7 (0.9)	0.89	2.9 (0.8)	3.0 (0.8)	0.003
	Tertile 1	1.7 (0.4)	1.7 (0.4)	0.02	1.9 (0.5)	1.9 (0.5)	0.85
	Tertile 2	2.9 (0.3)	2.9 (0.3)	0.23	3.0 (0.2)	3.0 (0.2)	0.22
	Tertile 3	3.8 (0.2)	3.8 (0.2)	0.42	3.8 (0.2)	3.8 (0.2)	0.37
Perpetrated violence in the last six months	No	76.9% (973)	71.0% (904)	<0.001	79.0% (486)	69.7% (417)	0.001
	Yes, both boys and girls	2.8% (35)	6.7% (85)		5.7% (35)	6.5% (39)
	Yes, a boy or boys	6.5% (81)	17.4% (222)		6.0% (37)	11.7% (70)
	Yes, a girl or girls	14.0% (177)	5.0% (63)		9.3% (57)	12.0% (72)
Victimized in the last six months	No	80.3% (1,016)	70.8% (903)	<0.001	67.3% (414)	57.2% (342)
	Yes, both boys and girls	1.4% (18)	5.7% (73)		7.8% (48)	6.9% (41)	<0.001
	Yes, a boy or boys	6.2% (79)	18.8% (240)		12.5% (77)	16.1% (96)
	Yes, a girl or girls	12.1% (153)	4.6% (58)		12.4% (76)	19.9% (119)
Peer violence composite	No violence or victimization	67.5% (854)	60.1% (765)	<0.001	61.1% (376)	49.5% (296)	<0.001
	Victimization only	9.4% (119)	10.9% (139)		17.9% (110)	20.2% (121)
	Perpetration only	12.8% (162)	10.8% (138)		6.2% (38)	7.7% (46)
	Victimization and perpetration	10.4% (131)	18.2% (232)		14.8% (91)	22.6% (135)

aData not available because these questions were not included in the questionnaire.

bParental closeness, parental monitoring and awareness, friend composition, and time spent with friends were collapsed and made into dichotomous variables based on the distribution of the data.

cGender stereotypical traits score ranged from 1 to 5, with higher scores indicating more unequal perceptions of gender norms.

Chi-squared tests were used to estimate gender difference within each site for all variables except gender norm perceptions and agency, which used unpaired t test.

ACEs, adverse childhood experience; DRC, Democratic Republic of Congo; SD, standard deviation.

Across sites, adolescents scored higher on scales of voice (range: 2.4 to 3.0 on a scale from 1 to 4) and decision-making (range: 2.7 to 3.0 on a scale from 1 to 4) than on freedom of movement (range: 1.4 to 2.3 on a scale from 1 to 4). Overall, 18.2% of boys in Kinshasa and 22.6% of boys in Blantyre reported experiencing a perpetrator–victim overlap in the last six months, while this was true of 10.4% to 14.8% of girls in each site, respectively. Gender differences in perpetration or victimization alone were negligible (Kinshasa perpetration: p = 0.12, Kinshasa victimization: p = 0.21, Blantyre perpetration p = 0.30, and Blantyre victimization p = 0.30). Peer violence mostly involved same gender peers in Kinshasa, for example, among adolescents who were victimized, 64% of boys were victimized by boys only, and 61% of girls were victimized by girls only. Likewise, among adolescents who perpetrated violence, 60% of boys committed aggression against boys only, and 60% of girls committed aggression against girls only. Experiences were more diversified in Blantyre. Among adolescents who were victimized, 38% of boys were victimized by boys only and 46% by girls only, and 38% of girls were victimized by girls only and 38% were victimized by boys only. Among adolescents who perpetrated violence, 39% boys committed aggression on boys only, and 40% committed aggression against girls only, while 44% of girls committed aggression against girls only, and 29% perpetrated violence against boys only.

Table 2 shows the proportion of peer violence experience according to an adolescent’s level of agency (in tertiles). Among boys in both sites and girls in Blantyre, there was a higher percentage of overlap of perpetration and victimization among those who expressed the greatest freedom of movement (24.0% among Kinshasa boys in Tertile 3 versus 12.5% in Tertile 1 p < 0.001; 33.2% among Blantyre boys in Tertile 3 versus 9.0% in Tertile 1 p < 0.001; and 26.7% among Blantyre girls in Tertile 3 versus 9.5% in Tertile 1 p < 0.001). Among boys in Kinshasa and girls in Blantyre, there were higher percentages of adolescents reporting a perpetrator–victim overlap among those with the most voice (22.2% among Kinshasa boys in Tertile 3 versus 11.2%, in Tertile 1 p = 0.002 and 16.4% among Blantyre girls in Tertile 3 versus 12.0% in Tertile 1 p = 0.05). Finally, among boys in Kinshasa and girls in both sites, there was a higher percentage of adolescents reporting a perpetrator–victim overlap among those with the most decision-making power (27.8% among Kinshasa boys in Tertile 3 versus 9.1% in Tertile 1 p < 0.001; 13.5% among Kinshasa girls in Tertile 3 versus 9.2% in Tertile 1 p = 0.04; and 16.6% among Blantyre girls in Tertile 3 versus 12.6% in Tertile 1 p = 0.04).

Table 2

Proportion of peer violence perpetration and/or victimization according to adolescent’s level of agency (freedom of movement, voice, and decision-making) by site and gender.

