Individual participant data network meta-analysis of psychosocial interventions for survivors of intimate partner violence: Study protocol

TL;DR

This study aims to compare the effectiveness of psychosocial interventions for survivors of intimate partner violence using advanced statistical methods to determine which interventions work best for specific individuals.

Contribution

The study introduces a novel approach combining individual participant data and aggregate data in a Bayesian framework to address limitations in data availability.

Findings

The study will use IPDNMA to identify which interventions are most effective for specific survivor characteristics.

A novel ML-NMR approach will be used to combine IPD and aggregate data, enhancing the use of available evidence.

Collaborations with RCT authors will be established to harmonize data and improve the accuracy of results.

Abstract

Many systematic reviews and meta-analyses have been conducted in the field of Intimate Partner Violence (IPV) and the evidence shows small to moderate effect sizes in improving mental health outcomes. However, there is considerable heterogeneity due to variation in participants, interventions and contexts. It is therefore important to establish which participant and intervention characteristics affect the different psychosocial outcomes in different contexts. Individual Participant Network Meta-analysis (IPDNMA) is a gold-standard method to estimate moderating effects, compare the effectiveness of different interventions and thus answer the question of which intervention is best-suited for whom. We will conduct an IPDNMA of randomised controlled trials (RCTs) of psychosocial interventions for IPV survivors aimed at improving mental health, psychosocial outcomes such as self-efficacy and quality of life, reducing IPV and increasing safety-behaviours and dropout from the intervention (as an indication of intervention acceptability) compared to any type of control (PROSPERO registration number: CRD42023488502). We aim to establish collaborations with the authors of eligible RCTs, to obtain and harmonise the Individual Participant Data of the trials. We will conduct one-stage IPDNMA under a Bayesian framework using the multinma package in R, after testing which characteristics of the participants and interventions are effect modifiers. We anticipate that not all study authors will provide access to IPD, which is a limitation of IPDNMA. We aim to address this by combining studies with aggregate data and studies with IPD using Multi-Level Network Meta-Regression (ML-NMR) implemented in the multinma R package. This approach is novel in the field and makes full use of available evidence to inform clinical and policy-related decision making.

Figures