Child Behavior Problems and Maltreatment Exposure

Abstract

Download video file ^{(36.8MB, mp4)}

DOI: 10.1542/6349321513112

Video Abstract

OBJECTIVES

Establish the longitudinal cross-lagged associations between maltreatment exposure and child behavior problems to promote screening and the type and timing of interventions needed.

METHODS

The Longitudinal Studies of Child Abuse and Neglect, a multiwave prospective cohort study of maltreatment exposure, enrolled children and caregivers (N = 1354) at approximately age 4 and followed them throughout childhood and adolescence. Families completed 7 waves of data collection with each wave occurring 2 years apart. Maltreatment was confirmed using official case records obtained from Child Protective Services. Six-month frequencies of behavior problems were assessed via caregiver-report. Two random-intercept, cross-lagged panel models tested the directional relations between maltreatment exposure and externalizing and internalizing behaviors.

RESULTS

Maltreatment exposure predicted increases in externalizing behaviors at ages 8 (b = 1.06; 95% confidence interval [CI] 0.14–1.98), 12 (b = 1.09; 95% CI 0.08–2.09), and 16 (b = 1.67; 95% CI 0.30–3.05) as well as internalizing behaviors at ages 6 (b = 0.66; 95% CI 0.03–1.29), 12 (b = 1.25; 95% CI 0.33–2.17), and 14 (b = 1.92; 95% CI 0.76–2.91). Increases in externalizing behaviors predicted maltreatment exposure at age 12 (odds ratio 1.02; 95% CI 1.00–1.05).

CONCLUSIONS

Maltreatment exposure is robustly associated with subsequent child behavior problems, strengthening inferences about the directionality of these relations. Early screening of externalizing behaviors in pediatric settings can identify children likely to benefit from intervention to reduce such behaviors as well as prevent maltreatment exposure at entry to adolescence.

What’s Known on This Subject:

Child maltreatment is a public health concern and an established risk factor for pediatric health outcomes during both childhood and adolescence, including externalizing and internalizing behavior problems, which are among the most common concerns in pediatric settings.

What This Study Adds:

For the first time, this study examines the directionality of associations between maltreatment exposure and behavior problems throughout childhood and adolescence, informing the type and timing of interventions for reducing behavior problems and preventing maltreatment exposure.

Child maltreatment is an act of commission or omission on the part of a caregiver that results in harm or risk for harm toward an individual under the age of 18 years, including acts of physical abuse, sexual abuse, psychological abuse, and neglect.¹ Approximately 12.5% to 37.4% of all children in the United States are exposed to maltreatment before age 18 years.^2–4 Exposure to maltreatment is not only associated with greater risks for a number of adverse pediatric outcomes^5,6 but also lifelong economic, societal, and health care costs estimated at $2 trillion dollars.⁷

Child behavior problems, including noncompliance with directives, depressed mood, hyperactivity, worry, and delinquency, are early transdiagnostic indicators of adverse health,^8,9 and are among the most common complaints in pediatric settings.^10–12 The association between child maltreatment and child behavior problems has been established and replicated in independent and well-characterized research.^13,14 A common conclusion from this research is that maltreatment exposure results in increased frequencies or severities of child behavior problems.¹⁵ However, higher levels or changes in child behavior problems over time can result in future maltreatment exposure.¹⁶ For example, children exhibiting greater frequencies of hyperactive or delinquent behaviors may lead some caregivers to adopt more severe strategies that potentially indicate maltreatment, such as spanking or physical aggression, to reduce the frequency of those behaviors. Similarly, caregivers may resort to yelling or harsh criticism to alter persistent worrying or mood instability. Existing research has so far not established the directional, cross-lagged relations between maltreatment exposure and child behavior problems, limiting conclusions about the directionality of cause-effect relations among these variables during childhood and adolescence.

Establishing the directionality of cause-effect relations between maltreatment exposure and child behavior problems, and the ages at which these associations occur, has the potential to inform pediatricians about the need for early screening and the type and timing of interventions to reduce child behavior problems and the risk for maltreatment. Using 2 separate, 7-wave, random-intercept, cross-lagged panel models with 1354 children spanning ages 3 to 18 years, the current study examined whether maltreatment exposure increased the risk for greater frequencies of subsequent behavior problems and, simultaneously, whether child behavior problems increased the risk for subsequent maltreatment exposure. Results aim to advance our understanding of the directional nature of these events and their potential translational value for clinic settings.