	Kinshasa, DRC					Blantyre, Malawi
	No violence (95% CI) (n)	Victim (95% CI) (n)	Perpetrator	Victim and perpetrator (95% CI) (n)	p-value	No violence (95% CI) (n)	Victim (95% CI) (n)	Perpetrator	Victim and perpetrator (95% CI) (n)	p-value
	(95% CI) (n)					(95% CI) (n)
Boys
Freedom of movement
Tertile 1	69.0%	8.5%	10.0%	12.5%		70.5%	18.0%	2.5%	9.0%
	(64 to 73.4)	(6 to 11.7)	(7.4 to 13.4)	(9.6 to 16.1)		(61.8 to 77.9)	(12.1 to 25.9)	(0.8 to 7.4)	(5.0 to 15.6)
	(276)	(34)	(40)	(50)		(86)	(22)	(3)	(11)
Tertile 2	65.7%	10.1%	9.2%	15.0%	<0.001	47.6%	25.5%	10.1%	16.8%	<0.001
	(60.2 to 70.8)	(7.2 to 14.1)	(6.4 to 12.9) (28)	(11.4 to19.5) (46)		(40.9 to 54.4)	(20.0 to 31.9)	(6.7 to 15.0)	(12.3 to 22.6)
	(201)	(31)				(99)	(53)	(21)	(35)
Tertile 3	50.7%	13.0%	12.3%	24.0%		41.4%	17.2%	8.2%	33.2%
	(46.6 to 54.8)	(10.5 to 16.0)	(9.9 to 15.3)	(20.6 to 27.6)		(35.6 to 47.4)	(13.1 to 22.2)	(5.5 to 12.2)	(27.8 to 39.1)
	(288)	(74)	(70)	(136)		(111)	(46)	(22)	(89)
Voice
Tertile 1	66.3%	10.9%	11.7%	11.2%		50.9%	17.5%	9.6%	22.0%
	(61.5 to 70.7)	(8.2 to 14.4)	(8.9 to 15.2)	(8.4 to 14.6)		(43.5 to 58.2)	(12.6 to 23.9)	(6.0 to 14)	(16.5 to 28.8)
	(267)	(44)	(47)	(15)		(90)	(31)	(17)	(39)
Tertile 2	57.7%	11.0%	10.7%	20.7%	0.002	46.0%	23.8%	8.9%	21.3%	0.38
	(52 to 62.4)	(8.2 to 14.4)	(8.1 to 14.1)	(17.0 to 24.9)		(39.3 to 53.0)	(18.4 to 30.1)	(5.7 to 13.7)	(16.2 to 27.5)
	(237)	(45)	(44)	(85)		(93)	(48)	(18)	(43)
Tertile 3	56.7%	10.9%	10.2%	22.2%		51.6%	19.2%	5.0%	24.2%
	(52.2 to 61.2)	(8.3 to 14.1)	(7.8 to 13.3)	(18.6 to 26.2)		(45.0 to 58.2)	(14.5 to 25.0)	(2.8 to 8.9)	(19.0 to 30.3)
	(261)	(50)	(47)	(102)		(113)	(42)	(11)	(53)
Decision-making
Tertile 1	66.4%	14.9%	9.6%	9.1%		54.8%	17.4%	9.6%	18.1%
	(61.6 to 70.9)	(11.7 to 18.8)	(7.1 to 12.9)	(6.6 to 12.4)		(47.2 to 62.3)	(12.4 to 24.0)	(6.0 to 15.2)	(12.9 to 24.7)
	(263)	(59)	(38)	(36)		(91)	(29)	(16)	(30)
Tertile 2	59.4%	9.1%	12.1%	19.4%	<0.001	47.7%	23.2%	8.3%	20.8%	0.15
	(55.4 to 63.4)	(7.0 to 11.7)	(9.6 to 15.0)	(16.4 to 22.9)		(41.1 to 54.4)	(18.0 to 29.3)	(5.3 to 12.9)	(15.9 to 26.8)
	(340)	(52)	(69)	(111)		(103)	(50)	(18)	(45)
Tertile 3	52.9%	9.2%	10.1%	27.8%		47.2%	19.4%	5.6%	27.8%
	(47.3 to 58.4)	(6.4 to 12.9)	(7.2 to 14.1)	(23.0 to 33.1)		(41.6 to 53.9)	(14.7 to 25.3)	(3.2 to 9.5)	(22.2 to 34.2)
	(162)	(28)	(31)	(85)		(102)	(42)	(12)	(60)
Girls
Freedom of movement
Tertile 1	68.7%	9.4%	12.0%	9.9%		71.3%	13.8%	5.5%	9.5%
	(64.8 to 72.3)	(7.3 to 12.1)	(9.6 to 14.8)	(7.8 to 12.6)		(65.4 to 76.5)	(10.0 to 18.6)	(3.3 to 9.1)	(6.4 to 13.7)
	(408)	(56)	(71)	(59)		(181)	(35)	(14)	(95)
Tertile 2	65.5%	9.4%	14.9%	10.2%	0.85	61.4%	18.7%	5.4%	14.5%	<0.001
	(60.6 to 70.1)	(6.9 to 12.8)	(11.7 to 18.8)	(7.5 to 13.6)		(55.1 to 67.4)	(14.2 to 24.1)	(3.2 to 9.1)	(11.0 to 19.6)
	(251)	(36)	(57)	(39)		(148)	(45)	(13)	(35)
Tertile 3	67.5%	9.3%	11.8%	11.4%		39.2%	25.0%	9.2%	26.7%
	(61.8 to 72.6)	(6.5 to 13.3)	(8.5 to 16.0)	(8.2 to 15.6)		(30.8 to 48.2)	(18.0 to 33.6)	(5.1 to 15.8)	(19.5 to 35.3)
	(195)	(27)	(34)	(33)		(47)	(30)	(11)	(32)
Voice
Tertile 1	65.7%	10.4%	12.2%	11.8%		67.1%	15.3%	5.6%	12.0%
	(61.3 to 69.7)	(8.0 to 13.3)	(9.6 to 15.4)	(9.2 to 15.0)		(60.6 to 73.1)	(11.1 to 20.7)	(3.2 to 9.5)	(8.3 to 17.1)
	(323)	(51)	(60)	(58)		(145)	(33)	(12)	(26)
Tertile 2	65.3%	9.9%	15.1%	9.7%	0.23	52.5%	24.0%	7.4%	16.2%	0.05
	(60.4 to70.0)	(7.4 to 13.3)	(12.0 to 19.0)	(7.1 to 13.0)		(45.6 to 59.2)	(18.6 to 30.4)	(4.5 to 11.9)	(11.7 to 21.9
	(263)	(40)	(61)	(39)		(107)	(49)	(15)	(33)
Tertile 3	72.2%	7.6%	11.1%	9.2%		63.6%	14.4%	5.6%	16.4%
	(67.4 to 76.6)	(5.3 to 10.7)	(8.2 to 14)	(6.6 to 12.6)		(56.6 to 70.1)	(10.1 to 20.0)	(3.1 to 9.9)	(11.8 to 22.3)
	(268)	(28)	(41)	(34)		(124)	(28)	(11)	(32)
Decision-making
Tertile 1	69.2%	9.0%	12.6%	9.2%		69.6%	11.1%	6.8%	12.6%
	(64.8 to 73.3)	(6.7 to 12.0)	(9.8 to 16.0)	(6.9 to 12.3)		(62.9 to 75.5)	(7.5 to 16.2)	(4.0 to 11.1)	(8.7 to 17.8)
	(308)	(40)	(56)	(41)		(144)	(23)	(14)	(26)
Tertile 2	71.0%	9.5%	10.5%	9.0%	0.04	57.1%	21.9%	5.6%	15.5%	0.04
	(66.7 to 74.9)	(7.1 to 12.5)	(8.1 to 13.7)	(6.7 to 12.0)		(50.6 to 63.3)	(17.0 to 27.7)	(3.3 to 9.4)	(11.3 to 20.7)
	(330)	(44)	(49)	(42)		(133)	(51)	(13)	(36)
Tertile 3	60.7%	9.8%	16.0%	13.5%		56.6%	20.6%	6.3%	16.6%
	(55.5 to 65.6)	(7.1 to 13.4)	(12.6 20.2)	(10.3 to 17.5)		(49.1 to 63.8)	(15.2 to 27.2)	(3.5 to 11.0)	(11.7 to 22.9)
	(216)	(35)	(57)	(48)		(99)	(36)	(11)	(29)

Chi-squared tests were used to assess differences in peer violence experiences by agency tertiles.

DRC, Democratic Republic of Congo.

In the multivariate analyses (Table 3) including all confounding variables, boys with the greatest freedom of movement had elevated RRR of a perpetrator–victim overlap relative to those with the lowest freedom of movement in both sites (Kinshasa Tertile 3 versus Tertile 1: RRR = 1.9 (95% CI: 1.2 to 2.8), p = 0.003; Blantyre Tertile 2 versus Tertile 1: RRR = 2.4 (1.1 to 5.5), p = 0.03; and Blantyre Tertile 3 versus Tertile 1: RRR = 3.8 (1.7 to 8.3), p = 0.001). The same was true for girls with greater freedom of movement in Blantyre (Tertile 3 versus Tertile 1: RRR = 2.4 (1.1 to 5.1), p = 0.03). In addition, boys in Kinshasa with the greatest freedom of movement had an elevated relative risk of being victimized relative to those with the least freedom of movement (Tertile 3 versus Tertile 1: RRR = 2.8 (1.7 to 4.6), p < 0.001).

Table 3

RRRs of violence victimization, violence perpetration, and violence victimization and perpetration by agency indicators: results from multivariate multinomial regression models 1.

	Kinshasa, DRC			Blantyre, Malawi
	Victimized RRR (95% CI) (p-value)	Perpetrated RRR (95% CI) (p-value)	Victimized and perpetrated RRR (95% CI) (p-value)	Victimized RRR (95% CI) (p-value)	Perpetrated RRR (95% CI) (p-value)	Victimized and perpetrated RRR (95% CI) (p-value)
Boys
Freedom of movement
Tertile 1	Ref	Ref	Ref	Ref	Ref	Ref
Tertile 2	1.5 (0.9 to 2.5) (p = 0.17)	0.9 (0.5 to 1.6) (p = 0.74)	1.1 (0.7 to 1.7)(p = 0.73)	2.1 (1.1 to 3.9)(p = 0.02)	7.2 (2.0 to 26.4)(p = 0.003)	2.4 (1.1 to 5.5)(p = 0.03)
Tertile 3	2.8 (1.7 to 4.6)(p < 0.001)	1.7 (1.1 to 2.6)(p = 0.03)	1.9 (1.2 to 2.8)(p = 0.003)	1.3 (0.7 to 2.5)(p = 0.45)	5.9 (1.6 to 21.9)(p = 0.009)	3.8 (1.7 to 8.3)(p = 0.001)
Voice
Tertile 1	Ref	Ref	Ref	Ref	Ref	Ref
Tertile 2	1.2 (0.8 to 2.0)(p = 0.42)	1.1 (0.7 to 1.8)(p = 0.67)	1.9 (1.2 to 2.9)(p = 0.003)	1.3 (0.7 to 2.3)(p = 0.45)	0.7 (0.3 to 1.7)(p = 0.44)	0.7 (0.4 to 1.4)(p = 0.37)
Tertile 3	1.2 (0.7 to 2.0)(p = 0.43)	1.2 (0.7 to 1.9)(p = 0.52)	1.8 (1.2 to 2.8)(p = 0.009)	1.0 (0.5 to 1.9)(p = 0.98)	0.4 (0.1 to 1.2)(p = 0.10)	0.7 (0.4 to 1.4)(p = 0.29)
Decision-making
Tertile 1	Ref	Ref	Ref	Ref	Ref	Ref
Tertile 2	0.6 (0.4 to 0.9)(p = 0.007)	1.3 (0.8 to 2.0)(p = 0.30)	2.1 (1.4 to 3.3)(p = 0.001)	1.5 (0.8 to 2.7)(p = 0.20)	0.9 (0.4 to 2.1)(p = 0.83)	1.2 (0.7 to 2.4)(p = 0.51)
Tertile 3	0.6 (0.4 to 1.0)(p = 0.07)	1.1 (0.6 to 1.9)(p = 0.70)	3.0 (1.8 to 4.8)(p < 0.001)	1.4 (0.7 to 2.6)(p = 0.36)	0.6 (0.2 to 1.7)(p = 0.36)	1.7 (0.8 to 3.3)(p = 0.14)
Girls
Freedom of movement
Tertile 1	Ref	Ref	Ref	Ref	Ref	Ref
Tertile 2	1.1 (0.7 to 1.8)(p = 0.63)	1.3 (0.9 to 2.0)(p = 0.20)	1.1 (0.7 to 1.8)(p = 0.58)	1.0 (0.6 to 1.8)(p = 0.89)	0.9 (0.4 to 2.0)(p = 0.77)	1.1 (0.6 to 2.0)(p = 0.84)
Tertile 3	1.0 (0.6 to 1.8)(p = 0.88)	0.9 (0.5 to 1.4)(p = 0.54)	1.1 (0.7 to 1.8)(p = 0.74)	1.9 (0.9 to 3.8)(p = 0.08)	2.5 (0.9 to 6.8)(p = 0.08)	2.4 (1.1 to 5.1)(p = 0.03)
Voice
Tertile 1	Ref	Ref	Ref	Ref	Ref	Ref
Tertile 2	0.8 (0.5 to 1.3)(p = 0.35)	1.2 (0.8 to 1.9)(p = 0.38)	0.8 (0.5 to 1.3)(p = 0.32)	1.9 (1.1 to 3.1)(p = 0.02)	1.8 (0.8 to 4.5)(p = 0.18)	2.0 (1.01 to 3.9)(p = 0.048)
Tertile 3	0.6 (0.3 to 1.0)(p = 0.052)	0.8 (0.5 to 1.3)(p = 0.30)	0.6 (0.4 to 1.0)(p = 0.07)	0.8 (0.4 to 1.5)(p = 0.47)	1.0 (0.4 to 2.7)(p = 0.98)	1.3 (0.7 to 2.7)(p = 0.40)
Decision-making
Tertile 1	Ref	Ref	Ref	Ref	Ref	Ref
Tertile 2	1.1 (0.7 to 1.7)(p = 0.81)	0.7 (0.5 to 1.1)(p = 017)	0.9 (0.5 to 1.4)(p = 0.61)	1.8 (1.0 to 3.2)(p = 0.06)	0.8 (0.3 to 1.8)(p = 0.52)	0.8 (0.4 to 1.6)(p = 0.59)
Tertile 3	1.5 (0.9 to 2.6)(p = 0.14)	1.3 (0.9 to 2.1)(p = 0.18)	1.7 (1.02 to 2.7)(p = 0.04)	1.7 (0.9 to 3.3)(p = 0.1)	0.8 (0.3 to 2.1)(p = 0.68)	1.0 (0.5 to 2.0)(p = 0.94)

