Now Reading: Environmental phenotypes for healthy weight in children using population-based linked environment and health data: a cross-sectional observational study

1
  • 01

    Environmental phenotypes for healthy weight in children using population-based linked environment and health data: a cross-sectional observational study

Light Dark

Environmental phenotypes for healthy weight in children using population-based linked environment and health data: a cross-sectional observational study

Environment Connectz2 hours ago6 Views

Abstract

Background

Childhood obesity is a major global health challenge, projected to affect one in three children worldwide by 2050. While individual and social factors contribute, increasing evidence highlights the built environment as a key determinant in shaping children’s behaviours and weight outcomes. Evidence suggests that neighbourhood design, greenspace access, and food retail availability influence diet and physical activity, but most studies rely on small samples or single-domain measures.

Methods

We linked nationwide geographic information systems (GIS) data describing residential neighbourhoods with objectively measured child weight from a national Welsh surveillance programme for children aged 4–5 years. Using multiple indicators—including housing type, garden size, neighbourhood greenness, walkability, access to recreational spaces, and food outlet density – latent class analysis was used to identify distinct “environmental phenotypes.” Associations between phenotypes and child weight status were examined using logistic regression.

Results

We identified discrete classes of residential environments characterised by varying combinations of built and food environment features. The model with five classes was the best fit overall, with percentage and number of households in each phenotypes: Rural, spacious and isolated 14% (24,266), Suburban 17% (29,324), Deprived and underserved 23% (39,227), Deprived and well-served 32% (53,210), and Dense, coastal and well-connected 13% (21,762). Children living in rural, spacious and isolated neighbourhoods, characterised as those with greater greenspace, private gardens, and walkable layouts, had significantly lower odds of overweight and obesity (OR = 0.89, CI = 0.86-0.93), whereas those in deprived and well-served neighbourhoods, characterised by high-density housing areas with limited greenspace and high fast-food outlet density, had elevated risk (OR = 1.09, CI = 1.06-1.13). These associations remained robust after adjustment for area-level deprivation and rurality.

Conclusion

Our findings highlight the importance of the residential environment in early childhood obesity risk. Nationally linked GIS and health data enable robust classification of obesogenic environments, informing urban planning and public health strategies to promote healthier, child-friendly neighbourhoods.

Highlights

  • •

    Nationwide linkage of GIS and health surveillance data to examine residential environments and childhood obesity at a population scale.

  • •

    Comprehensive environmental profiling using multiple indicators.

  • •

    Novel classification of “environmental phenotypes” capturing complex combinations of built and food environment features.

  • •

    Children living in rural, spacious and isolated neighbourhoods had significantly lower odds of overweight and obesity.

  • •

    Actionable insights for urban planning and public health to design child-friendly, obesity-preventive neighbourhoods.

1. Introduction

Childhood obesity is one of the most pressing global health challenges of the 21st century. The global prevalence of overweight and obesity in children and adolescents aged 5-19 rose from 8% in 1990 to 20% in 2022 (Obesity and overweight n). These trends are not only accelerating, but are projected to intensify – by 2050, nearly one in three children worldwide may be living with excess weight (GBD, 2021 Adolescent BMI Collaborators, 2025). The implications of this public health crisis are profound, increasing the risk of early-onset chronic diseases and placing a growing burden on already strained healthcare systems (Ling et al., 2023). While childhood weight status is influenced by a range of factors, growing attention has been directed towards the role of the built environment in shaping children’s health behaviours and outcomes.

Early childhood represents a critical period for growth and development, during which patterns of diet, physical activity and weight gain are established. Young children are also more dependent on their immediate residential environment than older children as their access to food, play spaces and transport are determined by their parents or caregivers. As such, a greater understanding of the neighbourhood-level environmental characteristics that influence childhood weight status is essential to inform effective prevention strategies during this critical developmental phase.

