Disparities of children’s obesogenic environments in Louisiana

Abstract

Background

Childhood obesity remains a pressing public health concern, with significant long-term implications for children’s growth and development, as well as an increased risk for various diseases. In recent years, the obesity prevalence among children has shown a concerning rise, particularly in the Deep South (e.g., Louisiana) exhibiting higher rates compared to other regions of the United States. Meanwhile, environmental factors play a critical role in shaping children’s health behaviors and outcomes. There has been an increasing body of research on the relation between obesogenic environment and health, but far less focused on child health.

Methods

This study provides a comprehensive measurement of obesogenic environments for children in Louisiana covering park accessibility, walkability, and food environment at the ZIP Code area level. Geographic information system (GIS) analyses revealed significant urban-rural disparities and heterogeneity by race and socioeconomic deprivation. A three-way analysis of variance (ANOVA) further revealed the main and interaction effects of urbanicity, Black population percentage, and area deprivation index (ADI) on obesogenic environment indicators.

Results

Regarding the disparities of walkability, not only the individual factors including urbanicity, Black population percentage, and ADI significantly affect walkability, but also the interaction terms among these factors, including both two-way and three-way interactions, are significant. Urban-rural differences and ADI levels contribute to disparities in the food environment. But no significant disparities are observed in park accessibility across different subgroups.

Conclusion

While park accessibility disparities are limited, the food environment varies significantly by urbanicity and ADI level, walkability shows complex socio-spatial interactions. The study sheds light on targeted public health interventions and urban planning to close the gaps in obesogenic environments for children’s health.

Introduction

Childhood obesity has become a global public health concern, with rising prevalence in both developed and developing countries. According to World Health Organization (WHO), from 1990 to 2020, the prevalence of overweight among 5–19-year-olds increased from 8% to 20%, affecting over 390 million children worldwide, while the prevalence of obesity rose from 2% to 8%, impacting 160 million young people [1]. International research further indicates that children’s health status and behaviors are influenced by the obesogenic environments, including physical activity infrastructure, food availability, and socioeconomic inequalities [2]. In the United States, childhood obesity remains a critical public health challenge. According to the National Health Statistics Reports, the obesity prevalence among children and adolescents aged 2–19 years increased from 17% in 2011–2014 to 19.7% in 2017–2020, with an even higher rate of 20.7% among school-age children aged 6–11 years during the same period [3, 4]. Obesity is more prevalent in racial or ethnic minorities from lower socioeconomic backgrounds, and disparities had worsened during the recent COVID-19 pandemic [5]. Childhood obesity is associated with numerous comorbidities, including hypertension, cardiovascular diseases, diabetes mellitus type 2 (T2DM), and other health conditions [6]. Therefore, it is important to establish behaviors and health patterns in early life through coordinated intervention efforts to prevent and control obesity and other health disorders in adulthood [4, 7, 8].

A growing body of research has emphasized the role of the obesogenic environment including the physical, social, and economic surroundings that promote unhealthy diet and physical inactivity in shaping childhood obesity risk [2, 9]. Built environment factors such as walkability, access to recreational facilities, and availability of healthy food options have been linked to children’s health outcomes [10]. However, most existing studies on the relationship between obesogenic environment and obesity have concerned more about adult populations, with limited research on children populations, who are more vulnerable to environmental exposures and less able to independently make health-related decisions [11–13]. In the Deep South states (i.e., Alabama, Georgia, Louisiana, Mississippi, and South Carolina), obesity prevalence is disproportionately higher than the national average [14]. According to the 2022–2023 National Survey of Children’s Health, the obesity rate among children aged 6 to 17 years in Louisiana was 23.1%, the third highest in the nation [15]. In Louisiana, structural and environmental inequalities, compounded by socioeconomic challenges, might have contributed to the increased vulnerability of children to obesity and related health conditions. Understanding and addressing these disparities is crucial for guiding effective place-based public health interventions.

This study addresses these gaps by focusing on the obesogenic environment for school-aged children (aged 5–14) in Louisiana. Specifically, it seeks to answer two research questions: (1) How can the obesogenic environment for children be measured in a comprehensive and objective manner? (2) What are the spatial and demographic disparities in children’s obesogenic environments across Louisiana? By developing robust measurement frameworks and analyzing regional disparities, this study contributes to the growing field of environmental health geography and offers policy-relevant insights for obesity prevention efforts in vulnerable children. Building on our previous work, which provided a comprehensive review of measurement approaches and identified significant disparities of obesogenic environments in a Louisiana region (i.e. East Baton Rouge Parish) using linear regression [16], the present study extends several advances. It integrates three key and objectively measured dimensions of the obesogenic environment, covering park accessibility, walkability, and food environment, which provides a more comprehensive assessment than studies focusing on only one domain. Meanwhile, measurements are tailored to school-aged children by incorporating the population aged 5–14 into park accessibility calculations, thereby enhancing the child-relevance of the exposure metrics. This study expands the geographic scope to the entire state of Louisiana, enabling a broader examination of spatial disparities. Beyond descriptive differences by single indicator, we investigate interactions among environmental indicators using three-way ANOVA, offering new insights into how multiple obesogenic exposures co-vary across subgroups.

Literature review

Obesogenic environment for children

Generally, the built environment is defined as a sum of artificially developed physical environments where people live and work (e.g., homes, buildings, streets, open spaces, urban spaces, walkways, roads, parks, and infrastructure) [12]. In early public health studies, obesogenic environment was defined as the sum of the environmental influence including surroundings, opportunities and life conditions that promote individual obesity [17]. The Centers for Disease Control and Prevention (CDC) listed several societal and community factors that could affect children obesity including childcare and school environments, neighborhood design, access to healthy, affordable foods and beverages, and access to safe and convenient places for physical activity (PA) [18]. In the context of children’s health, Kaczynski et al. defined the obesogenic environments as the combination of physical elements that promote sedentarism, restrict PA, and encourage unhealthy eating practices [19]. A recent review by Ortegon-Sanchez et al. summarized ten types of built environment factors, including residential or population density, intersection density (street connectivity), land-use diversity, walkability, street-level walking infrastructure, perceptions of street environments, accessibility to recreation and sports facilities, proximity to school for active transport, access to public open spaces, safety perceptions, motorized traffic levels, and social support and psychosocial factors [2]. These factors either facilitate or inhibit PA and dietary behaviors, which in turn continuously influence residents’ energy balance.

