Relation of Age of Onset to the Type and Severity of Child and Adolescent Conduct Problems

Relation of Age of Onset to the Type and Severity of Child and Adolescent Conduct Problems

Benjamin B. Lahey

Benjamin B. Lahey [1,8]

Sherryl H. Goodman [2]

Irwin D. Waldman [2]

Hector Bird [3]

Glorisa Canino [4]

Peter Jensen [5]

Darrel Regier [5]

Philip J. Leaf [6]

Rachel Gordon [1]

Brooks Applegate [7]

Received August 7, 1998; revision received March 22, 1999; accepted March 22, 1999

In a cross-sectional household sample of 9-through 17-year-old youths from 4 U.S. communities, youths with earlier ages of onset of conduct problems engaged in more conduct problems than youths with later ages of onset when current age and gender were controlled. Specifically, youths with earlier ages of onset were more likely to engage in several types of physical aggression, frequent lying, theft, and vandalism and were less likely to engage in only truancy. There also was an inverse relation between age of onset and level of functional impairment, mental health service use, and meeting diagnostic criteria for conduct disorder, attention-deficit hyperactivity disorder, and oppositional defiant disorder. Within the limits of cross-sectional data, these results support the hypothesis that key aspects of the heterogeneity of conduct problems among youths are related to the age of onset of conduct problems.

KEY WORDS: Conduct problems; age of onset; conduct disorder; attention-deficit hyperactivity disorder; oppositional defiant disorder.

Numerous researchers have reported a robust inverse relation between the age of a youth’s first conviction and his or her total number of convictions through early adulthood (Blumstein, Farrington, & Moitra, 1985; Glueck & Glueck, 1959; Loeber, 1982; Wadsworth, 1979; Wolfgang, Figlio, & Stellin, 1972). Youths who are first convicted earlier are convicted more times not only because they began their “criminal careers” earlier but also because they are convicted at higher rates at all ages into early adulthood. It is important to note that the same inverse association has been found between age of onset and self-reported delinquent behavior in several community samples. This is important, as self-reports of delinquency avoid the biases in detection, prosecution, and conviction that are inherent in official statistics (Loeber, 1987). Among 11- through 18-year-old boys who had engaged in any delinquent behavior, Tolan (1987) found that the half of the sample with younger ages of onset ([less than]12 years) reported higher levels of almost all types of delinquent behaviors during adolescence than the half of the sample with later ages of onset. Similarly, in a subset of female and male youths from the longitudinal National Youth Survey (Elliott, Huizinga, & Menard, 1988), Tolan and Thomas (1995) found that youths who reported first engaging in delinquent acts before age 12 were more likely to engage in serious offenses and to continue to engage in delinquent behavior during the 3 years following the onset of delinquent behavior. These differences were somewhat stronger for girls than boys, but were similar across genders.

These findings of an inverse relation between age of onset and the frequency, seriousness, and persistence of delinquency have led several theorists to distinguish two or more distinct “developmental pathways” of delinquent behavior (Hinshaw, Lahey, & Hart, 1993; Loeber, 1988; Moffitt, 1993). According to these developmental models, youths who engage in the most frequent, aggressive, and persistent delinquent behavior begin doing so during childhood. In contrast, youths who do not engage in delinquent behavior until adolescence are less likely to be aggressive, engage in fewer delinquent behaviors, and tend to desist prior to adulthood. Moffitt (1993) coined the terms “adolescent-limited” and “life-course persistent” delinquency for these two groups of youths. She hypothesized that youths who first engage in antisocial behavior during childhood do so for different reasons than youths who first engage in antisocial behavior during adolescence. Specifically, childhood-onset conduct problems result from early n europsychological deficits that cause cognitive delays, impulsivity, and difficult temperament. In the presence of adverse childrearing environments, these characteristics contribute to the origins of conduct problems. In contrast, the adolescent-onset group does not have predisposing neuropsychological dysfunction. Their delinquent behavior arises through the imitation of some of the nonaggressive antisocial behaviors of youths with childhood onsets. They do so during adolescence because it is a period of heightened peer influence and conflict regarding adult privileges.

Moffitt’s (1993) developmental model is based in part on her analysis of prospective data on boys from the Dunedin Study (McGee, Feehan, Williams, & Anderson, 1992). Boys who engaged in delinquent behavior at age 15 and who also met diagnostic criteria for DSM-III attention-deficit disorder (ADD; n = 19) showed relatively persistent conduct problems from the preschool period to the time of the final assessment. In contrast, boys who were delinquent at age 15, but who did not meet criteria for childhood ADD (n = 52), were relatively free of behavior problems until age 13. Delinquent boys with ADD had lower verbal intelligence and achievement scores and higher family adversity scores than other groups (Moffitt, 1990). The neuropsychological diathesis for delinquency in the early-onset group is seen in their ADD and cognitive deficits, and their unfavorable environments are seen in the their higher family adversity scores.

In another analysis of data on boys from the same study, Moffitt, Caspi, Dickson, Silva, and Stanton (1996) divided the sample into groups based on parent and teacher ratings of conduct problems at ages 5, 7, 9, and 11 years and self-reported delinquency at ages 15 and 18. Boys were assigned to the “life-course persistent” group based on delinquency scores that were at least 1 standard deviation above the mean on both parent and teacher ratings during most childhood assessments and at least 1 standard deviation above the mean on self-reported delinquency at either age 15 or age 18 (n = 32). Boys who met the delinquency criterion during adolescence but not during childhood were classed as adolescent-limited (n = 108). As a group, parent and teacher ratings of conduct problems for the adolescent-limited boys were at the mean for the sample during ages 5 through 11 years. The life-course persistent group was rated by parents as having more difficult temperaments at age 3 and had been convicted significantly mor e times for violent crimes through age 18, but did not have more total convictions than the adolescent-limited group. The life-course persistent group also reported feeling more distant from their family and more of them had dropped out of school, but they did not differ significantly from the adolescent-limited group on friendships with delinquent peers, unemployment, dangerous driving habits, or substance use during adolescence.

Based on similar considerations, two subtypes of conduct disorder (CD) were distinguished in DSM-IV based on the age of onset of the first CD behavior before age 10 (childhood onset) or 10 years of age or later (adolescent onset). In the DSM-IV field trials (Lahey et al., 1998), there was a nonlinear inverse relation between age of onset and the number of aggressive behaviors, with a sharp decline in aggression occurring around an age of onset of 10 years among youths who met criteria for CD, but age of onset was not significantly related to the number of nonaggressive behaviors. In addition, youths who met criteria for adolescent-onset CD were less likely to meet DSM-III-R criteria for oppositional defiant disorder (ODD), less likely to have a family history of antisocial behavior, and had a gender ratio closer to 1:1 than the childhood-onset group (which was predominantly male). The inverse curvilinear relation between age of onset of CD and aggression was replicated among the 53 boys and 21 girls who met criteria for CD in the MECA study household sample (Lahey et al., 1996), but age of onset was not significantly related to a concurrent diagnosis of ODD, parental antisocial behavior, or gender ratio in that sample.

Thus, the notion that there is an inverse relation between age of onset of antisocial behavior and the severity and persistence of antisocial behavior has had a major impact on theories of delinquent behavior and the taxonomy of CD. For many reasons, however, it would be premature to consider the relation between age of onset and conduct problems to be a closed question. First, the largest previous studies used official conviction records to define age of onset and persistence, whereas studies that used less biased parent, teacher, and youth reports of antisocial behavior had samples of more limited size. Due to their limited sample sizes, previous studies that used self-report measures of antisocial behavior may have reinforced an inaccurate view that there is a qualitative dichotomy between childhood-onset and adolescent-onset conduct problems by splitting their samples into two groups based on age of onset. This is potentially unfortunate as studies of the inverse relation between age of first conviction and the total number of convictions suggest that the relation is linear, without a marked (qualitative) distinction between childhood- and adolescent-onset groups (Loeber, 1982). It would be important, therefore, to examine age of onset using samples that are large enough to provide at least preliminary information on whether two distinct groups exist or whether there is simply a continuum of differences related to age of onset.