Evidence suggests that the built environment – encompassing housing type, neighbourhood design, land use, access to amenities, and foot outlet density – plays a crucial role in influencing children’s diet and physical activity and can either promote or inhibit healthy behaviours (Malacarne et al., 2022; Miller et al., 2014; Schalkwijk et al., 2018a; Maitland et al., 2013; Mizen et al., 2018; Mayor, 2014). Well-connected neighbourhoods with pedestrian-friendly routes can encourage active transport and outdoor play. In contrast, unsafe or poorly designed urban areas with limited access to designated greenspace can discourage physical activity, leading children to spend more time indoors engaging in sedentary behaviours (Daniels et al., 2021a; Datar et al., 2013). Several studies have examined these associations in young children (Nordbø et al., 2019; Reimers and Knapp, 2017; Jia et al., 2020; Terrón-Pérez et al., 2021) although evidence specific to the 4-5 year age group remains limited compared with studies of older children. Spence and colleagues (Spence et al., 2008) analysed 501 preschool-aged children in Edmonton, Canada and explored how measures of neighbourhood design related to overweight status, finding modest associations with built environment features in this age group. A systematic review of longitudinal studies across broader age ranges in the United States and in European settings suggest that built and social neighbourhood characteristics such as access to green space, parks, and recreational facilities may have beneficial effects on weight trajectories in children, whereas features of the food environment show mixed associations (Daniels et al., 2021b). Overall, while neighbourhood built environments appear to influence behaviours relevant to weight status in young children, the magnitudes of these associations are generally modest and vary across contexts (Miller et al., 2014; Mizen et al., 2018; Mayor, 2014; Daniels et al., 2021a; Datar et al., 2013)

Other environmental factors such as housing type, garden size, area-level deprivation, and rurality also play an important role. Schalkwijk and colleagues (Schalkwijk et al., 2018b) reported that access to private gardens and greater neighbourhood greenness were positively associated with healthier weight outcomes in children aged 0-7 years living in England, highlighting the importance of unstructured outdoor play environments in early childhood (Maitland et al., 2013). In contrast, high-density housing and neighbourhood deprivation have been associated with higher stress levels and limited access to health-promoting resources, such as safe play spaces and healthy food options, which may contribute to increased risk of overweight and obesity (Singh et al., 2010). Rural environments, while often offering more open space, may lack walkable infrastructure or access to healthy food outlets, presenting a different set of challenges (Guseman et al., 2022; Kegler et al., 2009). Together, these findings suggest that multiple aspects of the residential environment interact to shape opportunities for physical activity, play, and dietary behaviours in young children.

Despite the evidence, major gaps remain. Few studies have examined how multifaceted neighbourhood environments collectively influence early childhood weight status, and most rely on small sample sizes or single-domain measures. Reviews have highlighted the need for more integrated approaches that capture the complexity of built environments and their combined influence on health (Van Der Horst et al., 2007; Sallis et al., 2012; Giles-Corti et al., 2016). The concept of an environmental phenotype – the combination of physical, social, and infrastructural characteristics of a neighbourhood that may shape health – offers an opportunity to fill this evidence gap.

Using high-resolution geographic information systems (GIS) data linked to child weight data from a national measurement programme, this study classifies residential phenotypes and examines their associations with weight in children aged 4-5 years living in Wales, within the United Kingdom. By using latent class analysis to identify environmental profiles, we move beyond single-exposure approaches to capture the broader configuration of neighbourhood characteristics experienced by children. In doing so, we provide a more holistic understanding of how combinations of environmental features relate to early childhood weight outcomes.

To our knowledge, this is the first population-level study to link multifaceted environmental data with early childhood weight data in this way, providing novel insights for research, policy, and urban planning. We hypothesise that children who live in health-promoting neighbourhood phenotypes are more likely to have a healthy weight status at age four and five, compared to children living in obesogenic phenotypes.

2. Methods

2.1. Data and measures

This observational study used routinely collected administrative data held in the Secure Anonymised Information Linkage (SAIL) Databank, hosted at Swansea University (Lyons et al., 2009; Jones et al., 2014, 2019). The SAIL Databank is an internationally recognised trusted research environment that provides access to rich, longitudinal health, socioeconomic and environmental datasets. Personally identifiable information is removed and replaced with unique Anonymised Linkage Fields (ALFs), enabling secure individual-level record linkage across multiple data sources. A corresponding Residential ALF (RALF) is also used to link individuals to their place of residence, allowing for linkage of both licensed and opensource environmental data to examine the characteristics of an individual’s local environment. Full details of anonymisation and linkage methodology are published elsewhere (Lyons et al., 2009; Rodgers et al., 2009, 2012). Within this framework, multiple datasets were linked to characterise the residential locations of children and surrounding environmental features. These included the Child Measurement Programme (CMP) for measured child height and weight outcomes, the Welsh Demographic Service Dataset (WDSD) for residential location and demographic information, and a range of geospatial datasets to derive environmental characteristics surrounding each child’s home.

