Types of Childhood Exposure to Violence and Association With Caregiver Trauma in Peru

Abstract

This study examined typologies of childhood polyvictimization and the associations of profiles with demographic characteristics at the levels of child, household, and primary caregiver. This study evaluated a sample of children aged 4 to 17 years residing in San Juan de Lurigancho District, an urban setting on the edge of Lima, Peru (n = 384). An in-person interview of the primary caregiver of each child was conducted in June 2018, assessing the victimization of the child, the caregiver’s exposure to trauma and abuse, and general socioeconomic and demographic characteristics of the household. Latent class analysis was used to identify typologies of child victimization. Follow-up analysis was conducted to quantify differences between the classes which emerged, in terms of the child, caregiver, and household. Five classes emerged: High Peer victimization, Moderate Community victimization; High Community victimization; Verbal Abuse; High victimization across domains; and Low victimization across domains. Caregiver exposure to trauma was positively associated with membership in the high-exposure classes. This study offers a unique opportunity to more deeply understand childhood exposure to violence in Latin America, specifically in an urban setting in Peru, and to further understand how childhood victimization is associated with various characteristics of the child, caregiver, and household. These findings could inform interventions supporting children and families at risk of exposure to violence in Peru or globally.

Keywords

violence typology polyvictimization intergenerational victimization transmission international

Introduction

Childhood victimization occurs at alarming rates worldwide. In a recent systematic review examining violence against children in 96 countries, authors estimated a 50% prevalence of past-year exposure in Asia, Africa, and Northern America, and indicated that a minimum of 1 billion children worldwide are exposed to some form of violence within the past year (Hillis et al., 2016). Types of violence that children might experience include child maltreatment, including exposure to both abuse and neglect, bullying, adult intimate partner violence, and community and sociopolitical violence. There are ample data documenting short- and long-term detrimental effects of such exposure; childhood victimization heightens risk for psychopathology, sleep difficulties, low self-esteem, heightened fear responses, setbacks in educational outcomes, and myriad physical health problems (Breiding et al., 2014; Dillon et al., 2013; Evans et al., 2008; Fry et al., 2016; Graham-Bermann & Levendosky, 2011; Howell, 2011; Howell et al., 2016). Furthermore, early experiences of victimization appear to have particular potency, with more and earlier childhood victimization predicting more negative long-term outcomes for mental and physical health across the lifespan (Cloitre et al., 2009; Miller-Graff et al., 2016; Raposa et al., 2014). Children’s risk of experiencing mental health difficulties also appears to compound as victimization increases, making polyvictimization—the experience of multiple types of victimization experiences—of great concern (Turner et al., 2006).

Polyvictimization characterizes experiences of one in five children surveyed in the United States and is significantly more likely for children who live in high-risk communities or families (Finkelhor et al., 2007; Finkelhor et al., 2009). Despite mounting evidence of childhood victimization’s prevalence and adverse effects on children’s health and development, data on children’s diverse experiences of victimization are limited in many global contexts, especially in those countries that do not have systems to monitor such experiences regularly (Dubowitz, 2012).

Polyvictimization frameworks have rarely been used in psychological research conducted in Peru. Like much past research in other contexts, studies on children’s victimization experiences largely focus on a single type of violence exposure. There are several studies, for example, on bullying (Crookston et al., 2014; Lister et al., 2015) or child maltreatment in Peru (Morales & Singh, 2015), and a few on overlaps between two types of victimization experiences (e.g., maltreatment and intimate partner violence; Benavides et al., 2015). The dearth of literature on the overlap and co-occurrence of multiple types of child victimization experiences, however, represents a significant gap in the research literature in this context. In Peruvian schools, population-based surveys on lifetime trauma exposure estimate 57% to 68% of children and adolescents endorse psychological violence, 58% to 65% report physical violence, and around 35% indicated sexual violence (Ames et al., 2018; Fry et al., 2016). These studies provide valuable estimates of national prevalence, but do not detail the ways in which such types of violence overlap with other co-occurring types of victimization in this context.

Studies addressing polyvictimization may be especially important in urban contexts, where risk of exposure to violence can be even greater. For example, San Juan de Lurigancho District, considered one of Lima’s most dangerous neighborhoods, faces notably high rates of crime and victimization and historically high gang activity levels (Ciudad Nuestra, 2012). Youth in this area of Peru are vulnerable to violence exposure in their homes, communities, and schools. Nationally, 75% of Peruvian youth have been victims of physical or psychological violence perpetrated by classmates, and more than 80% of the adolescent population was victim of any psychological or physical violence (Instituto Nacional de Estdistica e Informatica [INEI], 2016). Students aged 11 to 13 years in San Juan de Lurigancho have been observed to demonstrate repetitive physical, verbal, and disruptive actions (Holguin Alvarez, 2017). In addition, youth face violence in the home, as a high proportion of parents in Peru believe it is necessary to shout at (42%), or physically punish (36%) children or adolescents to discipline them (INEI, 2016). Finally, Peruvian women and girls face additional risks of violence in the home. In Peru, 32% of women reported they been a victim of physical or sexual violence, and 64% experienced psychological or verbal violence from their partner (INEI, 2016). Several studies report additional anecdotal evidence that San Juan de Lurigancho, a densely populated district on the northernmost edge of Lima, consistently demonstrates higher incidence of violence than national averages (Balarin, 2015; Benavides et al., 2017). With significant international and multilateral funding attempting to address the issue of violence in Latin America, for both geo-political and migration reasons, understanding patterns and typologies of violence among youth and adolescents in Latin America remains a topic of special importance which merits more rigorous research to target programming for those who need it most.

