Predicting Adverse Childhood Experiences

Abstract

Few studies have examined multilevel effects of neighborhood context on childhood maltreatment. Less work has analyzed these effects with juvenile offenders, and no prior work has examined context effects of childhood maltreatment through the Adverse Childhood Experiences (ACEs) framework. ACEs include 10 indictors: emotional abuse, physical abuse, sexual abuse, emotional neglect, physical neglect, domestic violence toward the youth’s mother, household substance abuse, household mental illness, parental separation/divorce, and household member with a history of jail/imprisonment. Effects of concentrated disadvantage and affluence on ACE scores are examined in a statewide sample of more than 59,000 juvenile offenders, controlling for salient individual (including family and parenting) measures and demographics. Both disadvantage and affluence affect ACE exposure. Implications for research and policy are discussed.

Keywords

Adverse Childhood Experiences ACE trauma neighborhood context juvenile offenders

In celebration of the 40th anniversary of the Juvenile Justice and Delinquency Prevention Act, Office of Juvenile Justice and Delinquency Prevention (OJJDP; 2014) Administrator Robert Listenbee Jr. critiqued that while there was cause for celebration with declines in offending and residential placement of youth, there were several challenges and opportunities to further improve outcomes for youth while maintaining public safety. Toward the top of Administrator Listenbee’s list of reforms was recognizing the extensive trauma, abuse, and exposure to violence histories of justice-involved youth. Prior work has found higher prevalence of adversity and trauma for justice system-involved youth in comparison with the general population (Dierkhising et al., 2013). Justice-involved youth are more likely to have experienced multiple forms of trauma (Abram et al., 2004), with one third reporting exposure to multiple types of trauma each year (Dierkhising et al., 2013), and 50% reporting exposure to four or more types of trauma by age 18 (Baglivio et al., 2014). In the 2013 Sutherland Address, Widom (2014) stated “violence prevention policies and programs that target abused and neglected children are warranted, given the prominent role of [Child Abuse and Neglect] in the backgrounds of these violent offenders” (p. 313). While prior work has examined a limited number of individual types of abuse or neglect, few have analyzed the multilevel effects of neighborhoods on abuse and less have studied neighborhood effects on a comprehensive individual-level measure of multiple types of abuse and neglect. The present study aims to fill that void.

Abuse and Delinquent Youth

Among offenders, even after controlling for prior delinquency, experiencing childhood physical abuse and other maltreatment leads to higher self-reported total, violent, and property offending (Teague, Mazerolle, Legosz, & Sanderson, 2008). Experiencing child abuse/neglect has been shown to double the risk of arrest for violent offenses for girls (Maxfield & Widom, 1996). Controlling for socioeconomic status, physically abused youth had more violent offenses then nonabused youth (Lasford et al., 2007). Within disadvantaged families, childhood maltreatment predicts arrest (Mersky & Topitzes, 2010). Herrera and McCloskey (2001) found witnessing marital violence in childhood uniquely contributes to later behavioral problems and/or delinquency, and predicted referral to juvenile court. These findings support additional research, including meta-analytic work, indicating that exposure to domestic violence leads to internalizing and externalizing behavior problems (Evans, Davies, & DiLillo, 2008; Moylan et al., 2010). Parental divorce exhibits a strong association with delinquency (Amato, 2001) which has not decreased despite increased social acceptability and prevalence in recent decades (Amato, 2001; D’Onofrio et al., 2005), and which is not mediated by common genes when examining adoptive and biological families (Burt, Barnes, McGue, & Iacono, 2008). Exposure to parental incarceration is associated with delinquency and maladaptive behaviors (Geller, Garfinkel, Cooper, & Mincy, 2009; Murray & Farrington, 2008; Parke & Clarke-Stewart, 2002), even after controlling for other childhood risk factors (Murray & Farrington, 2005).

Adverse Childhood Experiences (ACEs)

A limitation of prior work has been examining only one, or a few, abuses at a time in a given study. One answer to this drawback has been the concept of the ACE score first described in 1998 in the seminal “Adverse Childhood Experiences (ACE) Study” (Felitti et al., 1998). The ACE concept acknowledges the complex and cumulative nature of risk factors through the process of summing risk factors and associating the composite score with relevant outcomes developed by Rutter (1983). ACEs refer to 10 childhood experiences, which include emotional abuse, physical abuse, sexual abuse, emotional neglect, physical neglect, domestic violence toward the youth’s mother, household substance abuse, household mental illness, parental separation/divorce, and household member with a history of jail/imprisonment. The ACE score is expressed as the sum of the 10 exposures, measured dichotomously (yes/no), and therefore ranges from 0 to 10. An exposure (e.g., sexual abuse) is counted as 1 point regardless of the number of incidents of the exposure (whether sexually abused 1 or 100 times). The implications of high ACE scores have been well documented in the medical literature (Anda, Butchart, Felitti, & Brown, 2010). Higher ACE scores have been shown to increase the odds of smoking, heavy drinking, incarceration, and morbid obesity, along with increased risk for poor educational and employment outcomes, and recent involvement in violence (Bellis, Lowey, Leckenby, Hughes, & Harrison, 2014). Higher ACE scores have also been shown to significantly increase the odds of developing some of the leading causes of death in adulthood, such as heart disease, cancer, chronic lung disease, skeletal fractures, and liver disease.

Prior studies have shown that for adults who have experienced four or more ACEs in childhood, the odds of having one of the above-mentioned negative health outcomes are up to 12 times greater than those of adults who have not had such exposure (Felitti et al., 1998). While first identified as risk factors for chronic disease, ACEs have more recently been identified with immediate negative consequences, such as chromosome damage (Shalev et al., 2013) and functional changes to the developing brain (Anda et al., 2010; Cicchetti, 2013; Danese & McEwen, 2012; Teicher et al., 2003). Furthermore, high ACE scores have been linked to sexually risky behaviors, such as having 50 or more sexual partners, intercourse before age 15 (Hillis, Anda, Felitti, & Marchbanks, 2001), and teenage pregnancy (Hillis et al., 2004). ACE studies of juvenile offenders have found prevalence rates 3 times higher than those reported in the original ACE study sample (Grevstad, 2010). In addition, juvenile offenders were found 13 times less likely to report zero ACEs and 4 times more likely to report ACE scores of 4 or more compared with the ACE Study’s private-insured population of mostly college-educated adults (Baglivio et al., 2014).

