Oppositional defiant disorder

G. Leonard Burns (1,3)

James A. Walsh (2)

Confirmatory factor analysis (CFA) has indicated that attention-deficit hyperactivity disorder-inattention (ADHD-IN), ADHD-hyperactivity/impulsivity (H/I), and oppositional defiant disorder (ODD) factors are highly correlated with each other (Beiser, Dion, & Gotowiec, 2000; Burns, Walsh, Owen, & Snell, 1997; Burns, Walsh, Patterson, et al., 1997; DuPaul et al., 1997, 1998; Fergusson & Horwood, 1993; Gomez, Harvey, Quick, Scharer, & Harris, 1999; Pillow, Pelham, Hoza, Molina, & Stultz, 1998; Wolraich, Feurer, Hannah, Baumgaertel, & Pinnock, 1998). DuPaul et al. (1997), for example, reported a correlation between the IN and H/I factors of .94 from teacher ratings, and DuPaul et al. (1998) reported a correlation of .92 from parent ratings. In a similar study in Australia, Gomez et al. (1999) found correlations of .75 and .68 between the IN and H/I factors for parent and teacher ratings, respectively. In the studies by Burns and colleagues, the correlation between the IN and H/I factors ranged from .74 to .80, be tween IN and ODD factors from .64 to .69, and between H/I and ODD factors from .69 to .80.

Although the high correlations among the IN, H/I, and ODD dimensions are partly due to common method variance as well as some of the symptoms lacking strong discriminant validity (Burns, Walsh, Owen, et al., 1997; Burns, Walsh, Patterson, et al., 1997; Fergusson & Horwood, 1993; Moura, Burns, & Walsh, 2001), and thereby inflating the factor correlations, a meaningful amount of the covariation among the factors is probably due to more substantial reasons. The high correlations among the factors, for example, could partly be due to the dimensions sharing common risk factors (McGee & Williams, 1999; Samudra & Cantwell, 1999; Sanson & Prior, 1999). Another possibility, and the focus of the current study, is that the high correlations among the factors are due to one factor influencing the development of another factor.

ADHD is the comorbid condition most commonly associated with CP [conduct problems], and is thought to precede the development of CP in the majority of the cases. In fact, some investigators consider ADHD (or, more specifically, the impulsivity or hyperactivity components of ADHD), to be the "motor" that drives the development of early-onset CP.... (McMahon & Estes, 1997, p. 133)

This hypothesis suggests that H/I behaviors foster the development of ODD behaviors across time (Lahey, McBumett, & Loeber, 2000; Loeber, Green, Lahey, Frick, & McBumett, 2000, Fig. 1). The basis for this hypothesis is that the H/I behaviors result in a child who is more difficult to parent, thereby increasing the likelihood of problematic parent-child interactions and the development of ODD behaviors (e.g., Dishion & Patterson, 1999; McMahon & Estes, 1997, pp. 134-140). In contrast, ODD behaviors are not considered to foster the development of H/I behaviors across time.

These are complex hypotheses to test and the resuits are inconsistent. Some studies have failed to find that ADHD predicts subsequent ODD/conduct disorder (CD) independent of earlier levels of ODD/CD (e.g., Biederman et al. 1996), whereas other studies indicate that ADHD is able to predict the development of ODD/CD independent of earlier levels of ODD/CD (Taylor, Chadwick, Heptinstall, & Danckaerts, 1996; see Lahey et al., 2000, and Loeber et al., 2000, for reviews of this literature). To our knowledge, there are no longitudinal studies, however, that have separated the IN and H/I aspects of ADHD to provide a more precise test of the hypothesis. In other words, though the hypothesis concems the influence of the hyperactivity and impulsivity aspects of ADHD in the development of ODD/CD, the various studies have investigated the influence of ADHD (inattention, hyperactivity, and impulsivity) on the development of ODD/CD (e.g., August, MacDonald, Realmuto, & Skare, 1996; August, Realmuto, Joyce, & Hektner, 1999; Biederman et al., 1996; Taylor et al., 1996). In addition, it is important to make a distinction between ODD and CD in the context of this hypothesis. It may be the case that H/I influences the development of ODD and, given the strong relation between ODD and CD (Lahey et al., 2000; Loeber et al., 2000), H/I does not predict future levels of CD after controlling for the relation between ODD and CD.

A longitudinal design in conjunction with structural equation procedures provides a way to test these hypotheses about the developmental relations between the H/I and ODD factors. In the current study, the IN, H/I, and ODD factors were measured at yearly intervals for 2 years in a sample of 752 children. These children were in kindergarten through fifth grade in the 1st year of the study and in the second through seventh grade in the 3rd year. Each child's behavior on the symptom dimensions was rated by a different teacher each year to reduce the influence of source effects. Our primary predictions were (1) that the H/I factor would longitudinally predict higher levels of the ODD factor and (2) that the ODD factor would not longitudinally predict higher levels of the H/I factor. We expected that the H/I factor would predict higher levels of ODD at subsequent assessments even after taking into account the ability of ODD to predict itself across time.