Methods

Sample

Participants (N = 1354) were children and caregivers from the Longitudinal Studies on Child Abuse and Neglect (LONGSCAN), a multiwave and multisite prospective cohort study on the causes and consequences of child maltreatment. LONGSCAN data collection commenced in 1991 and participants were recruited from 5 geographic sites across the United States: East, Midwest, South, Southwest, and Northwest. Each LONGSCAN site received approval from their respective institutional review boards and the LONGSAN Data Coordinating Center. Caregivers provided consent and children assent, respectively. Children were recruited and enrolled at or before age 4 years (M = 4.56, SD = 0.70; 51.5% female) and followed every 2 years until age 18 years. Caregivers reported on children’s race and ethnicity (a social construct, not a genetic or biological category) at the age 4 assessment. The current study modeled data collected consecutively from the LONGSCAN age 4 through age 16 assessments, representing seven measurement occasions. The attrition rate from the age 4 to age 16 assessment was 33.9%. Sample demographics at the age 4 assessment are provided in Table 1. Race and ethnicity information are provided to gauge the representativeness of the LONGSCAN sample relative to the current child welfare population,⁴ and therefore the generalizability of research findings, but are not used in statistical analyses.

TABLE 1.

Sociodemographic Characteristics at the Age 4 LONGSCAN Assessment

	N (%) or Median
Sex
Female	697 (51.5)
Male	657 (48.5)
Annual family income	$10 000–$14 999
Child race
White	354 (26.2)
Black	721 (53.3)
Hispanic	97 (7.2)
Native American	8 (0.6)
Asian American	4 (0.3)
Mixed race	161 (11.9)
Other	8 (0.6)
Caregiver education
Without high school diploma	389 (44.8)
No posthigh school education or some college	751 (64.5)
Vocational certificate or associate’s degree	384 (33.0)
Bachelor’s degree or more	30 (2.6)

Child race N (%) reported out of n = 1353 LONGSCAN participants. Caregiver education N (%) reported out of n = 868 respondents (without high school diploma) and n = 1165 (no posthigh school education or some college, vocational certificate or associate’s degree, bachelor’s degree or more). The other designation in child race denotes any individual child whose race was not represented by 1 of the other listed options.

Measures

Confirmed Child Maltreatment

Confirmed child maltreatment was determined via the Modified Maltreatment Classification System (MMCS).¹⁷ The MMCS is an objective rating system wherein a team of reliably trained, independent coders provide ratings on the presence or absence of child maltreatment using information obtained from official case reports of child maltreatment investigations. The MMCS provides standardized and prespecified definitions of child maltreatment to reduce statewide discrepancies across the United States. In the current study, child maltreatment indicators (1 = maltreated, 0 = not maltreated) were used for 7 periods of time: Birth to age 4 years, age 4 years to age 6 years, age 6 years to age 8 years, age 8 years to age 10 years, age 10 years to age 12 years, age 12 years to age 14 years, and age 14 years to age 16 years. Table 2 presents the frequency of confirmed child maltreatment during each of these 7 periods.

TABLE 2.

Prevalence of Confirmed Child Maltreatment in LONGSCAN

Age Range	Frequency of Confirmed Child Maltreatment
Birth–age 4 y	576 (42.5)
Age 4 y–age 6 y	117 (8.6)
Age 6 y–age 8 y	86 (6.4)
Age 8 y–age 10 y	81 (6.0)
Age 10 y–age 12 y	66 (4.9)
Age 12 y–age 14 y	34 (2.5)
Age 14 y–age 16 y	39 (2.9)

Confirmed child maltreatment as assessed via the MMCS. Values represent N (%) out of N = 1354 LONGSCAN participants.