The ACE research has shown that types of childhood abuse and neglect are common, highly interrelated, and exert a powerful cumulative effect on human development (Anda et al., 2010). This “cumulative stressor approach,” based on the co-occurrence and cumulative effect of these experiences, necessitates their examination as a collective composite, as opposed to the traditional approach of examining one or only a few adverse exposures, which misses the broader context in which they occur. The use of the ACE score as a measure of the cumulative effect of traumatic stress exposure during childhood is consistent with the latest understanding of the effects of traumatic stress on neurodevelopment (Anda et al., 2010; Anda et al., 2006).

Neighborhood Context and Abuse

Widom (2014) cautioned that any relationship between child abuse or neglect and later outcomes, such as delinquency, offending, or violence, is confounded by socioeconomic status. The concentration of child maltreatment in particular neighborhoods, predominately disadvantaged ones, is well established (Coulton, Crampton, Irwin, Spilsbury, & Korbin, 2007). In a review of the literature, Coulton and colleagues (2007) indicated neighborhood structural factors, and economic markers in particular, are most consistently linked to child abuse and neglect. One serious limitation of the studies reviewed, however, is their use of neighborhood characteristics and neighborhood maltreatment rates. Only 3 of the 25 studies reviewed used multilevel modeling, examining neighborhood-level measure associations with individual-level measures of childhood maltreatment (see Coulton, Korbin, & Su, 1999; Kim, 2004; Molnar, Buka, Brennan, Holton, & Earls, 2003). These multilevel designs have found smaller neighborhood effects than those aggregated to child maltreatment rates, with most of the variance being explained at the individual or family level (Coulton et al., 2007). Between 2% and 5% of the variance in abuse/neglect was found among neighborhoods in multilevel studies. An additional limitation in prior work has been the widespread use of official report data of child maltreatment from institutional sources (such as Child Protective Services) or a handful of studies that used self-report data from parents (Coulton et al., 2007). The danger in using official data for child maltreatment is of course in differences in how maltreatment is defined, recognized, and reported such that differences found are potentially variations in maltreatment reports, not necessarily in maltreatment behaviors. This has been termed the “definition, recognition, and reporting” path to accounting for the correlation between childhood maltreatment and neighborhood characteristics by Coulton and colleagues (2007). We are unaware of any published multilevel study using self-report of child maltreatment from the youth.

In addition, prior studies have found differences in the associations between neighborhoods and maltreatment by the type of abuse/neglect. Different neighborhood measures (such as poverty, low socioeconomic status [SES], high violent crime rates, residential stability, etc.) are differently associated with different measures of abuse and neglect (Coulton et al., 2007). However, these differences may miss the point as types of abuse/neglect are extremely interrelated such that exposure to one type increases the odds of exposure to additional types. Differences among prior studies may be differences of convenience or researcher choices (in what abuse measures were available) rather than true differences in effects. As the ACE literature argues against examining individual types, the current multilevel study examines neighborhood measure effects on the ACE composite score, a score empirically linked to a host of negative later life and health outcomes.

Prior multilevel work has strongly suggested the need to include examining the potential protective effects of affluent neighborhoods and not just the role of concentrated disadvantage (Morenoff, Sampson, & Raudenbush, 2001). The exiting of more wealthy families from middle-class areas, increased access to social and institution resources, and increased social control based on enhanced ability to mobilize resources afforded to affluent neighborhoods have been espoused as justification to include measures of the upper-end income distributions (Brooks-Gunn, Duncan, Kato, & Sealand, 1993; Morenoff et al., 2001; Sampson, Morenoff, & Earls, 1999). The current study includes concentrated disadvantage and the degree of concentrated affluence relative to the concentration of poverty in a given census tract (using the Index of Concentration at the Extremes [ICE] measure explained below).

Additional concerns in multilevel research involve the definition/identification of neighborhood dimensions, the range of neighborhoods available by measures examined, and the number of individuals within each neighborhood (Leventhal & Brooks-Gunn, 2000). The current study organizes neighborhoods by census tracts across a large and diverse state, ensuring a wide range of neighborhoods across contextual measures, and includes almost 60,000 youth. Census tract boundaries are composed around prominent physical features, as well as important social and ethnic divisions, that define neighborhoods (Duncan & Aber, 1997; Leventhal & Brooks-Gunn, 2000). For that reason, we utilize the data readily available at the census tract level to construct our neighborhood measures of concentrated disadvantage and affluence, described in detail below, to assess their impact on the youth outcome of interest (ACE scores).

Current Study

The current multilevel study is a statewide examination of neighborhood effects on the ACE score of juvenile offenders. While the literature on ACE scores and negative later life and health outcomes has been growing, and is well documented, the current study is the first to examine the effects of neighborhood context in a multilevel model on ACE scores. Furthermore, the vast majority of studies examining the neighborhood context and child maltreatment in general have not been multilevel in nature (Coulton et al., 2007). The current study aims to fill those gaps and move the ACE research forward by including consideration of the effects of both the disadvantage and affluence of neighborhoods, across an entire state, in addition to individual-level measures (including family characteristics) on the cumulative stressor quantitative ACE score.