Most research focuses primarily on two main aspects when exploring children’s obesogenic environments: the PA environment and the food/nutrition environment. At the neighborhood level, Saelens et al. developed several Geographic Information Systems (GIS) − based neighborhood metrics to measure supportive or unsupportive neighborhood environments conducive for PA and healthy eating [20]. By considering both high/low PA environments and high/low nutrition environments, Saelens et al. categorized neighborhoods into four types: High PA Environment/high Nutrition Environment (High PAE/high NE), High PAE/low NE, Low PAE/high NE, and Low PAE/low NE [20]. Deweese et al. utilized food and PA outlets to measure neighborhood characteristics [21]. Specifically, supportive factors for PA included walkability and access to high-quality parks, while the favorable nutrition environment is defined by the availability of supermarkets offering healthy food rather than fast-food outlets offering less healthy food [21, 22]. More recently, Kaczynski et al. developed the Childhood Obesogenic Environment Index (COEI), integrating diverse food and PA indicators, and also incorporating broader contextual factors including percent births at baby-friendly facilities, violent crime, and school proximity [19]. These studies highlighted the dual importance of PA and food environments, which form the foundation for more comprehensive measurement of children’s obesogenic environments. For clarity, we use the term “food environment” consistently, replacing “nutrition environment”.

Access to PA facilities is one of the top themes in recent research on children’s obesity and PA behavior [23]. Count, presence or absence, proximity of relevant parks and recreational facilities surrounding home, schools or route to/from school have been commonly employed to measure the PA environment [2, 22, 24]. Additionally, neighborhood amenity factors, including walkability, street connectivity, intersection density, street design interventions, local traffic patterns, crime rates, and safety measures, have been commonly considered in urban planning and the built environment-based health interventions [2, 25, 26]. The PA environment can thus be conceptualized as children’s exposure to PA resources and their potential access to these resources, including the presence, proximity, and quality of recreational facilities.

Food environment is another essential factor for children’s obesity and health. With wearable camera and GPS, Chambers et al. collected data on children aged 11–13 in New Zealand and found that children spent more time at food retail outlets than at structured sports or outdoor recreation locations combined [27]. School-age children occupy a unique developmental stage, as they gradually become less dependent on their families and face increased exposure to food in their neighborhood environment [8], while their food access is relatively limited to a shorter spatial range compared to adults [28]. However, the food environment surrounding children may not be particularly healthy or supportive. Engler-Stringer et al., in a study of a medium-sized city in Canada, found that areas with higher concentrations of children aged 5–14 years tended to be located farther from the nearest healthy food stores [29].

Regarding the food environment measurement, there were major discrepancies in how researchers and respondents defined the food environment and in using different geographical units of analysis. The differences in definitions of the food environment contribute to conflicting, inconsistent results and make it difficult to draw meaningful comparisons on the association between children’s health and food environment [8]. Current measurements of the food environment primarily rely on the Food Access Research Atlas (FARA) from the U.S. Department of Agriculture (USDA) and definitions provided by the CDC regarding healthy foods and healthy food environments. The FARA defines the healthy food vendor as supercenters, supermarkets, and large grocery stores, and measures the food environment as the distance to these three types of food providers [30]. Similarly, several studies adopted comparable approaches to measuring the food environment by using specific retail stores or restaurants such as supermarkets or fast food to represent healthy or less healthy environments, and quantified the environment through metrics such as distance, density, or count [31–34]. The CDC, on the other hand, classifies stores and restaurants as healthy or less healthy food retailers based on the North American Industry Classification System (NAICS). It uses the Modified Retail Food Environment Index (mRFEI), defined as the ratio of healthy food retailers to the total number of food retailers, to represent the food environment [35]. Conceptually, food environment is a component of the broader obesogenic environment and shares a common characteristic that it represents the sum of surrounding conditions and opportunities that may facilitate or hinder healthy behaviors. Different measurement approaches reflect the conception in various ways. For example, the USDA FARA emphasizes children’s accessibility to healthy food outlets, whereas the CDC’s mRFEI considers the relative presence of healthy versus less healthy food retailers within children’s living environments. Consequently, mRFEI is often preferred over the FARA measures for providing a relative measure of healthy food access [36], which has been widely adopted in related research.

Disparities of obesogenic environment

Building upon the measurements, disparities analysis from prior research reveals that the built environment varies across different geographic units, particularly with respect to urbanicity, race and socioeconomic factors, which also correspond to disparities in obesity rate [37]. At the national county level, Kaczynski et al. found that metropolitan counties generally had significantly more favorable Childhood Obesogenic Environment Index (COEI) scores, considering both PA and food environments, compared with micropolitan and rural counties [19]. Consistent with this, many studies have reported that children in rural areas have less access to healthy food outlets and PA environments [38, 39]. Rural areas also have been reported to have less friendly but high disparity PA environment [40, 41]. Wende et al. observed that disparities in food environments were relatively small, and rural counties demonstrated significantly better food environment scores [40]. One possible explanation was the use of population-weighted measures, the presence of even a limited number of food stores in rural areas with smaller populations can disproportionately elevate average scores. Furthermore, Wende et al. noted that previous research on rural–urban disparities lacked a consistent definition of urbanicity [39]. By applying multiple classification systems, they confirmed that childhood obesogenic environments exhibited significant rural–urban disparities across various urbanicity classifications [39].