Though not primary hypotheses, we also expected that the ODD factor would not longitudinally predict higher levels of the IN factor and that the IN factor would not longitudinally predict higher levels of the ODD factor. We did not, however, make predictions about the developmental relations among the IN and H/I factors. Given the high correlation between the IN and H/I factors in the CFA studies, each factor might influence the development of the other across time. On the other hand, each factor may be such a strong predictor of itself across time that the other factor has no influence above the ability of the factor to predict itself.

A longitudinal design with structural equation procedures allows for a specific test of these hypotheses. Structural equation procedures, however, require constructs (latent factors) with internal validity (Burns, Walsh, Patterson, et al., 1997; Moura et al., 2001; Waldman, Lilienfeld, & Lahey, 1995). Internal validity requires that each symptom assigned to a particular factor (e.g., the H/I factor) have a stronger relation with that factor than with the other two factors (i.e., the ODD and IN factors). As a first step, we evaluated the internal validity of individual IN, H/I, and ODD symptoms. Each symptom had to show significant internal validity to be included in the second step. The second step involved the evaluation of a three factor measurement model consisting of IN, H/I, and ODD factors. This measurement model had to provide a good fit in an absolute sense in Years 1, 2, and 3 prior to the evaluation of the structural model. The first two steps, the evaluation of the internal validity of the individu al symptoms and the evaluation of the measurement model, are critical for the evaluation of the hypotheses in the structural model (i.e., the measurement of each construct needs to be psychometrically sound prior to the evaluation of hypotheses in the structural model, Byrne, 1994, chap. 7). To our knowledge, this is the first study to use these steps prior to the use of structural equation procedures to evaluate the structural relations among the IN, H/I, and ODD constructs in a longitudinal study. As noted above, this study sought to evaluate these hypotheses with a population-based sample of kindergarten through fifth grade children.

METHOD

Measures

Sutter-Eyberg Student Behavior Inventoy (SESBI)

The SESBI measures 36 disruptive behaviors. The teacher indicates on a 7-point scale how often each behavior currently occurs; 1 (never), 2 and 3 (seldom), 4 (sometimes), 5 and 6 (often), and 7 (always). The teacher also indicates if the occurrence of the behavior is currently a problem by circling "yes" or "no" for each behavior. This results in a frequency score and a problem score for each item. Our analyses only used the frequency of occurrence score.

A series of studies indicate that SESBI scores have high internal consistency, interrater, and test-retest reliability (Burns & Owen, 1980; Bums, Sosna, & Ladish, 1992; Burns, Walsh, & Owen, 1995; Funderburk & Eyberg, 1989; McNeil, Eyberg, Eisenstadt, Newcomb, & Funderburk, 1991; Teegarden & Burns, 1993). These studies also provide support for the validity of the SESBI as a measure of disruptive behavior. Teegarden and Bums (1993), for example, showed that children with high SESBI scores relative to children with normal SESBI scores were observed by raters to be more disruptive and off-task during classroom activities 7 months after the initial SESBI ratings. In addition, in the McNeil et al. (1991, p. 141) study, high SESBI scores were associated with ADHD, ODD, and CD diagnoses. For the 10 children in their treatment group, 6 met criteria for ADHD and ODD, 3 for ADHD, ODD, and CD, and 1 for ADHD. The McNeil et al. (1991) study also showed that SESBI scores improved as a function of parent--child interaction therapy (see McMahon & Estes, 1997, pp. 150-152 for a more detailed review of the reliability and validity of the SESBI).

Child and Adolescent Disruptive Behavior Inventory--Teacher Scale (CADBI-TS)

The CADBI-TS measures 32 of the 36 DSM-III-R disruptive behavior disorder symptoms. The measure includes the 14 ADHD symptoms, the 9 ODD symptoms, and 9 of the 13 CD symptoms. The sexual assault, stealing with confrontation, cruelty to animals, and running away CD symptoms were not included because of their expected low frequency of occurrence in school settings. The wording of the symptoms on the scale was similar to the DSM-III-R description with one exception, the term "often" was not included in the description (e.g., the ODD symptom "often loses temper" was listed on the scale as "loses temper").

The teacher indicates how often each symptom has occurred in the past month on a 7-point scale; 1 (the behavior has not occurred in the past month), 2 (the behavior liar occurred one time in the past month), 3 (the behavior has occurred twice in the past month), 4 (the behavior has occurred one to four times per week in the past month), 5 (the behavior has occurred once per day in the past month), 6 (the behavior has occurred two to five times per day in the past month), and 7 (the behavior has occurred more than five times per day in the past month). The 32 items were randomly arranged in the scale.

Several studies provide support for the reliability and validity of the CADBI-TS as a measure of disruptive behavior (Bums, Walsh, Owen, et al., 1997; Skansgaard & Bums, 1998; Teegarden & Bums, 1999). Burns et al. (1995), for example, provided support for the structural validity of the scale through CFA. Skansgaard and Burns (1998), as an example of predictive validity, showed that teacher ratings on the IN, H/I, and ODD dimensions predicted the direct observation of the same dimensions in the classroom in a specific manner. This study also found test-retest values for the IN, H/I, and ODD dimensions of .86, .94, and .93, respectively, for an 11-week interval. The scale has thus demonstrated high levels of internal consistency, test-retest reliability, predictive, and structural validity in the earlier studies.

