Examining Ethnicity as a Moderator of Treatment Outcome for Children with ADHD Using Parent-Child Interaction (PCI) Ratings

Abstract

The vast parenting literature has shown differences in parenting practices across different ethnicities. However, parenting practices of parents of children with Attention-Deficit/Hyperactivity Disorder (ADHD) have primarily been studied in Caucasian middle-class families. Given that modifying parenting practices is often a primary target for psychosocial intervention for children with ADHD, ethnic differences in parenting may moderate treatment outcomes. Utilizing observations of 567 children with ADHD (ages 7–9) and their caregivers collected during the Multimodal Treatment Study of ADHD (MTA), the present study examined ethnic differences in observed parenting and child behavior at baseline and the moderating effects of ethnicity on the relationship between treatment and parenting and child behavior. The four MTA treatments were medication management, behavior therapy, combined medication management + behavior therapy, and community care. Baseline results indicated that both parenting practices and child behavior differed by ethnicity. However, ethnicity did not moderate the relationship between treatment and parenting and child behavior. Hence, children from different ethnicities did not differentially benefit from one treatment over another on observed ratings of parenting or child behavior. Implications for clinical care and future research are discussed.

Commonly used parenting practices have been found to differ among parents of different ethnic groups (Baumrind, 1966; Baumrind, 1972; Bradley, Corwyn, Pipes McAdoo, & Garcia-Coll, 2001; Brody & Flor, 1998; Durrett, O’Bryant, & Pennebacker, 1975; Florsheim, Tolan, & Gorman-Smith, 1996; Lamborn, Dornbusch, & Steinberg, 1996). For example, on average, African American parents are more likely than Caucasian parents to use physical discipline strategies (Bradley et al., 2001) and display lower levels of warmth towards their children (Kelley, Power, & Wimbush, 1992). These strategies are consistent with an authoritarian parenting style (e.g., high levels of parental control, lower levels of warmth; Baumrind, 1966). While most researchers have long assumed that authoritative parenting (e.g., high levels of parental warmth and the promotion of autonomy; Baumrind, 1966) is the most effective form of parenting in terms of promoting appropriate child behavior, research with African American families has challenged this view. Several studies have demonstrated that African American parenting strategies consistent with authoritarian parenting are associated with children who obtain better grades, demonstrate more social competence, and have less psychopathology (Brody & Flor, 1998; Lamborn et al., 1996; Portes, Dunham, & Williams, 1986). Also, physical discipline, which is thought to be an authoritarian parent behavior, may not be related to higher externalizing behavior ratings in African American children (Deater-Deckard, Dodge, Bates, & Pettit, 1996), but may be in Caucasian children (Baumrind, 1966; Lamborn, Mounts, Steinberg, & Dornbusch, 1991). Also, the warmth of the parent-child interaction may moderate the relationship between physical discipline and child externalizing behaviors (Deater-Deckard & Dodge, 1997).

Research on parenting practices in Latino/Hispanic families has been mixed. For instance, Mexican American parents have been rated as relatively permissive, nurturing, and egalitarian in some studies (e.g., Delgado, 1980; Durrett et al., 1975; Vega, 1990), but as authoritarian or authoritative in other studies (e.g., Chilman, 1993; Harrison, Wilson, Pine, Chan, & Buriel, 1990; Martinez, 1988). Furthermore, warm and supportive parenting in Latino/Hispanic families has been linked to fewer child conduct problems (Florsheim et al., 1996). Contrary to the research on African American and Latino/Hispanic parenting practices, research on Caucasian parent-child dyads seems to indicate that authoritative parenting may be the most successful form of parenting for Caucasian families in terms of being associated with few child behavior problems and positive child adjustment (Baumrind, 1966; Lamborn et al., 1991).

Much of the research examining linkages between parenting practices and child outcomes across ethnic groups has employed self-report of parenting practices. Self-report of parenting can be problematic when examining relations between parenting and child behavior since typically both child behavior and parenting behavior share method variance (i.e., parent report) which can confound observed relationships between parenting and child behavior. A more objective measurement strategy is to observe parents and children during parent-child interactions.

Studying ethnic differences in parenting styles using objective observation has been limited to the general population. Very little is known about ethnic differences in parenting practices in families with children with behavioral difficulties, such as those with Attention-Deficit/Hyperactivity Disorder (ADHD). It is well-documented that the presence of a child with ADHD has a significant impact on parenting and the parent-child relationship (for a review, see Johnston & Mash, 2001). Specifically, children with ADHD have been observed to be less compliant and more negative with their parents than children without ADHD as observed during structured interactions (Anderson, Hinshaw, & Simmel, 1994; Campbell, 1973; DuPaul, McGoey, Eckert, & VanBrakle, 2001). In turn, parents of children with ADHD are less structured and less warm with their children than parents of children without ADHD (Anderson et al., 1994; Cunningham & Barkley, 1979; DuPaul et al., 2001; Mash & Johnston, 1982). More specifically, mothers of children with ADHD have been observed to be more authoritarian with their children as evidenced by more negative and more directive behavior towards their children than mothers of children without ADHD (Cunningham & Barkley, 1979; Mash & Johnston, 1982). Parents of children with ADHD also perceive themselves as less warm and more controlling (Gerdes, Hoza, & Pelham, 2003; Gerdes et al., in press). In addition, mothers of boys with ADHD self-reported lower levels of authoritative parenting practices and beliefs than did mothers of a comparison group (Hinshaw, Zupan, Simmel, Nigg, & Melnick, 1997). Ethnic differences in these parenting behaviors have not been examined in parents of children with ADHD.