There are two primary approaches to the study of polyvictimization, both providing valuable, albeit different, types of information. Traditionally, victimization experiences have been studied using variable-centered analytic approaches. Such approaches generally create sum scores of various characteristics of victimization experiences (e.g., frequency, severity, type) and examine the relationship between such experiences and outcomes. This research has been useful in examining critical questions related to sensitizing effects of victimization (e.g., Elliott et al., 2009) and demonstrating evidence for both linear and nonlinear effects of traumatic experiences (e.g., Kira et al., 2012). Its empirical focus, however, is designed to focus on trends within the full sample under study. Yet, emerging evidence in polyvictimization research underscores that sum scores of victimization experiences used in variable-centered approaches may also mask important individual differences (Miller-Graff et al., 2015).

Latent class analysis (LCA) is a person-centered form of statistical analysis that presents an alternative strategy to understand victimization. LCA is an exploratory analytic method that allows for exploration of associations between observed variables to identify underlying “typologies” or “profiles” (Vermunt & Magidson, 2004). Model fit and profile characteristics are examined to determine if empirically and qualitatively distinctive subgroups emerge from the sample population. These profiles can provide valuable information about meaningful subgroups within larger data sets. Research using LCA has identified that although victimization experiences are highly correlated with one another, indicating that risk for one victimization is associated with risk for another (variable-centered), this risk is not equally distributed across persons. One recent study, for example, found that a college student sample in the United States exhibited four distinct victimization profiles—high victimization, low victimization, domestic victimization, and community victimization (Miller-Graff et al., 2015).

Person-centered analyses are particularly important in research on child victimization because variable-centered approaches examining core questions about the risk that victimization poses to children is clear: higher rates of victimization are linked with a host of negative developmental outcomes. What is not clear, however, is how victimization varies in its characterization across contexts, and implications of particular co-occurring experiences in the lives of children. Here, person-centered analytic methods can contribute substantially, not only to global literature on childhood polyvictimization, but may also provide information that more richly characterizes victimization in-context. This is an important step in creating robust and effective preventions and interventions that closely match local needs and capacities.

Although person-centered analysis of polyvictimization profiles provides important descriptive information about how different types of violence co-occur, it is also important to identify potential factors predicting such profiles. Sociodemographic risk factors, such as poverty, resource access and household overcrowding, are likely to contribute to higher risk for child polyvictimization (e.g., Ellonen & Salmi, 2011; Hu et al., 2018). Furthermore, children who have caregivers with a history of victimization may be at increased risk of victimization themselves (Berlin et al., 2011). This increased risk might be present through a variety of pathways, including ongoing family or community violence affecting both parents and children (Finkelhor et al., 2009).

The Current Study

While an abundance of research focuses on childhood victimization around the world, person-centered methods of analysis in polyvictimization research are still relatively rare, and few have been conducted in the context of Peru (for an exception, see Nguyen, 2016). To investigate these patterns, this study conducts LCA of a victimization scale among youth and adolescents in Peru. The LCA is followed by descriptive statistics on each latent class emerging from the analysis. Finally, multivariate regression analysis aims to understand differences across these emergent groups. Based on previous research, we hypothesize the following:

Hypothesis 1: There will emerge distinct latent classes, at least one of which is a “low-exposure” and another a “high-exposure” group.

Hypothesis 2: Household-level demographic risk factors (such as overcrowding in the home or low income) will positively correlate with likelihood of membership into the high-exposure group.

Hypothesis 3: Caregivers’ trauma history will positively correlate with likelihood of membership in the high-exposure group.

Method

Participants

Participants in this study included N = 384 dyads of caregivers and children in the most populous district of Lima, San Juan de Lurigancho. The majority of caregivers were female (91.18%), ranging in age between 17 and 82 (M = 42.04, SD = 13.49) years. Most caregivers were either the head of household (28.21%) or the spouse of the head of household (52.39%). Children in the sample were between 4 and 17 years old (M = 11.26, SD = 3.98) and were relatively balanced on sex (47.00% girls, 53.00% boys). Almost all were children (69.85%) or grandchildren (26.38%) of the head of household.

Procedure

Data for this study were drawn from a parish survey of San Juan de Lurigancho District conducted by a survey team from Universidad Cátolica de Sedes Sapientiae, in collaboration with researchers from the University of Notre Dame, with the aim of identifying needs of the parish to inform social service provision. A faith-based nonprofit organization, Instituto de Pastoral de la Familia, based at the parish, provides psychosocial supports and cultural programming to families living in San Juan de Lurigancho. INFAM commenced the survey detailed here for their strategic planning process, with goals to better understand experiences of their community’s families to inform programming decisions. IRB approval for the secondary use of the data for research purposes was obtained by the University of Notre Dame.

Two-staged probabilistic sampling was implemented from the survey team at Universidad Católica Sedes Sapientae in Lima, Peru. The primary sampling units were the enumeration areas used in Peru’s 2013 Demographic and Health Survey—these were stratified by socioeconomic status, and selected with probabilistic proportion to the number of dwellings in each area. In each of the 110 areas, six households were selected for inclusion in the interview, using a predetermined skip pattern which was proportional to the population in each area.

Within households, enumerators collected general information on the entire household, and then used a preprogrammed random number generator to select one caregiver–child dyad within each household. For the purposes of the survey, the caregiver was defined as a person living in the household who was primarily responsible for taking care of the child during the week. The enumerator asked for the name of the caregiver of each child in the household, referring to children by name. Then, the enumerator conducted the remainder of the interview with the caregiver of the randomly selected child, referring to both the selected caregiver and child by name. If the selected caregiver was not available, the enumerator returned at a convenient time to speak to that person. The caregiver answered questions about his or her own trauma history and depression, as well as indices about the child. In cases of households with no children, an adult in the household was interviewed using an abbreviated version of the survey.