The CMP for Wales provides robust public health surveillance data on the weight status of over 90% of children in state-maintained schools, using standardised height and weight measurements to estimate levels of overweight and obesity (The Child Measurement Programme for Wales Publication details Title, 2017). This study included children aged 4-5 with a measurement recorded in the CMP between 2012/2013 and 2018/19. Place of residence was captured for each child at the time of measurement. Children with missing identifiers or residential information were excluded.

2.2. Environmental phenotypes

Environmental variables were selected based on their potential relevance to young children’s daily activity spaces and behaviours, including opportunities for outdoor play, access to recreational spaces, and exposures to food environments. These variables were also selected based on their potential to influence healthy weight outcomes, either as an environmental insult (e.g. high availability of hot food takeaways) or an environmental benefit (e.g. good access to green space). These data were combined to characterise the overall pattern of environmental features surrounding each home in Wales, which we refer to as an “environmental phenotype”. By this, we mean the combination of environmental characteristics within an area, rather than any single attribute considered in isolation.

The environmental data were derived at area level or neighbourhood level dependent on data availability. Neighbourhood was defined as the area within 900m of each child’s residence, which was the closest to a 10-min walk we could achieve with the available data (Yang et al., 2021). A child-centred buffer was used to capture the local environment surrounding each child’s residence and reflects the area that families are likely to access in their daily activities. As a result, neighbourhood areas may overlap for children living in close proximity.

2.3. Area level characteristics

Area-level data were used for variables that were only available at administrative geographic scales. The small-area geographic unit (Census geographies) for each residence, which corresponds to the Lower Layer Super Output Area (LSOA) in Wales, was extracted from the WDSD. LSOAs are standard administrative units with populations of approximately 1500 residents, providing consistent spatial scale for area-level variables. These were linked to the 2019 Welsh Index of Multiple Deprivation (WIMD) to obtain measures of socioeconomic status (deprivation). The WIMD is a composite index that combines indicators across multiple domains, including income, employment, health, education, housing, access to services, and the physical environment. Overall WIMD scores were assigned as quintiles, where 1 = most deprived and 5 = least deprived.

The ONS urban rural classification (Office for National Statistics, 2013) was also linked to residences, to investigate whether there were any differences in urban morphologies, and was categorised as: urban; urban in a sparse setting; rural town; rural town in a sparse setting; rural village; and rural village in a sparse setting. Urban areas are typically defined as settlements with populations of 10,000 or more, while rural areas include smaller towns, villages, and dispersed settlements. “Sparse settings” refer to areas with relatively low surrounding population density and greater geographic isolation compared with similar settlement types in less sparse regions.

2.4. Home and neighbourhood characteristics

Characteristics of the built environment were derived from one time point for the whole study period as previous research indicates that the built environment does not change considerably over time (Robinson et al., 2024; Hirsch et al., 2016a). Food outlets and greenspace data were derived from 2014 to 2017 respectively to align as closely as possible with the study mid-point and linked with children’s residence at the time of CMP measurement.

Information on all points of interest (i.e., food outlets and greenspaces) within the neighbourhood were captured. Distance was categorised as hyper-local (0-300m), local (301-600m) or neighbourhood (601-900m).

2.4.1. Greenspaces

Greenspace data were sourced from the 2017 Ordnance Survey (OS) Open Greenspaces dataset (OS open greenspace data), which depicts the location of publicly accessible areas such as parks and sports facilities. Access to green space is particularly important for young children as they may support outdoor play and physical activity. We defined greenspace as play space, playing field, public park or garden, tennis court, and other sports facility. Play spaces located within schools or paid-for tourist attractions were not included. Distance was calculated in metres from each residence to greenspace access point (vehicle or pedestrian) within 900m, with each greenspace counted only once regardless of multiple access points. The total number of greenspaces within 900m of each residence was used to capture overall exposure to outdoor environments. In addition, counts of specific greenspace types were included as separate variables to reflect potential differences in their use and relevance for young children.