Data and Method

Data were drawn from the Florida Department of Juvenile Justice (FDJJ) archival data records and the U.S. Census Bureau. The FDJJ maintains a centralized database, the Juvenile Justice Information System (JJIS), that contains complete social, offense, placement, and risk assessment history data for all youth referred (equivalent to an adult arrest). The individual-level measures of interest were taken from the Positive Achievement Change Tool (PACT) risk/needs assessment panel used by the FDJJ. Data for this study included all the juvenile offenders under the care of FDJJ who turned 18 between January 1, 2007, and December 31, 2012, and who were assessed using the Full Positive Achievement Change Tool (C-PACT) risk/needs assessment (N = 64,329). Due to incomplete address data (a result of data entry error, or wards of the state incorrectly having state office addresses entered as home addresses), a total of 4,977 youth were excluded from the sample as a usable address could not be ascertained.¹ We used MapMarker software available from Pitney Bowes to match the home address information with 2010 census tract data to determine the Florida census tract in which each youth resided. There are a total of 4,245 census tracts in the state of Florida, but 1,397 had no youth contained in the current sample or were not residential tracts (e.g., parks or airports) and thus were excluded from the analyses presented here. Our final sample size includes 59,342 youth living in 3,948 neighborhoods. The average number of youth in each tract in the analyses presented was 15; however, this number ranged from 7 to 156.

Dependent Variable

ACE Scores

Although the purpose of the PACT assessment is to classify youth on risk to reoffend at their time of intake, certain PACT items relate to ACEs. These PACT items were used to derive ACE scores (see Baglivio et al., 2014, for specific items and responses). Ten individual ACEs were studied: parental separation or divorce; psychological abuse; physical abuse; sexual abuse; emotional neglect; physical neglect; household violence; living with household members who were substance abusers; household mentall illness; and household member jail/prison history. These are the same 10 ACEs used in the comprehensive Wave 2 of the original ACE Study (Dong et al., 2004). The outcome variable examined in the current study represents the sum (0-10) of all ACEs endured by youth at the time of their last contact with Department of Juvenile Justice (DJJ).

Independent Variables

Individual Level

Independent variables at both the individual level and neighborhood level were selected based on the relevant literature surrounding ACEs. At the individual level, the association between the total number of ACEs a youth has experienced by the time of their last contact with DJJ and a number of demographic, parental, and family variables is explored. Family support is measured as a dichotomous variable where the reference group is youth who reported they have little or no support outside their immediate family. Parental employment problem history is dichotomous where the reference group is youth who report no parental employment issues.

We also control for gender, race-ethnicity, and age because these demographic factors may also be associated with ACEs. Gender was measured using a dichotomous variable (0 = female, 1 = male). Race–ethnicity is measured using a set of dichotomous variables with 1 = Black, 1 = Hispanic, and 1 = Other (Haitian, Jamaican, Native American; in each category, non-Hispanic White is the reference group). Age is measured in years for each youth in the sample and reflects the age at which youth were last in contact with DJJ.

Neighborhood Level

Data used to construct the neighborhood-level measures were taken from the 2008-2012 American Community 5-year estimates for census tracts in the State of Florida (U.S. Census Bureau, 2014). Whether or not census tracts constitute neighborhoods is a long-standing issue within social science research (Coulton, Korbin, Chan, & Su, 2001). Tracts generally have stable boundaries, are designed to be relatively homogeneous with respect to population characteristics, and are a geographic unit for which data are readily available from the U.S. Census Bureau. Although imperfect, the use of census tracts as proxies for neighborhoods is common within most research which examines neighborhood effects (Morland, Wing, Diez Roux, & Poole, 2002; Mujahid, Diez Roux, Morenoff, & Raghunathan, 2007; Sampson, Morenoff, & Gannon-Rowley, 2002).

Six census tract variables were used to form the measure of neighborhood disadvantage used in the current study: the proportion of families living below the poverty line, median family income (logged and reverse coded), the proportion of female-headed households, the unemployment rate, the proportion of the population with a high school degree (reverse coded), and the proportion of households receiving public assistance. Previous studies have used some combination of these variables to assess the impact of community socioeconomic status on a variety of outcomes (Baumer et al., 1998; Kubrin & Stewart, 2006; Sampson, Radenbush, & Earls, 1997). These variables are strongly correlated to one another at the neighborhood level, and alpha factor analyses indicated that these variables loaded on a single factor in our sample (α = .920). The items were standardized and combined to form an additive index of neighborhood disadvantage.

In addition to assessing the impact of concentrated disadvantage on ACEs, we examine the effect of concentrated affluence on the prevalence of ACEs. Prior work has pointed out that researchers spend too much time exploring the consequences of disadvantage and little time focusing on affluence, which may generate a separate set of protective mechanisms, thereby reducing negative outcomes (Brooks-Gunn et al., 1993; Massey, 2001; Morenoff et al., 2001). To measure concentrated affluence, we use Massey’s (2001) ICE measure. ICE captures the degree to which affluence is concentrated, relative to the concentration of poverty in a neighborhood. Accordingly, it reflects the relative socioeconomic inequality in a community, rather than the absolute level of disadvantage. Consistent with prior research, the ICE index was calculated using the following formula: [(Number of affluent families − Number of poor families) / Total number of families] (Kubrin & Stewart, 2006). In the current analysis “affluent” is defined as families with incomes two standard deviations above the mean (M = US$60,362 σ = US$26,283, which equates to US$112,928), and “poor” is defined as families below the poverty line. This measure ranges from +1 to −1. A value of +1 indicates that all families in a given neighborhood are affluent; a value of −1 indicates all families are poor; and a value of 0 indicates an equal balance between affluent and poor families (Kubrin & Stewart, 2006; Massey, 2001; Morenoff et al., 2001).

Analytic Strategy

To assess the impact of neighborhood-level conditions on childhood experiences, multilevel linear regression analyses are fitted using maximum likelihood estimation in Stata 13. The hierarchical regression models used here are extensions of traditional regression models that account for the structuring of data across aggregate groupings, that is, they explicitly account for the nested nature of data across multiple levels of analysis (i.e., youth nested in neighborhoods; Raudenbush & Byrk, 2002). Because one of the assumptions of ordinary regression models is that the error terms are independent, such clustering would violate this core model assumption. Consequently, failure to account for nonindependence of observations can result in standard errors that are biased downward, increasing the chances of drawing incorrect conclusions (Raudenbush & Byrk, 2002).

