Table 1 CLD development and analysis

Step 1: setting system boundaries

Top-down conceptualisation; prior to iteration

The boundaries of the system were determined by the literature review examining socioecological determinants of dietary intake, assuming a conceptual model whereby individual, social, physical environmental, economic and political determinants drive food intake.

Boundaries: the scope of the system under study.

Step 2: identifying elements, feedback loops and structure

Data-driven development; first iteration (whole-system CLD)

Using the data matrix, variables that were reported to have an association with dietary intake and determinants of dietary intake were entered as elements of the CLD. Connections between elements were drawn based on reported presence, direction and polarity (positive, negative) of associations between elements. Feedback loops were identified only where drawn connections revealed a ‘closed’ loop, i.e. element A ➔ element B ➔ element C ➔ element A. This process to develop the initial CLD was led by AS and reviewed by KS and FvL.

Elements: included variables

Feedback loops: connections between elements in reinforcing loops denoting growth/decline and balancing loops denoting stabilisation.

Structure: clustering of connections between elements

Step 3: identifying paradigm and structure

Top-down review; second iteration (whole-system CLD)

An expert panel was formed of 12 authors and two additional researchers, with expertise in public health nutrition and systems thinking, who communicated via written comments, two face-to-face sessions and one video discussion; not all group members took part in each session. Each member of the review panel was presented with the CLD and provided with an in-depth explanation how the findings from the umbrella review were used to develop the system elements and connections.

First, the accuracy and saturation of the CLD was assessed. Accuracy (i.e. exclusion of errors emerging from data extraction and synthesis and verification of of relationships in the CLD as being interpreted from the umbrella review) was tested by reviewing individual connections between elements. Saturation (i.e. completeness in reflecting the evidence base [30, 31]) was assessed using expert knowledge from the expert panel of the evidence base, to overcome potential limitations of the literature search. Additional literature could be integrated at this stage if saturation was not apparent and there was available literature of sufficient quality. Iterations were only made if they were substantiated by empirical evidence.

Second, discussion of feedback loops and structure identified in step 2 elucidated key mechanisms operating within the system. These mechanisms were discussed in relation to four dimensions of the food environment, to explicate the perceived arrangement (temporal ordering and interaction) of mechanisms pertaining to accessibility, availability, affordability and acceptability. This exercise enabled the articulation of the paradigm and structure of sub-systems. Elements that were present in multiple sub-systems were manually highlighted as ‘link elements’, to demonstrate interconnection between sub-systems.

Paradigm: system’s deepest held beliefs derived from goals and sub-system structure.

Structure: sub-systems derived from feedback loops and connections between elements.

Step 4: identifying goals of sub-systems

Top-down conceptualisation; final iteration (overview, sub-system CLDs)

Following the organisation of sub-system CLDs and articulation of the system paradigm within the system overview, the goal of each sub-system was derived from feedback loops and sub-system structure.

Goals: aims of the system conforming to the system’s paradigm, derived from feedback loops and sub-system structure.