Exploring the Impact of Policies to Improve Geographic and Economic Access to Vegetables among Low-Income, Predominantly Latino Urban Residents: An Agent-Based Model

Abstract

Modifying the food environment of cities is a promising strategy for improving dietary behaviors, but using traditional empirical methods to test the effectiveness of these strategies remains challenging. We developed an agent-based model to simulate the food environment of Austin, Texas, USA, and to test the impact of different food access policies on vegetable consumption among low-income, predominantly Latino residents. The model was developed and calibrated using empirical data from the FRESH-Austin Study, a natural experiment. We simulated five policy scenarios: (1) business as usual; (2)–(4) expanding geographic and/or economic healthy food access via the Fresh for Less program (i.e., through farm stands, mobile markets, and healthy corner stores); and (5) expanding economic access to vegetables in supermarkets and small grocers. The model predicted that increasing geographic and/or economic access to healthy corner stores will not meaningfully improve vegetable intake, whilst implementing high discounts (>85%) on the cost of vegetables, or jointly increasing geographic and economic access to mobile markets or farm stands, will increase vegetable intake among low-income groups. Implementing discounts at supermarkets and small grocers is also predicted to be an effective policy for increasing vegetable consumption. This work highlights the utility of agent-based modeling for informing food access policies.

Agent-Based Modelling of Health Inequalities following the Complexity Turn in Public Health: A Systematic Review

“…Papers that used ABM to model the socioeconomic differences in dietary behaviours tested the impact of interventions on the consumption of sugar-sweetened beverages [ 19 ], the purchase of ultra-processed food [ 20 ], the consumption of fruits and vegetables [ 21 , 22 ], and the access to healthy food outlets [ 23 ]. The interventions were educational campaigns (e.g., nutrition warnings and school-based programmes), advertising campaigns, changes to tax, increasing access to vegetables, and reducing the cost of vegetables.…”

“…The only paper that did not include a spatial component to the model was set at a national level, and explored the impact of tax, nutrition warnings, and advertising on the purchase of ultra-processed food in Mexico [ 20 ]. The other six models used artificial grid space [ 24 ], a 1-dimensional linear township [ 25 ], a raster map to represent the spatial distribution of income [ 21 ], or actual geographic space, including GIS modelling of real-life cities [ 19 , 22 , 23 ]. Six of the models included agent–environment interactions which often captured how individual agents engage with food outlets [ 21 , 22 , 23 , 24 , 25 ].…”

“…The other six models used artificial grid space [ 24 ], a 1-dimensional linear township [ 25 ], a raster map to represent the spatial distribution of income [ 21 ], or actual geographic space, including GIS modelling of real-life cities [ 19 , 22 , 23 ]. Six of the models included agent–environment interactions which often captured how individual agents engage with food outlets [ 21 , 22 , 23 , 24 , 25 ]. Only two of the included papers modelled agent–agent interactions through dietary social norms operationalised via a social network which influenced the taste preferences and health beliefs [ 22 ], and the purchasing of ultra-processed foods [ 20 ].…”

“…Only two of the included papers modelled agent–agent interactions through dietary social norms operationalised via a social network which influenced the taste preferences and health beliefs [ 22 ], and the purchasing of ultra-processed foods [ 20 ]. Five of the models featured feedback loops, these included the update of social norms based on behaviour over the course of the simulation [ 20 , 22 ], and the food outlet responses to the agent’s behaviour by closing and opening outlets [ 23 ], changing the type of food available for sale [ 24 ], or an increasing appetite and overeating following the consumption of foods high in fat, sugar, and salt [ 21 ]. Only two of the papers had attempted validation by comparing the simulated outcomes to the ‘observed’ outcomes in real world data [ 19 , 22 ].…”

Healthy Diet Promotion through Evidence-Informed Policies

“…explore the relationship between the food environment, purchasing patterns, and dietary behavior. These studies underscore the importance of considering geographic and economic factors when designing interventions to improve access to healthy foods [5][6][7][8][9]. In particular, Salvo et al highlight the potential impact of policies that jointly increase geographic and economic access to fruit and vegetables among low-income, predominantly Latino urban residents [9].…”