In intervention studies of children with ADHD, ineffective parenting practices are a target of treatment (for a review, see Pelham, Wheeler, & Chronis, 1998). Therapists train parents to use more effective parenting strategies consistent with an authoritative parenting style, such as the increased use of labeled praise for compliance and non-physical discipline strategies for non-compliance. It is possible that such training may be differentially effective across ethnicity because of observed baseline differences in parenting practices, philosophies, and effects of parenting practices on child behavior by ethnicity. For example, Bradley et al. (2001) have shown that an authoritarian parenting style is related to low levels of child misbehavior among African American families. Hence, teaching African American families to be less authoritarian, and more authoritative, could possibly be a less effective strategy (maybe even counterproductive) for this ethnic group. On the other hand, in a study of 634 low-income Head Start parents of different ethnic groups, Reid, Webster-Stratton, and Beauchaine (2001) found that those parents who received parent training, regardless of ethnicity, were less critical and more positive with their children after parent training. Further, children of parents receiving intervention showed fewer behavior problems than children of control parents. The authors were not able to conclude that the parent training was differentially effective across the various ethnicities.

The largest treatment outcome study of children with ADHD is the Multimodal Treatment Study of ADHD (MTA; MTA Cooperative Group, 1999a; 1999b). The MTA study compared medication management (MedMgt), intensive behavioral treatment (Beh), combined medication management and intensive behavioral treatment (Comb), and community care (CC) for a large and diverse sample of children with ADHD-Combined Type. Analyses of self-reported parenting showed that negative discipline improved as a result of Beh, MedMgt, or Comb treatments compared to CC (Wells, Epstein, et al., 2000). In addition, reductions in negative discipline mediated improvements in children’s disruptive behavior and social skills at school, such that families receiving the Comb treatment and showing maximal improvement in such self-reported discipline practices had children with clinically significant improvement by teacher ratings (Hinshaw et al., 2000). Arnold and colleagues (2003) examined ethnicity as a moderator of treatment effects in the MTA study, focusing on reported child behavior as the primary treatment outcome. That report found that ethnic minority children as a group benefited significantly from the addition of behavioral treatment to medication management, but Caucasian children did not. Although there were differences in outcome between African American and Caucasian children on teacher reports of ADHD and ODD symptoms, there were no differences on parent reports of these same symptoms. Latino children also showed benefit from the addition of the behavioral treatment on parent ratings of ODD symptoms. The significant interactions of treatment and ethnicity on outcome attenuated to nonsignificance when socioeconomic status variables were covaried, except for the general difference between Caucasians and pooled minorities. The authors concluded that factors related to both socioeconomic status and ethnicity may influence treatment outcome for children with ADHD. However, only child behavior was examined in this study; ethnicity as a moderator of parenting outcomes in MTA treatments has not been examined. Examining the role of ethnicity on parenting behavior may help to explain the moderating role of ethnicity on treatment outcome for child behavior outcomes.

The goals of the current study are twofold. First, baseline ethnic differences in observed parenting practices and child behavior are described using the MTA sample of children with ADHD and their parents. Second, the moderating effects of ethnicity on the relations between MTA interventions and both observed parenting practices and child behavior is addressed. Both questions are examined using observational coding of parent and child behavior during analog laboratory settings.

Method

MTA Sample

Details of the MTA design and sample have been reported elsewhere (Arnold et al., 1997a, 1997b; Hinshaw et al., 2000; MTA Cooperative Group, 1999a). Briefly, 579 children ages 7–9.9 with rigorously diagnosed DSM-IV ADHD, combined type (American Psychiatric Association, 1994) were recruited from multiple sources at 6 sites. In the entry demographics form, parents were asked to designate the child’s ethnicity in standardized categories. Most (n = 567) of these children had observational data. Of the 567 who had observational data at baseline there were 348 Caucasians, 113 African Americans, 47 Latinos, and 59 belonging to “other ethnic groups.” See Table 1 for demographic details (e.g., gender, age, comorbidity) within each ethnicity. The MTA study was approved by Institutional Review Boards at each of the participating institutions.

Table 1.

Demographic Characteristics of Entire Sample

Demographic characteristic	Caucasian (n = 348)	African American (n = 113)	Latino (n = 47)	χ2 or F	p
Age, M (SD)	8.4 (0.9)	8.3 (0.8)	8.4 (0. 9)	1.5	.22
Maternal education, M (SD)	4.4 (1.1)	4.0 (1.1)	3.5 (0.9)	15.0	.00
% Male	79.3	79.6	89.4	2.7	.26
Single parent, %	19.0	52.2	36.2	49.7	.00
Public assistance, %	7.2	36.3	48.9	85.1	.00
Baseline comorbid anxiety disorder, %	31.0	34.5	44.7	4.7	.10
Baseline cormorbid ODD/CD, %	52.9	56.6	53.2	0.5	.78

Note. ODD = oppositional-defiant disorder; CD = conduct disorder; for Maternal education, a higher score indicates more education on the 6-level scale.