Size of the nearest greenspace type was recorded in m2. Where there were multiple greenspaces of the same type located at equal distance, the largest greenspace was allocated. Greenspace size was categorised into ‘small’, ‘medium’, ‘large’, and ‘none’ based on the distribution of values for each individual greenspace type. Thresholds were derived from the observed range of values for each variable to create approximately even groupings. This ensured that the categorisation reflected the relative magnitude for each specific type of greenspace. These groupings were chosen to facilitate interpretation and support model convergence in the latent class analysis.

2.4.2. Food outlets

The density of food outlets may influence dietary environments and exposure to energy-dense foods. Food outlet locations were derived from the 2014 OS Points of interest dataset and were included as a total number of food outlets within 900m of each residence and separate counts of specific outlet types. These included bakeries, cafes, confectioners (sweet/candy shop), convenience stores, hot food takeaways (including fish and chip shops), restaurants, and supermarkets. The total outlet measure was used to capture overall exposure to the local food environment, while individual outlet types were included to reflect potentially differing influences on dietary behaviours.

2.4.3. House type

House type was included as it reflects housing density and the availability of private outdoor space, which may influence opportunities for outdoor play and physical activity among young children. House type was derived from Address Base Premium (AddressBase premium data products OS), which provides the location and attributes of residential and business addresses in the UK. The categories used were detached, semi-detached, self-contained flat, terraced, ‘other’, and ‘missing’ for all residences where house type was unavailable (11.1% of dwellings).

2.4.4. Garden size

Access to outdoor space may provide safe spaces for outdoor play within the home environment. Garden polygons were extracted from OS MasterMap (OS MasterMap topography layer), which offers detailed vector mapping of land parcels, building footprints and other topographic features. Each garden polygon was assigned to the corresponding building using its Topographic Identifier (TOID), supplemented with land registry data to ensure accurate linkage between gardens and residential properties. Garden size was calculated in m2 and categorised into eight groups: 0-99.9 m2, 100-199.9 m2, 200-299.9 m2, 300-399.9 m2, 400-499.9 m2, 500 m2 +, no garden (for households with a garden size of zero) and ‘missing’ where no garden size data were available. Garden size data were unavailable for 9434 (5.6%) residential dwellings due to data linkage issues.

2.4.5. Greenness

We operationalised Landsat 8 satellite imagery (30-m resolution) to calculate mean Enhanced Vegetation Index (EVI) value for all homes in Wales, which were then linked to children’s addresses to capture a measure of residential greenness within 300m. EVI was derived from red, blue, and near-infrared (NIR) reflectance bands using the GRASS vegetation index tool in QGIS (QGIS manual. i). EVI work reported in this study was conducted as part of a wider longitudinal study (Garrett et al., 2023; Thompson et al., 2022; Mizen et al., 2019, 2024a) where a national level annual exposure variable was developed for 1.49 million households between 2008 and 2019. Full details of the EVI estimation methodology are described elsewhere (Thompson et al., 2022; Mizen et al., 2024a). EVI ranges from −1 to 1, where negative values generally correspond to water bodies, values near 0 indicate bare soil or built environments, and positive values indicate vegetation. Healthy vegetation is typically found in the 0·2 and 0·8 range (Huete et al., 2002; Mizen et al., 2024b).

2.4.6. Walkability

Walkability was measured using the Welsh Active Living Environments Index, produced by Mah and colleagues (Mah et al., 2022) using data from the Wal-ALE database. Walkability is a metric of how conducive an area is to walking, based on various measures such as connectivity (e.g. footpaths, intersection density, points of interest), land use, safety, and street quality. Walkable areas are likely to promote active travel and access to local amenities.

The walkability index was calculated at the small-area geographic level and linked to addresses. Walkability was divided into quintiles to create an ordered scale from 1 (most walkable) to 5 (least walkable).

2.4.7. Average distance to coast

The distance from the centroid of each small-area geographic unit to the nearest coastal location was applied to all addresses within that area to give an average distance to coast measure. Coastal environments may provide opportunities for outdoor recreation and physical activity, which promotes healthy behaviours in children.