A related problem is that statistical significance tests in ordinary regression models overestimate the influence of the neighborhood-level predictors as the degrees of freedom are not adjusted for the hierarchical nature of the data. Accordingly, in the models presented here, the degrees of freedom used in the statistical significance tests have been adjusted for the number of aggregate units in the data. This type of multilevel modeling provides insight into the extent that individual differences in ACEs are due to either individual-level characteristics or characteristics of the neighborhood in which the youth live (contextual effects).

A three-step strategy is used to assess the impact of both individual- and neighborhood-level factors on youth’s ACE scores. First, a null model (Model 1) is fitted, without any Level 1 or Level 2 predictors included. This model serves as a benchmark to which the more complex models are to be compared. The second model includes only the individual-level measures to determine to what extent youth-specific attributes explain variation in ACEs. Models 3 and 4 explore the contextual effects of neighborhood disadvantage and affluence separately, and Model 5 includes both predictors in a single model to assess their joint effects.

Results

Table 1 presents the descriptive statistics for the variables included in our analysis. Table 1 indicates that the mean number of ACEs in our sample of youth was 3.61, although this number ranged from 0 to a maximum of 10. Seventy-nine percent of the sample is male, with an average age of 17.67 at the time of their last PACT assessment. Blacks make up the largest proportion of our sample (43%), followed by Whites (38%), Hispanics (15%) and Other races (a total of 4%).

Table 1.

Summary Statistics for the Analysis of ACEs in Florida Youth.

	Description	M	Median	Minimum	Maximum	SD
Dependent variable (n = 59,342)
ACEs	Sum of ACEs	3.61	3	0	10	1.87
Youth-level independent variables
Age at assessment	Youth’s age when PACT assessment was completed	17.67	17.85	11.37	22.5	1.02
Male	Coded 0 for females, 1 for males	0.79	1	0	1	0.41
Black	Coded 1 for Black youth, 0 for all others	0.43	0	0	1	0.5
Hispanic	Coded 1 for Hispanic youth, 0 for all others	0.15	0	0	1	0.36
Other race	Coded 1 for Haitian, Jamaican, or native American; 0 for all others	0.04	0	0	1	0.19
Family support	Coded 0 for no support network, 1 for some support network, 2 for strong support network	1.35	1	0	2	0.61
Parental employment issues	Coded 0 for no parental employment problem history, 1 for parental employment problem history	0.14	0	0	1	0.35
Neighborhood-level independent variables (n = 3,958)
Concentrated disadvantage	Six-item index composed of % family poverty, logged median family income (reversed), % female-headed households, unemployment rate, % population with high school degree (reversed), and % households receiving public assistance (α = .920)	0.00	−0.08	−2.71	4.03	0.85
Concentrated affluence	ICE concentrated affluence index	−0.06	−0.06	−0.97	0.74	0.20

Note. ACEs = Adverse Childhood Experiences; PACT = Positive Achievement Change Tool; ICE = Index of Concentration at the Extremes.

Table 2 presents the bivariate correlations between the variables used in the current study. The correlations presented in Table 2 lend support to several hypotheses. In particular, the measures of neighborhood disadvantage, neighborhood affluence, and the individual-, parental- and familial-level variables are significantly associated with the sum of ACE scores. Females and Blacks are more likely to have endured a larger number of ACEs. Youth who have parents with substance abuse problems, mental health problems, or employment issues are also more likely to have higher ACE scores. To investigate these relationships more closely, we now turn to the multivariate results.

Table 2.

Bivariate Correlations for the Analysis of ACEs in Florida Youth (n = 59,342).

		(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)
(1)	ACEs	1.000
(2)	Age at assessment	.088**	1.000
(3)	Male	−.169**	.114**	1.000
(4)	Black	.019**	.059**	.005	1.000
(5)	Hispanic	−.112**	−.010*	.056**	−.374**	1.000
(6)	Other race	−.051**	.012**	.010*	−.169**	−.082**	1.000
(7)	Family support	−.196**	.039**	.056**	−.020**	.020**	−.010*	1.000
(8)	Parental employment issues	.254**	.052**	−.022**	.020**	−.021**	−.014**	−.049**	1.000
(9)	Concentrated disadvantage	.040**	.044**	.004	.392**	.005	.040**	−.059**	.073**	1.000
(10)	Concentrated affluence (ICE)	−.064**	−.037**	.004	−.349**	.001	−.029**	.068**	−.075**	−.824**	1.000

Note. ACEs = Adverse Childhood Experiences; ICE = Index of Concentration at the Extremes.

p < .05. **p < .01.

Unconditional Model

We began by assessing the degree to which ACEs vary across neighborhoods. To do this, we estimated an unconditional, random intercept model of variance (i.e., a model with no predictor variables). The results of this analysis are presented in Model 1 of Table 2. The significance of the grand mean intercept (3.571) corresponds to the mean level of ACE scores across neighborhoods. Also important is the finding of a significant random effects variance component .279 (χ² = 1,695.93, p < .01), which indicates that the sum of ACEs varies significantly across neighborhoods and therefore can be modeled. Figure 1 provides a graphical depiction of the degree of variation in the mean ACE score across neighborhoods in the current study. The figure shows that the average ACE score in a given neighborhood varies from 0 to greater than 8 and also important, the scores are fairly normally distributed.

Figure 1.

Distribution of ACE scores across neighborhoods.

Figure 1 raises the question of what predictors account for the differences in ACE scores across neighborhoods. One potential explanation is that these differences reflect individual-level characteristics of the youth in these neighborhoods. Alternatively, the prevalence of ACEs may be greater in some neighborhoods because youth living there are exposed to higher levels of concentrated disadvantage or inequality. In Models 2 to 5 of Table 3, we assess these possibilities in more detail.

Table 3.

Multilevel Multivariate Regression Results for the Analysis of Adverse Childhood Experiences in Florida Youth.