1All models estimate the relative risk of violence victimization of violence perpetration and of violence perpetration–victimization overlap relative to no violence according to tertiles of agency after adjusting for all other covariates (individual, family, peer, and community factors).

DRC, Democratic Republic of Congo; RRR, relative risk ratio.

Girls in Blantyre with a greater ability to be heard (voice) had an elevated relative risk of being victimized (Tertile 2 versus Tertile 1: RRR = 1.9 (1.1 to 3.5), p = 0.02) and an elevated risk of reporting an overlap of perpetration and victimization in comparison to girls who had the least voice (Tertile 2 versus Tertile 1: RRR = 1.9 (1.01 to 3.9), p = 0.048). Boys in Kinshasa with a greater ability to be heard had an elevated relative risk of a violence perpetration and victimization overlap relative to those who with the least voice (Tertile 2 versus Tertile 1: RRR = 1.9 (1.2 to 2.9), p = 0.003 and Tertile 3 versus Tertile 1: RRR = 1.8 (1.2 to 2.8), p = 0.009).

Adolescents with the highest decision-making power in Kinshasa had an elevated relative risk of a perpetrator–victim overlap relative to those with the least decision-making power (Boys Tertile 2 versus Tertile 1: RRR = 2.1 (1.3 to 3.2), p = 0.001; Boys Tertile 3 versus Tertile 1: RRR = 3.0 (1.8 to 4.8), p < 0.001; and Girls Tertile 3 versus Tertile 1: RRR = 1.7 (1.0 to 2.7), p = 0.04). While interpretation is limited given methodological differences by site, we also note significant differences in the associations between freedom of movement and peer violence that are stronger among girls in Blantyre compared to girls in Kinshasa girls (S2 Table), which remained even after excluding out-of-school adolescents in Kinshasa (S3 Table). No site interactions were noted for boys.

Other factors related to peer violence in the multivariate analyses are presented in S4 and S5 Tables. In both sites, older boys were less likely to report being victimized compared to younger boys (in Kinshasa: RRR = 0.6 (0.4 to 0.9), p = 0.03 and in Blantyre RRR = 0.4 (0.2 to 0.7), p < 0.001)). Likewise, older boys in Kinshasa were less likely to report perpetration–victimization overlap than younger boys (RRR = 0.6 (0.4 to 0.8), p = 0.001). Boys who were in a grade lower than age-expected grade were less likely to be victimized compared to out-of-school boys in Kinshasa (RRR = 0.5 (0.2 to 0.9), p = 0.03), and boys who were in their age expected grade or higher were less likely to be victimized than boys who were in a grade lower than age-expected grade in Blantyre (RRR = 0.5 (0.3 to 0.97), p = 0.04). In Kinshasa, greater wealth was associated with a lower risk of victimization (wealth quintile 2 versus 1: RRR = 0.6 (0.3 to 0.99), p = 0.048) and a lower risk of perpetration–victimization overlap (wealth quintile 2 versus 1: RRR = 0.5 (0.3 to 0.8), p = 0.005 and wealth quintile 4 versus 1: RRR = 0.4 (0.3 to 0.8), p = 0.004) compared to adolescents living in the poorest households. In both sites, adolescents who reported a history of 2 or more ACEs had elevated risk of all forms of peer violence experiences relative to those with no history of ACEs. Parent and peer relations, social cohesion, and social norms were related to peer violence experiences, although these associations differed by site. Among Kinshasa boys, being close to a parent was associated with an elevated relative risk of victimization (RRR = 1.6 (1.04 to 2.4), p = 0.03) and a lower risk of perpetration (RRR = 0.7 (0.5 to 0.99), p = 0.046) and of perpetration and victimization overlap (RRR = 0.6 (0.4 to 0.8), p = 0.001) compared to boys who were not close to their parents. Likewise, girls in Kinshasa who were close to their parents had a lower relative risk of violence perpetration (RRR = 0.5 (0.4 to 0.8), p = 0.001) and of perpetration–victimization overlap (RRR = 0.5 (0.3 to 0.8), p = 0.002) compared to girls who were not close to their parents. In Blantyre, boys with opposite gender friends had an elevated relative risk of perpetration relative to those with same gender friends (RRR = 3.9 (1.3 to 12.1), p = 0.02), while girls with opposite gender friends had an elevated relative risk of perpetration–victimization overlap (RRR = 2.1 (1.04 to 4.3), p = 0.04). In Kinshasa, high social cohesion was associated with a lower relative risk of perpetration among boys (RRR = 0.6 (95% CI: 0.4 to 0.9), p = 0.02); however, among girls, it was associated with an elevated relative risk of perpetration (RRR = 2.0 (1.3 to 3.0), p = 0.001) and of combined perpetration victimization (RRR = 2.1 (95% CI: 1.3 to 3.3), p = 0.003). In Blantyre, high social cohesion among boys was associated with an elevated relative risk of victimization (RRR = 1.7 (1.03 to 2.8), p = 0.04) and of a victimization and perpetration overlap (RRR = 3.2 (1.8 to 5.7), p < 0.001). Finally, higher perception of gender stereotypical traits was associated with an elevated relative risk of a victimization and perpetration overlap among girls in Kinshasa (RRR = 1.7 (95% CI: 1.1 to 2.6), p = 0.01).

Discussion

This study adds to the scarce literature exploring peer violence and individual agency among VYAs living in poor urban communities in two low-income countries. Boys in both settings were more likely to experience peer violence than girls, mostly due to higher rates of a perpetration and victimization overlap. Agency had different connections to peer violence experiences, by gender and site, although greater freedom of movement, voice, and decision-making power were generally related to increases in some form of peer violence experience. These associations were generally greater for boys than girls and often showed different associations in relation to victimization versus perpetration, when they occurred in isolation of each other.

The gender divide in peer violence exposure is well described in previous literature, reporting 3 to 12 times higher prevalence among boys compared to girls [9]. While mostly described in the form of female victimization and male perpetration [33,34], our results suggest more complex gender dynamics, as differences mostly originated from a greater perpetrator–victim overlap among boys than girls. These results are supported by other studies in Pakistan and Afghanistan indicating a larger perpetrator–victim overlap among boys than girls [35,36], suggesting boys may be more likely than girls to retaliate when victimized. Our study found that in Kinshasa, girls who had unequal perceptions of gender stereotypical traits, promoting male toughness over female vulnerability, were more likely to report both perpetration and victimization, highlighting the connection between gender norms and peer violence experiences in this age group. The fact that this was only true among girls in Kinshasa illustrates the fact that gender norms have different meanings across contexts and influence behaviors in different ways for boys and girls [30].