Second, except for Tolan and Thomas (1995), all previous studies of the relation between age of onset and conduct problems included only boys, or analyzed data on only boys even though girls were available in the sample. Because there is inconsistent evidence that the age of onset of conduct problems may tend to be later among girls than boys (Lahey et al., 1998; McGee et al., 1992), it is of great importance to compare the relation between age of onset and conduct problems in both girls and boys using the same methods. Third, only Tolan (1987) provided information on differences in rates of specific delinquent behaviors related to age of onset, beyond the simple distinction between violent and nonviolent crimes made in most studies. A detailed account of differences in specific conduct problems related to age of onset is essential if researchers are to fully describe and understand the ontogeny of conduct problems. Fourth, only Moffitt (1990) and Lahey et al. (1998) have related age of onset to other aspects of psychopathology, such as attention deficit/hyperactivity disorder (ADHD) and ODD.

The present study used data from the MECA study (Lahey et al., 1996) to examine the relation between age of onset and conduct problems. Unlike previous analyses of age of onset that focused on only the 74 youths in this sample who met criteria for CD (Lahey et al., 1998), we studied the correlates of age of onset among the much larger group of youths who were reported to exhibit one or more DSM-III-R CD behaviors. By using the number of conduct problems (CD behaviors) as the dependent variable, we placed our understanding of the diagnosis of CD in the larger context of the development of conduct problems. Because CD behaviors and delinquent acts refer to a similar set of behaviors, but are defined in somewhat different ways, the present study complements previous findings that used different definitions of conduct problems.

METHOD

The present analyses are based on the sample for the MECA study that was drawn from four communities in the United States selected on the basis of their socioeconomic, cultural, and ethnic diversity: (a) Hamden, East Haven, and West Haven, Connecticut; (b) DeKalb, Rockdale, and Henry counties, Georgia; (c) Westchester County, New York; and (d) San Juan, Puerto Rico (Lahey et al., 1996). The target population for the MECA survey included all youths aged 9 through 17 years residing in a housing unit within the four areas at the time of survey enumeration. Potential participants were excluded from the survey if their primary language was not English or Spanish. Youths were required to have resided at least half-time in their current household during at least the past 6 months so that the adult caretaker would be knowledgeable about the child’s behavior. If more than one eligible youth resided in the sample housing unit, one youth was randomly selected. If the youth’s mother (biological, step, or adoptive) had been living with the selected youth during at least the past 6 months, she was selected as the adult respondent. If not, the adult respondent was selected from among those present in the household for the past 6 months in a predetermined order. The interviewed adult caretakers were almost always parents (90% biological mothers, 3% bio-logical fathers, and 3% adoptive or stepmothers).

The response rate for the screen for eligibility exceeded 99%, and 84.4% of eligible pairs of youths and adult caretakers were interviewed (n = 1,285). The MECA sample cannot be said to be representative of all youths living in households in the United States, but the sampling procedures selected nearly equal numbers of youths at each age between 9 and 17 years and nearly equal numbers of girls and boys. Among the three mainland sites, the proportion of non-Hispanic White youths in the sample (68%) was somewhat lower, and the proportion of African American youths (20%) was somewhat higher than 1992 Bureau of the Census estimates. This resulted from deliberately selecting geographic areas where substantial numbers of ethnic minority youths resided to ensure diversity. Median family incomes of interviewed youths in all four regions were somewhat higher than for all families living in those regions, partly due to oversampling higher income African American families to help disentangle race –ethnicity and income (Lahey et al., 1996).

Measures

Version 2.3 of the National Institute of Mental Health Diagnostic Interview Scale for Children Shaffer et al., 1996) was used to obtain information on CD, ADHD, ODD, and other disorders not covered in these analyses. The DISC-2.3 interviews were administered to the youth and her or his parent in separate interviews by trained lay interviewers. Symptoms and diagnoses were generated from the interviews by computer algorithms. One-week test-retest reliability of the DISC-2.3 for the number of DSM-III-R CD behaviors was high for both parents and youths (intraclass correlations = .93 and .92, respectively) in the present sample (Shaffer et al., 1996). Test-retest kappa coefficients for the diagnosis of DSM-III-R CD were moderate (kappa = .56 for the parent and .64 for the youth; Schwab-Stone et al., 1996).

Based on previous studies showing that youth reports of symptoms of ADHD and ODD do not contribute to the validity of these diagnoses (Bird, Gould, & Staghezza, 1992; Hart, Lahey, Loeber, & Hanson, 1994), diagnoses of ADHD and ODD were based on parent reports as in previous studies (Lahey et al., 1998). One-week test-retest reliability of the DISC-2.3 for the DSM-III-R diagnoses of ADHD and ODD were moderate (kappa = .60 and .68, respectively) in the present sample (Schwab-Stone et al., 1996). Evidence of functional impairment was not required for diagnoses in these analyses to facilitate comparisons between numbers of behaviors and diagnoses, but the prevalences of the diagnoses of ADHD, ODD, and CD are influenced very little by the use of impairment criteria in this sample (Shaffer et al., 1996). Because the DISC-2.3 did not query some DSM-IV symptoms of ADHD, it was necessary to use DSM-III-R diagnostic criteria. In the case of ODD, it was possible to drop the DSM-III-R symptom of swearing and use the rev ised DSM-IV criteria for the diagnosis of ODD. The diagnosis of ODD was made whether or not the youth met criteria for CD.

Each conduct problem was defined by the DSM-III-R criteria for the corresponding CD symptom as operationalized by the DISC-2.3 (see Appendix). Because bullying was added to the DSM-IV criteria (American Psychiatric Association, 1994) for CD because of its strong diagnostic utility (Frick et al., 1994) and because it was assessed in the DISC-2.3, it was considered to be a conduct problem in these analyses. The DISC-2.3 queried the age of onset of every symptom of CD reported to be present in the past 12 months. The parent and youth reports of the age of onset of the first symptom were used separately to define age of onset for analyses of parent- and youth-reported conduct problems, respectively.

The parent was also administered the Service Utilization and Risk Factor (SURF) interview (Goodman et al., 1998; Leaf et al., 1996). In the SURF, the parent was shown a list of service settings and a list of service providers and asked if the youth had received mental health services in the past 12 months from a specialty mental health provider (psychologist, psychiatrist, counselor, or social worker) in any inpatient or outpatient mental health setting. In addition, the parent was asked if the youth had received any services for emotional, behavioral, or substance abuse problems in school during the past 12 months, including enrollment in special education. Service use was scored dichotomously as using or not using each type of services in the past 12 months. The parent SURF interview also obtained information on the youth’s lifetime history of police contacts (scored dichotomously). Independent ratings of the youth’s lowest level of functioning in the past 6 months were obtained from the parent and the two interviewers who interviewed the parent and youth using the nonclinician version of the Children’s Global Assessment Scale (CGAS; Setterberg, Bird, & Gould, 1992) after the DISC interviews were completed. The CGAS provides global ratings on a scale ranging from 1 to 100, with descriptions of levels of functioning provided at each decile.