Youth level	(1)	(2)	(3)	(4)	(5)
Intercept	3.571**	3.586**	3.604**	3.609**	3.601**
Intercept	(.012)	(.010)	(.010)	(.010)	(.010)
Age at assessment	—	0.188**	0.188**	0.188**	0.188**
Age at assessment	—	(.007)	(.007)	(.007)	(.007)
Male	—	−0 716**	−0.715**	−0.714**	−0.714**
Male	—	(.017)	(.017)	(.017)	(.017)
Black	—	−0.181**	−0.225**	−0.240**	−0.218**
Black	—	(.018)	(.019)	(.019)	(.019)
Hispanic	—	0−.512**	−0.542**	−0.552**	−0.537**
Hispanic	—	(.023)	(.023)	(.023)	(.023)
Other race	—	−0.537**	−0.576**	−0.588**	−0.569**
Other race	—	(.040)	(.041)	(.040)	(.041)
Family support	—	−0.524**	−0.521**	−0.519**	−0.519**
Family support	—	(.012)	(.012)	(.012)	(.012)
Parental employment issues	—	1.217**	1.210**	1.206**	1.208**
Parental employment issues	—	(.020)	(.020)	(.020)	(.020)
Neighborhood level
Concentrated disadvantage	—	—	0.100**	—	0.225**
Concentrated disadvantage	—	—	(.013)	—	(.032)
Concentrated affluence (ICE)	—	—	—	−0.595**	−1.395**
Concentrated affluence (ICE)	—	—	—	(.052)	(.123)
Random effects
σ²	0.279	0.154	0.149	0.143	0.138
χ²	1,695.93	903.01	880.32	832.19	792.39

Note. n = 59,342 youth within 3,948 neighborhoods. ICE = Index of Concentration at the Extremes.

p < .05. **p < .01.

Level 1 (Youth Level)

A comparison of the variance components from Models 1 and 2 indicates that individual-level characteristics account for roughly 45% of the variance in ACE scores within neighborhoods (.448 = .279 − .154 / .279). Thus, part of the explanation for why in some neighborhoods youth exhibit a larger number of ACEs is that some of the respondents have individual-level risk factors that increase ACEs. Specifically, youth who were older at the time of their release from DJJ supervision and those whose parents had employment issues reported higher ACE scores. Support from members outside of the youth’s immediate family was also associated with lower ACE scores. Despite the impact of the individual-level characteristics, the variance component in Model 2 indicates that a significant amount of variation in ACE scores still exists across neighborhoods, suggesting that additional factors contribute to ACE scores.

Level 2 (Neighborhood Level)

The vast literature on the importance of neighborhood effects described above suggests that living in a community characterized by poverty, inequality, and socioeconomic disadvantage can increase the risk of a number of negative life outcomes, including ACEs. Model 3 of Table 3 presents results that assess this possibility while accounting for the individual-level characteristics shown in Model 2 to have a significant effect on ACE scores. As shown, the relationship between neighborhood disadvantage and youth ACE scores is significant and positive. Consistent with our predictions, living in a disadvantaged neighborhood is a risk factor associated with a higher number of ACEs above and beyond youth-level attributes. Model 4 examines the effect of concentrated affluence on the sum of ACEs. This finding suggests that neighborhoods with large concentrations of affluent families (relative to poor families), or resource-rich neighborhoods, serve as a protective factor in reducing ACEs.

It is worth noting that neighborhood disadvantage accounted for roughly 3% of the variance in ACEs across neighborhoods, whereas the ICE measure accounted for roughly 7%. Combined, they accounted for a total of 10.4% of the variance in ACE scores across neighborhoods. Overall, Model 5 of Table 3 suggests that the individual- and neighborhood-level variables accounted just over half of the total variance in ACEs (.505 = .279 − .138 / .279) at the neighborhood level.

Although individual-level characteristics account for a large portion of the variance in ACEs observed, neighborhood disadvantage and the ICE measure of concentrated affluence are also significant predictors of ACE scores across neighborhoods. Figure 2 displays the predicted values for youth who reside in neighborhoods that differ on levels of disadvantage and inequality. The predicted values were computed using the coefficients from Model 5 of Table 3 and assume mean values for all other variables in the model (Raudenbush & Byrk, 2002). The predicted values associated with the estimated disadvantage effect show that the predicted ACE score of the juvenile offenders ranges from about 5.25 in neighborhoods with lower levels of disadvantage (−2σ) to about 6.0 in those with higher levels of disadvantage (+2σ), assuming the mean for all other variables. The results for the ICE measure show a slightly larger difference, where extreme inequality (−2σ) is associated with a predicted ACE score of 6.27, whereas a low degree of inequality (+2σ) is associated with lower ACE scores (5.15). Collectively, the results for neighborhood disadvantage and the ICE measure indicate community context is an important predictor of higher ACE scores and that childhood experiences do vary depending on neighborhood characteristics.

Figure 2.

Predicted ACE score given varying neighborhood socioeconomic contexts.

Discussion and Conclusion

The current study examined the effects of concentrated disadvantage and affluence on the ACE scores of high-risk juvenile offenders. Both disadvantage and affluence were significantly associated with ACE scores, in the expected directions, after controlling for demographics and family support and parental employment. The relationship between the concentrated affluence ICE measure and ACE scores was found more robust than for the disadvantage index. Results show the economic reality of the neighborhood in which a juvenile lives has an effect on the number of childhood maltreatment types the youth experiences. Childhood maltreatment, explicitly operationalized as an ACE score, has been linked to myriad of negative outcomes, including immediate chromosomal and functional brain damage (Anda et al., 2010; Cicchetti, 2013; Danese & McEwen, 2012; Shalev et al., 2013; Teicher et al., 2003), proximal behavioral issues (Bellis et al., 2014), and more distal negative health outcomes, including leading causes of death (Felitti et al., 1998). Studying the predictors of ACEs is paramount to understanding the prevention of the host of negative outcomes indicated in prior research. An understanding of the neighborhood factors that lead to higher abuse/neglect exposure indicates the contextual issues required to be addressed.

