Study 1
Brotman and colleagues (2013) conducted a cluster randomized controlled trial of ParentCorps in two school districts in disadvantaged minority and immigrant-dense New York City neighborhoods. To be included, schools were required to have pre-kindergarten (pre-K) programs (offered 1 year before kindergarten) with at least two classes, and a student population that was mostly Black (at least 80 percent) and of low income (at least 70 percent eligible for free lunch). The study aimed to enroll all pre-K students during 4 consecutive years (2005–2008). The only inclusion criteria for children was having an English-speaking caregiver.
Before randomization, schools were matched on size and split into pairs; within each pair, one school was randomly assigned to receive the ParentsCorps intervention, and the other school was the control group. The final sample included 1,050 children in 10 schools, with 561 in the intervention group and 489 in the control group. At baseline, 71.9 percent of the total sample of children were eligible for free lunch (indicating low income). Approximately 90.7 percent were Black and 6.2 percent were Latino. Almost half (49.3 percent) were male, and 44.7 percent came from single-parent homes. Over one third (36.4 percent) came from families in which a parent was unemployed, and 46.5 percent came from families where parents had a high school diploma or less. There were no statistically significant differences between the intervention and control groups at baseline on demographic characteristics.
The primary outcome of interest was achievement test scores assessed at the end of kindergarten and developmental trajectories of academic performance from pre-K through kindergarten. The Kaufman Test of Education Achievement (KTEA) Brief Form was administered to measure achievement test scores. Developmental trajectories of academic performance were measured with teacher reports on a global rating of academic problems.
Analyses were conducted on student-level data by using intent-to-treat principles. The intervention effect on achievement test scores was analyzed with a multivariate analysis of variance-type analysis using linear mixed-effects models. To account for potential correlations among outcomes of children, random effects for classes and schools were included in the model. The effect on trajectories of academic performance was estimated using mixed-effects models for longitudinal data. The study authors did not conduct subgroup analyses.
Study 2
The 2015 study by Dawson-McClure and colleagues used the same dataset as the study by Brotman and colleagues (2013) to examine the impact of ParentCorps on parenting skills and child conduct problems at the end of kindergarten. Of the original study sample (1,050), 92 percent in the intervention group and 93 percent in the control group remained in the study. Most of the parents/caregivers were mothers (88 percent), with an average age of 33.9 years. About two thirds (68 percent) were immigrants, and 85 percent were non-Latino Black (including Afro-Caribbean and African American).
The study examined teacher ratings of parent involvement and parent ratings of parenting and conduct problems. Pre-K and kindergarten teachers completed questionnaires at the beginning and end of each school year (four times). Parents were interviewed at the beginning and end of pre-K and at the end of kindergarten (three times). Parenting was assessed by multiple methods across three domains: Positive Behavior Support, Behavior Management, and Involvement in Early Learning. Knowledge of positive behavior support and effective behavior management were measured with the Effective Practices Test (EPT). Positive behavior support was measured with the Positive Reinforcement subscale of the Parenting Practices Interview (PPI). Harsh and inconsistent behavior management was measured with two subscales of the PPI (Harsh and Inconsistent Discipline). Parent report of parent involvement in early learning was based on the Involve Interview and Parent Perceptions of Parent Efficacy. Teacher report of parent involvement was obtained from a 6-item subscale of the Involve Interview. Parents rated child conduct problems during the past 4 weeks on the New York Rating Scale (NYRS).
Analyses used intent-to-treat principles. Standard errors were adjusted to account for repeated assessments and clustering of children within schools. The study authors did not conduct subgroup analyses.
Study 3
Brotman and colleagues (2016) conducted a 3-year follow up to the original randomized controlled trial discussed in their 2013 article. The original study enrolled children 4 years of age. This study included children at 5, 6, 7, and 8 years of age. Of the original sample (1,050 children), data was available on 792 children at follow up (432 in the intervention group and 369 in the control group). Of the 792, 409 children (51.6 percent) had complete data for all measures at all times. Almost all the children (768) had data for baseline and at least one follow up in teacher rating measures, and 615 had at least one follow up in KTEA scores. There were no statistically significant differences between children with data in second grade and those without data, on demographic characteristics, baseline teacher ratings, or school readiness measures.
The outcomes included teacher ratings of mental health problems and academic performance, and achievement test scores. Teacher ratings of mental health problems were obtained on externalizing scales (e.g., conduct problems, aggression, and hyperactivity) and internalizing scales (e.g., anxiety, depression, and somatization) of the Behavior Assessment System for Children. Ratings were made by teachers at baseline and at the end of each year from pre-K through second grade (five times with four different teachers). Teachers also rated academic performance annually. Reading and math achievement were assessed twice (at the end of kindergarten and second grade) with the KTEA Brief Form.
Intent-to-treat principles were again used in the analyses. Effects were estimated using linear mixed-effects models with all available data. For missing data, multiple imputations were used, assuming data were missing at random. The study authors did not conduct subgroup analyses.