Gender differences in peer violence experiences partly stem from a gender divide in freedom of movement. Boys enjoy greater freedom of movement than girls and spend more time socializing with their friends [28,37,38], due to normative expectations, prescribing more protection for girls and more independence for boys [21,30]. In addition, freedom of movement was more strongly related to violence experiences among boys than girls. A study of low-income Latino adolescents in two cities in the United States showed that more unstructured time with peers increased community violence and victimization among boys [39]. Together, these findings draw attention to gender differences in social environments and activities outside of the home contributing to greater male exposure and engagement in violence.

Decision-making power and voice, which are expressions of empowerment, are associated with reductions in gender-based violence against women, although greater empowerment may also lead to social backlash resulting in violence victimization [17]. In our study, we also suggest a complex relation between decision-making power, voice, and peer violence experiences, which differs by gender and site. For example, in Kinshasa, greater decision-making power was associated with lower risk of victimization for boys but not for girls. Greater voice among girls was associated with increased risk of peer violence in Blantyre but not in Kinshasa. These variations of effects have been described in a systematic review of interventions to prevent gender-based violence among adolescent girls, showing positive, null, or even negative effects of raising girls’ voice and agency in preventing violence against women, depending on context [17].

Across all groups, a history of childhood adverse events was associated with increased risk of violence victimization, perpetration, or both. These results are consistent with prior studies conducted in high-income and low-income settings, including studies using the current GEAS measure [40,41]. Other contextual factors related to parent and peer relations were also associated with young people’s violence experience, although these relations differed by site. Literature has shown that both family and peer factors affect adolescent violent behaviors [10,11].

Our results need to be interpreted with some limitations in mind. The cross-sectional design of this study limits our interpretation of the perpetrator–victim overlap, as well as the directionality of the associations described. Although we assume that participants with more agency are more prone to peer violence, it could be that those who engage in peer violence have more agency. While our results demonstrate that agency and peer violence are associated, we anticipate that the longitudinal data currently collected from the GEAS will provide further insights on the dynamics of peer violence experiences (from victimization to perpetration and vice versa) as well as the predictive effects of young people’s agency on subsequent violence outcomes. We also acknowledge the limitation of our peer violence measure, which only focuses on physical violence and does not account for gender-based violence behaviors. The small sample sizes prevented more refined analysis of same gender versus opposite gender violence experiences, and additional information would be needed to consider other types of peer violence (verbal, emotional, and cyber), which may also vary by gender and context. We also recognize the potential for residual confounding due to unmeasured confounders in our analysis, particularly as we investigated individual components of adolescent agency but were unable to assess the role of collective empowerment, a critical component of violence prevention programs for teens. Selection bias may also have impacted our findings, particularly if adolescents who are victims of violence or who have limited agency are less likely to be identified and enrolled in the study. If our study differentially selected participants who were less likely to experience violence, it is likely that our findings are conservative. Differences in data collection modalities between sites (in person in Kinshasa and Audio Computer-Assisted Self-Interview in Blantyre) may generate differences in the patterns of responses resulting from modality rather than true differences between sites. This is especially notable in Blantyre, which had a larger number of missing variables compared to Kinshasa. On the other hand, social desirability bias may have impacted responses in Kinshasa collected face to face [42]. We note, however, that violence perpetration was equally prevalent in Blantyre and Kinshasa, which may signal low social desirability bias. Another limitation is that we have little information about the context in which peer violence occurs, which is not only programmatically relevant for violence prevention, but may also inform the relationship between agency and peer violence. Future qualitative work could shed more light on the ways agency and violence interconnect, according to social context. Last, a purposive sample might not affect internal validity but affects the generalizability of our findings for the city and country.

Despite these limitations, this cross-cultural analysis allows a better understanding of peer violence experiences among boys and girls in early adolescence and the complex role of adolescent’s agency in shaping these experiences. Our results have important programmatic implications, raising the possibility that gender socialization may expose boys to more interpersonal violence as they enjoy more freedom of movement and are expected to protect themselves. They also draw attention to the potential for unforeseen consequences of empowerment, which was sometimes associated with increased peer violence experiences in our study. We suggest the following programmatic implications. First, programs that promote young people’s agency need to conjointly address gender norms and violence as normative expectations may lead those who gain power toward increased aggression as a means of assertion and self-protection. Second, the increase in violence perpetration and victimization among adolescents who have experienced ACEs as well as the extent of a perpetration and victimization overlap reaffirms the importance of a trauma-informed care to prevent teen violence. Third, strategies to prevent youth violence need to engage the broader social environment, starting with the family and extending to communities, to reduce violence exposure and perpetration. For instance, while many parents recognize the risk of violence for their sons, many fail to act considering that their sons need to rely on their own agency to protect themselves [21]. The role of the social context, illustrated in the differences in association by site, also stresses the importance of including young people and their social networks in the design and implementation of interventions in order to better address the social and structural drivers of teen violence. Training community members would ensure that interventions are trusted, followed, and sustained.

STROBE guideline checklist.

STROBE, Strengthening the Reporting of Observational Studies in Epidemiology.

(DOCX)

Click here for additional data file.

Baseline IRB Protocol for Kinshasa, DRC.

DRC, Democratic Republic of Congo; IRB, Institutional Review Board.

(DOCX)

Click here for additional data file.

Baseline IRB Protocol for Blantyre, Malawi.

IRB, Institutional Review Board.

(DOCX)

Click here for additional data file.

GEAS questions and measures.

GEAS, Global Early Adolescent Study.

(DOCX)

Click here for additional data file.

Differences in associations between agency and peer violence by site among adolescent boys and girls: results from pooled multivariate multinomial regression model across sites, including interaction terms by site.

(DOCX)

Click here for additional data file.

Overall model with interaction between site and agency variables and excluding out-of-school participants.

(DOCX)

Click here for additional data file.

Factors related to peer violence among boys and girls in Kinshasa: results from multivariate multinomial regression model.

(DOCX)

Click here for additional data file.

Factors related to peer violence among boys and girls in Malawi: results from multivariate multinomial regression model.

(DOCX)

Click here for additional data file.

Dataset used for analysis.

(CSV)

Click here for additional data file.

Codebook of the variables used for analysis.

(PDF)

Click here for additional data file.

16 Feb 2021

Dear Dr Ramaiya,

Thank you for submitting your manuscript entitled "Assessing the Relationship Between Agency and Peer-violence among Very Young Adolescents in Two African Settings: A Cross-Sectional Study" for consideration in PLOS Medicine’s Special Issue on Global Child Health.

Your manuscript has now been evaluated by the PLOS Medicine editorial staff as well as by the Special Issue Guest Editors, and I am writing to let you know that we would like to send your submission out for external peer review.

However, before we can send your manuscript to reviewers, we need you to complete your submission by providing the metadata that is required for full assessment. To this end, please login to Editorial Manager where you will find the paper in the 'Submissions Needing Revisions' folder on your homepage. Please click 'Revise Submission' from the Action Links and complete all additional questions in the submission questionnaire.

Please re-submit your manuscript within two working days, i.e. by .

Login to Editorial Manager here: https://www.editorialmanager.com/pmedicine

Once your full submission is complete, your paper will undergo a series of checks in preparation for peer review. Once your manuscript has passed all checks it will be sent out for review.

Feel free to email us at plosmedicine@plos.org if you have any queries relating to your submission.

Kind regards,

Caitlin Moyer, Ph.D.

Associate Editor

PLOS Medicine

2 Apr 2021

Dear Dr. Ramaiya,

Thank you very much for submitting your manuscript "Assessing the Relationship Between Agency and Peer-violence among Very Young Adolescents in Two African Settings: A Cross-Sectional Study" (PMEDICINE-D-21-00483R1) for consideration in PLOS Medicine’s Special Issue on Global Child Health.

Your paper was evaluated by a senior editor and discussed among all the editors here. It was also discussed with the special issue guest editors, and sent to three independent reviewers, including a statistical reviewer. The reviews are appended at the bottom of this email and any accompanying reviewer attachments can be seen via the link below:

[LINK]

In light of these reviews, I am afraid that we will not be able to accept the manuscript for publication in the journal in its current form, but we would like to consider a revised version that addresses the reviewers' and editors' comments. Obviously we cannot make any decision about publication until we have seen the revised manuscript and your response, and we plan to seek re-review by one or more of the reviewers.

In revising the manuscript for further consideration, your revisions should address the specific points made by each reviewer and the editors. Please also check the guidelines for revised papers at http://journals.plos.org/plosmedicine/s/revising-your-manuscript for any that apply to your paper. In your rebuttal letter you should indicate your response to the reviewers' and editors' comments, the changes you have made in the manuscript, and include either an excerpt of the revised text or the location (eg: page and line number) where each change can be found. Please submit a clean version of the paper as the main article file; a version with changes marked should be uploaded as a marked up manuscript.