This study is not without limitations. All youth had an offending history (at least one arrest). However, not all juvenile offenders were included, as only the PACT Full Assessment contained the required items to create the 10-item ACE score. While 64,329 youth who turned 18 during the study period were assessed with the PACT Full Assessment, an additional 136,691 youth who turned 18 during that time were only assessed with the PACT Pre-Screen, prohibiting the creation of ACE scores for those youth. While we captured ACE scores for all youth receiving a Full Assessment (approximately 32% of all juvenile offenders), caution should be used in generalizing the results to all Florida juvenile offenders. With this sample bias toward higher risk youth, however, it should be noted that 45% of the youth were classified as low or moderate risk to reoffend. Prior ACE work indicates that higher ACE scores are found in special populations (Baglivio et al., 2014; Dube et al., 2001). Accordingly, the results of this study are likely not typical of nonoffending juveniles.

An additional limitation is that of the temporal order of our ACE score and the neighborhood context measures of disadvantage and affluence. Although we used the address at the time of arrest and the ACE score at the time of arrest, all neighborhood context measures were based on the 2008-2012 American Community 5-year estimates regardless of when the arrest was made. Importantly, 18.7% of the youth were arrested during 2007 (not covered by the period for which our neighborhood-level measures were drawn). Although we chose to not use different sets of 5-year estimates for those youth who were arrested in 2007, given the relative stability of neighborhood conditions over short time periods, we anticipate this decision to have very little impact on the results presented (Rohe & Stewart, 1996; Temkin & Rohe, 1996). In addition, we did not want to limit our sample to only youth arrested during 2008 to 2012 as we strived to ensure the broadest range of census tracts, with the greatest number of youth per tract.

A second concern regarding temporal order stems from the inability to track the mobility of youth during their entire childhood. In the current study, we assessed the impact of neighborhood condition based on the youth’s neighborhood at the time of arrest. Accordingly, we are unable to say with certainty that the youth were living in a neighborhood characterized by these conditions during their entire childhood, the period during which many of the ACEs may have occurred. While it could be suggested that juvenile offenders may be a more transient population than nonoffending juveniles, and thus lived in different neighborhoods during their upbringing, the odds of a given youth moving substantially higher or lower in terms of disadvantage and/or affluence, we speculate, is quite rare. Future empirical efforts should test that speculation.

Future research should examine additional neighborhood context measures that may affect ACE exposure. Residential mobility, institutional resources, collective efficacy, and ethnic heterogeneity/immigration concentration are arguably theoretically linked to maltreatment. Couching ACE into a theoretical framework, such as the life-course perspective, would further advance the field. Exposure to multiple abuse types during childhood and adolescence is arguably a significant transition that may alter one’s life trajectory. Future endeavors should examine whether intervention programs can mitigate the effects of ACE on negative outcomes either by targeting individual factors (e.g., trauma-focused cognitive behavioral therapy [CBT]) or neighborhood context factors (such as vouchers to move to less disadvantaged areas). Examining interactions between individual and environmental characteristics, as well as the moderating effects of individual characteristics on the impact of affluence or disadvantage appears ripe for evaluation.

Neighborhood context clearly affects ACE exposure, as over 10% of the neighborhood-level variance in ACE was attributable to affluence and disadvantage. This is 2 to 5 times greater than the prior 2% to 5% variance explained found in prior context and childhood abuse work (see Coulton et al., 2007). Examining the components of our contextual measures leads to several policy implications through targeting those mechanisms. Our neighborhood disadvantage measure included unemployment rate, percent female-headed households, and percent with a high school diploma. Ripe policy targets include high school dropout prevention programs, teen pregnancy prevention programs, and job skills/life skills intervention and prevention both in schools and accessible community resources. Prevention services like the famed Nurse–Family Partnership may help curb the intergenerational transfer of childhood maltreatment. While the ACE score is a summation of abuse/trauma types, surely some communities are disproportionately plagued by some of those types over others. Targeted prevention to assessed community ACE scores, much like the process of individualizing case plans based on risk assessment results at the youth level, seems a productive avenue. A concerted and collaborative effort among youth and family services agencies to target existing family preservation, high school dropout and delinquency prevention, job readiness, and reentry services based on assessed community ACE scores should be considered. Specific areas may benefit from community revitalization projects, reentry assistance for members returning from prison, and domestic violence prevention.

Few studies have examined the effects of neighborhood characteristics on childhood maltreatment. No prior studies have examined those effects with juvenile offenders and maltreatment conceptualized using the ACE score. Marrying the literatures of multilevel neighborhood effects with the plethora of medical research on the negative effects of higher ACE scores advances discussion of how structural and contextual resources can be leveraged to improve health and justice system outcomes in truly disadvantaged and at-risk families and communities.

Footnotes

Authors’ Note

The views and opinions expressed in this article are those of the authors and not necessarily those of the Florida Department of Juvenile Justice.

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) received no financial support for the research, authorship, and/or publication of this article.

Notes

Author Biographies

Michael T. Baglivio, PhD, currently works as the director of Research and Program Development for G4S Youth Services, LLC, examinig the effectiveness of juvenile treatment programming on short- and long-term performance measures and outcomes. His research interests include criminological theory, risk assessment, and life-course criminology.

Kevin T. Wolff is an assistant professor at John Jay College of Criminal Justice in New York City. He earned his PhD from the College of Criminology and Criminal Justice at Florida State University. His research interests include the spatial patterning of crime, juvenile justice, criminological theory, and quantitative methods.

Nathan Epps, MS, completed his graduate studies in criminology in May 1986 at the Florida State University, School of Criminology. He currently works for the Bureau of Research and Planning at the FDJJ. His research interests include trauma-informed care, adverse childhood experiences, and the application of geographical information systems (GIS) software to juvenile justice data.

Randy Nelson, PhD, currently serves as the program coordinator for the Bethune Cookman University Criminal Justice Administration Graduate Program. As the president of 21st Century Research & Evaluations, Inc., he is currently working with the University of Florida and Florida’s Historically Black Colleges and Universities on the High-Risk Delinquent and Dependent Youth Research Project. His research interests include disproportionate minority contact, delinquency prevention, and community policing.

References

Abram

K. M.

Teplin

L. A.

Charles

D. R.

Longworth

S. L.

McClelland

G. M.

Dulcan

M. K.

(2004). Posttraumatic stress disorder and trauma in youth in juvenile detention. Archives of General Psychiatry, 61, 403-410.