Data Analysis

Tests of statistical significance were performed to estimate the likelihood that any differences found to be related to age of onset, gender, and other variables in this sample might have arisen by chance in the absence of true differences in the population of reference (Cohen, 1994). Preliminary comparisons of the proportions of girls and boys who engaged in antisocial behavior in the full sample of 1,285 youths were based on odds ratios, and preliminary comparisons of the mean age of onset of the first conduct problem in girls and boys were conducted using t-tests. Parent and youth reports of behavior and ages of onset were analyzed separately. Because there were few youths with ages of onset below 4 years of age and above 16 years of age, ages of onset of 4 years or less and ages of onset of 16 and 17 were combined.

All analyses involving age of onset were based on the subset of the MECA sample that engaged in one or more antisocial behaviors in the past 12 months. Valid ages of onset were reported for 98% of these youths. Parents reported ages of onset for 160 youths (104 boys and 56 girls) who engaged in one or more conduct problems according to parent reports. Similarly youths reported ages of onset for 284 youths (173 boys and 111 girls) who engaged in one or more conduct problems according to youth reports.

Analyses of the continuous CGAS ratings were conducted using least-squares regression. The number of conduct problems was treated as an ordinal “count” variable in poisson regression. Poisson regression is appropriate for ordinal count data such as numbers of symptoms in general population samples in which lower scores are much more prevalent than higher scores. When the dependent variable was dichotomous (i.e., presence or absence of each individual conduct problem, diagnosis, police contact, or use of mental health services), logistic regression was used.

When the results of these analyses were tabled, age of onset was divided into four levels based on quartiles for the sample to facilitate visual presentation, but age of onset was treated as a continuous variable in all statistical analyses. Conveniently, the best quartile breaks for both parent-reported and youth-reported ages of onset of the first conduct problem were the same for girls and boys. Thus, age of onset was divided into less than 8 years, 8-10 years, 11-13 years, and more than 13 years to facilitate the interpretation of some findings.

Because the youth’s age at the time of the assessment was correlated with age of onset in this sample (i.e., only older youths could have older ages of onset), the youth’s current age was considered as a possible covariate in each model. Because gender differences are common in the prevalence of conduct problems, gender and the Gender x Age-of-Onset interaction were also considered in all initial models, but they were retained in final models only if they were significant at the p [less than] .05 level.

RESULTS

In the full sample, boys were significantly more likely to exhibit one or more conduct problems in the past 12 months (15.6%) than girls (9.5%) according to parent report, odds ratio = 1.77 (95% CI 1.25-2.49). Similarly, 26% of boys and 19% of girls engaged in one or more conduct problems according to the youth report, odds ratio = 1.50(95% CI 1.15-1.96). There were no significant gender differences in the mean age of onset of conduct problems among girls and boys, however. The mean age of onset of the first conduct problem reported by parents was 11 years for girls (SD = 3.9) and 11 years for boys (SD = 3.7), t(135.9) = -0.04, ns. Similarly, the mean age of onset of the first conduct problem reported by youths was 10.2 years (SD = 4.2) for girls and 9.6 years (SD = 3.7) for boys, t(231.3) = 1.13, ns. As shown in Table I, the rank order of the median ages of onset for each conduct problem reported by the parent and youth was quite similar, Spearman’s r (12) = .89.

Age of Onset and Number of Conduct Problems

Among youths who engaged in at least one conduct problem in the past 12 months according to the parent report, poisson regression showed that the total number of conduct problems was inversely related to the parent report of age of onset of the first conduct problem, [X.sup.2](1, N = 160) = 12.89, p [less than] .0005 (Figure 1). When aggressive and nonaggressive conduct problems (see Appendix for definitions) were considered separately, the parent report of age of onset was inversely related to the number of parent-reported aggressive behaviors, [X.sup.2](1, N = 160) = 7.88, p [less than] .005, and was inversely related to the number of parent-reported nonaggressive conduct problems, [X.sup.2](1, N = 160) = 7.61, p [less than] .01.

As shown in Figure 2, poisson regression revealed that the total number of conduct problems reported by youths was inversely related to the youth report of age of onset, [X.sup.2](1, N = 284) = 43.04, p [less than] .0001. The youth report of age of onset was inversely related to the number of youth-reported aggressive behaviors, [X.sup.2](1, N = 284) = 17.84, p [less than] .0001, and was inversely related to the number of youth-reported non-aggressive conduct problems, [X.sup.2](1, N = 284) = 39.07, p [less than] .0001.

Although the numbers of aggressive and nonaggressive conduct problems were both inversely related to age of onset, there was a higher proportion of youths who engaged in only nonaggressive conduct problems in the past 12 months at older ages of onset according to both parent reports [X.sup.2](1, N = 160) = 13.41, p [less than] .0005, and youth reports [X.sup.2](1, N = 284) = 12.42, p [less than] .0005. Among youths whose parents reported at least one conduct problem in the past 12 months, the percentages of youths who engaged in at least one aggressive behavior in the four quartiles of age of onset (youngest to oldest) according to parent report were 47%, 44%, 32%, and 11%, respectively. For the youth report, the percentages of youths who engaged in at least one aggressive behavior in the four quartiles of age of onset (youngest to oldest) were 45%, 44%, 31%, and 19%, respectively.

Age of Onset and the Prevalence of Individual Conduct Problems

As shown in Table II, according to parent report, logistic regression revealed that age of onset was inversely related to the prevalence of three individual conduct problems (frequent lying, bullying, and stealing). In addition, frequent truancy was positively related to age of onset according to parent report. According to youth report (Table III), age of onset was inversely related to the prevalence of seven individual conduct problems (frequent 1ying, bullying, stealing, vandalizing, starting fights, using a weapon, and mugging), but no individual conduct problems were positively related to age of onset according to youth reports.

These findings on individual conduct problems underestimate the importance of frequent truancy in understanding the association between age of onset and individual differences in conduct problems. Frequent truancy was often the only conduct problem engaged in by youths with older ages of onset according to both informants. For the parent report, logistic regression revealed a significant Age of Onset x Gender interaction for the proportion of youths who only engaged in frequent truancy, [X.sup.2](1, N = 160) = 4.50, p [less than] .05. This interaction reflects somewhat different shapes of the functions relating age of onset to only engaging in frequent truancy (indicating a later increase in boys). The percentages of girls who only engaged in frequent truancy according to the parent report in the four quartiles of age of onset (youngest to oldest) were 0%, 9%, 38%, and 33%, respectively. For boys, the corresponding percentages were 0%, 0%, 18%, and 50%.

For youth reports, age of onset was also positively related to the proportion of youths who only engaged in frequent truancy, [X.sup.2](1, N = 284) = 18.57, p [less than] .0001, but the Age of Onset x Gender interaction was not significant. The percentages of girls who only engaged in frequent truancy according to the youth report in the four quartiles of age of onset (youngest to oldest) were 0%, 6%, 19%, and 50%, respectively. For boys, the corresponding percentages were 0%, 3%, 15%, and 38%.

Age of Onset and Diagnoses of CD, ODD, and ADHD

Among youths with at least one conduct problem, logistic regression showed that the parent report of age of onset was inversely related to the prevalence of meeting DSM–III–R criteria for the diagnosis of CD based on the parent report of symptoms, [X.sup.2](1, N = 160) = 6.48, p [less than] .05. The prevalence among girls of CD based on parent report of symptoms in the four quartiles of parent-reported age of onset (youngest through oldest) was 14.3%, 0%, 0%, and 0%, respectively. The corresponding percentages for boys were 33.3%, 14%, 5%, and 10%. The youth’s report of age of onset was also inversely related to the prevalence of DSM–III–R CD based on the youth report of symptoms, [X.sup.2](l, N = 281) = 30.15, p [less than] .0001. The prevalence among girls of CD according to youth report of symptoms in the four quartiles of youth-reported age of onset (youngest through oldest) was 24%, 6%, 11%, and 6%, respectively. The corresponding percentages for boys were 43%, 25%, 21%, and 5%.