Child Behavior Problems

Child behavior problems were assessed via caregiver-report on the Child Behavior Checklist (CBCL),¹⁸ which captures broadband indices of child externalizing (eg, noncompliance) and internalizing (eg, depressed mood) behaviors. The CBCL is a well-established measure of child behavior problems, wherein caregivers report on the frequency of such behaviors in the previous 6 months. The CBCL was administered at the age 4, age 6, age 8, age 10, age 12, age 14, and age 16 LONGSCAN assessments. The current study used externalizing and internalizing behavior T-scores in statistical models, which are standardized for child age and sex. The reliability of the CBCL externalizing and internalizing behaviors scales across all 7 LONGSCAN measurement occasions was Cronbach’s α = 0.81 to 0.94. Table 3 presents sample descriptives on externalizing and internalizing behavior problems.

TABLE 3.

Externalizing and Internalizing Behavior Problems in LONGSCAN

	Child Maltreatment Group (N = 716)			Comparison Group (N = 638)
LONGSCAN Assessment	M (SD)	Range	Clinical Range N (%)	M (SD)	Range	Clinical Range N (%)
Externalizing Behaviors
Age 4 y	56.10 (10.87)	30–84	162 (25.00)	54.47 (10.41)	30–89	113 (19.76)
Age 6 y	56.38 (11.21)	30–86	171 (26.80)	54.21 (10.41)	30–84	113 (19.48)
Age 8 y	55.47 (11.63)	30–95	153 (25.46)	52.77 (10.94)	30–83	94 (17.97)
Age 10 y	54.01 (11.68)	30–85	115 (20.68)	51.03 (11.79)	30–93	69 (15.03)
Age 12 y	56.27 (11.28)	30–87	128 (25.81)	53.04 (11.27)	30–90	85 (18.68)
Age 14 y	56.73 (11.82)	32–92	145 (30.53)	52.55 (11.41)	32–84	79 (17.36)
Age 16 y	54.84 (12.09)	32–93	105 (22.58)	51.48 (11.72)	32–91	60 (14.93)
Internalizing Behaviors
Age 4 y	48.85 (9.82)	33–78	49 (7.56)	49.44 (8.94)	33–80	42 (7.34)
Age 6 y	50.93 (10.19)	33–81	79 (12.38)	50.98 (9.51)	33–85	57 (9.83)
Age 8 y	52.00 (11.02)	33–80	94 (15.64)	51.13 (10.26)	33–88	59 (11.28)
Age 10 y	51.26 (11.32)	33–85	91 (16.37)	49.53 (11.33)	33–90	56 (12.20)
Age 12 y	51.95 (11.17)	31–83	81 (16.33)	50.10 (10.76)	31–82	50 (10.99)
Age 14 y	52.17 (11.42)	31–85	85 (17.89)	48.35 (11.24)	31–79	49 (10.77)
Age 16 y	50.22 (12.31)	31–93	76 (16.34)	47.34 (10.95)	31–79	34 (8.46)

Child maltreatment group represents those with any confirmed child maltreatment between birth and age 16 years. Comparison group represents those with no confirmed child maltreatment between birth and age 16 years. Child externalizing and internalizing behavior problems assessed via caregiver-report on the CBCL. Presented values are T-scores. Clinical range represents T-scores >63. Values represent N (%) of those providing data at each LONGSCAN assessment within each respective group.

Statistical Analysis

Random-intercept, cross-lagged panel models (RI-CLPMs)^19–21 were estimated in Mplus, version 8.4,²² to test the cross-lagged associations between maltreatment exposure and child behavior problems during childhood and adolescence. The RI-CLPM is an advancement from the traditional cross-lagged panel model because within-person concurrent (eg, X at time 1 and Y at time 1), autoregressive (eg, Y at time 1 and Y at time 2), and cross-lagged (eg, X at time 1 predicting Y at time 2) associations can be estimated while simultaneously adjusting for between-person differences in the levels of child behavior problems and chronicity of maltreatment exposure between birth and age 16 years (ie, random intercept). Here, chronicity refers to the number of time points in this study where a child was exposed to maltreatment (range 0–7). Taken together, the within-person component of the RI-CLPM approach meets the primary aim of the current study: estimate the directional, cross-lagged effects for a given child when they are exposed to maltreatment or when their externalizing or internalizing behaviors are higher than their own average and while adjusting for other within-person features (eg, concurrent, autoregressive) and between-person differences.

