Application of Research Design and Methods for Optimizing Prevention Science: Multiple Meta-Analytic Strategies to Identify Essential Components of Preventive Interventions

This abstract was presented at the 2018 Society for Prevention Research Annual Meeting which was held May 29 – June 1, 2018 in Washington, DC, US.

Frances Gardner University of Oxford

Patty Leijten University of Amsterdam; G.J. Melendez-Torres Cardiff University; Jamie Lachman University of Oxford; Wendy Knerr University of Oxford; Chris Mikton University of West England; Judy Hutchings Bangor University

Most preventive interventions deliver a complex package of skills, but rarely is it clear which of many components are necessary or effective. Elucidating essential components could help optimize programs so that they are briefer, more effective and efficient. 

This novel study employs multiple approaches for elucidating essential components of parenting interventions. Enhancing these interventions is vital, as they are prominent in both US and global policy recommendations for preventing violence to children, and child behavior disorders. This has resulted in widespread implementation efforts in low-resource settings (eg WHO, CDC, UNICEF, UNODC), where effectiveness, cost, scalability and sustainability are of paramount importance.

We systematically reviewed the literature for studies providing evidence from each of 4 methods we identified to test effective components of parenting interventions: 1: Meta-regression of associations between intervention components and outcomes (Kaminski et al, 2006); 2: Multi-arm dismantling trials, 3: Microtrials, i.e. focused randomized experiments (Howe et al, 2013) 4: Optimization studies, eg factorial designs. We found sufficient good quality trials for strategies 1 & 3, namely meta-regression and meta-analysis of microtrials - strategies with complementary strengths and weaknesses, in terms of causal inference vs. external validity. 

Meta-regression using robust variance estimation (156 RCTs; 386 effect sizes) tested if components aimed at enhancing the parent-child relationship (e.g. child directed play), and those aimed at behavior management (e.g. praise, rewards, timeout) predict greater reductions in disruptive child behaviour. Overall, we found that adding relationship components to behavior management yielded stronger effects treatment trials, but not in prevention trials (111 vs 45 RCTs respectively). Findings from meta-analysis of microtrials (20 RCTs; 78 effect sizes) partly concurred in finding that behavior management components were more strongly associated with child outcomes, however this applied only to ‘timeout’ and ‘ignore’, rather than positive discipline strategies, such as praise. 

Findings show we have more limited knowledge of essential components of preventive parenting interventions, raising the possibility that current practice to implement essentially the same parenting programs in treatment and prevention settings may be unwarranted. We discuss implications of these component analysis findings for practice and for understanding prevention mechanisms – findings which in many ways diverge from well-established practice; and implications for research, especially the need for factorial trials to maximise both external and internal validity.