In addition to the geographic disparities, socioeconomic disparities also play a critical role in influencing children’s obesogenic environments and health outcomes [10]. For the PA environment, White and urban adolescents tend to tolerate longer walking and driving distances compared with Black and non-urban peers, respectively [42]. Low socioeconomic position (SEP) children often face greater barriers to PA, including limited or poor-quality facilities, safety concerns, and inadequate public transportation, while higher-SEP families are more likely to prioritize safety and accessibility, particularly in rural settings [23]. Similarly, disparities also exist for the food environment. Proximity to fast-food outlets negatively impacted weight outcomes primarily among Black and Hispanic children, with less effect observed for White and Asian children [43]. Neighborhoods with higher minorities concentrations and low SEP had limited accessibility to healthy foods [38, 39]. Meanwhile, Black/African Americans may face unique dietary challenges, as their food preferences and choices often differ from Whites, and supermarkets do not always provide the culturally preferred healthy options [44].

Importantly, these disparities do not operate in isolation. While many studies have examined disparities by race, SEP, or urbanicity separately and discussed the potential interactions in interpreting results, few studies have systematically investigated how these factors interact. For instance, low-SEP areas showed that higher walkability may be related to the more urbanized areas, low-income and minority neighborhoods [37]. Through two-way analysis of variance (ANOVA), Kaczynski et al. found significant interactions of rurality with race and income regarding the disparities of obesogenic environments across US counties [45]. Low-income counties and those with more racial/ethnic minority residents, especially in rural areas, had less supportive PA and healthy eating environments for youth [45].

In summary, recent studies have employed various data sources and methodologies to measure and define children’s obesogenic environments. Some focused on isolated aspects of the built environment, either the PA or food environment. But childhood obesity is a complex social issue affected by many factors. A more comprehensive and multidimensional measurement of the obesogenic environment is crucial for better understanding and addressing childhood obesity [10]. Meanwhile, most existing research still relies on subjective questionnaires, lacking standardized objective metrics and comparability across different regions. While disparities research has increasingly considered socioeconomic factors, few studies have thoroughly examined the complex interactions among these variables. This study addresses these gaps by comprehensively and objectively measuring obesogenic environments for children, encompassing both PA and food environments. Furthermore, compared with the precious studies that discussed the disparities separately, this study adopts the three-way ANOVA to investigate the intricate interactions among spatial and socioeconomic factors contributing to disparities in children’s obesogenic environments. By comprehensively considering both PA and food environment, this study aims to better capture the spatial patterns of obesogenic environment across Louisiana. It also hypothesizes that the interactions among urbanicity, race and socioeconomic disadvantage level will significantly affect the obesogenic environment for children.

Study area

The study area encompasses the entire state of Louisiana. This study examines disparities of children’s obesity and built environment across areas from urban to rural areas, areas with various levels of Area Deprivation Index (ADI), and areas of various levels of racial minority (e.g., Black) distribution. ADI is considered a comprehensive metric that captures neighborhood socioeconomic disadvantages [46]. The built environment in this study is conceptualized to encompass two main domains: PA environment and food environment. Within the PA environment, this study considers access to parks and playgrounds, as well as walkability, representing the environmental factors influencing children’s general PA and active transportation, respectively. Food environment is captured by the comprehensive Modified Retail Food Environment Index (mRFEI).

According to the 2020 Census Urban and Rural Classification standard, an area is defined as urban area (UA) if density ≥ 425 housing units per square mile (HPSM), low-density fill zone (hereafter simply referred to as low density, LD) if density = 200 ~ 425 HPSM, and rural area (RA) if density < 200 HPSM [47]. The data for urbanicity is resourced from the 2020 census data [48]. In Louisiana, 75 ZIP Code areas are urban areas, 47 ZIP Code areas are low-density fill zones, and 418 ZIP Code areas are rural areas (Fig. 1). There are four major urban areas such as New Orleans, Baton Rouge, Lafayette and Shreveport. Some low-density areas are around them and others are scattered across the state, and most of the state is rural. The analysis unit of ZIP code area is commonly recognized by residents, and thus research results can be better publicized.

Fig. 1.

Urban, low-density, and rural areas in Louisiana at the ZIP Code area level

Based on the 2020 Decennial Census data, Louisiana’s population is approximately 60.62% White and 33.39% Black, with other racial or ethnic groups constituting a relatively small share of the state population [48]. Given this distribution, other racial or ethnic groups represent small populations (< 5%) and are therefore not included in this study. To conduct the analysis, this study used the 2021 American Community Survey (ACS) 5-year estimates [49], to classify the ZIP Code areas with various Black population percentage (BlackPop%) into 5 groups: 0–20, 20–40, 40–60, 60–80, and 80–100. The classification is considered semi-open intervals, where the left boundary was excluded and values less than or equal to the right boundary were included. As shown in Fig. 2, the percentage of Black population tend to coincide with urban areas to some degree, but also in rural areas in the far north of the state as well as the northern edge on the east part of the state.

Fig. 2.

Black population percentage across ZIP Code areas in Louisiana

Area deprivation index (ADI) was based on a measure created by the Health Resources and Services Administration over three decades ago and has since been refined, adapted, and validated at the 9-digit ZIP Code area level by the School of Medicine and Public Health, University of Wisconsin [46]. The state ranking values of 2021 ADI range from group 1 indicating the lowest disadvantaged level to group 10 representing the most disadvantaged level. ADI by 9-digit ZIP Codes is transformed into 5-digit ZIP Codes, with the average value being computed. Overall, the ADI range in Louisiana tends to be high, indicating a high level of disadvantage across the state. This trend is particularly pronounced in the northern and central regions (Fig. 3).

Fig. 3.

ADI in Louisiana at the 5-digit ZIP Code area level

Generally, with respect to urbanicity, Black population percentage, and ADI, Louisiana is characterized by large rural areas, communities with pronounced racial segregation, and overall high levels of socioeconomic deprivation.

Children obesogenic environment measures

Building on previous research and the literature on children’s obesogenic environments, this study focuses on two major domains: the physical activity (PA) environment and the food environment. The PA environment is evaluated through two components: (1) access to parks and playgrounds and (2) walkability. Park accessibility reflects potential opportunities for leisure and recreational physical activity1, whereas walkability primarily captures the feasibility of active commuting and routine walking behaviors among children. Thus, they address different dimensions and intensities of children’s PA. The food environment is measured using the Modified Retail Food Environment Index (mRFEI) values calculated based on the CDC's methodology. The data sources and methodological approaches utilized to quantify the children’s obesogenic environment are summarized in Table 1.