Measures

Demographics

One respondent in the household answered general demographic questions about each member of the household, including sex, age, income, years of education, and relation to head of household. They also answered general information on the household itself, including durable goods present in the home such as television and internet connection, which could serve as proxies for wealth.

Child victimization

The Juvenile Victimization Questionnaire (JVQ; Youth Lifetime, Reduced Item Version) is a 12-item caregiver-report measure assessing several types of child victimization experiences, including information about frequency and severity of events (Finkelhor et al., 2005). In this measure, victimization types include physical assault, sexual victimization, peer/sibling victimization, property crime, and witnessed/indirect victimization (Finkelhor et al., 2005). Caregivers endorsed victimization events or not, regarding children’s exposure (i.e., “yes” or “no”). This assessment has not previously been reviewed in Peru, but it was reviewed by the survey team, and items were deemed to be contextually appropriate. For this study, the JVQ was forward translated, and the bilingual research team discussed items to establish semantic correspondence across contexts. Because this measure assesses incident, separable events, internal reliability was not calculated.

Child adjustment

Child adjustment was measured by caregiver-report on the Strengths and Difficulties Questionnaire (SDQ; Goodman, 1997). The 25-item SDQ measures both positive and negative adjustment in children. Positive adjustment includes prosocial skills and behaviors, whereas negative adjustment incorporates emotional problems, conduct problems, inattention/hyperactivity, and peer relationship problems. For each item, caregivers responded on 3-point scale to indicate if each item is “Not true,” “Somewhat true,” or “Certainly true” of the child. A sum of responses from the negative adjustment subscales results in a total difficulties scale, without prosocial skills and behaviors. This measure has been translated into Spanish previously and used with Peruvian children (Millones et al., 2013; Millones et al., 2014). The SDQ has been found to be reliable and valid in prior research (Goodman, 2001; Millones et al., 2013; Millones et al., 2014), and reliability in this study is α = .78.

Caregivers’ trauma history

The Trauma History Questionnaire (THQ) is a 16-item measure which assesses lifetime history of exposure to trauma (Green, 1996). The questionnaire addresses various types of trauma exposure, including physical or sexual violence, disaster, or crime-related events. For each endorsed item, respondents report frequency of events and age at first occurrence. This module is scored by summing all events endorsed, with a total ranging from 0 to 16. Research supports the THQ as a reliable and valid measure in clinical and non-clinical study samples (Hooper et al., 2011). This measure has been translated and used in Spanish-speaking contexts, and its use in several countries have supported its validity across cultures (Fiszman et al., 2005; Hooper et al., 2011). Because this measure assesses incident, separable events, internal reliability was not calculated.

Data Analytic Plan

To address Hypothesis 1, LCA was conducted in STATA 16.1, using an LCA Stata plug-in (Lanza et al., 2018). For this analysis, all items assessing childhood exposure to violence were binary indicators, where 1 = yes and 2 = no. Bayesian information criterion (BIC), adjusted BIC, and Akaike information criterion (AIC) indices were utilized to compare model fit, with lower numbers indicating better fit. Entropy, a measure of class differentiation, was also considered, with higher values indicating better class differentiation. To address Hypotheses 2 and 3, we performed multivariate regression analysis, using children’s probability of belonging to the various groups as the dependent variables, and including a full-information maximum likelihood estimation method which accounted for missing responses. We examined characteristics of households, caregivers, and children which may emerge as potential risk factors. At the level of the household, we examined the sex of the household head, the daily income of the household, whether the household was nuclear or multi-generational, the size of the household (i.e., number of members), and durable goods present in the home such as television or internet connection. At the caregiver level, we included traumatic incidents endorsed by the caregiver, age, sex, and years of education of the caregiver. Finally, we examined the sex and age of the child, as well as the child’s score on various components of the strengths and difficulties questionnaire, as potential risk factors of victimization.

Missing Data

The total sample, where caregiver–child dyads were present, was 402 households. From this sample, five households were dropped because of errors in data entry. Because all items in the JVQ were necessary for inclusion in the LCA, those with missing item level data on the JVQ were dropped from the analysis leaving a total of 384 caregiver–child dyads included in this study. Missing data were also present in the caregiver trauma assessment; 47 of the 384 cases had missing responses to one or more items in the questionnaire. These families were retained in the analysis. Patterns in missingness between those participants with and without missing data on caregiver trauma history were evaluated; no differences were evident between those with and without missing data on the probability of class assignment (i.e., the dependent variable). Missing caregiver victimization data were also not related to logged household income, durable goods, multi-generational households, gender or years of education level of the caregiver, gender or age of the child, child adjustment or pro-social behavior score, but was significantly related to the indicator on female-headed households (p = .02). Based on these analyses, data were determined to be missing at random (MAR), which is an ignorable type of missing data that is best handled using full-information maximum likelihood estimation (Enders, 2010).

Results

First, we present descriptive statistics of households, caregivers, and children included in this sample. On average, 30% of households were headed by a female. The daily wage of the household was 115 soles (approximately US$34.69). Half of households were multigenerational, meaning more than two generations were living in the same household. On average, the caregiver endorsed almost three traumatic events. Caregivers were on average 42 years old, and overwhelmingly female. They had an average of 10 years of education. Youth means in the sample scored in the “normal” range for both Total Difficulties on the SDQ, and the Pro-Social Scale of the SDQ. Children were fairly balanced on sex, with 47% female and 53% male. The average age of children was 11 years (Table 1).

Table 1.

Summary Statistics of Sample.