Assignment of Items to Inattention, Hyperactivity/Impulsivity, and Oppositional Defiant Disorder Factors

Two PhD level clinical-child psychologists sorted the SESBI items into five categories. The five categories were (1) items similar to ADHD-IN symptoms, (2) items similar to ADHD H/I symptoms, (3) items similar to ODD symptoms, (4) items similar to CD symptoms, and (5) an other category (i.e., items that did not fit into the first four categories). The two judges agreed on 31 of the 36 items. The five disagreements occurred on the items dawdles in obeying rules or instructions (ADHD-IN vs. ODD),pouts (ODD vs. other), makes noises in class (ADHD-H/I vs. other), careless with books and other objects (ADHD-IN vs. other), and uncooperative in group activities (ODD vs. other). Table I shows IN, H/I, and ODD SESBI items that the two judges assigned to the same factor. There were four items assigned to the IN factor, four to the H/I factor, and the nine to the ODD factor. The study did not include the CD items because the occurrence of these symptoms was too low to use in the analyses.

Table I also shows the CADBI items used in the current study. For the 14 ADHD items, 6 items were assigned to the IN factor and 7 to the H/I factor. The final item, does dangerous things without thinking, was not included because this item was eliminated as a symptom for ADHD in DSM-IV. The DSM-III-R ODD symptom swears was also not included in the current analyses because this symptom was eliminated as a symptom for ODD in DSM-IV. This resulted in eight items representing the ODD factor. The CD items were not included because their occurrence was too low for use in the analyses.

Participants and Procedures

The SESBI was completed on the entire population of children in kindergarten through fifth grade in a rural eastern Washington school district in November of 1989 (40 teachers rated 1,116 children). In November of Year 2 (1990), the SESBI was again completed on the entire population of children in kindergarten through sixth grade in the same school district (46 teachers rated 1,286 children). In November of the 3rd year (1991), the CADBI was completed on the entire population of kindergarten through seventh grade children (54 teachers rated 1,445 children). The CADBI was introduced in the third year of the study because the school district wished to have a more direct measure of the DSM-III-R symptoms of ADHD, ODD, and CD than provided by the SESBI. In each year, the four kindergarten teachers taught a morning and afternoon class, thus Year 1 involved 45 classes, Year 2:50 classes, and Year 3:58 classes.

Teachers were asked to volunteer to participate in the study. In each year, every teacher volunteered to participate, thus resulting in ratings on the entire population of children in the grade range of the particular year. The ratings were anonymous (i.e., teachers and students were identified only with numbers with school district personnel also administrating the scales and then providing the scales to the researchers for analysis). Substitute teachers were used to provide the teachers with time away from classroom duties to complete the scales.

Of the original 1,116 children in Year 1 (kindergarten through fifth-grade children), there were 752 children who were rated by a different teacher in Years 2 and 3 (319 children moved out of the district and 45 children were rated by the same teacher). The 45 children rated by the same teacher were not included because our goal was to evaluate our hypotheses when each child was rated by a different teacher in each year of the study. Analyses were performed on measures from Year 1 to determine if the 752 children that remained in the study differed from the other two groups, the 319 children that moved from the district and the 45 children rated by the same teacher. A Group x Gender multivariate analysis of variance on the ODD, H/I, and IN SESBI dimensions for the 1st year indicated that only the multivariate effect for gender was significant, F(3, 1108) = 4.63, p = .003. Boys had significantly higher scores than girls had on each of the three dimensions. The three groups also did not differ in the percentage of boys and girls in each group, [chi square](2) = 3.17, p = .21. In addition, the three groups did not differ significantly in terms of age, F(2, 1113) = 3.11, p = .05, though there was tendency for the 45 children rated twice by the same teacher to be slightly older than the other two groups (i.e., approximately 0.60 years older). The 752 children that remained in the study thus did not differ significantly from the other two groups in terms of age, gender, or scores on the symptom dimensions at the initial assessment.

In Year 1, the 752 children were rated by 40 teachers, with each teacher rating an average of 18.80 children (SD = 6.25). In Year 2, the ratings were provided by 43 teachers, with each teacher rating an average of 17.49 children (SD = 3.79), and in Year 3,44 teachers rated an average of 17.09 children (SD = 4.53). A better design might have been for each teacher to rate only one child to ensure that each rating was independent of each other rating (i.e., 752 teachers rate 752 children), but in this situation the characteristics of the individual teachers and individual students would have been confounded. Having pairs of teachers rate the same students would have been ideal, but the organization of the classes did not allow for such. To evaluate the issue of each teacher rating multiple students, in an earlier study in this research program (Bums et al., 1995, p. 458) we compared the variance associated with systematic differences among students to the variance associated with systematic differences among tea chers (i.e., a comparison of variance components). These results indicated that much more of the total variance in the scores was due to systematic differences between the children as opposed to systematic differences between the teachers.