Table 1.

Data sources and description

	Variable Category	Variables	Data Source	Year
Obesogenic Environment	Park accessibility	4-15-year-old Children Population	American Community Survey (ACS)	2023_5Year
		USA Park	Esri	2024
	Walkability	National Walkability Index (NWI)	U.S. Environmental Protection Agency (EPA)	2021
	Food environment	Modified Retail Food Environment Index (mRFEI)	ZIP Business Patterns (ZBP)	2022

Park accessibility

Based on a survey of measures of spatial accessibility of resource or service, various methods strive to capture two elements as users value proximity to the service facilities around them and ease for securing that service, such as a ratio of supply capacity vs. demand [50]. The Two-Step Floating Catchment Area (2SFCA) method [51] is adopted as it integrates both elements.

A simple version of the 2SFCA model for accessibility A_i at a demand location (here, census block group) i is written as:

where the supply capacity of a park at location j is denoted by S_j (e.g., park area), the demand of census block group k (or i) is denoted by D_k (or D_i) (population, or precisely 5-14-years-old children in this study), and the spatial impedance between them is d_kj (or d_ij) measured as drive time in minutes2. The first step is for each supply (park) location (j) to search all demand locations (k) within a threshold drive time (d₀) from location j, and compute the ratio of supply (park areas) at j to the total demands (children) within that catchment area. The second step is for each block group location (i) to search all park locations (j) that are within the threshold drive time (d₀) from i, and sum up the previously derived areas-to-children ratios within its catchment area.

The resulting accessibility value is the ratio of supply to demand, here park areas per child, for each ZIP Code area. Three variables need to be defined as inputs to the method. First, estimated demand was the population of children aged 5–14 in each ZIP Code area, resourced from 2023 ACS 5-Year Estimates [54]. Secondly, supply was measured as park area with data obtained from Esri’s USA Parks dataset [55], which included information on park names, classifications, and area sizes. For missing values in some park area records, their areas were calibrated in ArcGIS Pro. As illustrated in Fig. 4, most parks and playgrounds are in urban areas, some in low-density areas, and very few in rural (also refer to Fig. 1). They are particularly concentrated around New Orleans, Baton Rouge and Shreveport. Thirdly, the travel time matrix between demand and supply locations via the road network was estimated by ArcGIS Network Analysis with the road network data sourced from the OpenStreetMap [56].

Fig. 4.

Parks in Louisiana

One important parameter in the 2SFCA method, catchment areas (d₀), needs to be defined. As the tolerance of drive range for park access may vary by children, this research conducted a sensitivity analysis by experimenting with various catchment areas (d₀) such as 5, 10, 15, and 20 min. As shown in Fig. 5a-d, accessibility is higher in major urban areas and declines towards surrounding and rural areas. As the catchment area size increases, parks become accessible for children in less favorable locations including those in rural areas, and therefore the rural-urban gap in accessibility is less pronounced. This effect is commonly referred to as stronger spatial smoothing with a larger catchment area size [57].

Fig. 5.

Park accessibility measured by 2SFCA with various catchment area sizes: (a) d₀ = 5 min, (b) d₀ = 10 min, (c) d₀ = 15 min, and (d) d₀ = 20 min

Table 2 represents the descriptive statistics of park accessibility scores across four travel time thresholds (d₀ = 5, 10, 15, and 20 min), at the ZIP Code area level. Two sets of statistics are reported: the original values and the values after excluding zeros and outliers. In the original data, the mean accessibility scores appeared to increase with larger catchment areas. However, the standard deviations were extremely large across all scenarios, and the medians remain very close to zero. This pattern indicates a highly skewed distribution, where a small number of ZIP Codes with exceptionally high accessibility scores, whereas most areas exhibit limited or no access to parks within the specified time thresholds. To better reflect the central trends and reduce the influence of extreme values, we generated a second summary by excluding zero values (i.e., ZIP Codes with no children population or no accessible park area within the catchment areas) and removing outliers, defined as values exceeding the 95th percentile among non-zero observations. After this filtering process, the mean accessibility scores increase with longer travel times, from 0.144 km²/1,000 children at d₀ = 5 min to 8.030 km²/1,000 children at d₀ = 20 min, reflecting the expansion of reachable park areas as catchment sizes grow. The standard deviation (SD) also increases with time, indicating greater variability in park accessibility among ZIP Codes as the travel radius expands. This is consistent with the finding from prior studies [51] that a larger catchment size leads to a stronger smoothing effect and thus a wider variance (or standard deviation). Meanwhile, the median values are consistently lower than the means, suggesting that the data remains right skewed, though much less so than in the original data.

Table 2.

Summary statistics for park accessibility within different catchment areas (km²/1000 children)

Catchment areas	Mean	SD	Min	Median	Max
d₀ = 5 min	6.126	140.164	0.000	0.000	3257.131
d₀ = 10 min	8.515	142.174	0.000	0.000	3257.131
d₀ = 15 min	145.540	2290.394	0.000	0.010	50052.100
d₀ = 20 min	205.300	2370.316	0.000	0.050	50052.100

As shown in Fig. 6a, after excluding zero values and extreme outliers (above the 95th percentile) and applying a log-transformation [log(accessibility + 1)], the density distributions of park accessibility scores demonstrate systematic changes across four catchment areas (d₀ = 5, 10, 15, and 20 min). Accessibility scores for shorter catchment areas (e.g., 5 min) exhibit a more concentrated, unimodal density distribution centered around lower values, indicating that most ZIP Codes have similar, relatively low accessibility within this restricted travel time. As the catchment area increases to 10, 15, and 20 min, the distributions become progressively more dispersed and develop bimodal characteristics, with accessibility scores spanning a wider range and showing two distinct peaks.

Fig. 6.