Variable	M	95% Confidence Interval
Household income, daily (2018 Peruvian Soles)	114.81	[103.98, 125.64]
Durable goods: television connection (%)	71.35	[66.81, 75.9]
Durable goods: internet connection (%)	49.22	[44.2, 54.24]
Female-headed household (%)	29.95	[25.35, 34.55]
Household size	5.24	[5.04, 5.44]
Multigenerational household (%)	49.74	[44.72, 54.76]
Traumatic incidents endorsed by caregiver	2.67	[2.37, 2.96]
Female caregiver (%)	91.15	[88.29, 94]
Years of education of caregiver	10.37	[10.02, 10.72]
Child’s score on Strengths and Difficulties Questionnaire (range = 0–40)	9.89	[9.28, 10.49]
Child’s pro-social behavior (range = 0–10)	7.94	[7.73, 8.15]
Female child (%)	47.40	[42.38, 52.41]
Age of child	11.28	[10.88, 11.68]

Table 2 shows results from the LCA. AIC demonstrated improvement as each additional class was included, up to the five-class model. BIC and adjusted BIC worsened as the number of classes increased beyond four. However, entropy did not reach acceptable levels until the five-class model. The five-class model also divided observations into groups of acceptable size (≥5% of the sample), indicating a likelihood of their stability and differentiation. The five-class model was therefore selected as the model with optimal fit (AIC = 447.83, BIC = 700.67, adjusted BIC = 497.60, Entropy = 0.81).

Table 2.

Summary of Model Results Used for Model Selection.

Model	AIC	BIC	Adjusted BIC	Entropy	LL	df
Two-class	493.7225	592.4886	513.1671	0.831068	−1,217.43	4,070
Three-class	468.6563	618.7808	498.2121	0.76211	−1,191.89	4,057
Four-class	450.6193	652.1021	490.2863	0.771788	−1,169.87	4,044
Five-class	447.8268	700.6679	497.6049	0.811295	−1,155.48	4,031
Six-class	455.4794	759.6789	515.3687	0.836926	−1,146.31	4,018

Note. AIC = Akaike information criterion; BIC = Bayesian information criterion; LL = log-likelihood; df = degrees of freedom.

A descriptive analysis of the five-class model indicates that the most common profile class is the Low-Exposed class (Class 5; n = 237, 62% of the sample). This class had low levels of exposure to violence across all domains. The next most common class experienced high rates of exposure primarily to verbal violence (Class 3; n = 77, 20% of the sample). Another class included 8% of the sample, and commonly endorsed several types of violence enacted by peers, as well as moderate community exposure. There emerged two High-Exposure Classes (Class 2, n = 20, 5% of the sample, and Class 4, n = 21, 5% of the sample). A comparison between these two high-exposure classes indicated that Class 2 exhibited high exposure to violence in communities, while Class 4 had the highest rates of endorsement of physical, verbal, and sexual violence. Youth in Class 4 were exposed to violence both by witnessing it and experiencing it themselves. Based on descriptive statistics of exposure within class, classes were therefore named as follows: Class 1 = High peer, Moderate community (HPMC); Class 2 = High Community (HC); Class 3 = Verbal Abuse (VA); Class 4 = High across domains (HAD); Class 5 = Low across domains (LAD). Further information is included in Table 3.

Table 3.

Percent of Sample Endorsing Each Item in JVQ.

Item of JVQ	Class 1: High Peer, Moderate Community Exposure	Class 2: High Community Exposure	Class 3: Verbal Abuse Exposure	Class 4: High Exposure	Class 5: Low Exposure
n	30	20	77	21	237
Experienced theft	31.03%	100.00%	7.89%	60.00%	5.63%
Attacked with an object or a weapon by anyone	16.67%	65.00%	1.30%	25.00%	1.72%
Attacked without a weapon or object by anyone	20.00%	45.00%	0.00%	66.67%	0.00%
Verbal abuse by adults	0.00%	60.00%	64.00%	95.24%	0.00%
Attacked by a group of children or gang	20.00%	40.00%	5.19%	28.57%	0.00%
Hit by another child	96.67%	40.00%	16.88%	71.43%	0.00%
Verbal abuse by children	13.33%	55.00%	70.67%	90.48%	0.00%
Sexual abuse by a known adult	0.00%	10.00%	1.32%	19.05%	0.00%
Sexual abuse by an unknown adult	0.00%	0.00%	1.32%	14.29%	0.00%
Witnessed a parent physically hurt by his/her partner	13.79%	15.00%	5.19%	71.43%	2.59%
Witnessed anyone physically attacked in public	17.86%	0.00%	1.30%	52.38%	0.86%
Witnessed people being shot, bombs or street riots	17.24%	0.00%	7.79%	71.43%	8.19%

Note. JVQ = Juvenile Victimization Questionnaire.

Next, we present results of our multivariate regression analysis (Table 4), which aims to address Hypotheses 2 and 3. Multivariate regressions predicting the probability of assignment to a specific class (i.e., a continuous score between 0 and 1) indicated that the models explained a significant amount of variance in probability of assignment. The percent of variance explained varied substantially across models. In this case, Class 3 (VA) was the poorest fitting (R² = .06) and Class 5 (LAD) was the best explained (R² = .24). First, we found that children raised in multigenerational households were less likely to be members of Class 3 (VA), meaning they were less likely to be exposed exclusively to verbal abuse than children not living in multigenerational households (β = −.09, p = .039). In multigenerational households, children’s primary caregivers were significantly more likely to be grandparents (p<.001, not shown). We found the strongest relationship between group membership and caregiver attributes. Specifically, the number of traumatic events endorsed by caregivers was negatively associated with membership in the low-exposure group (β = −.04, p < .001). Caregiver exposure to trauma was positively associated with probability of membership in Classes 1, 2, and 4 (i.e., groups which face higher levels of exposure to various types of violence; β = .02, p = .0002; β = .01, p < .001; β = .02, p < .001, respectively). In addition, years of education of the caregiver was negatively associated with membership into the group experiencing verbal abuse (β = −.01, p = .0167). Sex of the caregiver was negatively associated with probability of membership in Class 1 (HPMC; β = −.10, p = .0164), meaning children with female caregivers were more likely to be members of this group. Finally, child age was positively associated with probability of membership in the high-exposure class, implying that older children were more likely to be have experienced high levels of victimization in their lifetimes (β = .01, p ≤ .025). Child adjustment was associated with class designation in expected ways. Specifically, youth who were likely to be members in high exposure classes also had higher levels of adjustment problems (β = .01, p = .001, β = .01, p = .047; β = .01, p = .034), and those with lower scores are more likely to be members of the low exposure class (β = −.03, p < .001). However, children’s scores on the prosocial behavior scale were not associated with membership in any class.