Independent Variable: MTA Design and Treatment Groups

In a 4-group parallel design, children were randomly assigned to one of four treatment arms for 14 months of intervention: MedMgt, Beh, Comb, or CC. All interventions were delivered at each of the six MTA sites for 14 months. Each active MTA treatment arm was designed as a management strategy, sufficiently robust and flexible to stand on its own and respond to clinical needs. MedMgt began with a 4-week, double-blind, placebo-controlled titration period followed by monthly medication management (see Greenhill, Abikoff, Arnold, & Cantwell, 1996). Beh included 35 sessions of group and individual parent training, an intensive, child-focused 8-week summer treatment program (Pelham & Hoza, 1996), 12 weeks of a half-time paraprofessional aide in the classroom, and school-based teacher consultation, all delivered in a coordinated fashion (see Wells, Pelham, et al., 2000). Participants in the CC group received no active MTA treatments, but families were provided a report of their initial assessments, along with a list of community mental health services and resources. Periodic follow-up assessments have occurred over the ensuing years. Analyses reported in this study were performed on data obtained at the pre-treatment baseline (BL) and 14-month post-treatment assessments.

Observational Codings

The laboratory task

Laboratory observations of parent–child interactions (PCIs) were included in the MTA to assess parent–child relationships objectively. Identical procedures were followed at BL and 14 months; observations were made of interactions between the child and the child’s primary caregiver. Three separate segments of parent-child interaction were coded, including (a) the child’s sitting quietly while the caregiver was working (Parent Busy), (b) a schoolwork assignment (School Work), and (c) playroom clean-up (Clean-Up).

During the Parent Busy segment, a research assistant entered the room, handed the parent a work packet to complete and left before video-recording commenced. The parent was instructed to tell the child to stop playing and sit quietly on a chair in the corner (out of reach of the toys and games) without interrupting him/her. After 3 minutes, the child was temporarily escorted out of the room by another research assistant and the parent was given further instructions about the last 10 minutes of the PCI. In the School Work segment (5 minutes) the parent was instructed to give the child a work packet (consisting of age-appropriate spelling or math questions) and assist the child, if necessary, on the school work. After 5 minutes, the parent instructed the child to stop working on the school work and Clean-Up the room for 5 minutes by putting away the work packet, the toys, and some paper on the ground.

Codes and rater training

As described extensively in Wells et al. (2006), six trained research assistants from three MTA sites—who were blind to the diagnostic status of the participants, treatment assignment, and the hypotheses of the study—rated the parent and child PCI behaviors. The research assistants used a 6-point Likert metric (1 = Very Poor, 6 = Excellent; wording adjusted for different categories) to measure the quality of both parental (5 codes) and child (4 codes) behaviors during the PCI segments. These relatively global codes were based on past literature regarding the parental and child behaviors pertinent to ADHD, as discussed below; a more finely-grained, microanalytic system would have been unfeasible given the magnitude of the sample size.

The five parental behaviors included Setting Stage (parent prepares the child for upcoming events by giving a synopsis of what is going to happen and his/her expectations of the child), Behavior Management (parent set limits when child negative behavior is high and “backs off” when child is behaving appropriately), Annoyance (level of verbal and non-verbal anger and annoyance expressed by the parent toward the child; this item was reverse scored), Positive Reinforcement (rate of verbal and non-verbal positive reinforcement and praise expressed by the parent toward the child), and Warmth (level of positive emotional expression toward the child by the parent). Note that Setting Stage was not scored for the Clean Up segment because Setting Stage was not relevant for this activity. The child behaviors included Complaining/Whining (how often the child showed dissatisfaction and/or displeasure with the task, with the parent-child interaction, or any other aspects of the situation), Verbal Abuse (how often the child directed a negative verbal/non-verbal communication toward the parent such as a derogatory name, a hand gesture, or anything verbal/non-verbal that would usually elicit a clear indication of distress from the parent), Compliance (how often the child followed the parent’s directions), and Likable (how generally likable is the child). These parenting and child behaviors were chosen because research has shown the importance of balancing both limit-setting and positive reinforcement with hyperactive–aggressive children (Patterson, 1995) and the centrality of child defiance in parent–child discord in ADHD children (Barkley, Fischer, Edelbrock, & Smallish, 1990). Likable was included as a way to tap the general impression of the child by the raters as generally positive or generally negative.

Raters received extensive training in the coding system during a 2-day training seminar hosted by the MTA site that developed the PCI coding system (UC Berkeley). Prior to training, each rater was required to memorize the operational definitions of the 9 codes by passing a definition exam. Daily training consisted of group instruction on the PCI codes, re-coding criterion tapes (which had already been coded by the training staff), and daily feedback sessions. After the training seminar, all existing PCI tapes were randomly assigned to the 6 raters, with each PCI being rated at least twice, across different rater pairs.