Comparing 2SFCA park accessibility scores by various d₀: (a) probability density function curves for accessibility when d₀ = 5, 10, 15 and 20, and (b) correlation between accessibility scores d₀ = 10 vs. d₀ = 15

More importantly, accessibility is a measure of relative convenience of acquiring a service and only the relative difference across geographic areas is meaningful. As shown in Fig. 6b, accessibility scores across block groups are largely consistent between d₀ = 10 vs. 15 min. Therefore, in subsequent modeling, d₀ = 10 min is chosen to provide an appropriate balance between capturing meaningful accessibility variations while maintaining computational efficiency for the analysis that follows. The filtering of zero and extreme values is applied solely for visualization purposes, to make the density distributions and comparison more interpretable. All original data points, including the zero and extreme values, are included in the statistical analyses.

Walkability

Data for walkability was extracted from the National Walkability Index from the U.S. Environmental Protection Agency (EPA) [58]. Specifically, the index is a weighted average of multiple factors defined in the Smart Location Database (SLD), such as:

where w, x, y, and z are a block group’s ranked scores for intersection density, proximity to transit stops, employment, employment and household mix, respectively. The first two factors are weighed twice as much as the last two factors.

The block groups are assigned to their final National Walkability Index scores on a scale of 1 to 20, and correspond to least walkable, below average walkable, above average walkable, and most walkable (Table 3). As shown in Fig. 7a, most areas in Louisiana are in the least walkable or below average walkable categories. Only a few areas in urban areas are above average walkable, and a limited number of the areas around New Orleans reach the standard of most walkable. This reflects overall poor walkability throughout Louisiana.

Table 3.

Categories of National Walkability Index scores

National walkability index	Categories
1–5.75	Least walkable
5.76–10.5	Below average walkable
10.51–15.25	Above average walkable
15.26–20	Most walkable

Fig. 7.

National Walkability Index (NWI) in Louisiana

Census block groups and ZIP Code areas overlap but do not have a strict hierarchical relationship. This study assigned each block group to the ZIP Code area within which its geographic centroid fell, thereby establishing a linkage between block groups and ZIP Code areas. The data was then summarized by ZIP Code area by calculating the mean values of the block groups contained within each ZIP Code area, as shown in Fig. 7b. The results are consistent with those from the original data (Fig. 7a), indicating that higher walkability is mainly concentrated in urban areas, such as parts of New Orleans, while most rural areas exhibit the lowest levels of walkability.

Food environment

The measurement of mRFEI, the ratio of number of healthy food retails to the total number of food retails, was utilized to define the food environment across census tracts in the United States, based on the classification of NAICS codes by the CDC [35]. However, the data in 2011 were outdated and could not reflect the current situation. This study used updated data to measure mRFEI at the ZIP Code area level in Louisiana. The data source was the ZIP Code Business Patterns (ZBP) from the U.S Census Bureau [59], which included annual statistics for businesses with paid employees at the 2- through 6-digit NAICS industry levels. Starting with reference year 2019, the ZIP Code Business Patterns data has been made available as part of the County Business Patterns (CBP) API. Based on the 2020 NAICS codes, the data was collected from the 2022 County Business Patterns (CBP) APIs.

The measurement method of food environment is based on the mRFEI from the CDC [35] formulated as:

where healthy food includes supermarkets and larger grocery stores, fruit and vegetable market, while less healthy food retailers include fast food restaurants, small grocery stores and convenience stores. A higher mRFEI represents a higher percentage of healthy food in an area, and thus a healthier food environment.

In this study, food retail outlets were classified based on 2020 NAICS codes and resourced from using ZIP Business Patterns (ZBP) data [35, 59]. The food retailers of related classifications were spatially joined to ZIP Code boundaries via ArcGIS Pro and the mRFEI for each ZIP Code area was then calculated by aggregating the counts of healthy and less healthy retailers within each area.

Regarding to 2020 NAICS codes, healthy food retail outlets included NAICS codes = 445,230 and 452,311, and less healthy food retails included those with NAICS codes = 722,513 and 445,120. The NAICS code 445,110 includes both health (e.g., supermarkets and grocery stores) and less healthy food retailers (e.g., gas station stores) (Table 4). Specifically, as defined by the CDC’s mRFEI in 2011, outlets with 3 or fewer employees were classified as less healthy, and larger ones with 10 or more employees as healthy [35].

Table 4.

NAICS codes for food outlets in Louisiana

2007 NAICS code (CDC guide)	2022 ZBP NAICS code	Name	Categories	Count in Louisiana
445,110	445,110	Grocery stores	Supermarket or larger grocery stores with more than 10 employees are healthy food retails; small grocery stores with less than 3 employees are less healthy food retails.	686
445,230	445,230	Fruit & vegetable markets	Health	3
452,311	452,311	Warehouse clubs	Health	33
722,211	722,513	Limited-service restaurant	Less health	3,596
445,120	445,120	Convenience store	Less health	419
Total				4737

Figure 8 shows that the southern and urban areas of Louisiana exhibit higher mRFEI values and thus a healthier food environment. With comparison to Fig. 3, it is apparent that regions with lower ADI tend to have a healthier food environment. Such an observation needs to be confirmed by more rigorous analysis in the following section.

Fig. 8.

Food Environments across ZIP Code areas in Louisiana

Table 5 summarizes the key environmental measures used in this study, including Park Access, transferred NWI, and updated mRFEI. For park accessibility, scores are largely consistent across various scenarios of catchment areas (d₀) as discussed in the previous section, and therefore this study selects a middle-point value, i.e., d₀ = 10 min, as a representative case. Comparing the categories of NWI (Table 3), the mean NWI in Louisiana falls within the ‘below average’ to ‘least walkable’ categories. The range of mRFEI is from 0 to 100, and a higher mRFEI indicates a greater presence of healthy food retail stores. Based on the 2011 CDC national census tract level dataset, the national average mRFEI is 12.765 [35]. Based on these standards, the food environment indicators in Louisiana generally reflect poor conditions. This combination of limited access to healthy food options and low walkability overall may negatively impact children’s health.

Table 5.

