University of Florida. Evaluated generalization of treatment eflects from home to school setting in ten 2- to 7-year-old children who were referred for treatment of ...
Journal of Clinical Child Psychology 1991, Vol. 20, No. 2, 140-151
Copyright 1991 by Lawrence Erlbaum Associates, Inc.
Parent-Child Interaction Therapy With Behavior Problem Children: Generalization of Treatment Effects to the School Setting Cheryl Bodiford McNeil Department of Pediatrics University of Oklahoma Health Sciences Center
Sheila Eyberg Department of Clinical and Health Psychology University of Florida
Toni Hembree Eisenstadt University of Alabama
Katharine Newcomb and Beverly Funderburk Department of Clinical and Health Psychology University of Florida
Evaluated generalization of treatment eflectsfrom home to school setting in ten 2to 7-year-old children who were referred for treatment of severe conduct problem behaviors occurring both at home and in the classroom. Families received 14 weeks of parent-child interaction therapy. No direct classroom interventions were conducted. The treatment group displayed significantly greater improvements than two control groups on all measures of conduct problem behavior in the classroom. Results in the areas of hyperactivity/distractibility and social behavior were less supportive of generalization. Positive school generalization results contradict previous findings that children's behavior in the classroom either shows minimal improvement or worsens following parent training. Within the parent training literature, there have been relatively few studies examining school generalization, and these studies provide little evidence to suggest that school behavior improves without direct intervention in the classroom. In a controlled group study of the relative effectiveness of three treatment programs for child behavior problems, it was found that school generalization did not occur with group behavioral parent training, group selfcontrol training, or a combination of parent and self-control training (Horn, Ialongo, Popovich, & Paradotto, 1987). A second controlled group study found that a behavioral parent training program targeting specific problematic behaviors in the home actually resulted in an increase in problematic behaviors in the school setting, suggesting a "behavioral contrast effect" (Johnson, Bolstad, & Lobitz, 1976). More positive school generalization results Requests for reprints should be sent to Sheila Eyberg, Department of Clinical and Health Psychology, Box J-165 Health Sciences Center, University of Florida, Gainesville, FL 32610.
have been reported in several investigations using such interventions as social learning family therapy and parent discussion groups combined with videotaped modeling (Cox & Matthews, 1977; Sayger & Horne, 1987; Taylor & Hoedt, 1974; Webster-Stratton, Kolpacoff, & Hollinsworth, 1988). These results are difficult to interpret, however, because of such methodological difficulties as failure to obtain pretreatment evaluations, dependence on a single teacher report measure for assessment of classroom adjustment, or lack of appropriate comparison groups. Two additional investigations have addressed generalization of treatment effects to the school setting following an intervention based on the Hanf (1969) parent-child interaction model. More favorable classroom findings might be expected from this type of treatment because evidence suggests that other types of generalization occur following Hanf model interventions. However, in contrast to findings that treatment effects generalize across time (e.g., Baum & Forehand, 1981; Forehand et al.,
SCHOOL GENERALIZATION
1979; Webster-Stratton, 1984), to untreated siblings (Eyberg & Robinson, 1982; Humphreys, Forehand, McMahon, & Roberts, 1978), and from the clinic to the home setting (e.g., Boggs, 1990; Forehand & McMahon, 1981; Peed, Roberts, & Forehand, 1977), two controlled group studies of parent-child interaction have failed to support school generalization. Forehand et al. (1979) found that five of eight treated children, in contrast to four of eight untreated controls, demonstrated more inappropriate classroom behavior after a Hanf model intervention. These findings were interpreted as being suggestive of a behavioral contrast effect (Forehand et al., 1979). Breiner and Forehand (1981) showed that, following a parent-child interaction treatment approach, treated children did not display greater classroom behavioral improvement from pre- to posttreatment than did normal cointrol children. They concluded that "if school problems exist, they will not be reduced by treatment directed toward home problems. School problems will have to be directly programmed into an overall treatment package in order to successfully reduce them" (p. 41). This conclusion regarding school generalization may be premature due to two methodological issues of concern in these parent-child interaction studies. First, to evaluate cross-setting generalization, it is necessary to document that treatment resulted in clinically significant improvements in the primary setting. In the studies of school generalization just noted, child compliance in the home increased 5% (Breiner & Forehand, 1981,30% to 35%; Forehand et al., 1979,88% to 93%) and home deviant behavior decreased 3% (from 9% to 6%) after treatment (Breiner & Forehand, 1981). Because of the small magnitude of change in home behavior, it may be that the children's overall adjustment was not affected to such a degree that improvements could be noted across settings. Second, to evaluate school generalization, it is necessary to document that the treatment group children have conduct problems in the classroom to which the positive changes in home behavior can generalize. Children in both of these studies displayed normal classroom behavior before treatment, making it difficult to demonstrate significant improvements in the school setting. The purpose of our study was to evaluate school generalization in referred children who demonstrated severe conduct problem behalviors in both the home and school settings before Parent-Child Interaction Therapy (PCIT) and who demonstrated clinically significant improvements in home behavior after treatment. Comparisons of the treated children to both normal and deviant classroom controls were conducted on multiple measures to evaluate the clinical significance of school changes.
