Table 4 Initial learning for applying the Depth framework

Identify component-recipient combinations a priori

Prior to extracting data on framework constructs, we advise agreeing on the intervention component-recipient combinations. This reduced the number of disagreements at the extraction stage. We identified intervention components to the greatest level of granularity to improve agreement between researchers

Implicit vs explicit information

Not all relevant information is included in intervention reports. For example, some reports may not describe or have explored all mechanisms of action. The degree to which reviewers include only information explicitly included in the report or draw on implicit and wider topic knowledge will be dependent on the review aims and should be agreed a priori

Dealing with insufficient information

Intervention descriptions may not provide sufficient information to classify the framework constructs. In our review, where applicable, we chose to classify framework constructs as ‘insufficient information to code’ based on the intervention description. Other approaches may include seeking additional information from a wider range of sources (see below)

Information sources

The information sources included within reviews will influence what information is available. We used only data reported in papers that reported equity outcomes, but other approaches such as identifying linked papers, grey literature and speaking to study authors may be used to aid classification

Consistent application

The framework requires users to apply categorical classifications to constructs that lie on continua. Different users may draw these distinctions in different places. Nevertheless, distinctions should be agreed, reported and applied consistently