Table 4.

Multiple Regression Results using Probabilities.

Variable	Class 1: High Peer, Moderate Community Exposure	Class 2: High Community Exposure	Class 3: Verbal Abuse Exposure	Class 4: High Exposure	Class 5: Low Exposure
Household attributes
Household income (log)	.001	.000	.002	−.002	−.002
	(.003)	(.003)	(.005)	(.003)	(.006)
Durable goods: TV connection	−.012	−.012	.052	−.028	.000
	(.028)	(.022)	(.041)	(.025)	(.048)
Durable goods: internet connection	.031	.006	−.038	.017	−.015
	(.026)	(.020)	(.038)	(.024)	(.045)
Female headed household	−.013	.004	.008	−.008	.009
	(.026)	(.021)	(.039)	(.024)	(.045)
Household size	.011	.003	.000	−.008	−.007
	(.007)	(.006)	(.011)	(.007)	(.013)
Multigenerational household	−.018	−.009	−.087*	.038	.076
	(.029)	(.023)	(.042)	(.026)	(.050)
Caregiver attributes
Traumatic incidents endorsed by caregiver	.015*	.012*	−.011	.022*	−.038*
	(.004)	(.003)	(.006)	(.004)	(.007)
Sex of caregiver (reference category: male)	−.098*	.045	.040	.049	−.037
	(.041)	(.032)	(.060)	(.037)	(.070)
Years of education of caregiver	.006	.000	−.012*	.004	.002
	(.003)	(.003)	(.005)	(.003)	(.006)
Child attributes
Strengths and difficulties score	.003	.006*	.006*	.010*	−.025*
	(.002)	(.002)	(.003)	(.002)	(.004)
Prosocial behavior	−.007	.008	.016	.000	−.016
	(.006)	(.004)	(.008)	(.005)	(.010)
Sex of child (reference category: male)	−.023	−.015	−.013	.019	.031
	(.023)	(.018)	(.033)	(.021)	(.039)
Age of child	−.004	.001	−.001	.006*	−.003
	(.003)	(.002)	(.004)	(.003)	(.005)
Constant	.09	−.164*	.184	−.225*	1.118*
	(.089)	(.070)	(.131)	(.082)	(.154)
R ²	.092	.105	.057	.209	.238
N	384	384	384	384	384

Note. Standard errors in parentheses.

p < .05.

Discussion

This study examined patterns of polyvictimization in children living in a high-risk neighborhood in Lima, Peru. Consistent with Hypothesis 1, distinct latent classes emerged, one of which was high victimization and another which was low across most domains. Results suggested the presence of five latent classes: Class 1 = High peer, Moderate community (HPMC); Class 2 = High Community (HC); Class 3 = Verbal Abuse (VA); Class 4 = High across domains (HAD); and Class 5 = Low across domains (LAD). These results parallel recent person-centered analyses of polyvictimization in other contexts on the emergence of high-exposed and low-exposed groups, with membership in the latter being more common (Houston et al., 2011; Miller-Graff et al., 2015). Similar to Miller-Graff and colleagues’ (2015) analysis, the current study identified a group of individuals exposed to high levels of community violence, but not violence in interpersonal relationships. These data importantly point to the fact that many families, even those living in communities experiencing high rates of violence, are able to preserve and maintain safe family environments characterized by low levels of interpersonal violence.

Consistent with previous research on the adverse effects of childhood victimization (Cloitre et al., 2009; Miller-Graff et al., 2016; Raposa et al., 2014), membership in high exposure polyvictimization profiles was associated with higher levels of adjustment problems. In line with Hypothesis 3, we found that caregiver trauma was associated with an increased likelihood of children’s membership in all but the LAD and VA profiles (Classes 3 and 5). For the most part, this is consistent with intergenerational theories of trauma, whereby risk for victimization is passed from one generation to the next (Berlin et al., 2011; Pears & Capaldi, 2001; Widom & Wilson, 2015). It is interesting to note that this intergenerational transmission of risk did not appear to extend to children in the class experiencing primarily verbal abuse, suggesting that pathways of intergenerational transmission may vary in strength by type of victimization. Future research should explore this more deeply, and consider the intersection of victimization types between caregivers and children. Unique to this sample was a subgroup with high exposure to verbal abuse, but relatively low exposure to other forms of victimization (Class 3). In multivariate regression models examining predictors of class membership probability, children’s membership in this class was negatively associated with living in multigenerational households, such that children in these households were significantly less likely to belong to this group. The protective effect conferred by multigenerational households and grandparent caregivers is different from that of other contexts, such as the United States (e.g., Thomas et al., 2000). In these contexts, negative outcomes of grandparents-as-parents have been noted, but possibly due to the fact that rather than reflecting a culturally normative structure, the presence of grandparents-as-parents in the United States often related to significant family difficulties (e.g., kinship foster care; Thomas et al., 2000). However, studies on families in non-Western contexts have demonstrated the central role that grandparents (typically grandmothers) play in child nutrition and health status (Aubel, 2012; Bender & McCann, 2000). In Peru, one study found that maternal grandmother involvement buffered child risks of negative emotional reactivity and served as a protective factor against harsh parenting practices (Barnett et al., 2010). This body of research demonstrates the need to think beyond the caregiver–child dyad in contexts where the extended family has cultural and normative significance.