Two separate RI-CLPMs estimated cross-lagged associations: one estimating the associations between maltreatment exposure and externalizing behavior problems and another to estimate the associations between maltreatment exposure and internalizing behavior problems. Models were estimated using a weighted-least squares estimator with mean and variance adjustment²³ and theta parameterization²² to account for maltreatment exposure as a binary outcome. Adequate model fit was determined from a comparative fit index (CFI) >0.90 and a root-mean square error of approximation (RMSEA) with a 95% confidence interval (CI) that covered, or had an upper bound below, 0.05.²⁴ Consistent with previous research,²⁰ autoregressive and concurrent associations were fixed across time for both maltreatment exposure and child behavior problems because they were not of focal interest (eg, cross-lagged parameters). As such, these parameters reflect the average within-person concurrent and autoregressive associations. Where behavior problems were the outcome, path coefficients are reported as unstandardized estimates interpreted as the change in T-scores when a given individual experienced confirmed child maltreatment. Where maltreatment exposure was the outcome, path coefficients were exponentiated and interpreted as odds ratios (ORs). Statistically significant associations were determined from P values <.05 where externalizing or internalizing behaviors were the outcome and from a CI that did not cover 1.0 where maltreatment exposure was the outcome. Complete data were available for confirmed child maltreatment via official case reports. Missing data on externalizing and internalizing behavior problems ranged from 9.90% to 35.97% across the 7 assessment time points. Missing data were addressed in each RI-CLPM using full information maximum likelihood.

Results

Externalizing Behaviors

Figure 1 depicts the results of the RI-CLPM between maltreatment exposure and child externalizing behavior problems. Fit indices suggest the model fit the data well (CFI 0.97; RMSEA 0.03; 95% CI 0.02–0.04). There was a significant between-person association between maltreatment exposure and externalizing behaviors (r = 0.29; P = .01), which represents that, on average, children with a greater chronicity of maltreatment exposure had greater frequencies of externalizing behaviors.

FIGURE 1.

RI-CLPM of confirmed child maltreatment and externalizing behaviors. All estimates are unstandardized. Where behavior problems were the outcome, path coefficients are reported as unstandardized estimates interpreted as the change in T-scores (M = 50; SD = 10) when a given individual experienced confirmed child maltreatment. Dashed lines are nonsignificant paths. Bolded lines are significant cross-lagged paths. Fit statistics: χ²⁸³ = 179.76, P < .001; CFI 0.97; RMSEA 0.03, (95% CI 0.02–0.04). P < .05 was considered statistically significant. CM, confirmed child maltreatment; EXT, externalizing behavior problems. ^a P ≤ .05; ^b P ≤ .01; ^c P ≤ .001.

At the within-person level, there were no significant concurrent associations between maltreatment exposure and externalizing behaviors. There were significant, autoregressive effects for both maltreatment exposure (OR 1.23; 95% CI 1.11–1.36) and externalizing behaviors (b = 0.59; 95% CI 0.54–0.64). After accounting for the stability in the autoregressive effects, there were 3 statistically significant, cross-lagged associations wherein exposure to child maltreatment predicted greater subsequent externalizing behaviors: from age 6 years to age 8 years (b = 1.06; 95% CI 0.14–1.98), from age 10 years to age 12 years (b = 1.09; 95% CI 0.08–2.09), and from age 14 years to age 16 years (b = 1.67; 95% CI 0.30–3.05). Each of these directional, cross-lagged associations indicate that, while accounting for both the chronicity of maltreatment exposure and previous levels of an individual’s externalizing behaviors, exposure to maltreatment during these periods uniquely predicted an elevated level of subsequent externalizing behaviors beyond what would otherwise be expected for a given individual. There was also one statistically significant, cross-lagged association wherein externalizing behavior problems predicted subsequent maltreatment exposure. Externalizing behaviors at age 10 years were predictive of maltreatment exposure at age 12 years (OR 1.02; 95% CI 1.00–1.05). This cross-lagged association indicates that, while accounting for differences in the chronicity of maltreatment exposure, higher than average externalizing behaviors for a child at age 10 years increased the odds of exposure to maltreatment between ages 10 and 12 years.

Internalizing Behaviors

Figure 2 depicts the results of the RI-CLPM between maltreatment exposure and child internalizing behaviors. This model also fit the data well (CFI 0.93; RMSEA 0.04; 95% CI 0.03–0.04). The between-person association between maltreatment exposure and internalizing behaviors was not statistically significant (r = 0.19; P = .09). This means that, in this sample, the chronicity of maltreatment exposure was not associated with children’s average level of internalizing behaviors.