Amato

P. R.

(2001). Children of divorce in the 1990’s: An update of the Amato and Keith (1991) meta-analysis. Journal of Family Psychology, 15, 355-370.

Anda

R. F.

Butchart

Felitti

V. J.

Brown

D. W.

(2010). Building a framework for global surveillance of the public health implications of Adverse Childhood Experiences. American Journal of Preventive Medicine, 39, 93-98.

Anda

R. F.

Felitti

V. J.

Bremner

J. D.

Walker

J. D.

Whitfield

Perry

B. D.

. . . Giles

W. H.

(2006). The enduring effects of abuse and related adverse experiences in childhood: A convergence of evidence from neurobiology and epidemiology. European Archives of Psychiatry and Clinical Neuroscience, 256, 174-186.

Baglivio

M. T.

Epps

Swartz

Sayedul

Sheer

Hardt

N. S.

(2014). The prevalence of Adverse Childhood Experiences (ACE) in the lives of juvenile offenders. Journal of Juvenile Justice, 3(2), 1-23.

Baumer

E. P.

Messner

S. F.

Felson

(1998). Influence of crack cocaine on robbery, burglary, and homicide rates: A cross-city, longitudinal analysis. Journal of Crime and Delinquency, 33, 316-340.

Bellis

M. A.

Lowey

Leckenby

Hughes

Harrison

(2014). Adverse childhood experiences: Retrospective study to determine their impact on adult health behaviors and health outcomes in a UK population. Journal of Public Health, 36, 81-91. doi:10.1093/pubmed/fdt038

Brooks-Gunn

Duncan

Kato

Sealand

(1993). Do neighborhoods influence child and adolescent behavior? American Journal of Sociology, 99, 353-395.

Burt

A. S.

Barnes

A. R.

McGue

Iacono

W. G.

(2008). Parental divorce and adolescent delinquency: Ruling out the impact of common genes. Developmental Psychology, 44, 1668-1677.

10.

Cicchetti

(2013). Annual research review: Resilient functioning in maltreated children—Past, present, and future perspectives. Journal of Child Psychology and Psychiatry, 54, 402-422.

11.

Coulton

C. J.

Crampton

D. S.

Irwin

Spilsbury

J. C.

Korbin

J. E.

(2007). How neighborhoods influence child maltreatment: A review of the literature and alternative pathways. Child Abuse & Neglect, 31, 1117-1142.

12.

Coulton

C. J.

Korbin

J. E.

Chan

(2001). Mapping resident’s perceptions of neighborhood boundaries: A methodological note. American Journal of Community Psychology, 29, 371-383.

13.

Coulton

C. J.

Korbin

J. E.

(1999). Neighborhoods and child maltreatment: A multi-level study. Child Abuse & Neglect, 23, 1019-1040.

14.

Danese

McEwen

B. S.

(2012). Adverse childhood experiences, allostasis, allostatic load, and age-related disease. Physiology & Behavior, 106, 29-39.

15.

Dierkhising

C. B.

S. J.

Woods-Jaeger

Briggs

E. C.

Lee

Pynoos

R. S.

(2013). Trauma histories among justice-involved youth: Findings from the National Child Traumatic Stress Network. European Journal of Psychotraumatology, 4, 20274. Retrieved from https://dx-doi-org.web.bisu.edu.cn/10.3402/ejpt.v4i0.20274

16.

Dong

Anda

R. F.

Felitti

V. J.

Dube

S. R.

Williamson

D. F.

Thompson

T. J.

. . . Giles

W. H.

(2004). The interrelatedness of multiple forms of childhood abuse, neglect, and household dysfunction. Child Abuse & Neglect, 28, 771-784.

17.

D’Onofrio

B. M.

Turkheimer

Emery

R. E.

Slutske

W. S.

Heath

A. C.

Madden

P. A.

Martin

N. G.

(2005). A genetically informed study of marital instability and its association with offspring psychopathology. Journal of Abnormal Psychology, 114, 570-586.

18.

Dube

S. R.

Anda

R. F.

Felitti

V. J.

Croft

J. B.

Edwards

V. J.

Giles

W. H.

(2001). Growing up with parental alcohol abuse: Exposure to childhood abuse, neglect, and household dysfunction. Child Abuse & Neglect, 12, 1627-1640.

19.

Duncan

G. J.

Aber

J. L.

(1997). Neighborhood models and measures. In Brooks-Gunn

Duncan

G. J.

Aber

J. L.

(Eds.), Neighborhood poverty: Vol. 1. Context and consequences for children (pp. 62-78). New York, NY: Russell Sage.

20.

Evans

S. E.

Davies

DiLillo

(2008). Exposure to domestic violence: A meta-analysis of child and adolescent outcomes. Aggression and Violent Behavior, 13, 131-140.

21.

Felitti

V. J.

Anda

R. F.

Nordenberg

Williamson

D. F.

Spitz

A. M.

Edwards

. . .Marks

J. S.

(1998). Relationship of childhood abuse and household dysfunction to many of the leading causes of death in adults: The Adverse Childhood Experiences (ACE) Study. American Journal of Preventive Medicine, 14, 245-258.

22.

Geller

Garfinkel

Cooper

C. E.

Mincy

R. B.

(2009). Parental incarceration and child well-being: Implications for urban families. Social Science Quarterly, 90, 1186-1202.

23.

Grevstad

J. A.

(2010, June 8). Adverse childhood experiences and juvenile justice. PowerPoint delivered to Washington State Family Policy Council, Olympia: WA.

24.

Herrera

V. M.

McCloskey

L. A.

(2001). Gender differences in the risk for delinquency among youth exposed to family violence. Child Abuse & Neglect, 25, 1037-1051.

25.

Hillis

S. D.

Anda

R. F.

Dube

S. R.

Felitti

V. J.

Marchbanks

P. A.

Marks

J. S.

(2004). The association between adverse childhood experiences and adolescent pregnancy, long-term psychosocial consequences, and fetal death. Pediatrics, 2, 320-327.

