March 27, 2006

Aggression and the Family

Ovid: MILLER: J Am Acad Child Adolesc Psychiatry, Volume 45(3).March 2006.355-363

Family and Cognitive Factors: Modeling Risk for Aggression in Children With ADHD

Drs. Carlin Miller and Halperin are with the Department of Psychology, Queens College, City University of New York, Flushing; Drs. Scott Miller, Newcorn, and Halperin are with the Division of Child and Adolescent Psychiatry, Mount Sinai School of Medicine, New York; Dr. Halperin and Mr. Trampush are with the Neuropsychology Doctoral Program, Graduate School and University Center, City University of New York, New York; and Dr. McKay is with the Court Assessment Program, Westchester Jewish Community Services, Hartsdale, NY.
Accepted October 11, 2005.
This research was supported by National Institute of Mental Health grants RO1MH46448 and RO1MH60698, and The William T. Grant Foundation’s Faculty Scholar’s Award Program.
Correspondence to Dr. Carlin J. Miller, Department of Psychology, Queens College, City University of New York, 65-30 Kissena Boulevard, Flushing, NY 11367; e-mail:
Disclosure: Dr. Newcorn has had consulting/advisory relationships with Eli Lilly, McNeil, Novartis, Shire, UCB, Bristol-Myers Squibb, Cephalon, and Sanofi-Aventis; has received research support from Eli Lilly, McNeil, and Shire; and serves on speaker boards for Eli Lilly, McNeil, and Novartis.Dr. Halperin has consulted for Shire. The other authors have no financial relationships to disclose.

Objective: To explore the relationships of family and cognitive factors to aggression as reported by parents and teachers.

Method: Data regarding different types of aggressive behavior were collected from parents and teachers of 165 school-age (7-11 years old) children referred to a study of attention-deficit/hyperactivity disorder and disruptive behavior. Structural equation modeling was used to examine the differential relationships of family and cognitive factors to aggression in the home and school settings.

Results: Family risk factors influenced aggression reported at home and at school, whereas cognitive risk factors may exert their influence on aggression at school. Significant covariance between parent and teacher report of aggressive behavior was observed.

Conclusions: Intervention at the family level may serve to reduce aggressive behaviors in both home and school environments. J. Am. Acad. Child Adolesc. Psychiatry, 2006;45(3):355-363.

Aggressive behavior is among the most stable personality traits across the life span, such that aggression in childhood often predicts similar and perhaps escalating behavior over the life course (Farrington, 1991; Haapasalo and Tremblay, 1994; Loeber and Dishion, 1983; Loeber and Hay, 1997), resulting in peer rejection (Bagwell et al., 2001; Loeber et al., 1993), antisocial personality disorder (Cadoret and Cain, 1981; Stewart and Leone, 1978), domestic abuse (Fuller et al., 2003; Kolvin et al., 1988), and incarceration (Olweus, 1979; Schaeffer et al., 2003). Nevertheless, many aggressive children do not become aggressive adults.

A confluence of environmental and genetic factors is likely to affect the manifestation of aggression in children and the degree to which it persists/escalates with development. Severity of aggression appears to be the most important behavioral predictor of outcome. However, whether the behavior is situation specific or pervasive also has prognostic value. Thus, children who are aggressive at home or at school are likely to have a better outcome than those who are aggressive in both settings (Loeber and Hay, 1997; Loeber et al., 1993).