After initial training, continuous monitoring of rater reliability, fidelity, and drift were carried out through the coding of randomly assigned “gold standard” PCI tapes (50 families, approximately 5% of the total taped sessions). Over the next four months, these “gold standard” tapes were rated by the trainers and contrasted with those of the raters. Weekly phone supervision sessions were conducted by the trainers with the raters to discuss discrepancies, technical problems, and other conceptual questions related to the PCI coding system. Agreement between raters and trainers on the “gold standard” tapes ranged from.81 to .97 (alpha coefficients) for parenting behaviors across the 6 coders. For child behaviors, alpha coefficients ranged from .76 to .98 (Wells et al., 2006). We also assessed the proportion of instances in which raters diverged more than 2 points from the training staff ratings. Instances of large discrepancy were relatively infrequent during the 50 sessions. Fewer than 7% of the parent codes (range = 6–9%) and about 3% of the child codes (range = 0–7.5%) were discrepant by 2 or more points. For these “gold standard” rating sessions, any ratings that were discrepant by more than 1 point were resolved through a reviewing of the session and if discrepancies were still present, a written summary of the discrepant opinion was submitted to the training staff, and individual recalibration sessions were scheduled.

Formal inter-rater reliability between raters across all of the tapes was measured by the average intraclass correlation (ICC) coefficients (Fleiss, 1975). For each behavior across the three parts of the observational measure, reliabilities were adequate (reported in Wells et al., 2006)

Measures for Socioeconomic and Other Potential Confounders

Mother’s educational level, receipt of public assistance, and single-parent status were selected as measures of potential socioeconomic (SES) confounders, based on conceptual considerations and previous findings of effect on treatment outcome in this sample (bMTA Cooperative Group, 1999b; Rieppi et al., 2002). Mother’s education was coded on 6 levels from less than 7^th grade to graduate professional; receipt of public assistance was a dichotomous variable based on anyone in the family receiving public assistance; and single parent status was dichotomous, defined as other than living with 2 parent figures (natural, step, or common-law). Comorbidity was determined by a baseline Diagnostic Interview Schedule for Children–Parent version (DISC-P; Shaffer, Fisher, Dulcan, & Davies, 1996) and was found in previous analyses to affect treatment outcome (bMTA Cooperative Group, 1999b; Jensen et al., 2001; March et al., 2000). Baseline ADHD symptom severity was found in a previous analysis (Owens et al., 2003) to affect treatment outcome.

Overall Analytic Strategy

In order to compare baseline differences in parenting practices and child behavior across ethnicity, the entire sample of Caucasian (n = 352), African-American (n = 115), and Latino/Hispanics (n = 49) were used for analyses. Other ethnic groups (e.g., Asian-American) had too few participants to make meaningful comparisons.

Analysis for ethnic differences in treatment outcomes is complicated because although no site × treatment interactions were found for the primary MTA outcomes, there was a clear main effect of site on many treatment outcomes (aMTA Cooperative Group, 1999a), presumably due to site demographics and possibly subtle differences in selection, management, treatment, and evaluation during the MTA study. Because the ethnic distribution varied across sites, it would not be a fair comparison to use all participants from each ethnic group to compare treatment efficacy. Each participant must be compared only against other participants at the same site in the same treatment group. To resolve such problems at least partially, we adopted the strategy used by Arnold et al. (2003) in their examination of ethnic effects in the MTA sample. They used a matched-pair strategy, controlling for site, treatment group, and sex. Of the 115 African American participants, it was thus possible to match 92. Eight girls and 15 boys could not be matched because of insufficient Caucasians of the same sex and treatment group at their site. For our analyses, one African American-Caucasian pair had to be dropped due to missing baseline observational data, resulting in a total of 91 matches for the current analyses. Latino participants were similarly randomly matched with Caucasians, resulting in 37 matches out of 49 Latinos, with one girl and 11 boys unable to find Caucasian matches.

To examine the baseline differences on the observational task, we conducted univariate ANOVAs using ethnicity as the independent variable and the overall observation coding (averaged across the 3 settings) as the dependent variable. To examine the moderating effects of ethnicity on treatment outcomes, mixed model ANOVAs were conducted utilizing ethnicity and treatment as between-subjects variables and time as a within-subjects variable. For these calculations, we were interested in a significant interaction effect for the ethnicity × treatment × time term. For these analyses, an intent-to-treat analysis with the last observation carried forward for any missing data was used. For all statistically significant effects, Cohen’s d was calculated to help interpret the clinical significance of the findings.

As many factors are often correlated with minority status, such as low maternal education (e.g., Morrison, Rimm-Kauffman, & Pianta, 2003) and living below the poverty line (e.g., Kasarda, 1993), it is important to consider such factors when attempting to draw conclusions about differences in observed parenting behaviors. Consistent with those methods employed by Arnold et al. (2003), maternal education, single parenthood, public assistance/welfare status, baseline comorbid anxiety disorder, baseline comorbid Oppositional Defiant Disorder (ODD) or Conduct Disorder (CD), severity of ADHD symptoms, and severity of ODD symptoms were all examined as potential confounders. First, all ethnic comparisons were first performed on the dependent variables without the confounders. For any significant differences found, ethnic groups were compared on each confounder to identify significant group differences. Finally, for any ethnic comparisons (e.g., African American and Caucasian) comparisons that significantly differed on a confounder (e.g., single parenthood), the main analyses were repeated including that confounder as a covariate.