For the 752 children, 136 children were in kindergarten, 136 in the first grade, 140 in the second grade, 120 in the third grade, 94 in the fourth grade, and 126 in the fifth grade at the start of the study in the 1st year. A total of 372 of the children were girls and 380 boys. Approximately (information was available for the total sample at each year but not for specific children) 87% of the children were Caucasian, 8% Asian American, 3% African American, 2% Hispanic, and 1% Native American. In the 1st year, 4.92% of the children were receiving special services from the school (e.g., resource room for learning disorders), 15.16% services for reading problems, and 2.93% counseling from the school psychologist for behavioral problems. Similar percentages occurred in the 2nd and 3rd years. The study occurred in the Pullman School District. The town of Pullman had a population of approximately 23,000 in the 3rd year of the study, with about 16,000 of this total being students at Washington State University. The median family income for the town at the time of the 1990 census was $36,515.

RESULTS

Analytic Approach

The analyses involved three sequential steps. In the first step, the study determined the internal validity (convergent and discriminant) of the individual symptoms. For a symptom to have internal validity, the symptom had to have a significantly stronger correlation (corrected item-total correlation) with its own dimension than with the other two dimensions. If a symptom failed to show significant internal validity, the symptom was not included in the evaluation of the measurement and structural models. The purpose of this step was to ensure that each symptom on a particular dimension had significant internal validity, especially discriminant, prior to the evaluation of the measurement model.

CFA was then used in the second step to determine if a three factor model consisting of IN, H/I, and ODD factors provided a good fit. This measurement model was evaluated in the first, second, and third years. The purpose of the second step was to determine the adequacy of the measurement model prior to the evaluation of the structural model.

In the third step, structural equation procedures were used to determine the relations among the IN, H/I, and ODD factors across the 2-year interval. The first structural model began with the within factor paths from Time 1 to Time 2 (three paths) and the within factor paths from Time 2 to Time 3 (three paths). These paths deal with the ability of each factor to predict itself across each time interval (i.e., standardized partial regression coefficients). The Lagrange Multiplier (LM) test was then used to determine if any of 18 additional paths were significant and therefore should be added to the structural model. These 18 paths included the 6 diagonal paths from time one to time two, the 6 diagonal paths from time two to time three, and the six diagonal paths from time one to time three. The inclusion of the within factor paths in the model prior to the evaluation of the diagonal paths provided a more stringent test of the hypotheses (i.e., the diagonal paths have to make a contribution above and beyond the within factor paths to be significant). Our primary hypothesis was that the diagonal paths from the prior H/I factor to the subsequent ODD factor would be significant, whereas the diagonal paths from the prior ODD factor to the subsequent H/I factor would not be significant.

Internal Validity of the Individual ODD, H/I, and IN Symptoms

Meng, Rosenthal, and Rubin's test for dependent correlations was used to evaluate each symptom's internal validity (Meng, Rosenthal, & Rubin, 1992). For a symptom to demonstrate internal validity, the symptom was required to have a significantly higher corrected item-total correlation with its own dimension than with the other two dimensions. The per-comparison alpha was set at p <.00009 to maintain a familywise error rate of p <.01 for the 110 comparisons (i.e., 34 at Time 1, 34 at Time 2, and 42 at Time 3).

ODD Symptoms

The first segment of Table II shows the symptom validity results for the ODD symptoms at Time 1 and Time 2. The obey symptom failed to show discriminant validity with the H/I dimension at Time 1 and Time 2. The other eight ODD symptoms showed significant discriminant validity with the H/I and IN dimensions. The first segment of Table III shows the symptom validity results for the ODD symptoms at Time 3. The annoys symptom failed to show discriminant validity with the H/I dimension. The other seven ODD symptoms showed significant discriminant validity with H/I and IN dimensions.

H/I Symptoms

The second segment of Table II shows the results for the four H/I symptoms at Time 1 and Time 2. Each of these symptoms showed significant discriminant validity with the H/I and ODD dimensions. The second segment of Table III shows the symptom validity results for the H/I symptoms at Time 3. Here the turn symptom failed to show discriminant validity with the ODD dimension and the fidgets symptom failed to show discriminant validity with the IN dimension. The other five H/I symptoms showed significant discriminant validity with the IN and ODD dimensions.

IN Symptoms

The third sections of Tables II and III show the symptom validity results for the IN symptoms at Times 1, 2, and 3. Each of the IN symptoms showed significant discriminant validity with the H/I and ODD dimensions at each time period.

Preliminary Scale Analyses

Table IV shows the internal consistency (Cronbach's alpha), means, and standard deviations for the ODD, H/I, and IN dimensions for each time period. This descriptive information on the scales is based on the items that showed significant internal validity (i.e., if an item failed to show internal validity, the item was not included in the scale). Each of the scales had a similar level of high internal consistency across the three assessments (greater than .90). In terms of change in the dimensions across time, there was no significant change in H/I scores over the three assessments. For the ODD and IN dimensions, there was no significant change from the first to the second assessment. The ODD and IN dimensions, however, showed a significant decrease at the third assessment (per-comparison [alpha] < .001 to maintain a familywise error rate of p < .01 for the nine comparisons). The amount of variance accounted for by the time of measurement factor was 2% for the ODD and IN dimensions. The magnitude of this decr ease for the ODD and IN dimensions was small.