Considerable data indicate that aggression runs in families, and that both genetic and environmental factors play a role in the emergence and persistence of this maladaptive behavior in children. Twin studies consistently report higher concordance rates for aggression among monozygotic compared with dizygotic twins (Button et al., 2004; Hudziak et al., 2003), with heritability estimates for aggression ranging from 0.28 to 0.72. Yet, genes account for only a portion of the variance in aggression. A host of adverse environmental factors, including poverty, low socioeconomic status, marital discord, harsh parenting, poor supervision, parental psychopathology, and crowded living conditions are linked to the emergence of aggressive behavior in children and adolescents (Cadoret and Cain, 1981; Fergusson et al., 1996; Haapasalo and Tremblay, 1994; Vaden-Kiernan et al., 1995). It has been hypothesized that having limited financial and human resources increases risk because children are less likely to be closely supervised and the household is less structured by rules (Griffin et al., 2000; Haapasalo and Tremblay, 1994). Nevertheless, in naturalistic studies involving intact family units, it is difficult to parse out genetic and environmental effects. For example, studies of aggressive boys consistently indicate similar behaviors in their fathers (Pfiffner et al., 1999; Stewart and deBlois, 1983). Although these aggressive behaviors may represent the impact of genetic factors, it is likely that modeling and evocative effects are also involved. This hypothesis is supported by evidence indicating that the longer an aggressive or personality-disordered father is not in the home, the less effect his behavior has on the behavior of children in the home (Stewart and deBlois, 1983). Furthermore, when the genetic contribution of a father’s antisocial behavior is controlled, the antisocial parent’s behavior in the environment still affects the child’s aggressive behaviors (Jaffee et al., 2003). These findings are consistent with emerging data indicating interactive effects of genes and environment on the emergence of aggression in youths. Both adoption (Riggins-Caspers et al., 2003) and molecular genetic (Caspi et al., 2002) studies indicate greater susceptibility to adverse environmental factors in genetically vulnerable individuals.

In addition to home and family factors, a substantial literature indicates that groups of aggressive children (Coy et al., 2001; Vance et al., 2002) and adults (Brownlie et al., 2004; Huesmann et al., 2002) are characterized by cognitive deficits, primarily in the verbal domain. Thus, it is possible that childhood cognitive and learning problems lead to aggressive behaviors; however, findings have not been conclusive. Patterson and colleagues (1989) hypothesized that academic failure and rejection by normal peers in the presence of cognitive problems leads to a deviant peer group and further delinquency. The presence of inattentive behaviors, as seen in youth with attention-deficit/hyperactivitydisorder (ADHD), also increases the likelihood of aggressive behaviors, especially when inattention occurs together with low cognitive ability (Bellanti et al., 2000; Schaeffer et al., 2003). The specific relationship between learning problems and the development of aggression is also unclear. One study found that low intelligence and attention problems but not learning problems predicted delinquency (Manguin and Loeber, 1996). However, other studies have indicated that learning problems, particularly reading problems, increase the risk of aggressive behavior, especially in individuals with aggression problems that predate learning issues (Cornwall and Bawden, 1992; Vance et al., 2002).

Despite the fact that aggression reported by parents and teachers is different, there is little information regarding differential correlates of parent and teacher reports. Most studies examining both parent and teacher data have focused primarily on ADHD and/or oppositional-defiant behavior, rather than aggression. Furthermore, it is well known that aggression is multifactorial in nature, and it is not known how this may translate into reports from different informants. One approach to managing this problem is by defining and examining the correlates of latent factors that can better account for multiple components of complex constructs. The present study uses structural equation modeling of latent traits to examine the extent to which demographic factors and cognitive problems differentially correlate with aggression at home and school. It was hypothesized that family risk factors would predict aggression as reported by the parent and the teacher. In contrast, cognitive risk factors were hypothesized to predict aggression at school, where they are likely to engender greater distress, but not at home (Fig. 1).Consistent with past findings from Halperin and colleagues (2002, 2003), the correlation between home and school aggression was predicted to be statistically significant, but relatively modest in magnitude.

[Help with image viewing]
[Email Jumpstart To Image] Fig. 1 Model of hypothesized data relationships.