2.4.8. Nearest primary school

Proximity to primary schools may reflect neighbourhoods designed around family infrastructure and influence daily walking or travel patterns. Primary schools were identified from the Welsh Governments published school locations (Welsh Government). The network distance, using roads and footpaths, to the nearest primary school was calculated for each residence in metres and grouped into five categories: hyper-local, local, neighbourhood, >900m, and missing (0.2%).

2.5. Outcome measure – weight status

Children’s Body Mass Index (BMI) was calculated from CMP height and weight data and categorised using UK1990 clinical reference standards into four groups: “underweight” (BMI <2nd centile), “healthy weight” (≥2nd to <91st), “overweight” (≥91st to <98th), and “obese” (≥98th), based on sex- and age-standardised z-scores. Z-scores beyond five standard deviations were excluded. For analysis, categories were simplified into a binary measure: 0 for healthy weight and 1 for unhealthy weight (overweight or obese), aligning with public health standards. This approach facilitates clearer interpretation of results for policy and practice audiences. Children categorised as underweight were excluded due to low numbers.

2.6. Statistical analysis

Latent Class Analysis (LCA) was used to identify the unknown (latent) groups of childhood neighbourhoods. Analysis was performed in STATA MP v18 to group households into distinct environmental phenotypes based on similarities across the 35 area-level, home and neighbourhood variables. For the points of interest variables, all counts that occurred in less than 1% of the sample were grouped together into an upper category to ensure sufficient cell sizes and to support model convergence during analysis.

The LCA method involves a number of steps: 1) specification of different models across a range of a number of groups and compare model fit. LCA models with three classes through to nine were examined with model selection based on a combination of statistical fit indices (log-likelihood, AIC and BIC), entropy values, and consideration of class interpretability and plausibility. Models with lower information criterion indicate better fit and were plotted on a graph to identify points of diminishing return. In line with best practice, class size was also considered. A minimum threshold of 10% was applied to avoid identifying small, potentially unstable classes and to ensure the resulting phenotypes were meaningful at a population level. 2) select the model based on the optimal solution. Emerging classes from the selected model were evaluated within the study group to determine whether their characteristics were both qualitatively and quantitively distinct and meaningful in the context of children’s living environments, with phenotype labels assigned accordingly. Environmental phenotype labels were derived for each class based on a qualitative review of the characteristics and reflect the dominant or distinguishing features of each profile. 3) Assign each household to a distinct group. Class assignment for each residence was based on the latent class with the highest posterior probability score generated by the LCA model.

To visually show the distribution of classes across Wales, the percentage of households belonging to each class was calculated for every small-area unit. Each unit was then assigned to the class with the highest proportion of residences. Boundary shapefiles for Wales (2011) were obtained from the Office for National statistics Open Geography Portal. A thematic map was generated using the tmap package in R (tmap, 2018).

2.6.1. Logistic regression

Associations between environmental phenotype and weight status were analysed using logistic regression. Each child was linked to an environmental phenotype based on their place of residence at the time of their BMI measurement. We adjusted for age and sex to enable evaluation of differences between these groups. We assumed statistical independence between all children, including those registered to the same address. R2 value is reported as an indication of overall variation explained by neighbourhood phenotype membership.

3. Results

The environmental phenotypes were generated based on the environmental characteristics of 167,789 unique houses in which children resided. Full details of the physical environmental characteristics can be found in appendix 1.

3.1. Latent Class Analysis

The AIC and BIC model statistics for three through eight classes improved with increasing number of classes, however, the rate of improvement was minimal beyond five classes. This suggests that adding classes beyond this point would yield minimal additional explanatory value and may overfit the data. Additionally, models with six or more classes included at least one class with less than 10% of the sample. The model with nine classes failed to converge and deemed to be a poor fit. The model with five classes was therefore selected as the best fitting model. The five environmental phenotypes were subsequently labelled: Rural, spacious and isolated, Suburban, Deprived and underserved, Deprived and well-served, and Dense, coastal and well-connected.