Furthermore, logistic regression showed that the parent report of age of onset of the first conduct problem was inversely related to meeting criteria for DSM–IV ODD, [X.sup.2](1, N = 160) = 10.62, p [less than] .005. The prevalence of ODD among girls in the four quartiles of the parent report of age of onset (youngest through oldest) was 29%, 37%, 19%, and 7%, respectively. The corresponding percentages for boys were 46%, 25%, 18%, and 10%. Similarly, the parent report of age of onset of the first conduct problem was inversely related to meeting criteria for DSM–III–R ADHD, [X.sup.2](1, N = 159) = 12.10, p [less than] .0005. The prevalence of ADHD among girls in the four quartiles of the parent report of age of onset (youngest through oldest) was 36%, 0%, 19%, and 0%, respectively. The corresponding percentages for boys were 38%, 19%, 14%, and 7%. The youth report of the age of onset of conduct problems was not significantly related to the prevalence of either ODD or ADHD, however.

Age of Onset, Functional Impairment, and Service Use

Tables IV and V summarize analyses of the relation of age of onset to measures of functional impairment, mental health service use, and police contacts. In general, youths with earlier ages of onset were found to be more impaired and were more likely to use mental health services. Using least-squares regression, the parent report of age of onset was not found to be significantly related to the CGAS rating of adaptive functioning provided by the interviewer of the parent, but youths with later ages of onset were given ratings of more adaptive functioning on the CGAS rating by both the parent and the interviewer of the youth. Using logistic regression, parent report of age of onset was inversely related to the youth’s use of inpatient or outpatient specialty mental health services in the past 12 months, but not to the youth’s use of school-based services. Parent report of age of onset was not significantly related to the youth’s having one or more lifetime police contacts for offenses other than traffic violat ions.

Using least-squares regression, the youth report of age of onset was found to be positively related to the CGAS ratings provided by the interviewer of the youth, the CGAS ratings provided by the interviewer of the parent, and the parent’s own CGAS ratings of the youth’s adaptive functioning. Using logistic regression, the youth report of age of onset was also found to be inversely related to the youth’s use of both specialty mental health services and school-based mental health services in the past 12 months. All analyses of the use of mental health services were repeated adding the diagnoses ADHD and ODD as predictors to determine if the higher prevalence of these comorbid diagnoses among youths with earlier ages of onset explained differences in service use related to age of onset, but they did not change the result of any analysis. As for the parent report of age of onset, the youth report of age of onset was not significantly related to lifetime police contacts.

DISCUSSION

The present analyses confirm and extend previous findings and provide a detailed description of differences in the types and severity of conduct problems exhibited by girls and boys whose first conduct problem emerged at different ages. In the full sample of 1,285 youths, more boys than girls engaged in one or more conduct problems in the past 12 months according to both parent and youth informants, but there were no gender differences in the age of onset of conduct problems, and the Age of Onset x Gender interaction was significant in only one analysis.

Figures 1 and 2 and Tables II and III provide a comprehensive picture of the differences in the number and types of conduct problems that emerge at different ages of onset. According to both informants, the total number of conduct problems exhibited by girls and boys was lower at older ages of onset when current age and gender were controlled. The magnitude of age-adjusted differences in mean numbers of conduct problems related to age of onset was modest according to parent reports. According to youth reports, however, youths with ages of onset before 8 years of age engaged in an average of 2 to 3 times as many conduct problems as youths with ages of onset after 12 years of age, controlling for current age.

Although the prevalence of both aggressive and nonaggressive conduct problems was inversely related to age of onset, the proportion of youths who engaged only in nonaggressive conduct problems was higher at later ages of onset. Because the number of both kinds of conduct problems declined at older ages of onset, and the number of nonaggressive conduct problems was higher than the number of aggressive behaviors at all ages of onset, many youths with later ages of onset engaged in a small number of nonaggressive conduct problems, but no aggressive behavior.

There were also large differences in the prevalence of several individual conduct problems associated with age of onset. If confirmed in future prospective studies, these differences will reveal a great deal about the nature of conduct problems that emerge at different ages. According to both the parent and the youth, the prevalences of frequent lying, bullying, and stealing were markedly lower among youths who engaged in conduct problems for the first time at older ages. According to both informants, the prevalence of frequent lying and bullying in the oldest quartile of age of onset was one third or less of the prevalence in the youngest quartile of age of onset. Similarly, according to both informants, the prevalence of stealing in the oldest quartile was half of that in the youngest quartile of age of onset (adjusting for the current age of the youth informant). In addition, according to the youth informant only, the prevalence of vandalizing, starting fights, using a weapon, and mugging were each one ha lf as prevalent or less among youths in the oldest compared with the youngest quartile of age of onset.

Thus, across informants, youths who do not engage in conduct problems until older ages appear to be much less likely to engage in theft, frequent dishonesty, vandalism, and aggression than youths who first engage in conduct problems at earlier ages. The inverse association between aggression and age of onset has been understood for some time (Loeber, 1988), but it now appears that dishonesty, theft, and vandalism are also markedly less common among youths with older ages of onset. Furthermore, youths who engage in conduct problems for the first time at older ages appear to be more likely to engage in only the relatively benign behavior of frequent truancy. In the oldest quartile of age of onset, from one third to one half of youths who engaged in conduct problems only engaged in frequent truancy (depending on gender and informant), whereas no girls or boys in the youngest quartile of age of onset only engaged in frequent truancy according to either parent or youth reports. Thus, important aspects of the hete rogeneity of the behavior of youths who engage in conduct problems are correlated with the age of onset of their first conduct problem.

Given that the number of conduct problems is lower among youths with older ages of onset, it is not surprising that the proportion of youths who engage in enough conduct problems to meet DSM-III–R criteria for CD is lower at older ages of onset. According to youth reports of age of onset, youths with earlier ages of onset of the first conduct problem were also more likely to meet DSM–IV ODD at the time of the assessment and were more likely to meet DSM-III–R criteria for ADHD according to parent reports. The latter finding confirms the association between ADHD and earlier onsets of conduct problems reported by Moffitt (1990), but it appears that ODD may also be associated with earlier ages of onset of conduct problems.

An apparent conflict between the present findings and a previous report of analyses based on the MECA sample (Lahey et al., 1998) requires discussion. Among the 74 youths in the MECA sample who met DSM-III–R criteria for CD, the mean number of aggressive behaviors was similar at all ages of onset below age 10, but declined sharply among youths with older ages of onset, beginning around an age of onset of 10 years (Lahey et al., 1998). The number of nonaggressive CD behaviors did not vary as a function of age of onset among youths who met criteria for CD, however. These findings suggested a natural dichotomy between childhood-onset and adolescent-onset CD in terms of physical aggression (Lahey et al., 1998), but it is important to examine this conclusion in the light of the present findings on the age of onset of conduct problems. The diagnostic definition of CD requires that youths exhibit a minimum of three conduct problems in the past year. We reported earlier that the proportion of youths in the MECA sam ple who engaged in one or more conduct problems whose problem behaviors were limited to nonaggressive behaviors was higher at older ages of onset.

Therefore, youths who meet criteria for CD with older ages of onset of their first conduct problem will increasingly tend to exhibit three nonaggressive conduct problems and no symptoms of aggressive behavior. This creates the impression of a steep decline in the number of aggressive behaviors at older ages of onset. Differences in aggressive behavior related to the age of onset of the diagnosis of CD may still be meaningful for the nosology of subtypes of CD, but it is important to place these differences in the larger context of the development of conduct problems.