Measurement Models

EQS-Version 5.7b was used to perform the CFA (Bentler, 1995). Maximum likelihood estimation was used in conjunction with the robust estimation procedure. The EQS Comparative Fit Index (CFI), the EQS Robust Comparative Fit Index (RCFI), the standardized root mean squared residual (SRMR), and the root mean square error of approximation (RMSEA) were used to evaluate model fit. Table V shows the results of the CFA.

The three factor model consisting of ODD, H/I, and IN factors provided a good fit in an absolute sense as well as significantly better fit than two (ADHD and ODD) and one factor models at each time period. The multivariate LM test, however, indicated that a significant (p < .00001) and substantial improvement in model fit would occur with one set of correlated errors at each time period (i.e., interrupts1 with interrupts2 at Time 1 (.68), interrupts 1 with interrupts2 at Time 2 (.70), and blurts with interrupts (.61) at Time 3). The CFI, RCFI, SRMR, and RMSEA values for the three factor model with one correlated error at Time 1 were .95, .93, .059, and .092, at Time 2 94, .92, .057, and .101, and at Time 3 .93, .93, .056, and .087, respectively. At each of the time periods, the factor loadings were also significant (ps < .00001). There was good support for the three factor measurement model.

Structural Model

EQS was next used to determine the structural relations among the factors across time. For this analysis, given their substantial degree of interrelatedness, the three factors were allowed to correlate at Time 1, and at time periods 2 and 3 their disturbances were allowed to correlate. In addition, the same three error pairs were allowed to correlate as in the CFA.

As a first step, the structural analysis included the six within-factor paths (i.e., Time 1 to Time 2: ODD at Time 1 to ODD at Time 2, H/I at Time 1 to H/I at Time 2, IN at Time 1 to IN at Time 2; Time 2 to Time 3: ODD at Time 2 to ODD at Time 3; H/I at Time 2 to H/I at Time 3; IN at Time 2 to IN at Time 3). The LM test was then used to determine if any of 18 diagonal paths would result in a significant improvement in the model (i.e., the six diagonal paths from Time 1 to Time 2, the six diagonal paths from Time 2 to Time 3, and the six diagonal paths from Time 1 to Time 3). Given that 18 path coefficients were evaluated with the LM test, the per-comparison alpha was set at p < .0006 to maintain a familywise error rate of p < .01. The LM test indicated that 3 of the 18 paths were significant; the H/I at Time 1 to ODD at Time 2, the H/I at Time 2 to the ODD at Time 3, and the H/I at Time 1 to ODD at Time 3.

Figure 1 shows this model with the six within factor paths and the three paths from the H/I factor to the ODD factor. This structural model resulted in a good fit (i.e., CFI = .92, RCFI = .91,SRMR = .077,RMSEA = .059; 90 CI = .057-.061. The standardized partial regression coefficient from H/I at Time 1 to ODD at Time 2 was .18, the coefficient from H/I at Time 2 to ODD at Time 3 was .22, and the coefficient from H/I at Time 1 to ODD at Time 3 was .12. Even when taking into account the within factor paths, higher scores on the H/I factor still predicted higher scores on ODD across 1- and 2-year intervals. The strength of this prediction from the H/I factor at Year 2 to the ODD factor of Year 3 was similar in magnitude to the ability of ODD at Time 2 to predict ODD at Time 3 (.22 vs. .24 coefficients, respectively). In addition, even when taking into account the ability of H/I at Time 2 to predict ODD at Time 3, H/I at Time 1 still predicted ODD at Time 3. The loadings of the symptoms on their respective factor s in this structural model are shown for Time 1 and 2 in Table VI and for Time 3 in Table VII. These values were not put in the figure because the inclusion of the 50 variables would have made the figure difficult to read.

The structural model in Fig. 1 involved 121 free parameters for a subject to parameter ratio of 6.21 to 1. This represents an adequate ratio (Pillow et al., 1998, p. 298), especially given the use of the EQS robust estimation procedure (Raykov & Marcoulides, 2000, p. 27). Although the sample size was very marginal for the evaluation of the equivalence of the structural model across boys and girls (i.e., the subject to parameter ratio for boys was 3.14 and for girls 3.07), the model was still tested for invariant causal structure across gender. This analysis involved 62 constraints--the item-factor loadings at each time period, the factor correlations at Time 1, the disturbance correlations at Time 2, the disturbance correlations at Time 3, the six within factor path coefficients, and the three H/I to ODD path coefficients (see Byrne, 1994, chap. 11, for a description of such an analysis). The per-comparison alpha was set at p < .00016 to maintain a familywise error of p < .01 for the evaluation of the 62 cons traints.

The univariate and multivariate LM tests indicated that the same four of these constraints were significant. The four significant constraints involved the gets angry symptom at Time 1, the loses symptom at Time 3, the correlation of the H/I and ODD factors at Time 1, and the correlation of the IN and H/I disturbance at Time 3. These four parameters were not equivalent across gender. For the remaining 58 constraints, a total of 46 had per-comparison p values greater than .05. Thus, in addition to the six within factor and three H/I to ODD path coefficients being equivalent for boys and girls, each of these path coefficients were also significant for each gender (ps < .001). Although these results indicated that the structural model held for boys and girls, the results should be considered as tentative given the small subject to parameter ratio.

