Evaluating select factors and mechanisms influencing meat consumption in Baltimore City: an agent-based modeling study

Reducing meat consumption has been identified as an effective strategy for protecting human and planetary health. Policies and programs to reduce meat consumption have produced inconsistent results given the complexities of human behavior and food consumption. In this study, we used a systems approach to develop an agent-based model (ABM), explored key factors and mechanisms influencing meat consumption, and tested potential scenarios to reduce meat consumption. The ABM simulated food consumed at dinner, focusing on changes in meat consumption. For each simulated day, dinner decisions were driven based on individual characteristics, food environments, food preferences, and system-level driving factors. We selected Baltimore City, Maryland as a case study, to simulate virtual agents with independent decision-making abilities. Scenarios tested included the implementation of a non-meat marketing campaign, an increase in meat prices, an increase in the availability of non-meat options, and a combination of marketing and increased availability. Scenarios were also assessed by population sub-groups and geographic variability. The model simulated 596,377 adults (≥ 18 years) in Baltimore City (2019 ACS) and was parameterized using NHANES 2007–2018 (Cycles 2007–2008 to 2017–2018; N = 45,375 unique respondents) dinner recalls. The baseline scenario with no interventions showed that, on average, an agent’s dinner had a percent breakdown by weight as follows: 17.0% meat options, 79.8% non-meat options, and 3.1% fish. The largest reduction in meat consumption (12% decrease from baseline) was seen in the scenario where an increase in non-meat marketing and availability of non-meat options occurred. All scenarios showed the largest change in meat consumption among non-Hispanic White and Asian populations. Reduced meat consumption among Black individuals and individuals with a household income of $25k-$55k was relatively low across all scenarios. In each scenario, we observed significant variability in average meat consumption by zip code, with individuals living in higher income areas showing higher reductions in meat consumption. Overall, combination interventions can have synergistic effects and were shown to be more beneficial than each intervention alone in reducing meat consumption. Continued systems work on meat consumption models will allow researchers and practitioners to better understand the potential strengths and weaknesses of programs and policies across different populations and environments.