Mindful Schools

Study 1

Black, D. S., & Fernando, R. (2013). Mindfulness training and classroom behavior among lower-income and ethnic minority elementary school children. Journal of Child and Family Studies, 23, 1242-1246. http://doi.org/10.1007/s10826-013-9784-4

Evaluation Methodology

Design

Recruitment:

The sample was drawn from one elementary school in Richmond, CA. Data were collected from 17 teachers in 17 different classrooms who reported on the behaviors of 409 students in grades kindergarten through sixth.

Assignment:

Classrooms (N=17) were randomly assigned to receive either the Mindful Schools program or the Mindful Schools program plus seven additional once-weekly classes. There was no control group.

Attrition:

Some data were missing from each of the assessment periods: baseline (n=404 of 409), immediately post-intervention (n=364), and seven weeks post-intervention (n=403).

Sample:

The sample was drawn from a public elementary school comprised mainly of low income (83% enrolled in the free lunch program) and ethnic minority (95.7%; 52.3% Hispanic, 28.0% Black, 15.0% Asian, 4.3% White, and .4% other) children who tested at the national 44 percentile in English and 59 percentile in math. Details on the age range and gender composition of the sample were not reported.

Measures:

Teachers used a five-point Likert scale to rate each student on the four domains of the Student Behavior Rubric measure of classroom behavior: (1) paying attention (pays attention all of the time), (2) self-control (i.e., demonstrates calmness and self control), (3) participation in activities (physically engages and participates in all activities), and (4) caring and respect for others (shows care and respect for teachers and fellow students). The four summed items showed acceptable reliability (Cronbach's alpha: pretest = .83, post-test = .87, follow-up = .86; five-week test-retest reliability = .51, p<.001).

Analysis:

All analyses were performed at the individual level without adjustments for clustering within classrooms, the unit of assignment; however, the data showed no evidence of clustering (i.e., significant within-teacher variability, but not between-teacher variability). Estimated intervention effects were evaluated with a grade-adjusted mixed model including a between-subjects factor (group), a repeated measures factor (time: pre-intervention, post-intervention, and seven-weeks post-intervention), and their interaction. Simple effects were calculated separately for each group to examine continued improvements over time across the post-intervention period.

Intent-to-Treat:

All children with complete data were included in analyses.

Outcomes

Implementation Fidelity:

No quantitative measures of implementation fidelity were reported.

Baseline Equivalence:

At baseline, both groups were equivalent on average grade level and on scores for all four outcomes of classroom behavior. Group differences on baseline demographic characteristics were not reported.

Differential Attrition:

No tests of differential attrition were conducted.

Posttest:

Tests showed significant improvement over time but no group differences. From baseline to immediately post-intervention, teacher-reported student behavior scores in paying attention, self-control, participation, and respect for others, as well as the summed total score, significantly improved for both groups. Both groups continued to show significant improvements in all four domains at seven weeks post-intervention relative to baseline. Changes over time were statistically similar across groups, with the exception of attention, which continued to rise for students who received additional sessions in the seven-week post-intervention period, but remained unchanged among the students who did not receive the additional sessions.

Long-Term:

No long-term follow-up assessments were conducted.

Study 2

Design:

Recruitment: Participants were recruited from eight kindergarten classrooms at three schools in Toronto, Ontario. Initially, ten classrooms indicated interest in participating: three classrooms from School 1, four classrooms from School 2, and three classrooms from School 3. In order to have an even number of classes in each condition within each school, the researchers randomly selected one classroom from School 1 and one classroom from School 3 to be dropped from the evaluation. The remaining eight classrooms were randomly assigned to conditions within schools. All students in the assigned classrooms received the intervention or control programming, but data were only collected from students whose parents consented to participate in the study. Of the 219 children in the eight participating classrooms, consent was obtained from 127 (58%), with 62% enrollment in the treatment classrooms and 53% in the control classrooms.

Assignment: Eight classrooms were randomly assigned to the intervention or control conditions within the three schools. Specifically, one classroom was assigned to each condition in Schools 1 and 3 and two classrooms were assigned to each condition in School 2.