Summary for obesogenic environments in Louisiana

Obesogenic environment	Mean	SD	Min	Median	Max
Park Access (d₀ = 10 min)	8.515	142.174	0.0	0.000	3257.131
NWI	5.318	2.589	1.5	4.617	18.167
mRFEI	4.445	12.922	0.0	0.000	100.000

Disparities analysis

Sample distribution

Figure 9 presents a sample size distribution heatmap illustrating data availability across the intersection of rurality level, Black population percentage and ADI level. The predominance of rural area (RA) samples, particularly those with lower Black population percentages, reflects Louisiana’s geographic composition where rural areas constitute a significant portion of the state’s ZIP Codes. While rural areas show robust sample sizes across most demographic categories (indicated by green cells), low-density (LD) and urban areas (UA) exhibit more limited sample sizes, particularly in certain demographic segments (indicated by red cells). This uneven distribution suggests that findings regarding rural patterns can be interpreted with greater confidence, while conclusions about large town/fringe and urban contexts should be approached more cautiously due to smaller sample sizes.

Fig. 9.

Sample size distribution heatmap

Descriptive statistics

The analysis of disparities in built environment begins with summarizing the basic statistics for these variables and those in each subset of category such as urbanicity, black population percentage, and ADI. Summary statistics of park accessibility, walkability index, and food environment by urbanicity, Black population percentage, and ADI are presented in Table 6 and Fig. 10.

Table 6.

Park accessibility, walkability, and food environment by urbanicity, Black population percentage, and ADI

	Park access (d0=10 min)(km2/1000 children)		Walkability Index		Food environment(mRFEI)
	M	SD	M	SD	M	SD
Urbanicity
RA	10.800	162.000	4.490	1.750	3.430	13.700
LD	1.200	6.020	6.160	1.210	6.330	8.640
UA	0.315	1.260	9.380	3.100	8.930	9.470
Black Population Percentage
<20	12.600	200.000	4.850	2.350	3.520	12.600
20-40	4.080	34.100	5.360	2.260	6.620	15.300
40-60	7.660	45.500	6.000	2.950	3.700	6.660
60-80	2.070	9.250	6.580	3.680	6.250	18.500
80-100	0.941	3.810	6.000	2.470	2.960	6.370
ADI
1	40.000	360.000	5.480	3.320	0.000	0.000
2	0.412	0.599	8.450	3.600	10.300	10.600
3	12.000	64.700	6.090	3.080	7.720	17.800
4	10.700	62.200	6.190	3.240	4.610	9.510
5	0.929	4.510	5.680	1.860	8.260	17.300
6	3.550	14.300	5.720	2.730	3.140	5.910
7	0.126	0.667	5.200	2.110	5.960	11.500
8	0.128	0.627	4.390	1.390	5.130	17.800
9	2.200	13.700	4.560	1.910	5.520	18.000
10	0.278	1.860	4.390	1.400	1.490	6.460

RA, rural area; LD, low density; UA, urban area; mRFEI, Modified Retail Food Environment Index; ADI, area deprivation index.

Fig. 10.

Average values of built environment factors: (a-c) by urbanicity levels, (d-f) by Black population percentage, and (g-i) by ADI values

All three built environment factors exhibit a clear trend across urbanicity levels, including declining park accessibility, greater walkability and healthier food environment from rural, to low-density and then urban areas (Fig. 10a-c). The relationship between racial composition, ADI and built environment factors are more complex. Although park accessibility tends to decrease and walkability tends to increase from areas with lower to higher Black population percentage, fluctuations are observed across the distribution (Fig. 10d&e). Walkability decreases in areas where the Black population percentage reaches 80% to 100%. The food environment shows the most volatile pattern, with peak values in areas with 20–40% Black population and notably poor conditions in predominantly Black neighborhoods (Fig. 10f). Similarly, as the ADI value increases, indicating higher levels of socioeconomic disadvantage, there is an overall decrease trend in park accessibility, walkability, and food environment quality, although within this trend, fluctuations and non-linear changes are also observed (Fig. 10g-i). While the mean value illustrates general trends, determining whether the average value of a factor differs significantly between two categories or across multiple categories needs to be evaluated for statistical rigor.

Three-way ANOVA results

To further investigate spatial disparities in obesogenic environment indicators, we conducted three-way ANOVA to examine the main and interaction effects of urbanicity, Black population percentage, and ADI. As summarized in Tables 7, 8 and 9, the three-way ANOVA results incorporate the main effects of each factor (equivalent to one-way ANOVA results), the two-way interaction effects between pairs of factors, and the three-way interaction effect among all three factors. Therefore, separate two-way ANOVA results are not reported, as they are already included in the three-way ANOVA framework.

Table 7.

ANOVA Results for Park Accessibility by Racial Composition, ADI, and Urbanicity

Source of Variation	df	Sum of Squares	Mean Square	F Value	Pr(> F)
Urbanicity	2	16,941	8,470	0.353	0.702
BlackPop%	4	10,184	2,546	0.106	0.980
ADI	9	102,710	11,412	0.476	0.891
BlackPop% × ADI	32	106,266	3,321	0.139	1
BlackPop% × Urbanicity	8	8,646	1,081	0.045	1
ADI × Urbanicity	17	27,294	1,606	0.067	1
BlackPop% × ADI × Urbanicity	25	31,677	1,267	0.053	1
Residuals	442	10,591,353	23,962

*Significant at 0.05, ** significant at 0.01, *** significant at 0.001. BlackPop%, Black population percentage; ADI, area deprivation index

For park accessibility, the ANOVA result shows no significant main or interaction effects, implying limited disparities in park accessibility across ZIP Code areas in Louisiana (Table 7).

For mRFEI, significant main effects are observed for urbanicity and ADI, indicating notable differences across urbanicity levels or ADI levels but not Black population percentage, and no complex interactions between or among them (Table 8). Among the individual variables, urbanicity (F = 4.777, p = 0.009) and ADI (F = 2.649, p = 0.005) show statistically significant effects on mRFEI, indicating that food environment scores differ significantly by urban-rural classification and ADI level. In contrast, Black population percentage alone did not have a significant effect on mRFEI. Furthermore, none of the two-way or three-way interaction terms are statistically significant, suggesting that the joint influence of these factors does not produce additional variance in mRFEI beyond their main effects.

Table 8.