We did not find evidence to support Hypothesis 2. Other than family multigenerational status, demographic risk factors were not associated with children’s victimization profiles. It may be that these variables were not strongly predictive due to the lack of meaningful variation in the sample. Further research with samples that provide more variation on household demographics could provide clarity on this point. Alternatively, community-level factors may have superseded effects of household or demographic information, but more research is needed to investigate these possibilities.

Limitations and Conclusion

This study is subject to several limitations. First, as a cross-sectional study, it is difficult to determine timing or causality in this study’s correlations. However, this risk is mitigated because some events most likely occurred prior to included outcomes. For example, traumatic events or years of education of caregivers likely occurred before victimization events of the child. Second, although this study provides valuable findings from a large urban area in Lima, it is not representative of the entire country of Peru or even of all of Lima, which is a very diverse city in terms of level of income. This limitation also restricts our ability to address variation in household-level covariates, as most respondents to this survey were fairly homogeneous in income levels and other economic indicators. However, it also provides important context for this understudied, urban area in Peru.

In this study, LCA provided important distinctions of the types of victimization that youth experienced in urban areas of Peru. One group of youth were primarily exposed to verbal violence, while there emerged several groups exposed to varying combinations of multiple violent events. Finally, one group experienced low levels of all types of violence. Interventions could be designed which help youth deal with or avoid violence in communities or at the hands of their peers; these interventions could target the same youth. Meanwhile, other interventions could deal with the psychosocial and mental health repercussions of verbal abuse from family or friends.

The multivariate regression analysis points to several opportunities for interventions to reduce youth’s likelihood of victimization. These findings indicate characteristics of families which could benefit from additional support to reduce the risk of children’s polyvictimization. First, this study demonstrates the potential intergenerational nature of trauma or victimization supported in other literature on Peruvian families (Scheid, Miller-Graff, & Guzmán, in press). Specifically, children of caregivers who have experienced traumatic events may be more likely to experience multiple incidents of victimization. Interventions targeting these caregivers or their children to provide additional support could help to mitigate risk. Furthermore, associations between education level and risk of exposure to verbal violence demonstrate that caregivers with lower levels of education could benefit from additional support on positive parenting. Interventions could provide support to parents in single-generation households who perhaps lack the safety net of a larger family in the household caring for the children. Finally, the correlation between child adjustment and various levels of victimization demonstrates that adjustment problems may be useful in assessing youth when targeting a population for inclusion in programming. Overall, this study contributes valuable information on children’s polyvictimization that may help foster interventions tailored to specific households and types of exposure for Peruvian families.

Footnotes

Acknowledgments

The authors acknowledge and thank Holy Cross Family Ministries, and the Instituto de Pastoral de la Familia for facilitating access to the community and for connecting them to local enumerators. This research could not be completed without the Ford Program in Human Development and Solidarity Studies at the University of Notre Dame. The authors also acknowledge and thank Notre Dame undergraduate students Natalie Disher and Ellie Buerk for their support in the data collection and on-the-ground fieldwork in Peru. Local researchers Guido Maggi Poisetti and Rubí Rommy Espichán Parker provided crucial support during data collection. The authors also thank them for their review of the manuscripts’ preliminary analyses. Finally, we extend our gratitude to the families and respondents to this survey, for the gift of both their responses and their time.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Funding for this study was provided by Holy Cross Family Ministries, and the Ford Program in Human Development and Solidarity Studies at the University of Notre Dame.

ORCID iDs

Danice B. Guzmán

Laura E. Miller-Graff

Author Biographies

Danice B. Guzmán is a research associate with the University of Notre Dame, holding a joint position with the Pulte Institute for Global Development and the Ford Family Program in Human Development Studies and Solidarity, both within the Keough School for Global Affairs. She is a quantitative researcher who conducts cross-disciplinary research on issues of international development. She holds a Masters in Public Policy and a Certificate in International Development from Duke University.

Laura E. Miller-Graff is a clinical psychologist and assistant professor of psychology at the University of Notre Dame. Dr. Miller-Graff’s research examines the developmental effects of exposure to violence in childhood. With a focus on children who have multiple traumatic exposures, she investigates resulting patterns of resilience and psychopathology, including the development of posttraumatic stress symptoms. She holds a PhD from the University of Michigan and leads the BRAVE (Building Resilience After Violence Exposure) Research Lab at the University of Notre Dame.

Caroline R. Scheid is a doctoral student in clinical psychology at the University of Notre Dame. Caroline’s research interests concentrate on improving outcomes for children and families exposed to violence and trauma, with particular interests in intergenerational transmissions of trauma, multimethod measurement, and resilience. She focuses on basic and intervention research to mitigate mental health symptoms and other trauma cascades over time.

References

Ames

Anderson

Martin

Rodriguez

Potts

(2018). The multi-country study on the drivers of violence affecting children in Peru: The process and its outcomes. Vulnerable Children and Youth Studies, 13(Suppl. 1), 52–64.

Aubel

(2012). The role and influence of grandmothers on child nutrition: Culturally designated advisors and caregivers. Maternal & Child Nutrition, 8(1), 19–35. https://doi.org/10.1111/j.1740-8709.2011.00333.x

Balarin

(2015). The default privatization of Peruvian education and the rise of low-fee private schools: Better or worse opportunities for the poor? Privatization in Education Research Initiative, 65, 32.