The subject to parameter ratio was too small to test the model for invariance across the six grade groupings within the longitudinal design (i.e., kindergarten to second; first to third; second to fourth; third to fifth; fourth to sixth; and fifth to seventh grade). Here the ratios ranged from 0.78 to 1.16.

DISCUSSION

Our goal was to provide a more specific test of the hypothesis that H/I behaviors foster the development of ODD behaviors. As a first step, we used separate measures of the IN, H/I, and ODD dimensions. Earlier studies in this area have addressed the question without a separation of the IN and H/I dimensions. As a second step, we focused on the ability of the H/I dimension to predict ODD rather than CD given our hypothesis that H/I is more crucial to the development of ODD than CD. As a third step, we insured that the IN, H/I, and ODD dimensions had good discriminant validity prior to the test of the structural hypotheses. Discriminant validity required that each symptom have a stronger correlation with its own dimension than with the other two dimensions and that the three factor measurement model (IN, H/I, and ODD factors) provides a good fit in an absolute sense.

Structural equation procedures in conjunction with a longitudinal design were then used to determine the structural relations among the IN, H/I, and ODD factors. This methodology provides an ideal way to evaluate the hypotheses. The results showed that higher scores on the H/I factor in year one were associated with higher scores on the ODD factor in Years 2 and 3. In addition, higher scores on the H/I factor at Year 2 predicted higher scores on the ODD factor at Year 3. The H/I factor was able to predict higher scores on the ODD factor in subsequent years even after taking into account the ability of the ODD factor to predict itself across time (as well as the ability of the H/I and IN factors to predict themselves across time).

Finally, and perhaps the most interesting finding in this context, higher scores on the H/I factor at Time 1 were able to predict higher scores on the ODD factor at Time 3 even after taking into account the ability of the H/I factor at Time 2 to predict the ODD factor at Time 3. This relation also occurred after taking into account the ability of the ODD factor to predict itself across time. H/I at Time 1 was thus able to uniquely predict ODD at Time 3 even after taking into account the predictive paths closer in time to ODD at Time 3. In contrast to these results, the ODD factor did not predict higher levels of the H/I factor across the 2-year interval. Together these findings provided support for the hypothesis that H/I behaviors foster the development of ODD behaviors.

The study also found, as predicted, that the ODD factor did not predict higher levels of the IN factor at the subsequent assessments as well as that the IN factor did not predict higher levels of the ODD factor across time. In addition, though we did not make specific predictions, the IN and H/I factors did not predict each other across time after taking into account the ability of each factor to predict itself.

LIMITATIONS

One limitation of the study was that the assessment of the IN, H/I, and ODD factors was based on a single source, teachers. Although the children were rated by a different teacher in each year of the study, it would have been a stronger study if parent ratings had been available as well. Parent ratings would have also allowed us to perform the analyses at the level of summary scores rather than at the individual symptom level (i.e., each factor would have been represented by one parent and one teacher summary score). In addition, though our sample size was large enough to test the structural model in the total sample as well as provide tentative support for the invariance of the structural model across gender, the sample size was not large enough to evaluate the model within the various age groups. It will be important in subsequent research to determine if the hypothesis is supported for boys and girls within various age ranges.

Another issue was that the IN and H/I symptoms in the current study were not a perfect match to the DSM-IV IN and H/I symptoms. The match was close enough, especially for the H/I symptoms, that the results will probably generalize to the exact DSM-IV symptom set. Future studies with the DSM-IV symptoms will still need to insure that the symptoms used in the evaluation of the structural hypotheses have strong discriminant validity (Bums & Walsh, 2001; Bums, Walsh, Owen, et al., 1997; Bums, Walsh, Patterson, et al., 1997; Moura et al., 2001). In other words, tests of structural hypotheses should not occur without first evaluating the internal validity of the individual symptoms as well as the validity of the particular measurement model.

A final issue concerns the non-clinical nature of our sample. On the positive side, the sample involved the entire population of kindergarten through fifth grade children in one school district that remained in the district for two consecutive years. Within this context, support was found for the hypothesis that H/I behaviors foster the development of ODD behaviors in a general sample. The replication of the study with an at-risk sample will be an important future step. Children ages 3-4 years who score high on the H/I dimension could be considered a group of children at risk for the development of ODD. With multiple measures by multiple sources of the IN, H/I (Olson, Schilling, & Bates, 1999), and ODD constructs and frequent assessments for several years (e.g., every 3 or 6 months), it would be possible to determine the impact of H/I behaviors on the development of ODD behaviors in an at-risk sample. Such a study could also measure the critical parenting variables (e.g., Dishion & Patterson, 1999) assumed to facilitate the development of ODD. Finally, such a study could add the CD construct to test the hypothesis that it is ODD that predicts CD rather than H/I. Such designs have the potential to greatly increase the understanding of the developmental relations among the IN, H/I, ODD, and CD factors. Such designs also have the potential to facilitate our understanding of the impact of environmental processes on the constructs as well as the impact of the constructs on environmental processes (e.g., the influence of the parenting factor on H/I factor vs. the impact of H/I factor on the parenting factor). Finally, such designs have the added benefit of being able to test specific theoretical predictions. For these reasons, this methodology has a great deal of potential to further our understanding of the IN, H/I, and ODD constructs.

