Purpose:

Programmes addressing childhood obesity have long focused on targeting single determinants. Causes of childhood obesity are however diverse, complex and multiple.

Systems thinking embraces the complexity of  problems such as overweight and obesity and aids in understanding how factors in the system are interrelated and affect each other so that these can be targeted and produce favourable changes in a system.

 While there is a growing call for systems approaches in public health research, there is relatively little guidance on how to best develop and evaluate public health interventions in complex systems. This study therefore aimed to develop an evaluation framework using a complex systems approach.

Methods:

The LIKE (Lifestyle Innovations based on youths’ Knowledge and Experience) project was used as a case study. LIKE is part of the Amsterdam Healthy Weight Program and aims to create healthy habits amongst young adolescents in lower socio-economic and ethnically diverse neighbourhoods.

The evaluation framework served both as a tool for informing and supporting the development and implementation of the intervention programme and as way of generating generalizable knowledge on the impact of the programme so that it could be applied elsewhere.  

Results:

A developmental evaluation design was used at the start of the LIKE project as it is particularly well suited for programmes that target complex environments with high uncertainty. It assisted evaluators in developing an understanding of the system and in framing and adapting the intervention in real time as patterns of change emerged and as the intervention unfolded.

Besides supporting the intervention development, the evaluation framework also aimed to produce generalizable knowledge using a summative evaluation. Here the evaluation focuses on comparing pre-existing and follow-up systems and developing indicators on system level changes using the Intervention Level Framework.

Conclusions:

An evaluation framework was developed for the LIKE project that was informed by key principles of developmental and summative evaluation using a complexity perspective. This evaluation design can inspire future public health programmes in developing and evaluating interventions in complex systems.

Purpose:

Inequalities in obesity and related non-communicable diseases pertain in part to less healthy dietary intake in disadvantaged groups. Examining determinants of intake in low-income groups as a complex adaptive system – i.e. interconnected determinants exerting non-linear influence on outcomes - honours the complexity of the reality governing individuals’ dietary choices, behaviours and intake, and could enhance assessment of policies. However, it is not clear if existing literature on relevant determinants can be synthesised, and understood, as a complex system. 

This study aimed to use existing literature to map the complex system underlying dietary intake in low-income groups, in order to identify system structures and goals perpetuating poorer dietary outcomes.

Methods:

A systematic umbrella review was conducted on literature examining determinants of dietary outcomes in low-income children, adolescents and adults. Inclusion criteria:

·​Low-income (or related construct) sample or analysis by income;

·​(Non-)systematic, scoping, mapping reviews or meta-analysis of quantitative or qualitative, observational or intervention studies;

·​Higher/upper-middle-income countries;

·​Exposures: individual, sociocultural, physical, political determinants, effect modifiers;

The following outcomes were excluded: breastfeeding, alcohol and neophobia.

Data on determinants, associations and mechanisms were extracted and, using causal loop diagramming, embedded in a systems map of determinants underpinning dietary intake. The map was analysed in terms of system structure (e.g. subsystems) and goals (e.g. structure and feedback loops).

Results:

A systems map of hypothesised mechanisms underlying dietary intake in low-income groups was developed from 43 reviews and expert consensus. The system was interpreted as underpinned by cross-sectoral subsystems supporting goals around: commercial competitiveness, cost-efficient purchasing, use of food to indicate group membership and preference-based food selectivity. Goals may undermine opportunities for healthy intake; e.g., energy-dense food choices resulting from cost-efficiency and preferences determined by heightened exposure to energy-dense foods.

Conclusions:

Using an innovative but challenging systems approach, we developed a literature-based systems map which begins to articulate the systemic basis of dietary outcomes in low-income groups. Deeper understanding of identified system structures and goals will facilitate the development and assessment of effective and equitable policies.

Background Body weight perception is a product of interactions between an individual and their socio-cultural environment. We consider this a complex system inducing population-level obesity as an emergent property. The system consists of a micro-macro feedback loop between 1) social norms; 2) individuals’ body weight perception; and 3) individual lifestyle and BMI. Lifestyle affects norms while also being affected by norms. Given this feedback loop, it remains unclear whether addressing lifestyle only via health awareness can decrease obesity prevalence; norms might be a counteracting driver. Still, norms have been neglected in epidemiological research as quantitative methods cannot consider micro-macro feedback loops. System dynamics modelling (SDM) can be a solution to understand and simulate a system’s emergent behaviour.

Methods We used an expert-informed causal loop diagram as a template for six SDMs. These correspond to six socio-cultural groups (Dutch, Moroccan and South-Asian Surinamese men and women) from an Amsterdam-based cohort, which includes body weight (BMI) and weight perception measures. We validated the SDMs based on their performance on validation statements compared to random SDMs. We simulated the effect of three scenarios on population-level BMI: where individual lifestyle was driven 1) only by health awareness, 2) only by norms, 3) by their interaction (health awareness and norms).

Findings We operationalised norms in the SDMs based on population-level median BMI and data describing socio-cultural ideal BMI. The SDMs outperformed random SDMs. They showed that median BMI drops 10·8% (2·78 BMI points) in scenario 1), 5·3% (1·36 points) in 2), and 7·4% (1·91 points) in 3). The male groups demonstrated a 2·06 times larger drop in scenario 1) than in 3) than their female counterparts (1·12 larger drop).

Interpretation Social Norms diminish the potential effect of health awareness on BMI. This was consistent in all groups but the effect was stronger in males than females. Our findings imply that, since norms affect the lifestyle (and ultimately BMI) of female groups to a lesser extent, there must be additional underlying drivers of obesity in women.