In addition, we request that you upload any figures associated with your paper as individual TIF or EPS files with 300dpi resolution at resubmission; please read our figure guidelines for more information on our requirements: http://journals.plos.org/plosmedicine/s/figures. While revising your submission, please upload your figure files to the PACE digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at PLOSMedicine@plos.org.

We expect to receive your revised manuscript by . Please email us (plosmedicine@plos.org) if you have any questions or concerns.

***Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.***

We ask every co-author listed on the manuscript to fill in a contributing author statement, making sure to declare all competing interests. If any of the co-authors have not filled in the statement, we will remind them to do so when the paper is revised. If all statements are not completed in a timely fashion this could hold up the re-review process. If new competing interests are declared later in the revision process, this may also hold up the submission. Should there be a problem getting one of your co-authors to fill in a statement we will be in contact. YOU MUST NOT ADD OR REMOVE AUTHORS UNLESS YOU HAVE ALERTED THE EDITOR HANDLING THE MANUSCRIPT TO THE CHANGE AND THEY SPECIFICALLY HAVE AGREED TO IT. You can see our competing interests policy here: http://journals.plos.org/plosmedicine/s/competing-interests.

Please use the following link to submit the revised manuscript:

https://www.editorialmanager.com/pmedicine/

Your article can be found in the "Submissions Needing Revision" folder.

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. Additionally, PLOS ONE offers an option to publish peer-reviewed clinical study protocols. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols

Please ensure that the paper adheres to the PLOS Data Availability Policy (see http://journals.plos.org/plosmedicine/s/data-availability), which requires that all data underlying the study's findings be provided in a repository or as Supporting Information. For data residing with a third party, authors are required to provide instructions with contact information for obtaining the data. PLOS journals do not allow statements supported by "data not shown" or "unpublished results." For such statements, authors must provide supporting data or cite public sources that include it.

We look forward to receiving your revised manuscript.

Sincerely,

Caitlin Moyer, Ph.D.

Associate Editor

PLOS Medicine

plosmedicine.org

-----------------------------------------------------------

Requests from the editors:

1. Data availability statement: Thank you for noting the specific restrictions for access of your data. Please revise your statement, to include appropriate information allowing interested parties to access your study data (a dataset reference or accession number, a web link, and/or contact email addresses). Please note that the authors of the study (Eric Mafuta and William Stones) may not serve as contact points for providing data access.

2. When submitting your revised manuscript, please provide line numbers throughout the document.

3. Abstract: Please structure your abstract using the PLOS Medicine headings (Background, Methods and Findings, Conclusions; please combine methods and results into a single section “Methods and Findings”).

4. Abstract: Background: Please indicate Blantyre, Malawi,

5. Abstract: Methods and Findings: Please quantify the main results (with 95% CIs and p values).

6. Abstract: Methods and Findings: In the last sentence of the Abstract Methods and Findings section, please describe the main limitation(s) of the study's methodology.

7. Abstract: Conclusions: In the first sentence please address the specific implications of your study based on the results; the phrase "In this study, we observed ..." may be useful.

8. Author summary: At this stage, we ask that you include a short, non-technical Author Summary of your research to make findings accessible to a wide audience that includes both scientists and non-scientists. The Author Summary should immediately follow the Abstract in your revised manuscript. This text is subject to editorial change and should be distinct from the scientific abstract. Please see our author guidelines for more information: https://journals.plos.org/plosmedicine/s/revising-your-manuscript#loc-author-summary

9. Throughout the text: Please use square brackets for in-text citations, like this [1].

10. Methods: Reporting Checklist: Please ensure that the study is reported according to the STROBE guideline, and include the completed STROBE checklist as Supporting Information. When completing the checklist, please use section and paragraph numbers, rather than page numbers. Please add the following statement, or similar, to the Methods: "This study is reported as per the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline (S1 Checklist)."

The STROBE guideline can be found here: http://www.equator-network.org/reporting-guidelines/strobe/

11. Methods: Analysis Plan: Did your study have a prospective protocol or analysis plan? Please state this (either way) early in the Methods section.

a) If a prospective analysis plan (from your funding proposal, IRB or other ethics committee submission, study protocol, or other planning document written before analyzing the data) was used in designing the study, please include the relevant prospectively written document with your revised manuscript as a Supporting Information file to be published alongside your study, and cite it in the Methods section. A legend for this file should be included at the end of your manuscript.

b) If no such document exists, please make sure that the Methods section transparently describes when analyses were planned, and when/why any data-driven changes to analyses took place.

c) In either case, changes in the analysis-- including those made in response to peer review comments-- should be identified as such in the Methods section of the paper, with rationale.

12. Methods: Page 7: Please remove the trademark symbol from SurveyCTO®

13. Methods: Page 7: Please provide more detail on whether the description here applies to both child assent and parental consent, or whether these differed.

14. Methods: Please include the survey questions used to assess these three agency dimensions, and all covariates, as a supporting information file.

15. Methods: Page 9: Please clarify the sentence: “We first examined the incidence of violence perpetration and victimization in the last § months…”

16. Results: Please quantify results presented in the text as much as possible. On page 14, please present the risk ratio in the text along with the results from the multivariate analysis (95% CI and p values) for the findings discussed. Please clarify where differences between sites or sexes were quantified, as opposed to where differences are apparent at the qualitative level.

17. Discussion: Please slightly re-organize the Discussion as follows: a short, clear summary of the article's findings; what the study adds to existing research and where and why the results may differ from previous research; strengths and limitations of the study; implications and next steps for research, clinical practice, and/or public policy; one-paragraph conclusion.

18. Discussion and throughout: It is somewhat confusing to switch back and forth between “gender” and “sex” (even within a single sentence at the top of page 18). I understand the point you make on page 9 that sex is the relevant term in your study as you did not collect data on gender identity, and that your survey contains gender-related measures. Please make sure the two terms (which have different meaning) are used as consistently and clearly as possible throughout.

19. Discussion: In the limitations, please fully discuss limitations of the study design, including selection bias brought up by one reviewer, as well as issues related to residual confounding.

20. Table 1: Please provide the number for each percentage given (even though the total is given at the top of the columns). Please define the abbreviation ACES, SD, in the legend.

21. Figure 1: Please consider that this may be more clearly presented as a table. Please also provide the number in addition to the percentages for each value.

22. Table 2: Please provide the numbers for each percentage. Please report the full statistics for each comparison, along with the 95% CIs and p values, rather than noting significance with asterisks.

23. Table 3: Please report the full p values for each comparison, rather than indicating significance levels with asterisks.

24. References: Please use the "Vancouver" style for reference formatting, and see our website for other reference guidelines https://journals.plos.org/plosmedicine/s/submission-guidelines#loc-references

25. Appendix 1 and Appendix 2: Please provide the actual p values associated with the results presented, rather than levels of significance. In the legend, please briefly explain why parts of the table are grayed out. We suggest renaming these files as “S1_Table” and “S2_Table” respectively.

Comments from the reviewers:

Reviewer #1: This paper addresses the important topic of interpersonal violence among adolescents in Kinshasa and Blantyre. Overall, the study suffers from several limitations related to its cross sectional design, including the potential for reverse causation and selection bias. Moreover, the target population for the study is not explicitly stated, and the method used to sample, invite, and recruit participants was not clear. Therefore, it is difficult to understand to whom results are expected to apply. General and specific comments are provided below.

General comments:

1. The paper does not describe how participants were selected for inclusion. For example, how were the 2016 in school and 826 out of school adolescents in Kinshasa selected? How many were invited to participate? What was the refusal rate?

2. Are there circumstances in which experiences of violence would affect participation in the survey? Certainly this could be true in the unfortunate event that violence led to death. But violence could also affect school drop out, which could affect inclusion in the study in Malawi, and may affect recruitment in other ways. Moreover, agency (freedom of movement, voice, and decision-making) could also affect inclusion in the study. In settings where both exposure and outcome affect selection, the overall prevalence of reported outcomes as well as the associations between exposure and outcome could be distorted. This potential selection bias should be discussed.

3. The authors should comment on whether they believe that the difference in data collection techniques (ACASI vs in person interviews) could result in differences in results between the two countries.

4. The Methods section should clearly state the parameters of interest that will be interpreted in the Results section.

5. The paper acknowledges the limitations of working with cross sectional data in the Discussion section. However, more discussion is needed on the potential for reverse causation (i.e., in which violence affects the agency measures) and what that might mean for the interpretation of the results.

6. The items used to construct measures of agency should be described either in the text or supplementary materials.

Specific Comments:

1. Abstract: the first sentence of the methods states that the surveys were conducted in 2017 and 2020, yet the methods section states that data collection occurred between 2017 and 2020. I recommend providing more detail on specific dates of data collection in each site.