Participants were 163 boys (n = 141) and girls (n = 22), ages 7-11 years with ADHD and/or other disruptive behavior disorders, who participated in a research protocol investigating aggressive behaviors in individuals with ADHD. The mean age of the sample was 9.06 years (SD = 1.29 years), and all had a Verbal or Performance IQ score of at least 70. The sample was ethnically diverse: 40.9% were Latino, 29.6% were African American, 18.9% were white, and 10.7% were of mixed or other ancestry. All participants and their parents were English speaking. The majority lived in an urban environment within a major metropolitan area. Most were from lower middle-class homes (mean = 30.84; SD = 13.89), although there was a range of socioeconomic status in the sample; Hollingshead Index (Hollingshead,1975) scores ranged from 11 to 66. Referral was through local schools and medical care providers. Subjects were not paid for their participation and did not receive any intervention as part of the protocol. Approval from the medical center’s institutional review board was attained before the study began and informed consent was gathered from parents before participants were enrolled.

Participants were diagnosed for Axis I disorders using the Diagnostic Interview Schedule for Children (DISC), version 2.1 (Shaffer et al., 1989) or 2.3 (Shaffer, 1997) in concert with other measures. Trained graduate students administered the DISC to parents. Additional information was gathered through parent and teacher rating scales as well as clinical observations. Overall, 152 participants were diagnosed with ADHD. The remaining 13 had significant subsyndromal symptoms associated with ADHD. In addition to ADHD, 27.9% of the children met criteria for oppositional defiant disorder, 19.4% met criteria for conduct disorder, 12.6% met criteria for mood disorders, and 35.1% met criteria for anxiety disorders. Also, 18.8% of participants met criteria for two or more comorbid disorders in addition to ADHD.
Family Demographics Information.

Parents, most often mothers, were interviewed by trained graduate studentsabout the family environment and family risk factors. For the family risk factors latent construct, parents were asked about the number of siblings in the home (SIBS), number of parenting adults in the home (HOME), and whether the biological parents of the target child had a history of aggressive behavior (FHX). For the SIBS and the HOME variables, the reported number was entered in the database. The number of adults was typically one or two, but on occasion a third relative (e.g., grandparent) also served in this capacity. For the FHX variable, the number of biological parents with a history of aggression for each target child was entered.

The interaction variable (RISK) combined the biological parent’s history for aggressive behavior with whether the parent was living at home. In accordance with previous research (Stewart and deBlois, 1983; Stewart and Leone, 1978), having aggressive parents living in the home with the child was hypothesized as the highest risk and having nonaggressive parents in the home was seen as the lowest risk of child aggression. For the RISK variable, possible values ranged from 1 to 4 as follows: children with both parents living in the home, neither of whom had a history of aggression = 1; children with only one parent at home, neither having a history of aggression = 2; children with either one or two parents with a history of aggression but the aggressive parent was not living in the home = 3; children with either one or two parents with a history of aggression and an aggressive parent living in the home = 4.
WISC, Revised or Third Edition (WISC-R/WISC-III).

The Full Scale IQ score from the WISC-R (Wechsler,1974) or WISC-III (Wechsler,1992) was used as an indicator of general cognitive functioning/impairment and was included in the cognitive risk factors latent construct. The version of the WISC that was administered to the participants was determined by when they entered the study. The WISC series of tests are frequently used intelligence tests with acceptable psychometric properties. As would be expected by the Flynn effect (Flynn, 1999), scores from later published tests are often lower than those generated by previous versions. In light of the fact that Full Scale IQ scores generated by the WISC-III are on average 5.9 points lower than those from the older WISC-R (Wechsler, 1992), all WISC-III scores were adjusted upward 5.9 points, as suggested by the WISC-III manual.
Reading and Math Achievement.

The reading and math subtests from Wide Range Achievement Test, Revised Edition (Jastak and Wilkinson, 1984) or the Wechsler Individual Achievement Test (Wechsler,1992) were administered to each participant. The Wide Range Achievement Test and Wechsler Individual Achievement Test are frequently used in clinical settings, with considerable evidence of their reliability and validity. The standard score for word reading was used as an indicator of general reading achievement (READ) and the standard score for math calculation was used as an indicator of general mathematical achievement (MATH). Both MATH and READ were included in the cognitive factors latent construct. Administration of the particular achievement test was determined by when the participant entered the research protocol. In contrast to the IQ scores, there were no significant differences (p > .10) in academic achievement scores as a function of whether the child was administered the Wide Range Achievement Test or Wechsler Individual Achievement Test.
Children’s Aggression Scale-Parent and Teacher Versions (CAS-P, CAS-T).