Results

Baseline Differences

Observational codings

Baseline differences were examined using univariate ANOVA analyses. Table 2 displays a summary of these results. Significant effects for ethnicity were demonstrated for all parent and child behaviors, except Annoyance.

Table 2.

Baseline Means (and standard deviations) and Ethnic Differences on Parent and Child Observational Codings

Outcome Measure	Caucasian (n = 352)	African American (n = 115)	Latino (n = 49)	Ethnicity Main Effect F	Significant post-hoc effects
Parent Behaviors
Setting Stage	3.35 (0.61)	3.10 (0.63)	3.03 (0.50)	11.25**	C>A,L
Behavior Management	4.71 (0.04)	4.86 (0.06)	4.43 (0.82)	5.89*	A>L
Positive Reinforcement	1.89 (0.68)	1.49 (0.53)	1.42 (0.38)	24.97**	C>A,L
Warmth	3.55 (0.50)	3.23 (0.50)	3.23 (0.40)	24.16**	C>A,L
Annoyance	1.34 (0.47)	1.51 (0.70)	1.34 (0.34)	4.58	--
Child Behaviors
Complaining/Whining	2.31 (0.88)	1.80 (0.69)	2.20 (0.92)	15.41**	C,L>A
Verbal Abuse	1.30 (0.50)	1.10 (0.26)	1.28 (0.29)	9.73**	C >A
Compliance	5.19 (0.84)	5.63 (0.54)	5.14 (0.91)	14.38**	A>C,L
Likable	3.20 (0.73)	3.64 (0.53)	3.27 (0.76)	17.23**	A>C,L

Note. Range for all codings was 0–6; C=Caucasian, A=African-American, L=Latino; For Behavior Management, parents of Caucasian children did not differ from any of other parents. For Verbal Abuse, Latino children did not differ from any of the other children.

^*p<.01

^**p<.001

To further investigate this effect, Tukey post hoc tests comparing the different ethnicities were conducted. Caucasian parents had higher ratings on Setting Stage than African American (p < .001, d = .40) and Latino parents (p < .01, d = .57). Caucasian parents also were observed to use higher levels of Positive Reinforcement than African American (p < .001, d = .66) and Latino parents (p < .001, d = .85). Additionally, they were rated as using more Warmth than African American (p < .001, d = .64) and Latino parents (p < .001, d = .71). African American parents displayed higher levels of Behavior Management than Latino parents (p < .01, d = .74), but were not significantly different than Caucasian parents on this measure. Caucasian and Latino parents did not differ in their use of Behavior Management skills. African American parents were not significantly different from Latino parents on Setting Stage, Positive Reinforcement, or Warmth.

Using child behaviors as the dependent variables, significant effects were again detected for ethnicity. Tukey post hoc analyses showed that Caucasian and Latino children evidenced more Complaining/Whining (p < .001, d = .64 and p < .001, d = .49, respectively), less Compliance (p < .001, d = .62 and p < .001, d = .65, respectively), and were less likeable (p < .001, d = .69 and p < .01, d = .56, respectively) than African American children. African American children demonstrated less Verbal Abuse than Caucasian children (p<.001, d = .50), but did not significantly differ from Latino children. Caucasian and Latino children did not significantly differ on any outcome variables.

Correlational analyses

Correlations between the behavioral codings for Caucasian and African American and Latino children and their parents are presented in Table 3. Overall, there was a pattern of correlations showing relations between better child behavior (e.g., less complaining, less verbal abuse, more compliance) and increased parental Behavioral Management and decreased parental Annoyance. Positive Reinforcement did not show much of a relationship to child behavior. Parental Warmth was significantly related to child behavior but only moderately so, and primarily among Caucasian families. No correlations differed across ethnicities (all ps>.05), though there were patterns suggesting that parental Positive Reinforcement was positively related to child Complaining among African Americans (r=.22) but not among Caucasians (r=.02; difference between correlations p=.06). In addition, among Caucasians, parental Annoyance correlated with children’s Verbal Abuse scores (r=.34) while this correlation was much lower and non-significant among African-American families (r=.14; difference between correlations p=.0503).

Table 3.

Correlation Matrix for Behavioral Codings Reported Separately for Caucasian (n=352), African American (n=115), and Latino (n=49) Families

	2			3			4			5			6			7			8
	C	A	L	C	A	L	C	A	L	C	A	L	C	A	L	C	A	L	C	A	L
1. Behavior Management	.14	−.05	−.11	.37	.38	.24	−.36	−.49	−.24	−.46	−.45	−.59	−.36	−.28	−.55	.47	.41	.62	.53	.41	.69
2. Positive Reinforcement				.46	.45	.40	−.05	−.09	.00	.02	.22	.20	−.06	.14	.39	.07	.02	−.08	.10	−.02	−.14
3. Warmth							−.46	−.59	−.51	−.20	−.05	−.19	−.21	−.06	−.02	.15	.23	.15	.23	.16	.24
4. Annoyance										.36	.25	.29	.34	.14	.27	−.42	−.38	−.26	−.46	−.31	−.36
5. Complaining													.54	.50	.69	−.52	−.49	−.63	−.69	−.73	−.78
6. Verbal Abuse																−.57	−.50	−.61	−.70	−.69	−.77
7. Compliance																			.84	.82	.78
8. Likability

Note: Bolded typeface indicates p<.05; C=Caucasian; A=African American; L=Latino.