2. Page 9, last paragraph: the manuscript uses a symbol I don't understand here. What was the lookback period for incidence of violence?

3. Table 2: I appreciate that the paper shows absolute percentages experiencing each level of violence by tertile. However, because there are many categories, it is unclear what hypothesis the p-value is testing and what test was used to obtain the p-value. These point should be clarified if p-values are reported.

4. Page 13, first sentence: Table 2 appears to show absolute percentages of each tertile in each category of the outcome, not associations directly.

5. Page 14, first sentence: it is unclear to me what the term "relative risk ratio" represents here. Presumably the results of Table 3 come from the multiviariable multinomial logistic regression model mentioned in the Methods. However, that method would provide odds ratios, rather than risk ratios. If the "relative risk ratios" are calculated using some other approach, it should be detailed in the Methods section.

6. Moreover, to avoid confusion with "multivariate" models (typically used when more than one outcome is of interest), I would refer to the analysis in table 3 as the "adjusted analysis" rather than "multivariate analysis".

7. Page 14, second paragraph: the paper appears to commit the Table 2 fallacy, in which a multivariable model was used to assess, and interpret, relationships between all variables included in the model and the outcome. However, using results from a single multivariable model, one should not interpret all of the resulting regression coefficients because many of the characteristics affect, and are affected by, each other. This means that the interpretation of some resulting regression coefficients (or odds ratios) would be as "total effects" and others would be "direct effects after holding mediators constant" This phenomenon is nicely described in the reference below.

Daniel Westreich, Sander Greenland; The Table 2 Fallacy: Presenting and Interpreting Confounder and Modifier Coefficients, American Journal of Epidemiology, Volume 177, Issue 4, 15 February 2013, Pages 292-298, https://doi.org/10.1093/aje/kws412

8. Table 3, in the methods section and/or the footnote of the table, please state what p-values are comparing and how they were obtained.

9. Figure 1 was a bit difficult to read, I recommend larger text.

Reviewer #2: Review of Assessing the Relationship Between Agency and Peer-violence among Very Young Adolescents in Two African Settings: A Cross-Sectional Study

Overall, this is an interesting and competently written paper focusing on an important and often neglected age cohort - very young adolescents - in two urban populations in SSA. It makes the important argument that sex differences in the perpetration or victimisation of peer violence vary across context and to support interventions more precise understanding and operationalisation of individual adolescent empowerment is required.

In some ways, however, the article raises more questions than it answers and either additional analysis should be undertaken to address these or the limitations section expanded.

First, it would be good if the article could more clearly state whether peer violence is referring to violence experiences with young people only of the same sex or among girls and boys, and if so what proportion. For example, is the analysis of girls' experience of peer violence capturing violence by girls against girls or by boys or both and what is the balance of this experience. This matters in terms of the sorts of sensitisation and monitoring required.

Second, in what spaces is the violence occurring. The paper makes the argument that greater freedom of movement permitted of boys means that they are more exposed to peer violence experiences - but where is the violence happening? At school or in the community? And do we see differences in the Kinshasa sample based on school enrolment status?

Third, there are number of areas where the reader wants to know why. With only survey data -rather than mixed methods data (although my understanding is that GEAS does also collect qualitative data) - it is difficult to get purchase on for example why were sex differences in the perpetration or victimisation alone negligible? Why did girls with greater voice also find themselves still at risk of victimisation by peers -was this due to inter-personal communication skills (e.g. limited negotiation skills, gender norms that de-value girls' voices, participation in programming that focused on young people's empowerment but did not foster change in the wider enabling environment), gender-based harassment and violence, and so forth? If it is not possible to draw on qualitative data to triangulate the findings for this paper, then it would be good to add this to the limitations section.

Fourth, given the sensitivity of asking about experiences of violence, it would be important for the authors to reflect more on possible limitations of different survey data collection methods (i.e. with trained enumerators in Kinshasa vs self-reported in Blantyre). Could this potentially explain some of the level of difference across country?

Fifth, given that in Kinshasa there are both in-school and out-of-school adolescents, to what extent does this explain differences across country? Could the data be dis-aggregated and bar graphs added in to show the findings for these different samples? If not, it needs to be explicitly discussed in the limitations. Also the descriptive statistics show that there is a difference among those who at-age-for-grade and those who are over-age-for-grade - can the data be dis-aggregated to explore possible differences in findings among these sub-groups?

Finally, the discussion on implications for programming are quite general and there is a need for more nuanced discussion on how the proposed recommendations could be realistically achieved in such contexts. For example, "Second, while empowering young people is a critical component of positive youth development, promoting individual agency may also have unforeseen consequences, which should be anticipated and addressed to reduce violence experiences". As a reader I would like to know how this could be achieved and whether there are any examples of promising practice in this regard. Similarly, for this recommendation "Finally, the site differences in associations according to peer, family and neighborhood-context stress the need to ground interventions in the social context of young people's lives." - what does this mean in practice - more tailored case management?

With these revisions i am of the view that this paper can make a strong contribution to the literature on adolescent peer violence.

Reviewer #3: Referee Report: Assessing the relationship between agency and peer-violence among Very Young Adolescents in Two African Settings: A Cross-Sectional Study

1. Summary

This paper uses cross-sectional data on very young adolescents from two urban cities (Kinshasa (DRC) and Blantyre (Malawi) to assess the association between agency (operationalized with three scales: freedom from movement, voice, and decision making) and peer violence (defined as four categorical variables: no victimization/perpetration, victimization, perpetration, both victimization and perpetration). Analyzing results by sex and site, the authors find that adolescents reporting greater freedom of movement were more likely to experience offender-victim overlap, with more complicated associations for decision making and voice.

2. Major Comments

The paper was very interesting to read, and I appreciate the focus on defining peer violence looking at both victimization and perpetration (and the combination) with a gender lens. It also highlights the complex relationship between agency and peer violence and provides food for thought on the possible adverse consequences of programs that promote agency, without considering the environments adolescents live in.

I do think the authors could do more to motivate the research question, ensure the empirical analysis reflects the complexity laid out in the motivation, and to build more on the programmatic implications. Some specific suggestions below.

a. I am still not sure what to make of the research question. The introduction does a good job motivating the issue of peer violence and the issue of agency, but less on the association between the two. Why would we expect a relationship between the two in LMICs? And why is this likely to be gendered? What evidence do we already have on this association? If agency puts you at risk of exposure to violence, what does this mean for policy etc.? This comes out a bit clearer in the discussion, but I think the introduction could provide clearer motivation of the evidence base and what this paper hopes to add.

b. The research question is motivated within a socioecological model, and I am assuming Table 2 controls for all the variables that are in the appendix tables, is that correct? It would be great to unpack this relationship more, and perhaps explore moderators/mediators more explicitly. I personally think those appendix tables are important and would add value to the main text. One way to build on what is here is instead to show Table 1 as a regression (with no additional controls) and then add controls at different levels. Theory/existing evidence may also suggest the possibility of interacting the measures of agency with other relevant measures. Given the richness of the data, it seems like some additional analysis might shed further light on the complex findings.

c. The decision to create tertiles of the three dimensions of agency appears somewhat arbitrary. Is there a literature to justify this treatment of the data? Do results look similar if you have a continuous variable or above/below median? Are these tertiles created within gender/site? If you continue to use tertiles, the data in Table 1 should also be summarized this way. It would be good to know the mean (s.d.) freedom of movement for each tertile etc.

d. Are the differences between boys and girls significant? Are the differences across locations significant? Given that one site has only school kids and the other has both, does that explain the difference in findings across contexts?

3. Minor Comments

* Given that GEAS covers a lot more countries, and now has longitudinal data for some, why limit to these two countries in the cross-section? It also seems like a bid of an odd choice given the very different sampling frames for the two contexts. Perhaps only focusing on in school adolescents in the two contexts would allow for a clearer comparison? I am sure the authors have their reasons, but this needs at least a footnote to justify the choice.

* I suggest adding the data from Figure 1 to Table 1, the figure does not add that much value.

Any attachments provided with reviews can be seen via the following link:

[LINK]

Submitted filename: PLOS Medicine peer review.docx

Click here for additional data file.

26 Jul 2021

Submitted filename: Response to Reviewers.docx

Click here for additional data file.

14 Oct 2021

Dear Dr. Ramaiya,

Thank you very much for submitting your revised manuscript "Assessing the Relationship Between Agency and Peer-violence among Very Young Adolescents in Two African Settings: A Cross-Sectional Study" (PMEDICINE-D-21-00483R2) for consideration in PLOS Medicine’s Special Issue: Global Child Health: From Birth to Adolescence and Beyond.

Your paper was evaluated by a senior editor and discussed among all the editors here. It was also discussed with an academic editor with relevant expertise, and re-evaluated by one of the original reviewers. The reviews are appended at the bottom of this email and any accompanying reviewer attachments can be seen via the link below:

[LINK]

In light of these reviews, I am afraid that we will not be able to accept the manuscript for publication in the journal in its current form. The academic editor has noted several issues requiring further clarification, and we request clarification and more thorough description of the methods and analyses.