The CAS-P and CAS-T, respectively, are scales representing five domains of aggressive behavior, four of which were used in this study: verbal aggression (VERB), aggression against objects and animals (OBJ), provoked physical aggression (PROV), and unprovoked/initiated physical aggression (INIT). The use of weapons subscale was eliminated from the data analyses because it demonstrated limited discriminability effectiveness in this age range (Halperin et al., 2002, 2003). Psychometric data for the parent and teacher versions of the CAS have been published elsewhere and indicate good reliability and validity. Compared to other “aggression” rating scales, scores are less confounded by factors related to ADHD and oppositional-defiant behaviors (Halperin et al., 2002, 2003). Information from the CAS-P was entered into the model to indicate aggression in the home and nearby environments. CAS-T scores were used to indicate aggression displayed in the classroom and other school environments. Previous studies of the CAS-P and CAS-T (Halperin et al., 2002, 2003) suggest statistically significant but modest intercorrelations.
Statistical Methodology

Descriptive statistics and correlations for all indicator variables were calculated using SPSS 10.0 (SPSS Inc., 1999). LISREL 8.54 (Joreskog and Sorbom, 2003) was used to test the main hypotheses. This version of LISREL contains the full information maximum likelihood feature (FiML), which allows the use of all available data to estimate model parameters. Monte Carlo studies using FiML have found reliable parameter estimates when as much as 25% of the data is missing (Enders and Bandalos, 2001). The percentage of missing values in the present study was 7.28%. Two goodness-of-fit statistics were used in the present study: a FiML [chi]2 value and the root mean square error of approximation (RMSEA). The [chi]2 statistic tests the closeness of fit between the hypothesized model and the model implied by the data. The RMSEA represents the discrepancy per degree of freedom between the population data and the proposed model (Byrne, 1998). RMSEA values up to 0.06 (Hu and Bentler, 1999) have been used to indicate adequate fit.

A common structural equation modeling strategy for assessing the relative goodness of fit of competing models is the use of nested models. This approach begins with testing the fit of a baseline model, which provides a baseline [chi]2 value to which the [chi]2 values of subsequent models are compared. In the present study, the baseline model freely estimated all hypothesized paths with high levels of family risk factors predicting high levels of parent-reported aggression at home. Low levels of cognitive resources were hypothesized to predict high levels of aggression as reported by teachers. The path from family factors to school aggression was also hypothesized to be significant. The path from cognitive factors to home aggression was estimated but was not hypothesized to be significant. The covariance of home and school aggression, expected to be significant, was also estimated.

Descriptive statistics for all indicator variables are summarized in Table 1. Slightly more than 30% of participants lived in homes with one adult. Slightly more than 50% lived in homes with two adults, although in many cases, adults in the home were not necessarily the biological parents. Most participants had at least one sibling living with them at home, and nearly 14% had three or more siblings in the home with them. More than 35% of the participants had at least one biological parent with a history of aggressive behavior and nearly 13% had two biological parents with a history of aggressive behavior. Slightly more than 13% of participants were living with a parent who had a history of aggressive behavior, and slightly more than 25% living with biological parents with no history of aggressive behavior.

[Help with image viewing]
[Email Jumpstart To Image] TABLE 1 Descriptive Information for Indicator Variables

The mean level of intellectual functioning was in the low average range. Approximately one fourth of participants were functioning significantly below the average range of intellectual ability (i.e., Full Scale IQ .10) relative to the baseline model indicated that modeling cognitive factors as equally predictive of home and school aggression showed only slightly poorer fit to the data. Because the [chi]2 increase was not significant, the alternate hypothesis of cognitive factors predicting both home and school aggression ([beta] = -.11, t = -1.91, p


Assessing the effectiveness of treatment

Understanding the Effect Size of ADHD Medications

Create a free website or blog at