Establishing causality in complex interventions with moderators and mediators

Establishing causality is one of science’s core tasks, however it can be difficult to know how to ask the right research questions and use the most relevant methods to investigate these questions. Therefore, it is important to examine how to establish causality in complex interventions, both in regards to the effect of the intervention in and of itself but also the other active elements that can moderate and mediate a possible intervention effect.

Aim

The aim of this PhD study is threefold: 1) examining the construct validity of maternal sensitivity in an at-risk of postpartum depression sample, 2) a literature review of statistical analyses used to investigate causality, moderators, and mediators in interventional studies, and 3) a comparison of traditional statistical analyses (e.g. linear regression) with newer methods (e.g. conditional process analysis, invariant causal prediction, independent-based methods) to establish causality.

Method

This project is a subproject embedded in the larger project, the Copenhagen Infant Mental Health Project (CIMHP). 314 parents at risk of postpartum depression and their children (1-11 months old) were randomized to two intervention groups: Circle of Security – Parenting (COS-P) and care as usual (CAU). The parents filled in questionnaires and were observed in a short free-play situation with their child at baseline before intervention and at follow-up after end of treatment. Data was collected from 2016-2020.

Contribution

The results will contribute with insights of how to analyse and evaluate quantitatively complex interventions. These results will help expand the statistical analytical methods researchers usually use to establish causality by investigating whether these newer methods are appropriate in psychological intervention research. Furthermore, the results will contribute with insights on how and for whom the parent intervention COS-P works in a sample with postpartum depression by focusing on possible moderators and mediators.