We will consider a revised version that addresses the reviewers' and editors' comments. Obviously we cannot make any decision about publication until we have seen the revised manuscript and your response, and we may seek re-review by one or more of the reviewers.

In revising the manuscript for further consideration, your revisions should address the specific points made by each reviewer and the editors. Please also check the guidelines for revised papers at http://journals.plos.org/plosmedicine/s/revising-your-manuscript for any that apply to your paper. In your rebuttal letter you should indicate your response to the reviewers' and editors' comments, the changes you have made in the manuscript, and include either an excerpt of the revised text or the location (eg: page and line number) where each change can be found. Please submit a clean version of the paper as the main article file; a version with changes marked should be uploaded as a marked up manuscript.

In addition, we request that you upload any figures associated with your paper as individual TIF or EPS files with 300dpi resolution at resubmission; please read our figure guidelines for more information on our requirements: http://journals.plos.org/plosmedicine/s/figures. While revising your submission, please upload your figure files to the PACE digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at PLOSMedicine@plos.org.

We expect to receive your revised manuscript by Nov 04 2021 11:59PM. Please email us (plosmedicine@plos.org) if you have any questions or concerns.

***Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.***

We ask every co-author listed on the manuscript to fill in a contributing author statement, making sure to declare all competing interests. If any of the co-authors have not filled in the statement, we will remind them to do so when the paper is revised. If all statements are not completed in a timely fashion this could hold up the re-review process. If new competing interests are declared later in the revision process, this may also hold up the submission. Should there be a problem getting one of your co-authors to fill in a statement we will be in contact. YOU MUST NOT ADD OR REMOVE AUTHORS UNLESS YOU HAVE ALERTED THE EDITOR HANDLING THE MANUSCRIPT TO THE CHANGE AND THEY SPECIFICALLY HAVE AGREED TO IT. You can see our competing interests policy here: http://journals.plos.org/plosmedicine/s/competing-interests.

Please use the following link to submit the revised manuscript:

https://www.editorialmanager.com/pmedicine/

Your article can be found in the "Submissions Needing Revision" folder.

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. Additionally, PLOS ONE offers an option to publish peer-reviewed clinical study protocols. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols

Please ensure that the paper adheres to the PLOS Data Availability Policy (see http://journals.plos.org/plosmedicine/s/data-availability), which requires that all data underlying the study's findings be provided in a repository or as Supporting Information. For data residing with a third party, authors are required to provide instructions with contact information for obtaining the data. PLOS journals do not allow statements supported by "data not shown" or "unpublished results." For such statements, authors must provide supporting data or cite public sources that include it.

We look forward to receiving your revised manuscript.

Sincerely,

Caitlin Moyer, Ph.D.

Associate Editor

PLOS Medicine

plosmedicine.org

-----------------------------------------------------------

Requests from the editors:

1. From the Academic Editor: Please address the following comments:

-The measure of victimization/perpetration is based on only 2 questions focused on physical perpetration/victimization. This type of behavior may be gender-specific, other types of peer-violence (e.g. relational) may be gender-specific as well.

- The temporal ordering between outcome (over the prior 6 months) and primary explanatory variables (agency, at the time of survey) is problematic in the context of the cross-sectional nature of the data.

-The description of agency-related variables is limited and warrants more detail.

-Sex differences are referred to and it seems as if these should be referred to as gender differences.

-The discussion of the analytical strategy is limited, and it is not clear how the nested nature of the data was handled.

Other editorial points:

2. Throughout: Sex vs Gender terminology: It seems as if “sex” was established via the GEAS questions and measure (Supporting File 4) question: “Are you a boy/girl” and we suggest changing “sex” to “gender” throughout. Please also see, https://www.who.int/health-topics/gender.

3. Abstract: Background: It might be helpful to include a sentence briefly describing personal agency.

4. Abstract: Methods and Findings: Please include the percent breakdown by gender for each survey site.

5. Abstract: Methods and Findings: Please report exact p values, unless p<0.001.

6. Abstract: Methods and Findings: Please clarify that T represents tertile at first use of the abbreviation.

7. Abstract: Line 60-61: We suggest revising to “The main limitation of the study is that due to the cross-sectional nature…”

8. Author summary: Line 85-86: Here, and throughout the manuscript, please avoid language implying causality: “...higher agency resulted in higher likelihood of being in the perpetrator-victim overlap group…”

9. Introduction: Line 135-140: Please clarify the final sentence of this paragraph. You mention that empowerment of women/girls is a mechanism to reduce violence against women and challenge hegemonic forms of masculinity, but in the final sentence mention that empowerment programs could backfire in threatening male hegemony in relationships.

10. Methods: Line 174-175: Please provide some more detail on how participants were selected, including selection of schools, participants within schools, participants not enrolled in school.

11. Methods: Line 199-200: Please note that this was parental consent and adolescent assent.

12. Methods: Line 228: Wealth quintile is reported in the S5 and S6 Tables, but the assessment of wealth is not described in the Methods.

13. Methods: Line 259-267: Please consider if the description of your study hypotheses would be better presented in the last paragraph of the Introduction.

14. Results: Please report the numbers of participants at each site, and proportions by gender. present numerators and denominators for percentages.

15. Results: Line 281: Please clarify if this should be boys/men and girls/women.

16. Results: The table shows the overall difference for male vs female in terms of victimization/perpetration but comparisons for perpetration alone or victimiazation alone, for example, are not shown. Please present these: “Sex-differences in perpetration or victimization alone were negligible.”

17. Results: Line 286-290: Please present the results on sex differences in perpetration/victimization (same sex/opposite sex) in a table (such as the one presented in the response to reviewer comments).

18. Results: Line 299-311: Please clarify if the p values reported represent the overall chi-square test (including all three tertiles) or if the results reported reflect comparisons between T1 and T3 specifically. Please describe this analysis more completely in the methods, including whether correction for multiple testing was applied.

19. Results: Line 334-343: Please present the complete results (this can be in a table) for tests done to examine the main effect of site and interactions.

20. Results: Line 342-343: Please present the results of the analyses with out of school students excluded: “These differences remained after excluding out of school adolescent in Kinshasa.”

21. Results: Line 344-361: Please quantify the results in the text, if possible. Please avoid the use of language implying causality in this section (e.g. “...high parental connectedness relative to low connectedness increased the RRR…”

22. Discussion: Line 440: Please expand on the statement that social desirability bias may have factored in to responses from Kinshasa.

23. Table 1: Please note in the legend the statistical tests use to investigate sex differences.

24. Table 2: Please adjust the formatting of the table, as the column widths lead to the text being difficult to read. Please note in the legend the statistical tests used.

25. Table S5 and S6: Instead of using gray boxes, we suggest using different symbols with footnotes to describe the specific reasons why data were unavailable for the analyses.

26. STROBE checklist: Thank you for including the checklist. Please use section and paragraph numbers to identify locations within the text, rather than page numbers.

Comments from the reviewers:

Reviewer #1: Overall, this revision is very responsive to my previous comments. A few issues regarding clarity remain:

In my earlier comments, I was not clear enough about what I meant by "clarifying the Parameters of interest". In the statistical methods section, I recommend that the authors tell the readers the parameter, or estimand, that they will be using and how to interpret it. For example, the main results in table 3 consist of the "relative risk ratio", which, not being a Stata user myself, is an unfamiliar term to me. This measure ought to be defined in the statistical methods section and interpreted for the readers (beyond stating that the relative risk ratio is a measure that comes out of such and such model). What does it mean? If standard multinomial logistic regression is used, this is the odds ratio for being the dependent variable being in the given category rather than the reference category for a one-unit change in the independent variable.

Any attachments provided with reviews can be seen via the following link:

[LINK]

2 Nov 2021

Submitted filename: Response to Reviewers - R2.docx

Click here for additional data file.

18 Nov 2021

Dear Dr. Ramaiya,

Thank you very much for re-submitting your manuscript "Assessing the Relationship Between Agency and Peer-violence among Very Young Adolescents in Two African Settings: A Cross-Sectional Study" (PMEDICINE-D-21-00483R3) for consideration in PLOS Medicine’s Special Issue: Global Child Health: From Birth to Adolescence and Beyond.

I have discussed the paper with my colleagues and the academic editor. I am pleased to say that provided the remaining editorial and production issues are dealt with we are planning to accept the paper for publication in the journal.

The remaining issues that need to be addressed are listed at the end of this email. Any accompanying reviewer attachments can be seen via the link below. Please take these into account before resubmitting your manuscript:

[LINK]

***Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.***

In revising the manuscript for further consideration here, please ensure you address the specific points made by each reviewer and the editors. In your rebuttal letter you should indicate your response to the reviewers' and editors' comments and the changes you have made in the manuscript. Please submit a clean version of the paper as the main article file. A version with changes marked must also be uploaded as a marked up manuscript file.