The Rural, spacious and isolated phenotype represented 14% (24,266) of residences and contained predominantly detached homes, the largest gardens and the highest proportion of neighbourhood vegetation. This profile had very limited access to designated greenspaces, food outlets and primary schools. The Suburban phenotype represented 17% (29,324) of residences and was characterised most notably by the presence of semi-detached housing with low levels of deprivation. It also had good access to the coast and to greenspaces, and very low counts of food outlets. The Deprived and underserved phenotype represented 23% (39,227) of residences and was predominantly urban with terraced housing in deprived areas. Households in this group also had limited access to greenspace overall but had higher counts of children’s play spaces in the hyper-local area. The Deprived and well-served phenotype represented 32% (53,210) of residences and was similar to profile three except having better access to primary schools, and higher counts of food outlets and greenspaces. The final phenotype, Dense, coastal and well-connected represented 13% (21,762) of residences and had predominantly terraced housing in deprived, urban areas with the smallest gardens. This group had higher counts of hot food takeaways in their local neighbourhood compared with households in other phenotypes. Households were also closest to the coast, primary schools and had better access to the largest greenspaces. Summary characteristics of the five environmental phenotypes can be seen in Table 1 and a full breakdown in Appendix 1.

Table 1.

Summary characteristics of environmental phenotypes.

PhenotypeSocioeconomic deprivationGreenspaceFood environmentUrbanicityHousing/gardens
Rural, spacious and isolatedVery lowVery low accessVery lowRural/sparseLarge gardens, detached housing
SuburbanLowModerateLowMixed suburbanMedium gardens, semi-detached
Deprived and underservedHighLimited (few large spaces, some play spaces)ModerateUrbanMedium gardens, more terraced housing
Deprived and well-servedHighGood access (larger parks, facilities)HighUrbanMedium gardens, dense housing
Dense, coastal and well-connectedHighVery high access; greater access to the largest greenspacesVery high (especially hot food takeaways)Highly urban/closest to the coastSmallest gardens, high-density terraced housing

Fig. 1 illustrates the geographic distribution of phenotypes across Wales to highlight the spatial variation in neighbourhood environments experienced by children. An inset map of the capital city, Cardiff, is also included to provide greater detail of the urbanised metropolitan area. Fig. 2 provides representative examples of the characteristics associated with each phenotype, illustrating how combinations of features such as housing type, greenspace availability, and local amenities differ across neighbourhood types.

Fig. 1.

Fig. 1

Phenotype distribution by small-area geography across Wales, United Kingdom.

Fig. 2.

Fig. 2

Representative example of how each environmental phenotype would look geographically.

3.2. Association between residential environment and weight status

The study included 214,947 children aged 4-5 years, of whom 18% were classified as having an unhealthy weight (Table 2). The sample was evenly distributed by gender (51% boys, 49% girls), and most children were aged 5 years (58%). Almost one third (32%) of children lived in the phenotype Deprived and well served. Overall, the distribution of weight status was broadly similar across demographic groups.

Table 2.

Children in the Child Measurement Programme (CMP) by weight category (n = 214,947).

N (%)Healthy weight (%)Unhealthy weight (%)
N214,947(100)176,301(82)38,646(18)



Gender
Boys109,688(51)89,548(82)20,140(18)
Girls105,259(49)86,753(82)18,506(18)



Age
4 years91,312(42)74,563(82)16,749(18)
5 years123,635(58)101,738(82)21,897(18)



Environmental Phenotype
Rural, spacious and isolated31,248(15)26,180(84)5068(16)
Suburban37,544(17)30,782(82)6762(18)
Deprived and underserved50,150(23)41,227(82)8923(18)
Deprived and well-served68,163(32)55,123(81)13,040(19)
Dense, coastal and well-connected27,842(13)22,989(83)4853(17)

After adjusting for age and sex, children who lived in Deprived and well-served environments had the highest odds of being an unhealthy weight and were 9% (OR = 1·09; 95% CI: 1·06 – 1·13) more likely to be living with an unhealthy weight compared to children living in Deprived and under-served environments. Children living in Rural, spacious and isolated environments, had the lowest odds of being an unhealthy weight. These children were 11% (OR = 0·89; 95% CI: 0·86 – 0·93) less likely to have an unhealthy weight compared to children living in class three.

As shown in Table 3, girls were less likely to be an unhealthy weight compared to boys (OR = 0.95; 95% CI: 0.93-0.97) and children aged five had a lower odds of being an unhealthy weight compared to children aged four (OR = 0.93; 95% CI: 0.90-0.96).

Table 3.

Logistic regression estimates for the association between environmental phenotypes and children’s weight category (healthy weight vs. unhealthy weight).