ANOVA Results for mRFEI by Racial Composition, ADI, and Urbanicity

Source of Variation	df	Sum of Squares	Mean Square	F Value	Pr(> F)
Urbanicity	2	1653.000	826.500	4.777	0.009**
BlackPop%	4	1069.000	267.400	1.545	0.188
ADI	9	4125.000	458.300	2.649	0.005**
BlackPop% × ADI	32	3681.000	115.000	0.665	0.921
BlackPop% × Urbanicity	8	238.000	29.800	0.172	0.994
ADI × Urbanicity	17	2007.000	118.100	0.682	0.821
BlackPop% × ADI × Urbanicity	25	750.000	30.000	0.173	1.000
Residuals	442	76477.000	173.000

*Significant at 0.05, ** significant at 0.01, *** significant at 0.001. BlackPop%, Black population percentage; ADI, area deprivation index

For walkability, the three-way ANOVA results are summarized in Table 9. All three individual predictors, including urbanicity (F = 173.375, p < 0.001), Black population percentage (F = 13.278, p < 0.001), and ADI (F = 16.892, p < 0.001), have statistically significant effects on walkability, suggesting that walkability levels differ significantly across these social and spatial dimensions. In addition to the individual main effects, the interaction terms are also statistically significant. These complex interactions demonstrate that the relationship between any two factors varies substantially depending on the third factor, indicating that walkability disparities cannot be understood by examining single characteristics in isolation.

Table 9.

Three-way ANOVA Results for Walkability by Racial Composition, ADI, and Urbanicity

Source of Variation	df	Sum of Squares	Mean Square	F value	Pr(> F)
Urbanicity	2	1082.100	541.000	173.375	< 0.001 ***
BlackPop%	4	165.700	41.400	13.278	< 0.001 ***
ADI	9	474.400	52.700	16.892	< 0.001 ***
BlackPop% × Urbanicity	8	50.300	6.300	2.014	0.043*
ADI × Urbanicity	17	110.700	6.500	2.086	0.007**
BlackPop% × ADI	32	147.400	4.600	1.476	0.048*
BlackPop% × ADI × Urbanicity	25	203.800	8.200	2.613	< 0.001 ***
Residuals	442	1379.300	3.100

*Significant at 0.05, ** significant at 0.01, *** significant at 0.001

BlackPop%, Black population percentage; ADI, area deprivation index

The two-way interactions are shown as Fig. 11. In the interaction between urbanicity type and Black population percentage, walkability consistently follows the pattern UA > LD > RA across all levels of Black population percentage. However, the relationship between Black population percentage and walkability varies significantly by urbanicity type. In rural areas (RA), walkability shows an inverted U-shaped pattern, peaking at moderate Black population levels (40–60%) before declining. Urban areas (UA) exhibit a similar inverted U-shaped pattern but with much higher overall walkability scores. In contrast, low-density areas (LD) demonstrate a declining trend as Black population percentage increases, showing the most pronounced disparities. Similarly, in the interaction between ADI and urbanicity, the overall pattern remains consistent, with UA > LD > RA across most ADI levels. This suggests that the general urban advantage in walkability is robust regardless of area deprivation levels. However, within each urbanicity category, walkability does not change linearly with ADI. Particularly in LD and UA, non-linear peaks are observed at ADI levels 2–4, 6, and 8–9, indicating more complex spatial dynamics. The interaction between ADI and Black population percentage reveals the most complex patterns. In less disadvantaged areas (lower ADI), areas with 40–60% Black population percentage show the highest walkability, while areas with the highest Black percentage (80–100%) demonstrate the poorest walkability. However, this pattern reverses in highly disadvantaged areas (higher ADI), where predominantly Black areas (80–100%) show the best walkability. This finding underscores that walkability disparities cannot be understood by examining either disadvantage level or racial composition alone, highlighting the importance of contextual and spatial heterogeneity.

Fig. 11.

Two-Way Interactions in Walkability Index Disparities

The three-way interaction effects are illustrated in Fig. 12, displaying complex patterns that vary across urban contexts. In rural areas (RA) (Fig. 12a, left), walkability scores are generally low and relatively stable across all combinations of ADI and black population percentage, ranging between 4 and 8. Differences between Black population categories are minimal, and although walkability slightly decreases with increasing disadvantage, the changes are modest and not pronounced. But in low-density areas (LD) (Fig. 12a, middle), more variation begins to emerge. Particularly, neighborhoods with 40–60% and 60–80% Black populations show an upward trend in walkability as ADI increases, especially in the most disadvantaged areas. In urban areas (UA) (Fig. 12a, right), the interaction becomes most dramatic. With low Black population (0–20%), walkability declines steadily with increasing ADI. In contrast, areas with higher Black percentage (60–100%) show the opposite trend that walkability scores decrease sharply with higher ADI level. These patterns reflect a transitional zone between rural stability and urban complexity, suggesting moderate interactions shaped by both Black percentage and deprivation.

Fig. 12.

Three-way interaction effects on walkability

Figure 12b provides a summary visualization of these disparities by comparing walkability gaps between the most and least disadvantaged areas (ADI 8–10 vs. ADI 1–3) across racial and urban contexts. Rural areas maintain minimal disparities regardless of Black population levels. In contrast, urban areas exhibit larger disparities. Among neighborhoods with 40–80% Black population percentage areas, walkability is higher in less disadvantaged areas. However, this pattern is reversed in neighborhoods with 80–100% Black population, where walkability is notably higher in the most disadvantaged areas than in the least disadvantaged ones. This inversion underscores a complex interaction where high disadvantage does not always equate to poor built environment quality, particularly in predominantly Black urban neighborhoods.

Discussion

Key findings

Unlike prior studies that typically focus on a single dimension, this study integrates multiple components related to children’s health: park accessibility, walkability, and food environment.

Overall, children across Louisiana are exposed to relatively poor obesogenic conditions, with pronounced disparities between urban and rural areas. Urban centers such as Baton Rouge and New Orleans tend to offer more supportive environments with higher walkability scores and better access to healthy food outlets, while rural communities consistently suffer worse situations. However, areas with high walkability and favorable food environments often exhibit poor park accessibility. Very few ZIP code areas exhibit consistently supportive conditions across both physical activity (PA) and food environments.