FIGURE 2.

RI-CLPM of confirmed child maltreatment and internalizing behaviors. All estimates are unstandardized. Where behavior problems were the outcome, path coefficients are reported as unstandardized estimates interpreted as the change in T-scores (M = 50; SD = 10) when a given individual experienced confirmed child maltreatment. Dashed lines are nonsignificant paths. Bolded lines are significant cross-lagged paths. Fit statistics: χ²⁸³ = 235.40, P < .001; CFI 0.93; RMSEA 0.04, (95% CI 0.03–0.04). P < .05 was considered statistically significant. CM, confirmed child maltreatment; INT, internalizing behavior problems. ^a P ≤ .05; ^b P ≤ .01; ^c P ≤ .001.

At the within-person level, there was one significant concurrent association between maltreatment exposure and internalizing behaviors at age 4 years (σ² = −1.47). There were also significant autoregressive effects for both maltreatment exposure (OR 1.19; 95% CI 1.08–1.32) and internalizing behaviors (b = 0.58; 95% CI 0.54–0.63). After accounting for this within-person stability, there were 3 significant, cross-lagged associations wherein exposure to maltreatment predicted more frequent internalizing behaviors: from age 4 years to age 6 years (b = 0.66; 95% CI 0.03–1.29), from age 10 years to age 12 years (b = 1.25; 95% CI 0.33–2.17), and from age 12 years to age 14 years (b = 1.92; 95% CI 0.76–2.91). There were no significant within-person, cross-lagged associations wherein internalizing behaviors predicted subsequent maltreatment exposure.

Discussion

Child maltreatment and child behavior problems are highly prevalent concerns relevant to the day-to-day operations of pediatricians. Extensive research has demonstrated a well-established relation between maltreatment exposure and child behavior problems.¹⁵ However, is it maltreatment exposure that leads to subsequent behavior problems or is it child behavior problems that lead to subsequent maltreatment? Answering this question is important for establishing the directionality of cause-effect relations between maltreatment exposure and child behavior problems and has translational implications for identifying which event or outcome is more important to screen or target with intervention. The current study provided an opportunity to answer this question and identify the directionality of associations between maltreatment exposure and child behavior problems given a unique sample, the prospective and longitudinal design incorporating 7 repeated measurements of maltreatment and child behavior problems throughout childhood and adolescence, and a statistical modeling approach that generates accurate directional, cross-lagged effects of key variables of interest.²⁵

Maltreatment exposure during childhood and adolescence demonstrated robust, cross-lagged relations with subsequent externalizing and internalizing behaviors after adjusting for key within- (eg, autoregressive) and between-person (eg, chronicity of maltreatment) components. After controlling for differences in the chronicity of child maltreatment and levels of child behavior problems, an individual’s exposure to maltreatment during childhood and adolescence was associated with subsequent increases in both externalizing and internalizing behavior problems. This provides important evidence that maltreatment directionally affects changes in externalizing and internalizing behaviors, a critical component for promoting causal inference about how maltreatment and behavior problems are related.²⁶ For example, a wealth of child maltreatment research has examined the relations between maltreatment exposure and child behavior problems at different ages during childhood and adolescence.^13,14,27 However, simultaneous tests of whether child behavior problems are also directionally related to child maltreatment, a key test to rule out reverse causality,^28,29 are lacking and, to our knowledge, no previous study has examined each of these cross-lagged relations within the age range of this study.

There was one occasion where child behavior problems, specifically externalizing behaviors at the entry to adolescence, were directionally associated with subsequent exposure to child maltreatment. To our knowledge, this is the first time in which externalizing behaviors demonstrated directional effects on the risk for subsequent maltreatment exposure in adolescence after adjusting for both between- and within-person effects. Maltreatment exposure was directionally associated with externalizing behaviors during this same developmental window, suggesting a bidirectional relation where both maltreatment exposure and externalizing behaviors are simultaneously related to one another from ages 10 to 12 years and highlighting a specific age range where screening and intervention hold considerable promise. This bidirectional association does not detract from the robust directional associations observed for child maltreatment given the number of significant associations and the small effect size observed for externalizing behaviors (OR = 1.02). However, it does add specificity to the current set of results in that higher levels of externalizing behaviors at the transition to adolescence also increased the risk for subsequent maltreatment exposure.