Not only was age of onset related to individual differences in conduct problem behaviors, age of onset was found to be positively related to global ratings of adaptive functioning using the CGAS. Given these differences in level of functioning and the higher number of conduct problems at earlier ages of onset, it is not surprising that youths with earlier ages of onset of conduct problems were significantly more likely to have used inpatient or outpatient mental health services in the past 12 months. It is important to note that these differences in the use of mental health services were not explained by differences in rates of comorbid ADHD and ODD among youths with different ages of onset of conduct problems. Consistent with other findings (Moffitt, 1990), there was not a significant relation between the age of onset of conduct problems and contacts with police when current age was controlled. This may indicate that conduct problems that emerge at all ages are significant at least in the sense of putting t he youth at risk for legal difficulties.

One difference should be noted between the present findings and Moffitt’s earlier findings from the Dunedin Study. Moffitt et al. (1996) found that delinquent adolescent boys with ages of onset after age 12 were over three times more prevalent than delinquent adolescent boys with earlier ages of onset. In the present sample, in contrast, Tables II and III show that approximately equal numbers of youths with conduct problems had ages of onset before and after age 12. This was true for both genders and both informants. This may reflect the difference between cross-sectional and prospective designs or may reflect differences in the definitions of conduct problems and delinquency, or both. Because the DISC was used in the present study, the number of conduct problems were defined as the symptoms of CD, which often set high thresholds of frequency of occurrence for individual behaviors. In the Dunedin Study, in contrast, youths with a single occurrence of the same behaviors would be classified as delinquent. Thus , the present study may underestimate the prevalence of youths with mild conduct problems with later ages of onset.

The present findings also provide some information on similarities and differences between parent and youth informants. In the present study, the median ages of onset of the first conduct problem to emerge were similar for the two informants, and the correlation between parents and youths on the rank order of the ages of onset of each individual conduct problem was very high. Moreover, findings regarding the relations of age of onset with the number and types of conduct problems were quite similar across informants. There were some differences in findings, but because youths tended to report more conduct problems than their parents in this sample, and because statistical power for detecting differences in proportions is less for smaller proportions (Fleiss, 1981), there was less statistical power for detecting differences associated with age of onset for parent reports than youth reports. Thus, some apparent differences between informants may be an artifact of differences in statistical power.

Limitations of the Present Study

The use of cross-sectional data to study the relation of age of onset with the nature of conduct problems is inherently limited by reliance on retrospective recall. The structure of the questions regarding conduct problems used in the present study raises particular concern about the measurement of the age of onset of the youth’s first conduct problem. Parents and youths were asked about the presence or absence of each conduct problem in the past 12 months and then were asked to recall the age of onset of each behavior that was present. Thus, it is likely that some youths had engaged in problem behaviors in the past that were not currently in their repertoire. If those behaviors had earlier ages of onset than any conduct problem exhibited in the past 12 months, the estimate of the age of onset of the first conduct problem would bebiased upwards. Any such bias would add noise to the data and could bias findings. Fortunately, if the current methods resulted in some youths with earlier ages of onset being said to have later ages of onset, such error would reduce differences related to age of onset and have the effect of making statistical tests of such differences more conservative.

In addition, it seems probable that youths who were older at the time of their assessment would be more likely than younger youths to have a biased report of age of onset because they had engaged in a conduct problem at younger ages that is no longer in their repertoire. It is important, therefore, that age at the time of the assessment was positively related to the number of conduct problems in many analyses in this article and never inversely related to any measures of conduct problems. If ages of onset were sometimes biased upwards in older youths who have had time to develop more conduct problems that would tend to artificially increase the positive association between age of onset and number of conduct problems, whereas age of onset was found to be inversely associated with number of conduct problems in this study. Thus, any biases due to the method of assessing age of onset would seem unlikely to have created an artifactual inverse association between age of onset and the severity of conduct problems i f such a relation did not actually exist.

Three other considerations also lend support to the present findings. First, although retrospective recall is always suspect, studies of the validity of retrospective recall suggest that it is acceptable for assessing the relative standing of individuals in distributions, such as the distribution of ages of onset in the present study (Henry, Moffitt, Caspi, Langley, & Silva, 1994). Second, reports of age of onset by parents and youths were highly correlated, and findings regarding the relation of age of onset with the severity of conduct problems were very similar for each informant. Thus, if the present findings reflect retrospective biases, these biases must operate similarly for parents and youths. Third, the inverse relation between age of onset and the severity of conduct problems has been found in both cross-sectional studies (Tolan, 1987) and prospective longitudinal studies (Moffit et al., 1996; Tolan & Thomas, 1995). Such consistency in findings would be unlikely if the present results were merely a n artifact of biased retrospective recall of age of onset. Overall, however, the present cross-sectional findings are of value primarily because they confirm the findings of more defensible previous prospective studies.

Theoretical Implications

The differences in conduct problems, impairment, and comorbidity related to age of onset found in the present study support the model advanced by Moffitt (1993). The present findings were not entirely consistent with Moffitt’s hypothesis of a “developmental taxonomy,” however, which postulates two distinct types (taxons) of conduct problems that emerge before and after puberty (Moffit, 1993). Rather, the present findings are more consistent with the view that the quantity and quality of conduct problems vary gradually across the continuum of ages of onset (Lahey, Waldman, & McBurnett, in press). It should be noted, however, that our ability to provide a test of the hypothesis that there are two distinct groups of youths with conduct problems based on their ages of onset was limited for two important reasons. First, the number of participants needed to distinguish between continuous versus discrete distributions on statistical grounds is considerably larger than the present sample size. Thus, it is possible t hat there are discrete differences in conduct problems based on age of onset that appear to be continuous in the present sample because of a combination of measurement error and lack of statistical power. Second, Moffitt (1993) stated her hypothesis that there are two distinct groups of antisocial youths in terms of ages of onset before and after puberty, whereas we measured age of onset in terms of age. Because ages of pubertal maturation vary among individuals, our use of age as an index of development may have obscured discrete differences based on age of onset before and after puberty. Again, only prospective studies beginning in early childhood with considerable sample size can address this and other key questions with confidence.

ACKNOWLEDGMENTS

The MECA Program is an epidemiologic methodology study performed by four independent research teams in collaboration with staff of the Division of Clinical Research, which was reorganized in 1992 with components now in the Division of Epidemiology and Services Research and the Division of Clinical and Treatment Research of the NIMH, in Rockville, Maryland. The NIMH Principal Collaborators are Darrel A. Regier, M.D., M.P.H., Ben Z. Locke, M.S.P.H., Peter S. Jensen, M.D., William E. Narrow, M.D., M.P.H., and Donald S. Rae, M.A.; the NIMH Project Officer was William J. Huber. The principal investigators and coinvestigators from the four sites Emory University, Atlanta, Georgia, U01 MH46725: Mina K. Dulcan, M.D., Benjamin B. Lahey, Ph.D., Donna J. Brogan, Ph.D., Sherryl H. Goodman, Ph.D., and Elaine Flagg, Ph.D.; Research Foundation for Mental Hygiene at New York State Psychiatric Institute, New York, New York, U01 MH46718: Hector R. Bird, M.D., David Shaffer, M.D., Myrna Weissman, Ph.D., Patricia Cohen, Ph.D., Denise Kandel, Ph.D., Christina Hoven, Ph.D., Mark Davies M.P.H., Madelyn S. Gould, Ph.D., and Agnes Whitaker, M.D.; Yale University, New Haven, Connecticut, U01 MH46717: Mary Schwab-Stone, M.D., Philip J. Leaf, Ph.D., Sarah Horwitz, Ph.D., and Judith Lichtman, M.P.H.; University of Puerto Rico, San Juan, Puerto Rico, U01 MH46732: Glorisa Canino, Ph.D., Maritza Rubio-Stipec, M.A., Milagros Bravo, Ph.D., Margarita Alegria, Ph.D., Julio Ribera, Ph.D., Sara Huertas, M.D., and Michel Woodbury, M.D.