Confounder analyses

To identify confounders, chi-square goodness of fit tests and ANOVAs were conducted to examine ethnic group differences. Significant ethnic group differences emerged for maternal education [χ²(10) = 43.74, p < .001], single parenthood [χ²(2) = 50.14, p < .001], public assistance [χ²(2) = 91.60, p < .001], and comorbid anxiety disorder [χ²(2) = 6.43, p < .05]. See Table 1. Ethnic groups did not differ on severity of ADHD symptoms, ODD symptoms, or comorbid ODD/CD. When the main analyses were repeated with those confounding variables that differed across ethnicity included as covariates, the reported pattern of results remained the same. All significant ethnicity effects remained significant after entering potential confounding variables.

Treatment Differences

Analyses of MTA treatment outcomes using the post-treatment objective PCI data showed improved parenting behaviors for the groups receiving the psychosocial treatments (Beh and Comb; Wells et al., 2006). In the current study, we examined whether ethnicity may have been a moderating variable for any effect of treatment on parent or child behavior. This was explored using 3 (ethnicity) × 4 (treatment) × 2 (time) ANOVA analyses. There were no significant interaction effects of ethnicity × treatment × time on any parent behaviors or child behaviors for either the African American-Caucasian or Latino-Caucasian matched pairs (Table 4). See Table 5 for scores across treatment groups. In order to assess whether the lack of significant effects was due to the complexity of the interaction effect being modeled, we conducted an additional set of analyses using change scores and dropping Time from the analyses. This paralleled the analyses conducted by Arnold et al. (2003), in which significant effects of ethnicity were demonstrated. There continued to be no significant interaction between ethnicity and treatment for any of the outcomes.

Table 4.

Three-way Interaction (Ethnicity × Time × Treatment) Results Examining Ethnicity as a Moderator of Treatment Outcome

	3-way Interaction
	C-A	C-L
Outcome Measures	F	F
Parent Behaviors
Setting Stage	1.44	0.18
Behavior Management	2.75	2.65
Positive Reinforcement	0.65	0.74
Warmth	0.89	0.87
Annoyance	2.52	1.89
Child Behaviors
Complaining/Whining	1.96	0.35
Verbal Abuse	1.43	1.55
Compliance	1.11	1.45
Likable	2.07	0.09

Note: C-A = Caucasian-African American matched pairs; C-L = Caucasian-Latino matched pairs.

Table 5.

Baseline and 14-month Follow-up Observational Codings across Treatment Groups for Each Ethnicity^*