Received October 27, 2000; revision received April 2, 2001; accepted August 16,2001

(1.) Department of Psychology, Washington State University, Pullman, Washington.

(2.) Department of Psychology, University of Montana, Missoula, Montana.

(3.) Address all correspondence to G. Leonard Burns, Department of Psychology, Washington State University, Pullman, Washington 99164; e-mail: glburns@mail.wsu.edu.

REFERENCES

August, G. J., MacDonald, A. W., Realmuto, G. M., & Skare, S. S. (1996). Hyperactive and aggressive pathways: Effects of demographic, family, and child characteristics on children's adaptive functioning. Journal of Clinical Child Psychology, 25,341-351.

August, G. J., Realmuto, G. M., Joyce, T., & Hektner, J. M. (1999). Persistence and desistance of oppositional defiant disorder in a community sample of children with ADHD. Journal of the American Academy of Child and Adolescent Psychiatry, 38, 1262-1270.

Beiser, M., Dion, R., & Gotowiec, A. (2000). The structure of attentiondeficit and hyperactivity symptoms among native and non-native elementary school children. Journal of Abnormal Child Psychology, 28, 425-437.

Bentler, P. M. (1995). EQS: Structural equations programs manual. Encino, CA: Multivariate Software.

Biederman, J., Faraone, S., Milberger, S., Guite, J., Mick, E., Chen, L., et al. (1996). A prospective 4-year follow-up study of attention-deficit hyperactivity and related disorders. Archives of General Psychiatry, 53, 437-446.

Burns, G. L., & Owen, S. M. (1980). Disruptive behaviors in the classroom: Initial standardization data on a new teacher rating scale. Journal of Abnormal Child Psychology, 18, 515-525.

Burns, G. L., Sosna, T. D., & Ladish, C. (1992). Distinction between well-standardized norms and the psychometric properties of a measure: Measurement of disruptive behaviors with the Sutter-Eyberg Student Behavior Inventory. Child and Family Behavior Therapy, 14, 43-54.

Burns, G. L., & Walsh, J. A. (2001). Importance of the discriminant validity of the individual ADHD. ODD, and CD symptoms for the external validity of the three disorders: A reexamination of the symptom utility estimates from the DSM-IV Field Trial. Manuscript submitted for publication.

Burns, G. L., Walsh, J. A., & Owen, S. M. (1995). Twelve month stability of disruptive classroom behavior as measured by the Sutter-Eyberg Student Behavior Inventory. Journal of Clinical Child Psychology, 24,453-462.

Burns, G. L., Walsh, J. A., Owen, S. M. & Snell, J. (1997). Internal validity of attention deficit hyperactivity Disorder, oppositional defiant disorder, and overt conduct disorder symptoms in young children: Implications from teacher ratings for a dimensional approach to symptom validity. Journal of Clinical Child Psychology, 26, 266-275.

Burns, G. L., Walsh, J. A., Patterson, D. R., Holte, C. S., Sommers-Flanagan, R., & Parker, C. M. (1997). Internal validity of the disruptive behavior disorder symptoms: Implications from parent ratings for a dimensional approach to symptom validity. Journal of Abnormal Child Psychology, 25, 307-319.

Byrne, B. M. (1994). Structural equation modeling with EQS and EQS/Windows: Basic concepts, applications, and programming. Thousand Oaks, CA: Sage.

Dishion, T. J., & Patterson, G. R. (1999). Model building in developmental psychopathology: A pragmatic approach to understanding and intervention. Journal of Clinical Child Psychology, 28, 502-512.

DuPaul, G. J., Anastopoulos, A. D., Power, T. J., Reid, R., McGoey, K. E. & Ikeda, M. J. (1998). Parent ratings of ADHD symptoms: Factor structure, normative data, and psychometric properties. Journal of Psychopathology and Behavioral Assessment, 20, 83-102.

DuPaul, G. J., Power, T. J., Anastopoulos, A. D., Reid, R., McGoey, K. E., & Ikeda, M. J. (1997). Teacher ratings of attention deficit hyperactivity disorder symptoms: Factor structure and normative data. Psychological Assessment, 9, 436-444.

Fergusson, D. M., & Horwood, L. J. (1993). The structure, stability and correlations of the trait components of conduct disorder, attention deficit and anxiety/withdrawal report. Journal of Child Psychology and Psychiatry, 34, 749-766.

Funderburk, B. W., & Eyberg, S. M. (1989). Psychometric characteristics of the Sutter-Eyberg Student Behavior Inventory: A school behavior rating scale for use with preschool children. Behavioral Assessment, 11,297-313.