Odds ratio95% Confidence Interval
Profile
 Rural, spacious and isolated0.890.86 – 0.93
 Suburban1.020.98 – 1.05
 Deprived and under-served (reference)––
 Deprived and well-served1.091.06 – 1.13
 Dense, coastal and well-connected0.980.94 – 1.01
Age
 4 years (reference)––
 5 years0.930.90 – 0.96
Sex
 Boys (reference)––
 Girls0.950.93 – 0.97



R2= 0.0009

4. Discussion

This study is the first to use population-level health and multi-faceted environmental exposure data to examine associations between environmental phenotypes and early childhood weight status. Consistent with previous research (Schalkwijk et al., 2018b; Jia et al., 2021; Van Der Zwaard et al., 2018), findings suggest that environments associated with healthy weight at age four and five were characterised by rurality, low fast-food density and opportunities for both structured and unstructured play in private gardens. In contrast, children in more urban, deprived settings with high fast-food density benefitted from protective features such as accessible greenspace and coastal areas, highlighting the importance of publicly available play opportunities when private outdoor space is limited.

The Deprived and underserved and Deprived and well-served phenotypes showed similar levels of deprivation, however, the Deprived and well-served group had greater access to food outlets and greenspaces overall, whereas the Deprived and underserved group had fewer neighbourhood-level greenspaces but a higher density of hyper-local play spaces alongside fewer hot food takeaways. While total food outlet density captures overall exposure to the local food environment, it is important to recognise that this comprises a heterogeneous mix of outlet types, including both potentially health-promoting (e.g. supermarkets) and less healthy options (e.g. hot food takeaways). In this study, both total outlet density and counts of specific outlet types were included, allowing us to capture variation in overall exposure as well as differences in exposure to specific food outlet types. Consistent with previous research (Daniels et al., 2021a; Welsh Government; tmap, 2018; Jia et al., 2021; Van Der Zwaard et al., 2018), results suggest that even in highly deprived settings where there is often an accumulation of obesogenic features (Jenkin G et al., 2015), the availability of accessible, child-focused play spaces combined with reduced exposure to food outlets may provide important protective factors for maintaining healthy weight in early childhood. This supports the growing evidence that it is not single exposures, but combinations of built environment features that collectively shape health outcomes. Policy interventions which target areas based on deprivation in addition to other phenotype characteristics such as increased play spaces and fewer food outlets may be more effective.

By applying a latent class approach, we were able to define distinct environmental phenotypes rather than assessing individual environmental features in isolation. This multidimensional perspective offers a more holistic understanding of how contextual factors interact to influence childhood weight status. Although existing evidence indicates that associations between environmental features and childhood weight status are generally weak, prior studies have relied on narrow exposure measures (Poole and Moon, 2017) or area-level proxies (Beynon et al., 2021). While the observed differences in unhealthy weight prevalence across environmental phenotypes were relatively modest (16%–19%), this is broadly consistent with findings from other population-level studies examining neighbourhood influences on early childhood weight (Daniels et al., 2021a; Jia et al., 2021). These findings can also be understood within ecological frameworks of child development, such as Bronfenbrenner’s (Urie, 1979) ecological systems theory, which emphasises the interacting influence of multiple environments including family, neighbourhoods, and wider policy contexts. Within this framework, neighbourhood characteristics represent one layer of influence that may shape opportunities for physical activity, access to food, and outdoor play. The modest differences observed across phenotypes are consistent with the idea that childhood weight is shaped by multiple, interacting systems rather than any single environmental factor. Nevertheless, understanding environmental contributions remains important for identifying population-level interventions and planning health-promoting neighbourhoods. Using a phenotype-based approach allows us to capture the complexity of children’s lived environments and how these configurations promote or inhibit healthy behaviours.

4.1. Strengths and limitations

A key strength of this study is the linkage of a national dataset (CMP) with objectively measured environmental exposures via the SAIL Databank, enabling the first population-level classification of home environment phenotypes across Wales. This data-driven approach allowed us to integrate multiple exposures and identify meaningful patterns associated with weight status.