	Baseline scores						14-month follow-up scores
	Caucasian		African American		Latino		Caucasian		African American		Latino
Outcome Measures	M	SD	M	SD	M	SD	M	SD	M	SD	M	SD
BEHAVIORAL TREATMENT
Parent Behaviors
Setting Stage	3.44	0.64	3.06	0.50	3.08	0.41	3.60	0.69	3.17	0.45	3.02	0.51
Behavior Management	4.54	0.74	4.79	0.63	4.45	0.89	5.28	0.60	5.20	0.46	4.85	0.65
Positive Reinforcement	2.07	0.72	1.41	0.28	1.45	0.50	2.35	1.03	1.65	0.75	1.83	0.75
Warmth	3.60	0.64	3.16	0.40	3.20	0.49	3.74	0.53	3.23	0.40	3.20	0.37
Annoyance	1.45	0.58	1.65	0.92	1.40	0.26	1.14	0.38	1.30	0.40	1.48	0.36
Child Behaviors
Complaining/Whining	2.48	0.59	1.87	0.80	2.39	1.29	1.61	0.78	1.85	1.10	2.20	1.20
Verbal Abuse	1.28	0.39	1.20	0.39	1.38	0.39	1.12	0.26	1.14	0.32	1.46	0.53
Compliance	5.17	0.68	5.60	0.69	5.06	1.10	5.88	0.24	5.76	0.42	5.40	0.64
Likable	3.09	0.57	3.55	0.64	3.01	1.01	3.84	0.35	3.64	0.56	3.29	0.85
MEDICATION MANAGEMENT
Parent Behaviors
Setting Stage	3.45	0.74	2.95	0.82	3.04	0.50	3.05	0.58	2.96	0.56	2.96	0.25
Behavior Management	4.77	0.92	4.81	0.67	4.57	0.70	5.13	0.68	5.13	0.46	5.09	0.73
Positive Reinforcement	1.72	0.61	1.43	0.44	1.34	0.23	1.98	0.81	1.41	0.36	1.47	0.43
Warmth	3.51	0.52	3.19	0.61	3.06	0.50	3.48	0.50	3.24	0.33	3.14	0.24
Annoyance	1.56	0.74	1.48	0.76	1.45	0.58	1.25	0.48	1.26	0.31	1.25	0.38
Child Behaviors
Complaining/Whining	2.18	0.80	1.85	0.71	2.15	0.95	1.78	0.65	1.59	0.62	1.76	0.65
Verbal Abuse	1.54	0.86	1.12	0.29	1.17	0.15	1.12	0.16	1.07	0.20	1.21	0.38
Compliance	4.92	1.05	5.56	0.57	5.37	0.69	5.63	0.55	5.81	0.32	5.67	0.38
Likable	2.99	0.94	3.65	0.64	3.38	0.77	3.66	0.40	3.77	0.37	3.74	0.29
COMBINED TREATMENT
Parent Behaviors
Setting Stage	3.20	0.39	3.06	0.64	3.07	0.45	3.26	0.39	3.32	0.76	3.25	0.38
Behavior Management	4.94	0.46	4.72	0.72	4.21	0.71	5.15	0.72	5.32	0.44	4.99	0.74
Positive Reinforcement	1.86	0.77	1.56	0.63	1.53	0.54	1.96	0.67	1.68	0.78	1.90	0.95
Warmth	3.59	0.43	3.24	0.60	3.36	0.31	3.45	0.33	3.41	0.43	3.46	0.49
Annoyance	1.08	0.12	1.56	0.65	1.22	0.23	1.28	0.45	1.16	0.31	1.21	0.26
Child Behaviors
Complaining/Whining	1.97	0.72	1.70	0.64	1.87	0.53	1.73	0.66	1.48	0.60	2.17	1.41
Verbal Abuse	1.19	0.40	1.10	0.21	1.26	0.25	1.15	0.23	1.04	0.09	1.46	1.09
Compliance	5.35	0.70	5.65	0.41	4.85	1.23	5.58	0.48	5.86	0.22	5.39	0.79
Likable	3.40	0.67	3.65	0.49	3.34	0.64	3.64	0.48	3.91	0.22
COMMUNITY COMPARISON
Parent Behaviors	3.26	0.72	3.31	0.58	2.96	0.80	3.14	0.58	3.07	0.58	2.98	0.28
Setting Stage	4.57	0.97	5.19	0.60	4.19	1.18	5.13	0.58	5.16	0.51	5.12	0.26
Behavior Management	1.71	0.57	1.78	0.85	1.44	0.39	1.81	0.61	1.47	0.49	1.69	0.39
Positive Reinforcement	3.55	0.60	3.36	0.48	3.19	0.37	3.46	0.56	3.25	0.54	3.42	0.48
Warmth	1.22	0.37	1.36	0.44	1.36	0.19	1.23	0.33	1.44	0.45	1.46	0.55
Annoyance
Child Behaviors	2.59	1.46	1.85	0.66	2.25	0.94	1.85	0.90	1.76	0.68	1.57	0.65
Complaining/Whining	1.31	0.43	1.05	0.09	1.30	0.27	1.31	0.31	1.07	0.12	1.12	0.13
Verbal Abuse	5.10	1.14	5.78	0.30	5.44	0.53	5.73	0.26	5.89	0.13	5.62	0.22
Compliance	3.13	0.83	3.74	0.35			3.72	0.36	3.75	3.35	3.29	0.90

^*Means and standard deviations reported for Caucasians and African Americans are from the Caucasian-African American pairs analyses. Means and standard deviations reported for Latinos are from the Caucasian-Latino pairs analyses.

Discussion

Using coded interactions between children with ADHD and their parents, ethnic differences in observed parenting and child behavior were examined. Differences were found between the different ethnic groups on most baseline measures of observed parent and child behavior. However, ethnicity did not moderate treatment effects.

Several significant baseline ethnic differences between parent-child dyads were found. Parents of Caucasian children in the MTA, on average, demonstrated higher levels of Setting Stage, Positive Reinforcement, and Warmth than parents of children of other ethnicities. All of these behaviors are consistent with parenting strategies comparable to the authoritative parenting dimension (Baumrind, 1966), such as providing warmth and positive reinforcement. This finding is consistent with the literature on rates of authoritative parenting in Caucasian families (Park & Bauer, 2002). Parents of African American children in the MTA, on average, were observed to use higher levels of Behavior Management than parents of children of other ethnicities. This may be consistent with the reliance on more unilateral and authoritarian strategies in African American families (Baumrind, 1972; Brody & Flor, 1998; Lamborn et al., 1996; Portes et al., 1986). Although the literature on parenting in Latino families has been mixed, the current results indicate that parents of Latino children may parent similarly to parents of African American children, at least in an analog situation as used in this study.

Despite higher levels of parental Warmth, Positive Reinforcement, and Setting Stage, Caucasian children with ADHD demonstrated significantly more negative behaviors than African American children with ADHD, even though comparable rates of ODD/CD diagnoses were documented across these ethnic groups. This finding was unexpected given that children from this same sample of African American children exhibited more misbehavior than Caucasian children on observational codings collected in the school setting (Epstein et al., 2005). Without codings of teacher behavior in these classrooms, we can only speculate what contextual factors promote greater rates of misbehavior among African American children in the classroom setting and less in the laboratory setting. It is possible that the analog setting itself, either defined by its novelty or its structure, may have had a beneficial effect on the behavior of African American children or a detrimental effect on the Caucasian and Latino children. Alternatively, it could be that school settings have a detrimental effect on the behavior of African American children or a beneficial effect on Caucasian children. Another explanation for the present set of findings is that the mere presence of parents, regardless of parenting techniques, has more of a positive behavioral influence on African American children than for other racial groupings.