Please also check the guidelines for revised papers at http://journals.plos.org/plosmedicine/s/revising-your-manuscript for any that apply to your paper. If you haven't already, we ask that you provide a short, non-technical Author Summary of your research to make findings accessible to a wide audience that includes both scientists and non-scientists. The Author Summary should immediately follow the Abstract in your revised manuscript. This text is subject to editorial change and should be distinct from the scientific abstract.

We expect to receive your revised manuscript within 1 week. Please email us (plosmedicine@plos.org) if you have any questions or concerns.

We ask every co-author listed on the manuscript to fill in a contributing author statement. If any of the co-authors have not filled in the statement, we will remind them to do so when the paper is revised. If all statements are not completed in a timely fashion this could hold up the re-review process. Should there be a problem getting one of your co-authors to fill in a statement we will be in contact. YOU MUST NOT ADD OR REMOVE AUTHORS UNLESS YOU HAVE ALERTED THE EDITOR HANDLING THE MANUSCRIPT TO THE CHANGE AND THEY SPECIFICALLY HAVE AGREED TO IT.

Please ensure that the paper adheres to the PLOS Data Availability Policy (see http://journals.plos.org/plosmedicine/s/data-availability), which requires that all data underlying the study's findings be provided in a repository or as Supporting Information. For data residing with a third party, authors are required to provide instructions with contact information for obtaining the data. PLOS journals do not allow statements supported by "data not shown" or "unpublished results." For such statements, authors must provide supporting data or cite public sources that include it.

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. Additionally, PLOS ONE offers an option to publish peer-reviewed clinical study protocols. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols

Please review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the rebuttal letter that accompanies your revised manuscript.

Please note, when your manuscript is accepted, an uncorrected proof of your manuscript will be published online ahead of the final version, unless you've already opted out via the online submission form. If, for any reason, you do not want an earlier version of your manuscript published online or are unsure if you have already indicated as such, please let the journal staff know immediately at plosmedicine@plos.org.

If you have any questions in the meantime, please contact me or the journal staff on plosmedicine@plos.org.

We look forward to receiving the revised manuscript by Nov 25 2021 11:59PM.

Sincerely,

Caitlin Moyer, Ph.D.

Associate Editor

PLOS Medicine

plosmedicine.org

------------------------------------------------------------

Requests from Editors:

1. Title: Please revise to include the age range and please use sentence-case capitalization: “Assessing the relationship between agency and peer-violence among adolescents aged 10-14 years in Kinshasa, Democratic Republic of Congo and Blantyre, Malawi: A Cross-Sectional Study”

2. Throughout: Please replace “sex” with “gender” where appropriate, including in the Tables.

3. Abstract: Line 44: Please note that covariates adjusted for include individual, family, peer, and community factors.

4. Author summary: Line 101: We suggest revising to: “both genders” in this point.

5. Author summary: Line 107: Please fully spell out the abbreviation “ACEs” at first use in the text.

6. Introduction: Line 132: Please clarify the direction of relationship between social skills and risk of violence in this sentence.

7. Methods: Line 234, 264: Please change “sex” to “gender” in these sentences.

8. Results and Tables: Please report p values to two decimal places if p>0.01, three decimal places if p>0.001, and for smaller values please report p<0.001.

9. Results: Line 315 and 400-401: Please change “sex” to “gender” if accurate.

10. Results: Line 381-382: Please clarify if this should be “Boys who were in a grade lower than their age-expected grade…”

11. Discussion: Line 478: Please change “sex” to “gender” in this sentence.

12. Discussion: Line 503: We suggest revising to: “...raising the possibility that gender socialization may expose boys to more interpersonal violence as they enjoy more freedom of movement…” or similar.

13. References: Please check each citation for completeness and accuracy. For example, please provide the complete information for Reference 7. For guidance, please use to the "Vancouver" style for reference formatting, and see our website for other reference guidelines https://journals.plos.org/plosmedicine/s/submission-guidelines#loc-references

14. Table 1: Please revise “Peer sex composition” and “Any opposite sex friends” to use “gender” if accurate to report this way.

15. Table 1: Only one category of “Time spent with friends” is reported, though there are 5 categories for this question, please note for this category that responses were collapsed (e.g. please clarify this in the table legend, and in the Methods at line 265). Please also clarify this for “parent connectedness” and any other questions for which responses were collapsed/dichotomized.

16. STROBE Checklist: Please revise the paragraph, using section and paragraph numbers to refer to locations in the text. Please do not refer to page numbers (e.g. Methods, paragraph 2). For “Funding” please refer to the “Financial Disclosures” section.

17. Supplement 9: Thank you for including the data underlying the study as a supporting information file. Please note that we do not require that authors submit their entire data set if only a portion of the data was used in the reported study. The included file contains a large number of variables. A file explaining the nature of each variable would be help researchers to use the dataset.

Prior to sharing human research participant data, authors should consult with an ethics committee to ensure data are shared in accordance with participant consent and all applicable local laws.

Data sharing should never compromise participant privacy. It is therefore not appropriate to publicly share personally identifiable data on human research participants. The following are examples of data that should not be shared: Name, initials, physical address; Internet protocol (IP) address; Specific dates (birth dates, death dates, examination dates, etc.); Contact information such as phone number or email address; Location data.

Data that are not directly identifying may also be inappropriate to share, as in combination they can become identifying. For example, data collected from a small group of participants, vulnerable populations, or private groups should not be shared if they involve indirect identifiers (such as sex, ethnicity, location, etc.) that may risk the identification of study participants.

Steps necessary to protect privacy may include de-identifying data, adding noise, or blocking portions of the database.

Please check the file and remove any potentially identifying information “e.g. deviceid, subscriberid, simid, devicephonenumber, etc.

18. Supplement 2: Kinshasa IRB: Please confirm that the appropriate usage rights apply to the use of the maps on page 3 and 4 of the protocol. PLOS applies the Creative Commons Attribution (CC BY) license to all the works we publish, and the use of figures must be compatible with this license. Please see our guidelines for map images: https://journals.plos.org/plosmedicine/s/figures#loc-maps

Please note relevant permissions and license/copyright information in the legends.

Any attachments provided with reviews can be seen via the following link:

[LINK]

19 Nov 2021

Submitted filename: Response to Reviewers - R3.docx

Click here for additional data file.

29 Nov 2021

Dear Dr Ramaiya,

On behalf of my colleagues and the Academic Editor, Kathryn Yount, I am pleased to inform you that we have agreed to publish your manuscript "Assessing the relationship between agency and peer-violence among adolescents aged 10-14 years in Kinshasa, Democratic Republic of Congo and Blantyre, Malawi: A cross-sectional study" (PMEDICINE-D-21-00483R4) in PLOS Medicine’s Special Issue: Global Child Health: From Birth to Adolescence and Beyond.

Before your manuscript can be formally accepted you will need to complete some formatting changes, which you will receive in a follow up email. Please be aware that it may take several days for you to receive this email; during this time no action is required by you. Once you have received these formatting requests, please note that your manuscript will not be scheduled for publication until you have made the required changes.

Please also address the following editorial requests:

1. Abstract: Lines 57-58: Please report exact p values, unless p<0.001, instead of reporting “p<0.01” or “p<0.05” in these sentences.

2. Methods: Line 269: It seems that “graffiti” should not be capitalized in this sentence.

3. Reference 11: Please abbreviate the journal title as: Int J Adolesc Med Health.

4. Reference 33: Please abbreviate the journal title as: Lancet Public Health.

5. Supplement 3: Blantyre IRB: Please include a clean version of the document, without comments/tracked changes.

6. S5 Table and S6 Table: In the legend, please note the meaning behind the # symbol used in the table (e.g. “#Malawi tertile 2”).

In the meantime, please log into Editorial Manager at http://www.editorialmanager.com/pmedicine/, click the "Update My Information" link at the top of the page, and update your user information to ensure an efficient production process.

PRESS

We frequently collaborate with press offices. If your institution or institutions have a press office, please notify them about your upcoming paper at this point, to enable them to help maximise its impact. If the press office is planning to promote your findings, we would be grateful if they could coordinate with medicinepress@plos.org. If you have not yet opted out of the early version process, we ask that you notify us immediately of any press plans so that we may do so on your behalf.

We also ask that you take this opportunity to read our Embargo Policy regarding the discussion, promotion and media coverage of work that is yet to be published by PLOS. As your manuscript is not yet published, it is bound by the conditions of our Embargo Policy. Please be aware that this policy is in place both to ensure that any press coverage of your article is fully substantiated and to provide a direct link between such coverage and the published work. For full details of our Embargo Policy, please visit http://www.plos.org/about/media-inquiries/embargo-policy/.

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. Additionally, PLOS ONE offers an option to publish peer-reviewed clinical study protocols. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols

Thank you again for submitting to PLOS Medicine. We look forward to publishing your paper.

Sincerely,

Caitlin Moyer, Ph.D.

Associate Editor

PLOS Medicine