Barnett

M. A.

Scaramella

L. V.

Neppl

T. K.

Ontai

L. L.

Conger

R. D.

(2010). Grandmother involvement as a protective factor for early childhood social adjustment. Journal of Family Psychology, 24(5), 635–645. https://doi.org/10.1037/a0020829

Benavides

Almonte

J. L. J.

de Leon Marquina

M. P.

(2015). The Co-occurrence of domestic and child violence in urban Peru: Evidence from three regions. Journal of Family Violence, 30(8), 1045–1053.

Benavides

Bellatín

Cavagnoud

(2017). Social protection systems and domestic violence in poor urban contexts: The case of San Juan de Lurigancho (Nopoor Working Paper 56). https://www.grade.org.pe/en/publicaciones/social-protection-systems-and-domestic-violence-in-poor-urban-contexts-the-case-of-san-juan-de-lurigancho/

Bender

D. E.

McCann

M. F.

(2000). The influence of maternal intergenerational education on health behaviors of women in peri-urban Bolivia. Social Science & Medicine, 50(9), 1189–1196. https://doi.org/10.1016/S0277-9536(99)00369-X

Berlin

L. J.

Appleyard

Dodge

K. A.

(2011). Intergenerational continuity in child maltreatment: Mediating mechanisms and implications for prevention. Child Development, 82(1), 162–176.

Breiding

Smith

Basile

Walters

Chen

Merrick

M. T.

(2014). Prevalence and characteristics of sexual violence, stalking, and intimate partner violence victimization—National intimate partner and sexual violence survey, United States, 2011. MMWR Surveillance Summaries, 63(8), 1–21. http://search.proquest.com.proxy.library.nd.edu/publication/publications_41804?accountid=12874

10.

Ciudad Nuestra. (2012). Presenta su segunda Encuesta Metropolitana de Victimización [Presents their second Metropolitan Survey of Victimization]. http://www.limacomovamos.org/seguridad-ciudadana/ciudad-nuestra-presenta-su-segunda-encuesta-metropolitana-de-victimizacion-2012/

11.

Cloitre

Stolbach

B. C.

Herman

J. L.

Kolk

B. V. D.

Pynoos

Wang

Petkova

(2009). A developmental approach to complex PTSD: Childhood and adult cumulative trauma as predictors of symptom complexity. Journal of Traumatic Stress, 22(5), 399–408.

12.

Crookston

B. T.

Merrill

R. M.

Hedges

Lister

West

J. H.

Hall

P. C.

(2014). Victimization of Peruvian adolescents and health risk behaviors: Young lives cohort. BMC Public Health, 14(1), Article 85.

13.

Dillon

Hussain

Loxton

Rahman

(2013). Mental and physical health and intimate partner violence against women: A review of the literature. International Journal of Family Medicine, 2013, Article 313909.

14.

Dubowitz

(Ed.). (2012). World perspectives on child abuse. International Society for Prevention of Child Abuse and Neglect.

15.

Elliott

A. N.

Alexander

A. A.

Pierce

T. W.

Aspelmeier

J. E.

Richmond

J. M.

(2009). Childhood victimization, poly-victimization, and adjustment to college in women. Child Maltreatment, 14(4), 330–343.

16.

Ellonen

Salmi

(2011). Poly-victimization as a life condition: Correlates of poly-victimization among Finnish children. Journal of Scandinavian Studies in Criminology and Crime Prevention, 12(1), 20–44.

17.

Enders

C. K.

(2010). Applied missing data analysis. Guilford Press.

18.

Evans

S. E.

Davies

DiLillo

(2008). Exposure to domestic violence: A meta-analysis of child and adolescent outcomes. Aggression and Violent Behavior, 13(2), 131–140. https://doi.org/10.1016/j.avb.2008.02.005

19.

Finkelhor

Hamby

S. L.

Ormrod

Turner

(2005). The Juvenile Victimization Questionnaire: Reliability, validity, and national norms. Child Abuse & Neglect, 29(4), 383–412. https://doi.org/10.1016/j.chiabu.2004.11.001

20.

Finkelhor

Ormrod

R. K.

Turner

H. A.

(2007). Poly-victimization: A neglected component in child victimization. Child Abuse & Neglect, 31(1), 7–26.

21.

Finkelhor

Ormrod

R. K.

Turner

H. A.

Holt

(2009). Pathways to poly-victimization. Child Maltreatment, 14(4), 316–329.

22.

Fiszman

Cabizuca

Lanfredi

Figueira

(2005). The cross-cultural adaptation to Portuguese of the Trauma History Questionnaire to identify traumatic experiences. Brazilian Journal of Psychiatry, 27(1), 63–66.

23.

Fry

Anderson

Hildalgo

R. J. T.

Elizalde

Casey

Rodriguez

… Fang

(2016). Prevalence of violence in childhood and adolescence and the impact on educational outcomes: Evidence from the 2013 Peruvian National Survey on Social Relations. International Health, 8(1), 44–52. https://doi.org/10.1093/inthealth/ihv075

24.

Goodman

(1997). The Strengths and Difficulties Questionnaire: A research note. Journal of Child Psychology and Psychiatry, 38(5), 581–586.

25.

Goodman

(2001). Psychometric properties of the strengths and difficulties questionnaire. Journal of the American Academy of Child & Adolescent Psychiatry, 40(11), 1337–1345.

26.

Green

B. L.

(1996). Trauma History Questionnaire. In Stamm

B. H.

(Ed.), Measurement of stress, trauma, and adaptation (pp. 366–369). Sidran Press.

27.

