Large-scale diet tracking data reveal disparate associations between food environment and diet

Unhealthy diets are a leading cause of chronic diseases globally. Limited access to healthy food options is considered a contributing factor. However, prior research on the impact of food environments on diet has yielded mixed results due to limitations in sample size, geographic scope, study population, and measurement inconsistencies. This study addresses these limitations by employing a massive dataset of self-reported dietary information from a large number of smartphone users of the MyFitnessPal app. This dataset, encompassing 1,164,926 US participants and 2.3 billion food entries, allows for a nationwide investigation of the relationship between food environment factors (grocery and fast-food access), socioeconomic status (income and education), and dietary health outcomes (consumption of fresh fruits and vegetables, fast food, and soda; body mass index). This represents the largest nationwide study of its kind to date, offering an unprecedented opportunity to examine geographical and other forms of heterogeneity in the effects of food environments on diet.

Previous studies exploring the link between the food environment and diet have produced inconsistent results. These inconsistencies are often attributed to methodological limitations, including small sample sizes, localized studies, and variations in study design and measurement across studies. The authors review existing literature highlighting these methodological limitations and the resulting inconsistencies in findings regarding the effect of factors like access to grocery stores and fast food on diet quality and obesity. They note the scarcity of studies using nationally representative samples. The review sets the stage for the current study, which aims to address these shortcomings by analyzing a large, nationwide dataset.

This cross-sectional study utilized anonymized, retrospective data from MyFitnessPal (MFP), a popular smartphone application for tracking food intake and BMI. The data included 2.3 billion food entries from 1,164,926 US participants across 9822 zip codes, spanning 1730 counties (55% of all US counties). The study only included zip codes with at least 30 participants to ensure data reliability and privacy. Data validation involved comparing the MFP data to existing large-scale datasets, such as the Behavioral Risk Factor Surveillance System (BRFSS) survey data and Nielsen purchase data, demonstrating high correlations. The MFP food entries were categorized into fresh fruits and vegetables, fast food, and soda. Demographic and socioeconomic data (income, education level, racial demographics) were obtained from CensusReporter, using data aggregated at the zip code level. Grocery store access was determined based on the USDA Food Access Research Atlas, employing a 0.5-mile threshold regardless of urban or rural setting. Fast food access was measured by analyzing Yelp data, considering the nearest 1000 businesses within a 40km radius of each zip code's center. The statistical analysis used a matching-based approach, employing genetic matching to create matched pairs of zip codes with above-median and below-median values for each factor of interest (income, education, grocery store access, fast food access). This method aimed to control for confounding factors and isolate the individual effects of each variable. The analysis was repeated for zip codes with predominantly Black, Hispanic, and non-Hispanic white populations to explore disparities across ethnic groups. Multiple validation experiments were carried out to evaluate the accuracy of the food classification and the robustness of the analytical methods.

The study found that higher income, higher education, higher grocery store access, and lower fast-food access were generally associated with healthier diets across the entire US population. Specifically, in zip codes with above-median grocery store access, participants logged 3.4% more fruits and vegetables, 7.6% less fast food, 6.4% less soda, and were 2.4% less likely to be obese (all p < 0.001). Similar trends were observed for lower fast-food access and higher education. However, income had a slight (0.6%), non-significant positive association with obesity. When the analysis was stratified by predominant ethnic group in the zip code, significant disparities emerged. In zip codes predominantly Black, higher income showed an inverse relationship with healthy food consumption, leading to lower fruit and vegetable consumption and a higher likelihood of obesity. In contrast, in predominantly Hispanic zip codes, higher income was positively associated with fruit and vegetable consumption. High educational attainment demonstrated the strongest positive associations with diet health across all ethnic groups. Grocery store access showed a substantially larger positive association with fruit and vegetable consumption in zip codes with predominantly Hispanic (7.4% difference) and Black populations (10.2% difference) compared to predominantly white populations (1.7% difference). The results indicate that the associations between food environment, socioeconomic factors, and dietary health outcomes vary considerably across different racial/ethnic groups.

The findings highlight the complex interplay between food environments, socioeconomic factors, and dietary behaviors across different racial and ethnic groups. The significant disparities observed across ethnic groups underscore the need for targeted interventions tailored to specific contexts and populations. The study's large scale and nationwide scope provides strong evidence of these disparities, indicating that a one-size-fits-all approach to improving dietary health is likely to be ineffective. Higher educational attainment consistently showed the strongest effect on healthier outcomes, suggesting a critical role of education in promoting healthy food choices. Future research should delve into the underlying mechanisms driving these disparities and explore the efficacy of community-based interventions that address both access to healthy food and education levels.

This study, using a massive dataset of smartphone-based diet logs, demonstrates significant associations between food environment, socioeconomic status, and dietary health outcomes across the US. However, these associations vary substantially across different ethnic groups. While higher grocery store access and lower fast food access showed overall positive correlations with healthier diets, the impact differed across racial demographics, highlighting the importance of targeted interventions for specific populations. Higher educational attainment is shown to be a key predictor of healthier diets and lower obesity prevalence across all ethnicities. Future research needs to incorporate longitudinal data to establish causality and investigate the efficacy of various targeted intervention strategies.

The cross-sectional nature of the study limits causal inferences. The use of self-reported data via the MyFitnessPal app, while providing a large-scale dataset, introduces potential biases, including underreporting and selection bias due to the sample not being fully representative of the US population. The study also lacks granular data on the specific types of foods consumed, relying primarily on broader categories, limiting the depth of analysis. Furthermore, the study does not directly measure the quality of food offerings in grocery stores and fast-food restaurants.