Exploring spillover effects beyond workplace dietary interventions using agent-based modelling

OBJECTIVE: We applied an innovative agent-based modelling approach to investigate how and to what extent spillover effects occur beyond dietary interventions. DESIGN: We developed an agent-based model in NetLogo to simulate the dietary practices of the Dutch adult population during lunch and dinner, representing participants in the Dutch National Food Consumption Survey. SETTING: The model assessed spillover effects beyond a hypothetical workplace canteen intervention targeting increased plant-based protein consumption during lunch. Direct effects were defined as the percentage of non-vegetarian employees consuming a vegetarian lunch. Spillover effects were indicated by the percentage of plant-based protein consumed at dinner, influenced by dietary identity, habituation, social networks and eating location. Scenarios with varying direct effect sizes were simulated to compare plant-based protein consumption at dinner. RESULTS: Simulations demonstrated a positive spillover effect at the population level, with plant-based protein consumption at dinner increasing over 3 years compared with baseline (no intervention). At direct intervention effect sizes up to 0·5, there was a linear increase in plant-based protein consumption at dinner (from 23·0 % (95 % CI 22·9, 23·2) at baseline to 32·3 % (95 % CI 31·8, 32·8)). Spillover effects plateaued beyond an effect size of 0·5, with no significant further increases observed. CONCLUSIONS: Our application of agent-based modelling revealed that positive spillover effects of dietary interventions from the workplace to home can be achieved. Despite challenges, agent-based modelling might complement other approaches for studying spillover effects and supporting dietary behaviour change towards more sustainable diets.