The major findings of this study have important practice implications for the type and timing of interventions for addressing maltreatment exposure and child behavior problems. One, screening and preventing child maltreatment continues to be a significant public health priority to reduce the onset of many adverse pediatric health outcomes, including child behavior problems. Universal, selective, and indicated forms of child maltreatment prevention exist, are widely available, and are even being integrated in pediatric health settings.^30–33 Increased uptake of these existing interventions holds considerable promise in preventing new and repeated instances of child maltreatment. Two, screening for elevations in externalizing behaviors at the transition to adolescence can identify potential interventions to reduce these behaviors. This suggestion is consistent with recent calls for the screening of behavioral concerns in pediatric settings.³⁴ Well-established behavioral interventions that directly address child externalizing behaviors in the child maltreatment population exist,³⁵ including Parent–Child Interaction Therapy,³⁶ which targets caregivers and children up to age 10 years with demonstrated efficacy for reducing externalizing behaviors and preventing future child maltreatment.^37,38 Alternatives for Families: A Cognitive-Behavioral Therapy³⁹ is also an established and widely-available intervention reducing externalizing behaviors for children and adolescents at risk for and exposed to maltreatment.⁴⁰ Screening of externalizing behaviors allows pediatricians to have a discussion with families about the need and value of such programs. Of course, externalizing behaviors are not the only risk factor for child maltreatment, and other established maltreatment risk factors, including caregiver drug and alcohol use, mental health concerns, family violence, and poverty,^4,41 should still be considered.

There are important limitations to this study. One, different dimensions of maltreatment exposure, such as timing and chronicity,¹⁷ have been raised as important features to examine in research on the effects of child maltreatment.⁴² The RI-CLPM approach allows for more accurate estimation of the unique effects of maltreatment exposure on child behavior problems, at specific ages, because it can account for differences among children who may have been exposed to more maltreatment relative to another child with a different maltreatment history. Although the RI-CLPMs in the current study accounted for the timing and chronicity of maltreatment exposure, they did not include other dimensions of maltreatment exposure, such as severity, that may further explain results or improve effect size magnitudes for behavior problems across child and adolescent development. Two, and related, the current study did not examine directional relations among different types of child maltreatment and resulting child behaviors. Therefore, conclusions are limited to broader categorizations of maltreatment exposure. The decision to model the cross-lagged associations for the broad category of child maltreatment, instead of individual types, was deliberate and made on the basis of evidence demonstrating that multiple types of child maltreatment are similarly associated with more frequent child behavior problems.^43–45 Different types of child maltreatment also cooccur,^46–48 making it difficult to parse the unique effects of one type of maltreatment on child behavior problems relative to another type. Finally, the current study measured child behavior problems via caregiver-report to ensure consistent measurement of the outcomes across all measurement occasions, a range of time where these same outcomes were not assessed by other reporters, such as children. However, caregivers and children may report behavior problems differently⁴⁹ and future studies may wish to use a multiinformant approach when possible.

Conclusions

Establishing the directional associations between maltreatment exposure and behavior problems across childhood and adolescence has important scientific and practice implications. The results suggest that exposure to child maltreatment during childhood and adolescence leads to subsequent increases in both externalizing and internalizing behaviors. Our results also suggest that, at the transition to adolescence, elevated levels of externalizing behaviors may put children at risk for future child maltreatment. The transition to adolescence may represent an important period for child maltreatment prevention and the targeting of child externalizing behavior problems. Screening for maltreatment exposure and externalizing behaviors can assist pediatricians and families given the available and effective behavioral interventions for reducing the risk for maltreatment and child behavior problems.

Glossary

CBCL

Child Behavior Checklist

CFI

comparative fit index

CI

confidence interval

LONGSCAN

Longitudinal Studies on Child Abuse and Neglect

MMCS

Modified Maltreatment Classification System

OR

odds ratio

RI-CLPM

random-intercept, cross-lagged panel model

RMSEA

Root-Mean Square Error of Approximation