Graham-Bermann

S. A.

Levendosky

A. A.

(Eds.). (2011). How intimate partner violence affects children. American Psychological Association.

28.

Hillis

Mercy

Amobi

Kress

(2016). Global prevalence of past-year violence against children: A systematic review and minimum estimates. Pediatrics, 137(3), Article e20154079.

29.

Holguin Alvarez

J. A

. (2017). Efectos de conductas proactivas y prosociales en incidentes críticos de escolares limeños [Effects of proactive and prosocial behavoirs in critical incidents of schoolchildren in Lima]. Propósitos y Representaciones, 5(2). https://doi.org/10.20511/pyr2017.v5n2.172

30.

Hooper

L. M.

Stockton

Krupnick

J. L.

Green

B. L.

(2011). Development, use, and psychometric properties of the Trauma History Questionnaire. Journal of Loss & Trauma, 16(3), 258–283.

31.

Houston

J. E.

Shevlin

Adamson

Murphy

(2011). A person-centred approach to modelling population experiences of trauma and mental illness. Social Psychiatry & Psychiatric Epidemiology, 46(2), 149–157.

32.

Howell

K. H.

(2011). Resilience and psychopathology in children exposed to family violence. Aggression and Violent Behavior, 16(6), 562–569.

33.

Howell

K. H.

Barnes

S. E.

Miller

L. E.

Graham-Bermann

S. A.

(2016). Developmental variations in the impact of intimate partner violence exposure during childhood. Journal of Injury & Violence Research, 8(1), 43–57.

34.

Zhu

Jiang

Zheng

… Shang

(2018). The association and mediating mechanism between poverty and poly-victimization of left-behind children in rural China. Children and Youth Services Review, 91, 22–29.

35.

Instituto Nacional de Estdistica e Informatica. (2016). Más del 40% de las niñas y niños fueron víctimas de violencia física o psicológica en los últimos 12 meses [More than 40% of girls and boys were victims of physical or psychological violence in the last 12 months] (No. 124). https://www.inei.gob.pe/prensa/noticias/mas-del-40-de-las-ninas-y-ninos-fueron-victimas-de-violencia-fisica-o-psicologica-en-los-ultimos-12-meses-9191/

36.

Kira

Lewandowski

Yoon

Somers

Chiodo

(2012). The linear and nonlinear associations between multiple types of trauma and IQ discrepancy indexes in African American and Iraqi refugee adolescents. Journal of Child & Adolescent Trauma, 5(1), 47–62.

37.

Lanza

S. T.

Dziak

J. J.

Huang

Wagner

A. T.

Collins

L. M.

(2018). LCA Stata plugin users’ guide (Version 1.2.1). The Methodology Center, The Pennsylvania State University. Methodology.psu.edu.

38.

Lister

C. E.

Merrill

R. M.

Vance

D. L.

West

J. H.

Hall

P. C.

Crookston

B. T.

(2015). Victimization among Peruvian adolescents: Insights into mental/emotional health from the young lives study. Journal of School Health, 85(7), 433–440.

39.

Miller-Graff

L. E.

Howell

K. H.

Martinez-Torteya

Hunter

E. C.

(2015). Typologies of childhood exposure to violence: Associations with college student mental health. Journal of American College Health, 63(8), 539–549.

40.

Miller-Graff

L. E.

Scrafford

Rice

(2016). Conditional and indirect effects of age of first exposure on PTSD symptoms. Child Abuse & Neglect, 51, 303–312.

41.

Millones

D. L. M.

Ghesquière

Van Leeuwen

(2014). Evaluation of a parental behavior scale in a Peruvian context. Journal of Child and Family Studies, 23(5), 885–894.

42.

Millones

D. L. M.

Van Leeuwen

Ghesquière

(2013). Associations between psychosocial functioning and academic achievement: The Peruvian case. Universitas Psychologica, 12(3), 725–737.

43.

Morales

Singh

(2015). The effects of child physical maltreatment on nutritional outcomes: Evidence from Peru. The Journal of Development Studies, 51(7), 826–850.

44.

Nguyen

A. J.

(2016). A latent class approach to understanding experiences of bullying victimization among youth in four low-resource settings. Johns Hopkins University. https://jscholarship.library.jhu.edu/bitstream/handle/1774.2/60247/NGUYEN-DISSERTATION-2016.pdf?sequence=1&isAllowed=y

45.

Pears

K. C.

Capaldi

D. M.

(2001). Intergenerational transmission of abuse: A two generational prospective study of an at-risk sample. Child Abuse & Neglect, 25(11), 1439–1461.

46.

Raposa

E. B.

Hammen

C. L.

’Brennan

P. A. O.

Callaghan

Najman

J. M.

(2014). Early adversity and health outcomes in young adulthood: The role of ongoing stress. Health Psychology, 33(5), 410–418.

47.

Scheid

C. R.

Miller-Graff

L. E.

Guzmán

D. B.

(in press). Parenting practices and intergenerational cycle of victimization in Peru. Development and Psychopathology.

48.

Thomas

J. L.

Sperry

Yarbrough

M. S.

(2000). Grandparents as parents: Research findings and policy recommendations. Child Psychiatry & Human Development, 31(1), 3–22.

49.

Turner

H. A.

Finkelhor

Ormrod

(2006). The effect of lifetime victimization on the mental health of children and adolescents. Social Science & Medicine, 62(1), 13–27.

50.

Vermunt

J. K.

Magidson

(2004). Latent class analysis. The Sage Encyclopedia of Social Sciences Research Methods, 2, 549–553.

51.

Widom

C. S.

Wilson

H. W.

(2015). Intergenerational transmission of violence. In Lindert

Levav

(Eds.), Violence and mental health (pp. 27–45). Springer.