26.

Hillis

S. D.

Anda

R. F.

Felitti

V. J.

Marchbanks

P. A.

(2001). Adverse childhood experiences and sexual risk behaviors in women: A retrospective cohort study. Family Planning Perspective, 5, 206-211.

27.

Kim

J. S.

(2004). Neighborhood effects on the etiology of child maltreatment: A multilevel study. Dissertation Abstracts International, 65, 3157B.

28.

Kubrin

C. E.

Stewart

E. A.

(2006). Predicting who reoffends: The neglected role of neighborhood context in recidivism studies. Criminology, 44, 165-197.

29.

Lasford

J. E.

Miller-Johnson

Berlin

L. J.

Dodge

K. A.

Bates

J. E.

Pettit

G. S.

(2007). Early physical abuse and later violent delinquency: A prospective longitudinal study. Child Maltreatment, 12, 233-245.

30.

Leventhal

Brooks-Gunn

(2000). The neighborhoods they live in: The effects of neighborhood residence on child and adolescent outcomes. Psychological Bulletin, 126, 309-337.

31.

Massey

D. S.

(2001). The prodigal paradigm returns: Ecology comes back to sociology. In Booth

Crouter

(Eds.), Does it take a village? Community effects on children, adolescents, and families (pp.41-48). Mahwah, NJ: Lawrence Erlbaum.

32.

Maxfield

Widom

C. S.

(1996). The cycle of violence: Revisited six years later. Archives of Pediatrics & Adolescent Medicine, 150, 390-395.

33.

Mersky

J. P.

Topitzes

(2010). Comparing early adult outcomes of maltreated and non-maltreated children: A prospective longitudinal investigation. Children and Youth Services Review, 32, 1086-1096.

34.

Molnar

B. E.

Buka

S. L.

Brennan

R. T.

Holton

J. K.

Earls

(2003). A multi-level study of neighborhoods and parent-to-child physical aggression: Results from the project on human development in Chicago neighborhoods. Child Maltreatment, 8, 84-97.

35.

Morenoff

J. D.

Sampson

R. J.

Raudenbush

S. W.

(2001). Neighborhood inequality, collective efficacy, and the spatial dynamics of urban violence. Criminology, 39, 517-559.

36.

Morland

Wing

Diez Roux

Poole

(2002). Neighborhood characteristics associated with the location of food stores and food service places. American Journal of Preventive Medicine, 22, 23-29.

37.

Moylan

C. A.

Herrenkohl

T. I.

Sousa

Tajima

E. A.

Herrenkohl

R. C.

Russo

M. J.

(2010). The effects of child abuse exposure to domestic violence on adolescent internalizing and externalizing behavior problems. Journal of Family Violence, 25, 53-63.

38.

Mujahid

M. S.

Diez Roux

A. V.

Morenoff

J. D.

Raghunathan

(2007). Assessing the measurement properties of neighborhood scales: From psychometrics to econometrics. American Journal of Epidemiology, 165, 858-867.

39.

Murray

Farrington

D. P.

(2005). Parental imprisonment: Effects on boys’ antisocial behavior and delinquency through the life-course. Journal of Child Psychology and Psychiatry, 46, 1269-1278.

40.

Murray

Farrington

D. P.

(2008). The effects of parental imprisonment on children. In Morris

Tonry

(Eds.), Crime and justice: A review of research (Vol. 37, pp. 133-206). Chicago, IL: University of Chicago Press.

41.

Office of Juvenile Justice and Delinquency Prevention. (2014). Celebrating 40 years of working together for youth justice and safety [The Justice Blog]. Department of Justice. Retrieved from http://www.justice.gov/opa/blog/celebrating-40-years-working-together-youth-justice-and-safety

42.

Parke

Clarke-Stewart

(2002). Effects of parental incarceration on young children. Washington, DC: U.S. Department of Health and Human Services.

43.

Raudenbush

S. W.

Bryk

A. S.

(2002). Hierarchical linear models: Applications and data analysis methods. Thousand Oaks, CA: Sage.

44.

Rohe

Stewart

L. S.

(1996). Homeownership and neighborhood stability. Housing Policy Debate, 7, 37-81.

45.

Rutter

(1983). Statistical and personal interactions: Facets and perspectives. In Magnusson

Allen

(Eds.), Human development: An interactional perspective (pp. 295-320). New York, NY: Academic Press.

46.

Sampson

Morenoff

Earls

(1999). Beyond social capital: Spatial dynamics of collective efficacy for children. American Sociological Review, 64, 633-660.

47.

Sampson

Morenoff

Gannon-Rowley

(2002). Assessing neighborhood effects: Social processes and new directions in research. Annual Review of Sociology, 28, 443-478.

48.

Sampson

Raudenbush

S. W.

Earls

(1997). Neighborhoods and violent crime. Science, 277, 918-924.

49.

Shalev

Moffitt

Sugden

Williams

Houts

R. M.

Danese

. . .Caspi

(2013). Exposure to violence during childhood is associated with telomere erosion from 5 to 10 years of age: A longitudinal study. Molecular Psychiatry, 18, 576-581.

50.

Teague

Mazerolle

Legosz

Sanderson

(2008). Linking childhood exposure to physical abuse and adult offending: Examining mediating factors and gendered relationships. Justice Quarterly, 25, 313-348.

51.

Teicher

M. H.

Andersen

S. L.

Polcari

Anderson

C. M.

Navalta

C. P.

Kim

D. M.

(2003). The neurobiological consequences of early stress and childhood maltreatment. Neuroscience & Biobehavioral Reviews, 27, 33-44.

52.

Temkin

Rohe

(1996). Neighborhood change and urban policy. Journal of Planning Education and Research, 15, 159-170.

53.

U.S. Census Bureau. (2014). American Community Survey 2008–2012 (5-year summary file technical documentation). Retrieved from http://www2.census.gov/acs20135yr/summaryfile/ACS2013SFTechDoc.pdf

54.

Widom

C. S.

(2014). The 2013 Sutherland Address: Varieties of violent behavior. Criminology, 52, 313-344.