(1.) University of Chicago, and National Opinion Research Center at the University of Chicago, Chicago, Illinois.

(2.) Emory University, Atlanta, Georgia.

(3.) Columbia University, New York, New York.

(4.) University of Puerto Rico, San Juan, Puerto Rico.

(5.) National Institute of Mental Health, Bethesda, Maryland.

(6.) Johns Hopkins University, Baltimore, Maryland.

(7.) University of Miami, Coral Gables, Florida.

(8.) Address all Correspondence concerning this article to Benjamin B. Lahey, Department of Psychiatry (MC 3077), University of Chicago, 5841 South Maryland Avenue, Chicago, IL 60637. E-mail: blahey@yoda.bsd.uchicago.edu.

REFERENCES

Bird, H. R., Gould, M. S., & Staghezza, B. (1992). Aggregating data from multiple informants in child psychiatry epidemiological re-search. Journal of the American Academy of Child and Adolescent Psychiatry 31, 78-85.

Blumstein, A., Farrington, D. P., & Moitra, S. (1985). Delinquency careers: Innocents, desisters, and persisters. In M. Tonry & N. Morris (Eds.), Crime and justice (pp. 187-219). Chicago: University of Chicago Press.

Cohen, J. (1994). The earth is round (p [less than] .05). American Psychologist, 49, 997-1003.

Elliott, D. S., Huizinga, D., & Menard, S. (1988). Multiple problem youth: Delinquency, substance use, and mental health problems. New York: Springer-Verlag.

Fleiss, J. (1981). Statistical methods for rates and proportions. New York: Wiley.

Frick, P. J., Lahey, B. B., Applegate, B., Kerdyck, L., Ollendick, T., Hynd, G. W., Garfinkel, B., Greenhill, L., Biederman, J., Barkley, R. A., McBurnett, K., Newcorn, J., & Waldman, I. (1994). DSM-IV Field Trials for the Disruptive Behavior Disorders: Symptom utility estimates. Journal of the American Academy of Child and Adolescent Psychiatry 33,529-539.

Glueck, S., & Glueck, E. (1959). Predicting delinquency and crime. Cambridge, MA: Harvard University Press.

Goodman, S. H., Hoven, C. W., Narrow, W. E., Cohen, P., Fielding, B., Alegria, M., Leaf, P. J., Kandel, D., Horwitz, S. M., Bravo, M., Moore, R., Dulcan, M. K. (1998). Measurement of risk for mental disorders and competence in a psychiatric epidemiologic community survey: The NIMH Methods for the Epidemiology of Child and Adolescent Mental Disorders (MECA) Study. Social Psychiatry and Psychiatric Epidemiology, 33, 162-173.

Hart, E. L., Lahey, B. B., Loeber, R., & Hanson, K. S. (1994). Criterion validity of informants in the diagnosis of disruptive behavior disorders in children: A preliminary study. Journal of Consulting and Clinical Psychology, 62, 410-414.

Henry, B., Moffitt, T, E., Caspi, A., Langley, J. A., & Silva, P. A. (1994). On the “remembrance of things past”: A longitudinal evaluation of the retrospective method. Psychological Assessment, 6, 92-101.

Hinshaw, S. P., Lahey, B. B., & Hart, E. L. (1993). Issues of taxonomy and comorbidity in the development of conduct disorder. Development and Psychopathology, 5, 31-50.

Lahey, B. B., Flagg, E. W., Bird, H. R., Schwab-Stone, M,, Canino, G,, Dulcan, M. K., Leaf, P. J., Davies, M., Brogan, D., Bourdon, K,, Horwitz, S. M., Rubio-Stipec, M., Freeman, D. H., Lichtman, J., Shaffer, D., Goodman, S. H., Narrow, W. E., Weissman, M. M., Kandel, D. B., Jensen, P. S., Richters, J. E., & Regier, D. A. (1996). The NIMH Methods for the Epidemiology of Child and Adolescent Mental Disorders (MECA) Study: Background and methodology. Journal of the American Academy of Child and Adolescent Psychiatry 35, 855-864.

Lahey, B. B., Loeber, R., Quay, H. C., Applegate, B., Shaffer, D., Waldman, I., Hart, E. L., McBumett, K., Frick, P. J., Jensen, P., Dulcan, M,, Canino, G., & Bird, H. (1998). Validity of DSM-IV subtypes of conduct disorder based on age of onset. Journal of the American Academy of Child and Adolescent Psychiatry 37, 435-442.

Lahey, B. B., Waldman, I. D., & McBurnett, K. (in press). The development of antisocial behavior: An integrative causal model. Journal of Child Psychology and Psychiatry.

Leaf, P. J., Alegria, M., Cohen, P., Goodman, S. H., Horwitz, S. M., Hoven, C. W., Narrow, W. E., Vanden-Kiernan, M., & Regier, D. (1996). Mental health service use in the community and schools: Results from the four-community MECA study. Journal of the American Academy of Child and Adolescent Psychiatry 35, 889-897.

Loeber, R. (1982). The stability of antisocial and delinquent child behavior: A review, Child Development, 53, 1431-1446.

Loeber, R. (1987), The prevalence, correlates, and continuity of serious conduct problems in elementary school children. Criminology, 25, 615-642.

Loeber, R. (1988). Natural histories of conduct problems, delinquency, and associated substance abuse: Evidence for developmental progressions. In B. B. Lahey & A. E. Kazdin (Eds.), Advances in clinical child psychology (Vol. 11, pp. 73-124). New York: Plenum.

McGee, R., Feehan, M., Williams, S., & Anderson, J. (1992), DSM-III disorders from age 11 to age 15 years. Journal of the American Academy of Child and Adolescent Psychiatry 31, 50-59.

McGee, R., Feehan, M., Williams, S., Partridge, F., Silva, P. A., & Kelly, J. (1990). DSM-III disorders in a large sample of adolescents. Journal of the American Academy of Child and Adolescent Psychiatry 29, 611-619.

Moffitt, T. E. (1990). Juvenile delinquency and attention deficit disorder: Boys’ developmental trajectories from age 3 to 15. Child Development, 61, 893 910.

Moffit, T. E. (1993). Adolescence-limited and life-course-persistent antisocial behavior: A developmental taxonomy. Psychological Review, 100, 674-701.

Moffitt, T. E., Caspi, A., Dickson, N., Silva, P., & Stanton, W. (1996). Childhood-onset versus adolescent-onset antisocial conduct problems in males: Natural history from ages 3 to 18 years. Development and Psychopathology, 8, 399-424.

Schwab-Stone, M., Shaffer, D., Dulcan, M., Jensen, P., Fisher, P., Bird, H., Goodman, S. H., Lahey, B. B., Lichtman, J. H., Canino, G., Rubio-Stipec, M., & Rae, D. S. (1996). Criterion validity of the NIMH Diagnostic Interview Schedule for Children (DISC 2.3). Journal of the American Academy of Child and Adolescent Psychiatry, 35, 878-888.

Setterberg, S., Bird, H., & Gould, M. (1992). Parent and interviewer versions of the Children’s Global Assessment Scale. New York: Columbia University.