The article contains conflicting information regarding the number of participants in each condition within each school at assignment. The CONSORT diagram (Figure 1) lists 19 intervention and 13 control students in School 1 (n=32), 38 intervention and 31 control students in School 2 (n=69), and 15 intervention and 11 control students in School 3 (n=26). However, the text reports that "…29 children were from School 1 (19 in the Mindfulness Group and 10 in the Control Group); 28 children were from School 2 (15 in the Mindfulness Group and 13 in the Control Group) and 70 children were from School 3 (38 in the Mindfulness Group and 32 in the Control Group)." Both sets of information yield an initial sample of N=127 with 72 in the intervention and 55 in the control, but the distribution across classrooms and schools cannot be discerned from the article.

Assessments/Attrition: Assessments were conducted before (at baseline) and after (at posttest) the six-week program. The initial sample included N=127 participants (72 in the intervention, 55 in the control). There was no attrition; though attrition was not reported in the article, it was confirmed by the author that all 127 participants were used in analysis.

Sample:

Participants included kindergartners between the ages of four years, three months and six years, three months (mean age in months = 62.32, SD = 7.5 months for the intervention group and 61.36, SD = 7.1 months for the control group). The majority of participants spoke English. All three schools scored high on a measure of "external challenges" faced by students relative to other schools in the same district, indicating that the children in the study may have had more difficulties at school and may have benefited more from the intervention than the general population of children in the area.

Measures:

A total of six outcome measures were used. The first measure was the Head-Toes-Knees-Shoulders (HTKS) task, a performance-based measure of self-regulation that was administered by trained research assistants who were blind to conditions. This measure has demonstrated high inter-rater reliability (chi = .90) and internal consistency (α = .80) in prior samples. The remaining five measures were subscales on the Strengths and Difficulties questionnaire (SDQ), a questionnaire measure completed by teachers who were not blind to conditions. The five subscales include (1) prosocial behavior, (2) hyperactivity, (3) conduct problems, (4) emotional symptoms, and (5) peer problems. Reliability estimates for the subscales in this sample ranged from α = .72 to .91, indicating high reliability.

Analysis:

All analyses were conducted at the individual level despite classroom-level randomization. The researchers reported that they explored for the presence of the nesting effect by conducting variance components analyses using multilevel modeling, and that the "proportion of variance at the classroom level for different outcome variables ranged between 3% and 16% and was non-significant for all of them," thus justifying their use of individual-level analyses. Statistics for these tests were not reported.

Self-regulation scores ranged from 0-60 and were analyzed using ANCOVA models predicting posttest scores from condition with baseline scores as a covariate. For teacher-reported behavioral outcomes, subscale scores were not normally distributed and thus analyzed using non-parametric Mann-Whitney U tests that tested whether "the total change scores for each of the 5 scales of the SDQ were different in the Mindfulness Group compared to the Control Group at Time 2" (page 1155). It is unclear whether baseline outcome levels were adjusted in these analyses. A Bonferroni correction was used to adjust for multiple tests of the teacher-reported behavioral questionnaire for each of the five subscales (adjusted alpha = .01, taken by dividing the familywise alpha of .05 by the number of analyses).

Additionally, correlations between baseline scores and change scores were computed for each outcome to examine whether intervention effects were related to baseline levels.

Intent-to-Treat: All data used in analysis.

Outcomes

Implementation Fidelity:

No quantitative fidelity information was provided.

Baseline Equivalence:

Groups were equivalent at baseline on 5 of 6 outcome measures, with one difference favoring the control group. Specifically, children in the control condition had significantly higher prosocial behavior scores at pretest relative to children in the intervention condition. The article also reports that conditions were equivalent at baseline on "child characteristics," which included age and gender as reported by the author. Significance tests were not reported in the article but results of equivalence tests were confirmed by the author via email.

All data were used in analyses, indicating no attrition overall or differential attrition.

Posttest:

At posttest after the six-week intervention, the researchers found three significant differences of six tests. Specifically, the intervention condition had a significantly higher adjusted mean score for self-regulation relative to the control condition after controlling for baseline self-regulation. Additionally, children in the intervention condition had significantly greater prosocial behavior and significantly lower hyperactivity at posttest relative to children in the control condition, who did not appear to change from baseline to posttest on either outcome (based on the plots in Figure 2). Moderation analyses indicated these patterns were greater for children with lower scores on the prosocial behavior subscale and higher scores on the hyperactivity subscale at pretest. No significant differences were found for the remaining measures of emotional symptoms, conduct problems, and peer problems.

Long-Term:

Not examined.