The use of broad molar codings as opposed to a microanalytic behavioral coding system (e.g., Snyder, 1977) makes sequential analyses of parent-child interaction impossible. Correlations were conducted among these behavioral codings in order to get some insight into parent-child behavior associations and to determine whether parent and child behavior were differentially associated across ethnicities. Overall, there was a pattern towards increased parental behavioral management relating to better child behavior. Also, those parents who were more annoyed had children who complained more often. These correlations are consistent with research demonstrating that an authoritative style of parenting is related to low rates of child misbehavior (Baumrind, 1966; Querido, Warner, & Eyberg, 2002). None of the relations between parent and child behaviors showed differential association across ethnicities suggesting that children respond to parental cues and parents respond to child behavior similarly across the ethnic groups.

Given the findings from Arnold et al. (2003) indicating that Comb was more effective for ethnic minority children than for Caucasian children in reducing reported ADHD symptoms and the findings from Wells et al. (2006) showing treatment-related changes on observed parent-child interactions, we examined whether there were any moderating effects of ethnicity on treatment as measured by observational codings. No such effects were found for observed parent or child behaviors. In other words, children and parents of different ethnicities did not differentially benefit from one treatment over another on the observed measures. This is somewhat surprising as Arnold et al. (2003) reported that, for ODD symptoms, both African American and Latino children benefited more from the behavioral treatments (Beh over MedMgt for African American children and Comb over CC for Latino children) than Caucasian children. As ODD symptoms seem to overlap conceptually with several of the child behaviors (e.g., Complaining/Whining), it is interesting that there was no benefit of the treatments with a behavioral component over the others for African American and Latino children. Back to the explanation offered above, it may be that contextual variables inherent to the analog PCI setting produced an effect on parenting behaviors that exceeded any intervention effects.

The study has several limitations. First, it is difficult to draw conclusions about Latino parenting and child behavior due to the small number of Latino participants examined. Second, ethnicities of parents, therapists, and observers were not obtained during the course of the study thereby preventing an examination of the effects of the ethnicities of parents, therapists, and/or observers on the observed findings. Indeed, there are data to suggest that the ethnic status of observers can affect observational ratings (Gonzales, Cauce, & Mason, 1996; Mueller et al., 1995). Another important limitation lies in the sample of families considered. The MTA was an intensive clinical trial. As such, the data analyzed herein reflects only those families agreeing to participate in such a trial. As ethnic differences exist in treatment acceptability (e.g., Krain, Kendall, & Power, 2005) and more broadly in research participation, the sample included in this study is a select group of families, and results may not be representative of families recruited for effectiveness studies. Also, critics could argue that the controlled and observational nature of the analog design of the parent-child interactions is not comparable to real-life parenting. These interactions were designed to be as ecologically valid as possible, with parents interacting with a child during times when the parent was busy, while the child completed homework, and during clean-up periods. However, these parent-child interactions took place in a laboratory setting in which the parent and child knew that they were videotaped, and their interactions were periodically interrupted by research assistants giving instructions. These methodological limitations may have influenced child and parent behavior, possibly restricting the application of these findings. Additionally, observational coders are typically blind to variables that may influence their behavior ratings, such as treatment status or child diagnosis. However, it is not possible for coders to remain completely blind to ethnicity, as people make judgments about ethnicity based on appearance and behavior. Unfortunately, this limitation was unavoidable for the current study. Finally, the observational codings used in the present study reflect ratings of the overall quality of parent and child behavior. They do not reflect actual behaviors or sequences of behaviors. Without these more molecular codings, the temporal relationships between child and parent behaviors can not be examined. Hence, it is impossible to derive any conclusions as to the causal or temporal relationships between parent behavior and child behavior.

The clinical implications of baseline parenting differences are important to consider in light of the lack of any moderating effect of ethnicity on treatment outcomes. Should clinical researchers concentrate on creating different treatments for different ethnicities? Results from this study do not support that conclusion. However, it may be fruitful to consider how we can deliver the empirically-supported treatments for ADHD, particularly those that focus on teaching parenting strategies (e.g., behavioral parent training) in a culturally-responsible way. Researchers may need to consider how to build on the successful authoritarian parenting in a way to make it more effective as well as warmer and more supportive (i.e. can a velvet glove be fitted to the iron fist?). Parents who demonstrated more Annoyance had children who demonstrated more Complaining/Whining. Also, as parents of African American children exhibit lower baseline levels of Positive Reinforcement and Warmth than parents of Caucasian parenting, perhaps much more time should be spent during behavioral parent training with African American families on these components to ensure that the strategies are well-implemented. As ethnic minority parents are more likely to drop out of treatment (Kazdin, Holland, & Crowley, 1997), future research on such subtle, but formal adaptations of typical parent training has been suggested by some (e.g., Forehand & Kotchick, 1996) to examine whether these adaptations are more amenable to these families and subsequently result in better outcomes.