Method Subjects Subjects were 30 children between the ages of 2 and 7 years old. Twenty-four of the children attended preschool, 3 were in kindergarten, and 3 attended first grade. There were three subject groups: the treatment group (TG; rz = lo), nonnal classroorn controls (NC; n = lo), and untreated deviant classroom controls (DC; n = 10). The TG consisted of 10 parent-child dyads referred far treatment of child behavior problems at home and at school. All TG children met the following criteria: (a) received parent ratings 011the Eyberg Child Behavior Inventory (ECBI; Eyberg, 1974) above the published cutoff scores for child deviancy (intensity score > 127; problem score > 11; Eyberg & Ross, 1978); (b) demonstrated a compliance ratio (averaged arcross three semistructured situations) less than that typically shown by nonreferred children ( .05). Thus, the results did not demonstrate improved social behavior in the treatment children as a result of decreased levels of deviant behavior in the presence of peers.
Table 2. Comparisons of Treatment Groupato Untreated Deviant Controlsband Normal Controlscon Observational Measures of Classroom Behavior
Post
Pre
Pre
Post
Pre
Post
Measure
M
SD
M
SD
M
SD
M
SD
M
SD
M
SD
F
% appropriate 5% compliance % on task
65 54 69
14 23 14
87 87 82
14 16 16
78 64 80
12 18 16
89 75 80
5 15 10
89 73 86
8 14 6
90 80 84
9 13 16
8.09** 4.24* 1.75
Note: Pairwise comparisons for % appropriate and % compliance showed that the treatment group's scores were greater than those of both the DCs and NCs, all p s < .05, There were no differences between the two latter groups. an = 10. bn = 8. 'n = 9. *p < .05. **p < .01.
Table 3. Comparisons of Treatment Groupato Untreated Deviant Controlsband Normal Controlscon Teacher Report Measures of Classroom Behavior TG
DC
Pre Measure RCTRS Conduct Problem HyperactivityIndex SESBI Intensity Problem
M 1.6 1.9 155 20
Post SD .4 .6 21 6
M 1.0 1.4 116 10
NC
Pre SD .6 .8
34 7
M 1.3 1.5 126 13
Post SD .7 .5 36 8
M 1.2 1.3 115 10
Pre SD .6 .7
35 9
M .35 .28 57 .67
Post SD .2 .3 15 1
M .33 .36 67 1.1
SD .2 .4 21 2
F 13.90**~ 4.51*e 9.8~**~ 14.00**~
an = 10. bn = 8. Cn = 9. d ~ e s u l t of s pairwise comparisons: The TG's scores were greater than those of both the DCs and NCs, all ps < .05. There were no significant differences between the latter two groups. eResults of painvise comparisons: The TG's score was greater than that of the NC, p < .05, but did not differ from that of the DC. The DC's score was greater than that of the NC, p < .05. *p < .05. **p .01.
SCHOOL GENERALIZATION
Table 4. Summary of School Generalization Results for Individual Children in the Treatment Group on Measures of Behavior Problems in the Classroom Subject Measure Observational measures Appropriate Compliance On task Teacher report measures SESBI Intensity Problem RCTRS Conduct problem Hyperactivity Index
1
1!
+ + +
444-
+ + +
+ t-
3
+
t-
4
+ + + +
5
6
7
8
9
10
+
Note: + signifies improvement of greater tha.n 30% relative to pretreatment. - signifies worsening by more than 30% relative to pretreatment.
In these initial group comparison results, it is important to note that classroom comparison children were less deviant at pretreatment than children referred for treatment of school behavior problems, and the NC group was close to ceiling on some measures, which can confound interaction effects. To minimize potential regression artifacts, the TG and DC groups were compared using analyses of covariance (ANCOVA) with pretreatment scores used as covariates. The results of the 3 X 2 ANOVA were replicated using the ANCOVA for the following variables: SESBI problem score, F(l, 15) = 5.03, p = .038; RCTRS conduct problem score, F(l, 15) = 4.71, p = .044; RCTRS Hyperactivity Index, F(1, 15) = 2.63, ns; percentage of compliance in the classroom, F (1, 15) = 4 . 5 9 , ~= .047; percentage of time on task in the classroom, F(l, 15) = 0.37, ns; and all Walker-McConnell scales, ns. The SESBI intensity score approached significance, F(l, 15) = 4.05, p = .059. The only ANCOVA result in direct contradiction with that obtained using ANOVA was for percentage of time appropriate in tlhe classroom. Although the ANOVA results indicated that the TG children improved significantly more than the DC children on this measure of disruptive behavior, when baseline between-group differences were controlled for in the ANCOVA, the result failed to obtain statistical significance, F(1, 15) = 2.24, ns. Qualitative analyses sf the behavior of the individual TG children were conducted to evaluate whether a subgroup of children displayed either no improvement or behavioral decrements that could not be detected in the group comparisons. Because Eyberg and Johnson (1974) found a 30% change in pre- to posttreatment scores to be an effective cutoff value for determining the clinical significance of behavioral change in individual families, the 30% cri-
terion was used in our study to interpret the classroom behavioral changes of individual TC; children. Summary results of the pre- to posttreatment changes in classroom conduct problem behaviors for individual TG children are presented in 'Table 4. Inspection of the table in~dicatesthat all of the TG children improved on at least one of the seven measures of classroom behavior problems, and over half of the children improved on at least three of these measures. This suggests that -thepositive sch.001generalization results in the area of oppositional/conduct problem behavior are due to relatively consistent improvements across individual TG children, rather than to a positive response for only a small subgroup of the sample,.Inspection of Tab1.e 4 also 'indicates that a behavioral decrement was evident for only one child and on only one measure, providing little evidence to support the presence of ix behavioral contrast effect.
Discussion Results of this study indicate that the successful treatment of home behavior problems using PCIT is associated with improvem~entsin certain bethaviors in the school setting. This is the first cointrolled group study of parent management training in which both classroom observations and teacher report measures have shown significant pre- it0 posttreatment improvements in conduct problem1 behaviors. As such, our findings (contradicttwo pr