Shaffer, D., Fisher, P., Dulcan, M., Davies, M., Piacentini, J., Schwab-Stone M., Lahey, B. B., Bourdon, K., Jensen, P., Bird, H., Canino, G., & Regier, D. (1996). The NIMH Diagnostic Interview Schedule for Children (DISC 2.3): Description, acceptability, prevalences, and performance in the MECA study. Journal of the American Academy of Child and Adolescent Psychiatry, 35, 865-877.

Tolan, P. H, (1987). Implications of age of onset for delinquency risk. Journal of Abnormal Child Psychology, 15, 47-65.

Tolan, P. H., & Thomas, P. (1995). The implications of age of onset for delinquency risk: II. Longitudinal data. Journal of Abnormal Child Psychology, 23, 157-181.

Wadsworth, M. (1979). Roots of delinquency. London: Martin Robinson.

Wolfgang, M. E., Figlio, R. M., & Stellin, T. (1972). Delinquency in a birth cohort. Chicago: University of Chicago Press.

Median Age of Onset in Years of Conduct

Problems Based on Youth and Parent Reports

Conduct problem Youth report Parent report

Frequent lying 7 8

Bullying 10 10

Starting fights 10 12

Stealing 11 10.5

Hurting animals 11.5 12

Using a weapon 11.5 9.5

Vandalizing 12 11

Physical cruelty 12 11

Mugs 13 13

Running away 13 14

Breaking and entering 14 12

Setting fires [*] — 12

Frequent truancy 14 14

Note. Data are presented in rank order according to youth report.

(*.)Behavior not reported by youths. Forced sex was not reported

by either informant.

Specific Conduct Problems by Quartiles of

Parent-Reported Age of Onset of First Conduct

Problem

Less than 8 years 8 through 10 years

N = 38 N = 39

Frequent lying (79%) [**] Frequent lying (54%) [**]

Frequent truancy (l3%) [t] Frequent truancy (l0%) [t]

Bullying (37%) [**] Bullying (21%) [**]

Stealing (29%) [**] Stealing (13%) [**]

Vandalizing (11%) Vandalizing (21%)

Running away (8%) Running away (3%)

Starting fights (5%) Starting fights (13%)

Hurting animals (5%) Hurting animals (3%)

Breaking/entering (5%) Breaking/entering (8%)

Physical cruelty (5%) Physical cruelty (8%)

Using a weapon (3%) Using a weapon (3%)

Mugging (0%) Mugging (3%)

Setting fires (3%) Setting fires (0%)

Less than 8 years 11 through 13 years

N = 38 N = 38

Frequent lying (79%) [**] Frequent lying (16%) [**]

Frequent truancy (l3%) [t] Frequent truancy (34%) [t]

Bullying (37%) [**] Bullying (21%) [**]

Stealing (29%) [**] Stealing (5%) [*]

Vandalizing (11%) Vandalizing (13%)

Running away (8%) Running away (11%)

Starting fights (5%) Starting fights (3%)

Hurting animals (5%) Hurting animals (13%)

Breaking/entering (5%) Breaking/entering (3%)

Physical cruelty (5%) Physical cruelty (3%)

Using a weapon (3%) Using a weapon (0%)

Mugging (0%) Mugging (0%)

Setting fires (3%) Setting fires (0%)

Less than 8 years Greater than 13 years

N = 38 N = 45

Frequent lying (79%) [**] Frequent lying (16%) [**]

Frequent truancy (l3%) [t] Frequent truancy (69%) [t]

Bullying (37%) [**] Bullying (4%) [**]

Stealing (29%) [**] Stealing (13%) [**]

Vandalizing (11%) Vandalizing (4%)

Running away (8%) Running away (20%)

Starting fights (5%) Starting fights (7%)

Hurting animals (5%) Hurting animals (2%)

Breaking/entering (5%) Breaking/entering (0%)

Physical cruelty (5%) Physical cruelty (0%)

Using a weapon (3%) Using a weapon (0%)

Mugging (0%) Mugging (2%)

Setting fires (3%) Setting fires (0%)

Note. Aggressive behaviors are presented in bold. Among girls and boys who exhibited at least one conduct problem, the prevalence of each parent-reported behavior at the time of the assessment among youths in the four quartiles of age of onset of according to the parent report of age of onset (behaviors are presented in rank order of prevalence for the parent informant across all quartiles of age of onset).

(1.)The age of onset-by-gender interaction was significant for frequent truancy, p = .03, controlling for a positive association with age at the time of the assessment, p [less than] .0001. The association of truancy with age of onset was positive for both genders. Age-adjusted prevalences for frequent truancy in each quartile of age of onset (lowest to highest) for girls were 23%, 32%, 40%, and 33%; the corresponding age-adjusted prevalences of frequent truancy for boys were 21%, 19%, 28%, and 62%.

(**.)Significant inverse ordinal relation of age of onset with prevalence, p [less than or equal to] .01 (age and gender not significant at p [less than] .05).

Specific Conduct Problems by Quartiles of Youth-reported

Age of Onset of First Conduct Problem

Less than 8 years 8 through 10 years

N = 85 N = 54

Frequent truancy (25%) Frequent truancy (19%)

Frequent lying (75%) [**] Frequent lying (52%) [**]

Stealing (32%) [1] Stealing (24%) [1]

Vandalizing (22%) [2] Vandalizing (15%) [2]

Hurting animals (18%) Hurting animals (15%)

Starting fights (23%) [**] Starting fights (15%) [**]

Breaking/entering (9%) Breaking/entering (9%)

Running away (7%) Running away (4%)

Using a weapon (10%) [3] Using a weapon (9%) [3]

Bullying (13%) [**] Bullying (7%) [**]

Mugging (9%) [**] Mugging (9%) [**]

Physical cruelty (7%) Physical cruelty (4%)

Setting fires (0%) Setting fires (0%)

Less than 8 years 11 through 13 years

N = 85 N = 75

Frequent truancy (25%) Frequent truancy (50%)

Frequent lying (75%) [**] Frequent lying (9%) [**]

Stealing (32%) [1] Stealing (25%) [1]

Vandalizing (22%) [2] Vandalizing (28%) [2]

Hurting animals (18%) Hurting animals (12%)

Starting fights (23%) [**] Starting fights (11%) [**]

Breaking/entering (9%) Breaking/entering (9%)

Running away (7%) Running away (15%)

Using a weapon (10%) [3] Using a weapon (8%) [3]

Bullying (13%) [**] Bullying (1%) [**]

Mugging (9%) [**] Mugging (5%) [**]

Physical cruelty (7%) Physical cruelty (4%)

Setting fires (0%) Setting fires (0%)

Less than 8 years Greater than 13 years

N = 85 N = 70

Frequent truancy (25%) Frequent truancy (65%)

Frequent lying (75%) [**] Frequent lying (9%) [**]

Stealing (32%) [1] Stealing (23%) [1]

Vandalizing (22%) [2] Vandalizing (11%) [2]

Hurting animals (18%) Hurting animals (9%)

Starting fights (23%) [**] Starting fights (3%) [**]

Breaking/entering (9%) Breaking/entering (9%)

Running away (7%) Running away (6%)

Using a weapon (10%) [3] Using a weapon (1%) [3]

Bullying (13%) [**] Bullying (4%) [**]

Mugging (9%) [**] Mugging (0%) [**]

Physical cruelty (7%) Physical cruelty (3%)

Setting fires (0%) Setting fires (0%)

Note. Aggressive behaviors are presented in bold. Among girls and boys who exhibited at least one conduct problem, the prevalence of each youth-reported behavior at the time of the assessment among youths in the four quartiles of age of onset of according to the youth report of age of onset (behaviors are presented in rank order of prevalence for the youth informant across all quartiles of age of onset).

