Integrating human activity into food environments can better predict cardiometabolic diseases in the United States

The United States faces a significant burden of cardiometabolic diseases (CMDs), including obesity, hypertension, and high cholesterol, impacting a substantial portion of the adult population. These diseases are strongly linked to the country's evolving food system, marked by the proliferation of fast-food chains, increased consumption of ultra-processed foods, and the decentralization of residential communities leading to uneven access to healthy food options. Existing research on the connection between food environments and CMDs has yielded inconsistent results, with studies showing positive, negative, null, and mixed relationships between fast-food restaurant accessibility and obesity. This inconsistency highlights the limitations of traditional food environment measures which primarily focus on the density and proximity of food retailers within a fixed geographic area, neglecting the dynamic aspect of human mobility and food procurement behaviors. Individuals frequently travel beyond their immediate neighborhoods to acquire food, and their choices are influenced by various factors including cultural preferences, health literacy, and food security. This study addresses these limitations by integrating human mobility data into a novel food environment index, allowing for a more comprehensive understanding of the relationship between food environments and CMDs.

Numerous studies have explored the link between food environments and cardiometabolic health outcomes, employing measures such as the Food Access Research Atlas and the modified retail food environment index (mRFEI). However, the results have been inconsistent, partly due to variations in measurement approaches, confounding factors, and the limited consideration of human mobility. While some studies linked fast-food restaurant accessibility to obesity, others reported null or negative correlations. This highlights the importance of considering human behavior and travel patterns when assessing food environment impacts on health. The inconsistent findings suggest that location-based measures alone may not fully capture the complexity of how individuals interact with and navigate their food environments.

This nationwide observational study developed a novel retail food activity index (RFAI) at the census tract level, incorporating human mobility data from SafeGraph's Core Places and Patterns datasets. This dataset encompasses over 94 million aggregated visit records to approximately 359,000 food retailers across the US over two years (2018-2019). The RFAI is defined as the percentage of visits to healthy food retailers (supermarkets, grocery stores, warehouse clubs, fruit and vegetable markets) relative to total visits to all food retailers (including convenience stores and limited-service restaurants). This differs from the location-based mRFEI, which considers only the presence of food retailers within a census tract. The study included analysis of food retailer visit patterns, examining the distances traveled to food retailers and variations by urban status and sociodemographic groups. Multivariate linear regression was used to assess the associations between sociodemographic characteristics and both the RFAI and mRFEI. Further, multivariate linear regressions investigated the associations between both indices (RFAI and mRFEI) and the prevalence of five CMDs (obesity, high blood pressure, high cholesterol, diabetes, and coronary heart disease) at the census tract level, adjusting for various covariates. Several sensitivity analyses were conducted to ensure the robustness of the findings, including exploring non-linear relationships, comparing against an updated location-based index, performing county-level analysis, and addressing spatial autocorrelations. Data on CMD prevalence were obtained from the CDC's PLACES data, while sociodemographic variables were derived from the American Community Survey and the CDC's Social Vulnerability Index.

The analysis revealed that the majority (approximately 80%) of food retailer visits occurred outside residents' immediate neighborhoods, highlighting the limitations of solely location-based measures. The RFAI demonstrated a weak correlation with the mRFEI, indicating distinct patterns in activity-based versus location-based assessments. Socioeconomically deprived census tracts exhibited lower RFAIs. Census tracts with predominantly Hispanic populations had significantly higher RFAIs compared to other groups, a pattern not entirely mirrored by the mRFEI. Importantly, the RFAI showed stronger associations with the prevalence of obesity, high cholesterol, and high blood pressure than the mRFEI. For instance, a one interquartile range increase in RFAI was associated with a 0.629% lower prevalence of obesity, compared to 0.174% for mRFEI. These effects remained largely linear and robust across various sensitivity analyses, indicating the RFAI's stronger predictive power for multiple CMDs. The study also found that residents in lower population density areas and non-urban areas traveled longer distances for food, consistent with existing literature. Racial and ethnic disparities were also observed in food retailer visit patterns and distances traveled.

This study's findings challenge the reliance on traditional location-based measures of food environments, demonstrating that human mobility and activity patterns are crucial factors in understanding the relationship between food access and cardiometabolic health. The RFAI, by integrating human mobility data, provides a more nuanced and accurate assessment of food environments and their impact on health. The stronger association between RFAI and CMD prevalence compared to mRFEI highlights the importance of considering behavioral aspects of food acquisition. The observed racial and ethnic disparities in food retailer visit patterns underscore the need for culturally sensitive interventions. The results suggest that policies focusing solely on improving local food access may have limited effectiveness, and interventions should consider broader geographic access and individual behaviors.

This study introduced the RFAI, a novel activity-based index that outperforms location-based measures in predicting CMD prevalence. The RFAI offers valuable insights for policymakers seeking to create more effective food policies and health interventions. Future research could explore the individual-level determinants of food procurement behavior and refine the RFAI to incorporate additional factors such as food quality and dietary practices. Developing an interactive analysis platform for the RFAI would enhance its accessibility and applicability.

The study's limitations include the ecological fallacy inherent in aggregating mobility data at the census tract level, potential underrepresentation of certain demographic groups in the SafeGraph data, the observational nature precluding causal inferences, and potential minor differences in the definitions of healthy and unhealthy food retailers between the RFAI and mRFEI. Future research should employ longitudinal studies with individual-level data and explore ways to mitigate biases in the mobility data.