Previous studies have examined disparities in neighborhood environments by urbanicity, race, socioeconomic position (SEP), and other attributes, but few have considered these factors simultaneously. Most studies overlooked the interaction effects among these factors. As an exception, Kaczynski et al. [45] observed, using a 2-way ANOVA, that significant interactions existed among urbanicity, income, and race for environmental factors. This study employed a 3-way ANOVA to explore more complex disparities by revealing significant differences in obesogenic environments across urbanicity, percentage of Black population, and ADI, as well as their complex interaction effects. While park accessibility does not show significant associations with these variables, the food environment is primarily shaped by the main effects of urbanicity and ADI. In contrast, walkability is influenced not only by individual factors but also by their multifaceted interactions. The findings underscore that disparities in the built environment cannot be fully understood without accounting for the joint effects of urbanicity, racial composition, and socioeconomic disadvantages, which has been rarely explored in previous research. These interactions exhibit divergent spatial patterns, particularly between low-density and urban areas, with some socioeconomically disadvantaged Black neighborhoods in urban contexts demonstrating relatively higher levels of walkability.

Limitations and future research

This study also has several limitations and challenges that suggest directions for future research. First, due to data constraints, the environmental measures are limited to areas within the state of Louisiana. However, residents living near state borders may also access or be impacted by the built environments in neighboring states. Future studies should consider cross-border spatial factors to more accurately capture environmental exposures. Regarding the measurement of park accessibility, this study conducted a sensitivity analysis and found general consistency across different driving time thresholds, but noticeable differences still existed. Future research could combine surveys or interviews to better understand driving time preferences and travel behaviors specific to children and families in Louisiana and adopt a catchment area size that truly reflects their behaviors. Furthermore, this study uses park area size as a key component in measuring accessibility. While parks in rural areas tend to be larger than those in urban settings, this may partly explain the decline in park accessibility from rural to urban areas. However, park quality may not be captured solely by area size. Some small parks may offer diverse facilities and programming, while large parks may lack child-friendly amenities. Future research should consider additional variables to assess the actual suitability of parks for promoting children’s PA.

Secondly, the three-way ANOVA used in this study assumes independence among ZIP-code observations. Spatial diagnostics indicated that residuals for walkability exhibited significant positive spatial autocorrelation (Moran’s I = 0.150, p < 0.001), whereas residuals for the food environment (Moran’s I = 0.006, p = 0.396) and park accessibility (Moran’s I ≈ 0.0005, p = 0.441) did not show spatial dependence. This suggests that the ANOVA results for walkability should be interpreted with caution, and future research needs to apply spatial statistical modeling to appropriately account for spatial dependence.

Finally, it is important to note that accessibility reflects potential access rather than actual use. Given the generally poor walkability across Louisiana, children’s park use may still depend heavily on parental transportation. Therefore, in areas with high park accessibility but low walkability, children’s actual utilization of park resources may remain limited. Addressing these nuances is essential for developing more accurate assessments of obesogenic environments.

Conclusion

Building upon previous research, this study provides a comprehensive and objective measurement of the children obesogenic environment across Louisiana. Children’s obesogenic environments cannot be adequately characterized using a single indicator. Therefore, this study collectively measured and characterized the distribution of multiple indicators including park accessibility, walkability, and the food environment, to better capture children’s opportunities for physical activity and healthy food.

In addition, our findings further reveal substantial spatial and demographic disparities in children’s obesogenic environments across Louisiana. Overall, the obesogenic built environment in Louisiana remains suboptimal and unevenly distributed. These disparities manifest not only in the spatial patterns of individual factors but also in their lack of consistency that few areas simultaneously support both physical activity and healthy dietary access for children. Moreover, this study explores the disparities of children’s obesogenic, especially for walkability, which is shaped by the complex interaction among urbanicity, race, and socioeconomic disadvantage, rather than by any single determinant.

This study extends the existing research on disparity research on children’s obesogenic environments and provides evidence and guide for future policy makers. It should not be simply assumed that rural or disadvantaged areas are healthier or unhealthier. Instead, policymakers and planners should design interventions that account for the complex and context-specific interactions among urbanicity, race, and socioeconomic conditions.

Future research should combine additional factors that more accurately capture children’s real-world accessibility to PA and food environment to provide a more comprehensive understanding of how obesogenic environment shape children’s outcomes. For example, future studies could examine how children and their families utilize these environmental resources, and how individual or household behaviors may facilitate or constrain children’s exposure to obesogenic conditions. With advanced approaches like multilevel model and child-level health outcomes, future research could further interpret the links between obesogenic environment and children’s obesity.

Author contributions

YZ collected the data, built the models, and drafted the manuscript; FW conceptualized and designed the study, supervised the research, and revised the manuscript; SC provided project supervision and guidance, and revised the manuscript; SC and FW secured funding resources for the project. All authors read and approved the final manuscript.

Funding

This study was supported by the Public University Partnership Program at the Louisiana Department of Health, Bureau of Health Services Financing; the Louisiana State University Provost’s Fund; and the Louisiana State University Foundation.

Data availability

This study used publicly available datasets from the U.S. Census Bureau (Decennial Census, American Community Survey, and ZIP Business Patterns ( https://www.census.gov), the Area Deprivation Index ( https://www.neighborhoodatlas.medicine.wisc.edu/ , accessed 07/23/2023), and the U.S. Environmental Protection Agency’s National Walkability Index ( https://www.epa.gov/smartgrowth/smart-location-mapping#walkability , accessed 10/10/2023), USA parks data from Esri ( https://www.arcgis.com/home/item.html?id=578968f975774d3fab79fe56c8c90941 , accessed 01/28/2025).

Declarations

Ethics approval and consent to participate

Not applicated.

Consent for publication

Not applicated.

Disclaimer

The content presented in this paper is the sole responsibility of the authors and does not necessarily represent the official views of the funding sponsors.

Competing interests

The authors declare no competing interests.