A methodological review of network meta-analyses applied to complex interventions for public health and a typology of...

Network meta-analysis (NMA) methods allow simultaneous comparison of multiple interventions. Public health interventions are typically complex, being composed of multiple components. Methodological choices are therefore necessary when planning NMAs of complex interventions (eg, 'splitting' vs 'lumping'; intervention- vs component-level analysis). This node-making activity is often poorly reported and has received limited attention within methodological guidelines. This review aims to describe the use of NMA methods in this context and investigate the elements involved in node-making to develop a typology.

We conducted a systematic review to identify application of NMA methods to complex interventions applicable to public health (health promotion and primary prevention). We extracted the characteristics of the included reviews and networks, and the statistical tools applied therein. Thematic network analysis of text extracts related to node-making within the included reviews was applied to develop a typology of elements potentially involved in the formation of network nodes.

We included 89 reviews and 102 networks applied to complex interventions relevant to public health. Included reviews were distributed across a breadth of topics including universal, selective, and indicated primary prevention, and involved outcomes including both physical and mental health. Almost all reviews applied standard NMA models, while just six reported additive component NMA models. Network nodes were formed by grouping similar interventions or intervention types (65/102), comparing named interventions (6/102), defining nodes as combinations of intervention components (26/102), or using an underlying component classification to group interventions into nodes (5/102). We developed a typology of node-making elements, identifying seven sets of potential considerations (Approach, Ask, Aim, Appraise, Apply, Adapt, and Assess).

NMA methods are applied widely to evaluate complex interventions across health promotion and primary prevention. The proposed typology of these elements can inform future networks and facilitate improved reporting of methodological steps.

Public health interventions - such as programs to prevent smoking or increase physical activity - are often complex and made up of many parts ('components'). Deciding between multiple different interventions, or understanding which combination of intervention components works best, can require methods such as "network meta-analysis" (NMA). How to group interventions or components for NMA is known as 'node-making' but it is not clear what node-making approaches are currently used in NMA of public health interventions. In this paper, we reviewed 102 networks, identified from 89 reviews, and have summarized the range of methods used, with particular focus on the steps taken by reviewers to reach their node-making decisions. We arranged the identified node-making elements into seven categories, creating a classification which might be used to help researchers make decisions and report their methods in the future.