(1.)Age of onset was inversely related to the prevalence of stealing, p = .001, controlling for a positive association with age at the time of the assessment, p [less than] .0001, and gender, p = .05. Age-adjusted prevalences for stealing in each quartile of age of onset (lowest to highest) for girls were 25.2%, 20.4%, 19.3%, and 10.9%; the corresponding age-adjusted prevalences of stealing for boys were 46.7%, 39.3%, 24.6%, and 11.2%.

(2.)Age of onset was inversely related to the prevalence of vandalizing, p = .03, controlling for a positive association with age at the time of the assessment, p [less than] .001, and gender, p = .02. Age-adjusted prevalences for vandalizing in each quartile of age of onset (lowest to highest) for girls were 14.1%, 13.5%, 24.4%, and -2.6%; the corresponding age-adjusted prevalences of vandalizing for boys were 34.3%, 23.3%, 27.6%, and 10.5%.

(3.)Age of onset was inversely related to the prevalence of using a weapon, p = .01, controlling for a positive association with age at the time of the assessment, p = .05. Age-adjusted prevalences for using a weapon in each quartile of age of onset (lowest to highest) for girls and boys were 11.7%, 12.8%, 7.0%, and -2.8%.

(**.)Significant inverse ordinal relation of age of onset with prevalence, p [greater than or equal to] .01 (age and gender not significant at p [less than] .05).

Impairment and Service Use by Quartiles of Parent-reported

Age of Onset of First Conduct Problem (in Years)

Less than 8 through 11 through

8 years 10 Years 13 Years

N = 38 N = 39 N = 38

Parent CGAS

Unadjusted means 64.92 69.72 77.87

Interviewer of parent CGAS

Unadjusted means 64.53 68.82 72.21

Interviewer of youth CGAS

Unadjusted means 65.68 74.38 72.57

Adjusted means [1] 63.65 71.32 72.93

12-month mental health service use

Unadjusted % 29.00 12.82 5.26

Adjusted % [1] 34.34 21.00 4.52

12-month school mental health services use

Unadjusted % 36.84 15.38 13.16

Lifetime one or more police contact

Unadjusted % 23.68 17.95 26.32

Greater than

13 Years Regression

N = 45 t or [X.sup.2]

Parent CGAS

Unadjusted means 75.00 t = 3.28

Interviewer of parent CGAS

Unadjusted means 70.75 t = 1.82

Interviewer of youth CGAS

Unadjusted means 71.38 t = 3.18

Adjusted means [1] 75.46

12-month mental health service use

Unadjusted % 26.27 [X.sup.2] = 5.98

Adjusted % [1] 15.64

12-month school mental health services use

Unadjusted % 28.29 [X.sup.2] = 3.08

Lifetime one or more police contact

Unadjusted % 35.36 [X.sup.2] = 1.49

p[less than or equal to]

Parent CGAS

Unadjusted means .005

Interviewer of parent CGAS

Unadjusted means ns

Interviewer of youth CGAS

Unadjusted means .005

Adjusted means [1]

12-month mental health service use

Unadjusted % .01

Adjusted % [1]

12-month school mental health services use

Unadjusted % ns

Lifetime one or more police contact

Unadjusted % ns

Note. Higher CGAS ratings indicate more adaptive functioning. Among youths who exhibited at least one conduct problem, ratings on the Children’s Global Assessment Scale (CGAS) by the parent, interviewer of the parent, and interviewer of the youth, and the parent report of 12-month use of mental health services and lifetime police contacts as a function of the parent report of age of onset of the first conduct problem.

(1.)Adjusted for current age.

Impairment and Service Use by Quartiles of Youth-reported

Age of Onset of First Conduct Problem (in Years)

Less than 8 through 11 through

8 years 10 years 13 years

N = 85 N = 54 N = 75

Parent CGAS

Unadjusted means 77.78 78.78 79.93

Adjusted means [1] 76.43 76.69 80.52

Interviewer of parent CGAS

Unadjusted means 76.09 78.00 76.03

Adjusted means [1] 74.74 75.84 76.62

Interviewer of youth CGAS

Unadjusted means 68.38 70.94 70.61

Adjusted means [1] 66.17 67.51 71.58

12-month mental health service use

Unadjusted % 21.18 11.11 10.67

12-month school mental health services use

Unadjusted % 17.65 12.96 21.33

Adjusted % [1] 21.21 18.51 19.76

Lifetime one or more police contact

Unadjusted % 21.18 12.96 17.33

Greater than

13 Years Regression

N = 70 t or [X.sup.2]

Parent CGAS

Unadjusted means 83.69 t = 3.07

Adjusted means [1] 86.30

Interviewer of parent CGAS

Unadjusted means 80.27 t = 1.93

Adjusted means [1] 82.91

Interviewer of youth CGAS

Unadjusted means 77.10 t = 4.67

Adjusted means [1] 81.39

12-month mental health service use

Unadjusted % 7.14 [X.sup.2] = 6.25

12-month school mental health services use

Unadjusted % 4.29 [X.sup.2] = 6.27

Adjusted % [1] -2.65

Lifetime one or more police contact

Unadjusted % 18.57 [X.sup.2] = 2.48

p[less than or equal to]

Parent CGAS

Unadjusted means .005

Adjusted means [1]

Interviewer of parent CGAS

Unadjusted means .05

Adjusted means [1]

Interviewer of youth CGAS

Unadjusted means .0001

Adjusted means [1]

12-month mental health service use

Unadjusted % .01

12-month school mental health services use

Unadjusted % .01

Adjusted % [1]

Lifetime one or more police contact

Unadjusted % ns

Note. Higher CGAS ratings indicate more adaptive functioning. Among youths who exhibited at least one conduct problem, ratings on the Children’s Global Assessment Scale (CGAS) by the parent, interviewer of the parent, and interviewer of the youth, and the parent report of 12-month use of mental health services and lifetime police contacts as a function of the youth report of age of onset of the first conduct problem.

(1.)Adjusted for current age.

Appendix. Definitions and Abbreviations of Conduct Problems

Aggressive conduct problems

Start fights – Started at least four fights in which there was hitting in the past 12 months

Bully – Threatened, picked on, or hurt other children who do not fight back at least once a week during the past 12 months

Use weapon – Used a weapon like a bat, brick, bottle, knife, or gun in more than one fight in the past 12 months

Cruel – Been physically cruel or tried to cause someone pain (not during a fight) one or more times in the past 12 months

Hurt animals – Tortured animals or hurt them on purpose one or more times in the past 12 months

Forced sex [a] – Forced someone to do something sexual with her or him one or more times in the past 12 months

Mug – Snatched a purse, held up, or threatened someone to steal one or more times in the past 12 months

Nonaggressive Conduct Problems

Steal – Shoplifted or stole without confronting the victim inside or outside the home more than once in the past 12 months

Run – Ran away from home overnight at least twice or once for two weeks in the past 12 months

Frequent lying – Told lies that caused trouble at least once a month for the past 12 months

Set fires – Set at least one fire with the intention of causing damage or hurting someone in the past 12 months

Frequent truancy – Skipped class or school without permission at least four times in the past 12 months

Break and enter – Broke into a building, house, or car in the past 12 months

Vandalize – Broke something or messed up some place on purpose, like breaking windows, writing on a building, or slashing tires in the past 12 months

(a.) No occurrences in this sample reported by either parents or youths.

COPYRIGHT 1999 Plenum Publishing Corporation

COPYRIGHT 2000 Gale Group