| ||Literature Coverage Dates||Number of Studies||Number of Study Participants|
|Meta-Analysis 1||1973 - 2009||142||0|
|Meta-Analysis 2||1977 - 2017||91||0|
|Meta-Analysis 3||1981 - 2014||14||0|
Pinquart and Teubert (2010) conducted a meta-analysis to examine the effects of interventions for new and expectant parents that are designed to prevent child maltreatment. The researchers searched electronic databases for all studies available up until 2009. Studies were eligible for inclusion if 1) they used a randomized controlled design, with the control group receiving no intervention or only minimal intervention; 2) they included a parenting education component; 3) the intervention was initiated during pregnancy or in the first 6 months after childbirth; and 4) effect sizes were reported or could be computed from the available information, such as from means and standard deviations. Studies were excluded if they 1) were not targeted at improving parenting (e.g., programs that limited their focus to improving the couple relationship), 2) did not provide enough information for extracting effect sizes, and 3) focused exclusively on treatment or prevention of recurrence of psychological disorders, such as postnatal depression in parents (usually mothers), or were programs designed solely for parents of chronically ill or disabled children.
The search yielded 142 eligible studies from 1973 to 2009. All studies were randomized controlled trials. Of the eligible studies, 137 were published in a peer-reviewed journal. The average number of participants per intervention was 108. More than half (N
= 82) of the interventions worked with families at risk, and 107 included only mothers. Most interventions began after the baby was born, 38 were conducted during both pregnancy and post-pregnancy, and 10 were held only during pregnancy. The average length of the intervention was 15 months, but could range from 1 day to 60 months, with participants attending an average of 29 meetings. Most interventions (N
= 84) were delivered at home, 16 interventions were held in hospitals, 6 were held in the community, and a further 26 combined home visits with other locations (e.g., support group meetings in the community). The participating parents were, on average, 24.3 years old; approximately 79 percent were expecting or had just given birth to their first child. About 58 percent were married, and 21 percent were cohabiting. In addition, 59 percent were ethnic minorities, and 56 percent had completed high school education. Eighty-nine percent of the participants were mothers, and 51 percent of the infants were girls.
Outcome data were analyzed using random effects models and iterative maximum-likelihood estimations. Effect sizes were calculated using Cohen’s d
, in which the differences in the posttreatment measures between the intervention and control conditions were divided by the pooled standard deviation (Lipsey and Wilson 2001). The authors averaged effect sizes for studies that reported results from one dataset for different outcome variables (e.g., for different parenting behaviors). They also averaged results on couples’ interventions in which separate results were reported for mothers and fathers. Weighted mean effect sizes were also computed.Meta-Analysis 2
Van der Put and colleagues (2017) conducted a meta-analysis to examine the effectiveness of interventions designed to prevent child maltreatment. The authors searched electronic databases and conducted a manual search of reference sections of retrieved articles, reviews, and book chapters for studies published in 2013 or earlier. The authors also contacted fellow researchers to request studies and unpublished manuscripts that may have been relevant for inclusion. Studies were eligible if they 1) were preventive programs that targeted the general population, 2) were randomized controlled trials or quasi-experimental design studies, and 3) reported at least one effect size or provided sufficient information to calculate at least one effect size.
A total of 91 studies published between 1977 and 2017, comprising 63 randomized controlled trials and 29 quasi-experimental designs, were eligible for inclusion. Sixty-two of the studies collected outcome data based on self-report measures completed by the parent(s), 41 used official reports, and 7 percent used another collection method. Of the 91 studies, 83 focused on at-risk families, and the remaining 8 delivered the preventive programming to parents who were representative of the general population. At-risk families were identified using risk factors that included teenage parents, parents with substance use problems, or multi-problem families; however, “at-risk” did not refer to families with a history of child maltreatment.
The authors used a random effects model and estimated the overall effects using Cohen’s d
in separate three-level, intercept-only models, consisting of 1) a random sampling variation of observed effect sizes, 2) variance within studies, and 3) variance between studies. Effect sizes were weighted by the inverse of their variance (i.e., sampling error).Meta-Analysis 3
Vlahovicova and colleagues (2017) conducted a meta-analysis to evaluate the effectiveness of preventive programming in reducing re-abuse in families with a history of child maltreatment. Databases were searched for studies conducted prior to 2015. To be eligible for inclusion, studies had to 1) be randomized controlled trials or quasi-experimental designs; and 2) have parents as the participants (i.e., mothers, fathers, or other primary caregivers of children ages 0 to18), for whom there was a suspected or substantiated report of child physical abuse. Maltreatment history had to be supported by a police report, child protection referral, or other official agency report; the self-report of an abusive parent or abused child; or an above-threshold score in standardized instruments used for detection of child physical abuse.
A total of 14 studies were included in the review. All 14 studies that met the inclusion criteria used randomized controlled trials to evaluate eight different behavioral, parent-training programs that were based on social learning theory. The number of participants in the study ranged from 26 to 2,176. Seven of the studies included only physically abusive parents, and the others ranged from between 23 and 63 percent of parents who were physically abusive. In seven of the studies, data were collected from official reports to child protective services or similar agencies. The other seven studies used data collected through parent and child self-reports.
The researchers used a random effects model to analyze the data, which was presented as Cohen’s d