Gomez, R., Harvey, J., Quick, C., Scharer, I., & Harris, G. (1999). DSM-IV AD/HD: Confirmatory factor models. prevalence, and gender and age differences based on parent and teacher ratings of Australian primary school children. Journal of Child Psychology and Psychiatry and Allied Disciplines, 40, 265-274.

Lahey, B. B., McBurnett, K., & Loeber, R. (2000). Are attentiondeficit/hyperactivity disorder and oppositional defiant disorder developmental precursors to conduct disorder. In A. J. Sameroff, M. Lewis, & S. M. Miller (Eds.), Handbook of developmental psychopathology (2nd ed., pp.43 1-446). New York: Kluwer.

Loeber, R., Green, S. M., Lahey, B. B., Frick, P. J., & McBurnett, K. (2000). Findings on disruptive behavior disorders from the first decade of the developmental trends study. Clinical Child and Family Psychology Review, 3, 37-60.

McGee, R., & Williams, S. (1999). Environmental risk factors in oppositional-defiant disorder. In H. C. Quay & A. E. Hogan (Eds.), Handbook of disruptive behavior disorders (pp. 419-440). New York: Kluwer.

McMahon, R. J., & Estes, A. M. (1997). Conduct problems. In E. J. Mash & L. G. Terdal (Eds.), Assessment of childhood disorders (3rd ed., pp. 130-193). New York: Guilford.

McNeil, C. B., Eyberg, S., Eisenstadt, T. H., Newcomb, K., & Funderburk, B. (1991). Parent-child interaction therapy with behavior problem children: Generalization of treatment effects to the school setting. Journal of Clinical Child Psychology, 20, 140-151.

Meng, X.-L., Rosenthal, R., & Rubin, D. B. (1992). Comparing correlated correlation coefficients. Psychological Bulletin, 111, 172-175.

Moura, M. A., Bums, G. L., & Walsh, J. A. (2001). Internal validity of the ADHD and ODD symptoms in Brazilian children: What confirmatory factor analysis does not tell us about the internal validity of individual symptoms. Manuscript submitted for publication.

Olson, S. L., Schilling, E. M., & Bates, J. E. (1999). Measurement of impulsivity: Construct coherence, longitudinal stability, and relationship with externalizing problems in middle childhood and adolescence. Journal of Abnormal Child Psychology, 27, 151-165.

Pillow, D. R., Pelham, W. E., Hoza, B., Molina, B. S. G., & Stultz, C. H. (1998). Confirmatory factor analyses examining attention deficit hyperactivity disorder symptoms and other childhood disruptive behaviors. Journal of Abnormal Child Psychology, 26,293-309.

Raykov, T., & Marcoulides, G. A. (2000). A first course in structural equation modeling. Mahwah, NJ: Erlbaum.

Samudra, K., & Cantwell, D. P. (1999). Risk factors for attentiondeficit/hyperactivity disorder. In H. C. Quay & A. E. Hogan (Eds.), Handbook of disruptive behavior disorders (pp. 199-220). New York: Kluwer.

Sanson, A., & Prior, M. (1999). Temperament and behavioral precursors to Oppositional defiant disorder and conduct disorder. In H. C. Quay & A. E. Hogan (Eds.), Handbook of disruptive behavior disorders (pp. 397-418). New York: Kluwer.

Skansgaard, E. P., & Burns, G. L. (1998). Comparison of DSM-IV ADHD combined type and predominantly inattentive types: Correspondence between teacher ratings and direct observations of inattention, hyperactivity/impulsivity, slow cognitive tempo, oppositional defiant, and overt conduct disorder symptoms. Child and Family Behavior Therapy, 20, 1-14.

Taylor, E., Chadwick, O., Heptinstall, E., & Danckaerts, M. (1996). Hyperactivity and conduct problems as risk factors for adolescent development. Journal of the American Academy of Child & Adolescent Psychiatry, 35, 1213-1226.

Teegarden, L. A., & Bums, G. L. (1993). Construct validity of the Sutter-Eyberg Student Behavior Inventory: Relation between teacher perception of disruptive behavior and direct observation of problem classroom behavior over a seven month interval. Child ad Family Behavior Therapy, 15, 43-58.

Teegarden, L. A., & Burns, G. L. (1999). 12-month stability of ADHD, oppositional defiant disorder, and conduct disorder symptoms in kindergarten through fifth grade children based on a single information source: Usefulness of teacher ratings for the creation and study of the ADHD subtypes. Child and Family Behavior Therapy, 21,53-70.

Waldman, I. D., Lilienfeld, S. O., & Lahey, B. B. (1995). Toward construct validity in the childhood disruptive behavior disorders: Classification and diagnosis in DSM-IV and beyond. In T. H. Ollendick & R. J. Prinz (Eds.), Advances in clinical child psychology (Vol. 17, pp. 323-363). New York: Plenum.

Wolraich, M. L., Feurer, I. D., Hannah, J. N., Baumgaertel, A., & Pinnock, T. Y. (1998). Obtaining systematic teacher reports of disruptive behavior disorders utilizing DSM-IV. Journal of Abnormal Child Psychology, 26. 141-152.

COPYRIGHT 2002 Plenum Publishing Corporation
COPYRIGHT 2002 Gale Group