The child measurement programme is conducted in state-maintained schools and therefore does not include children attending private schools or those who are homeschooled. While coverage is high, this may result in a small proportion of the child population not being represented. We also acknowledge the limitations of using BMI as an indicator of overweight and obesity. Whilst it is a widely used and practical measure of weight status in children, it does not distinguish between body fat and lean mass and may not fully capture differences in body composition across subpopulations.

Other limitations include reliance on proximity-based measures, which may not reflect utilisation or the quality of spaces accessible to children. Further, young children’s behaviours are dictated by parental preferences which is also true for dietary behaviours of the children and household. This study is observational in nature, meaning that findings should be interpreted as associations rather than causal effects. We were unable to account for potential sources of endogeneity, including residential selection into neighbourhood phenotypes. Although we adjusted for key sociodemographic factors, residual confounding and unmeasured selection effects may remain. While approaches such as comparing children within smaller geographic areas may help to address unobserved heterogeneity, the area-based nature of the environmental phenotypes limits the feasibility of such analyses within the current framework, as this would substantially reduce variability in exposures. It should also be noted that deprivation was included as an indicator in the latent class analysis, meaning that some environmental characteristics may still be patterned by socioeconomic context.

Points of interest and greenspace data were derived from a single time point to align with the study mid-point and as a result, temporal changes to the built environment before and after this period were not captured. Additionally, home environment phenotypes were assigned based on place of residence at the time of CMP measurement and do not reflect longitudinal exposures. Despite these limitations, research indicates that the built environment does not change considerably over time (Hirsch et al., 2016b) and that families in Wales tend to remain in similar areas in terms of environmental characteristics (Davies et al., 2024). While differences in characteristics are likely to be minimal, we acknowledge that employing a longitudinal design would allow changes in exposures to be measured over time and better linked to child health trajectories.

We assumed independence between observations as a reliable family-level identifier was not available to account for clustering of children within households. As a result, children living in the same household either concurrently or at different time points, were assigned to the same latent class and may share behavioural similarities that influence weight. Future work could apply more advanced methodology to account for within-household clustering and shared behaviours.

Although the logistic regression model demonstrated statistically significant associations between built environment phenotypes and childhood weight status, the overall explanatory power was limited (as indicated by a low pseudo-R2 value). This is not unexpected, as childhood obesity is a multifactorial outcome influenced by a wide range of determinants; however, the results demonstrate that environmental phenotypes play an important role in shaping differences in population level childhood weight status.

4.2. Directions for future research

Future research should integrate behavioural and lifestyle data, such as information captured by supermarket loyalty schemes and digital devices, to better quantify the use of environmental features. Analyses should also be extended to include older children as they gain greater independent mobility, and additional exposures should be considered, such as access to play spaces in schools. Future work could also examine how mobility between environmental phenotypes influences childhood weight status. While our study focussed on healthy weight of children aged 4-5, our method provides a categorisation of environmental phenotypes that can be used to investigate associations with any number of health outcomes.

Our study demonstrates the value of defining environmental phenotypes for evaluating the role of the place around children’s homes on their physical health and highlights the importance of conducting individual-level data linkages. Given the public health challenge of childhood obesity, identifying modifiable risk factors is crucial. By uncovering the neighbourhood characteristics associated with childhood weight status, this research provides policy makers with an important evidence-base to design targeted, place-based interventions that can help transform neighbourhoods into environments that support healthier weight trajectories for children.

Funding

This work is part of the Built Environment and Child Health in Wales and Australia (BEACHES) project which is a joint initiative between Telethon Kids Institute, University of Western Australia and Swansea University. The BEACHES Project is funded by the UKRI-NHMRC Built Environment Prevention Research Scheme (grant number GNT1192764 and MR/T039329/1). Administrative Data Research (ADR) Wales also supported this research, which forms part of the ADR UK investment that unites research expertise from Swansea University Medical School and WISERD (Wales Institute of Social and Economic Research and Data) at Cardiff University with analysts from Welsh Government. ADR UK is funded by the Economic and Social Research Council (ESRC), part of UK Research and Innovation. Hayley Christian is supported by an Australian National Heart Foundation Future Leader Fellowship (102549) and partially supported by the Australian Government through the Australian Research Council’s Centre of Excellence for Children and Families over the Life Course (Project ID CE200100025).

CRediT authorship contribution statement

Jo Davies: Conceptualization, Formal analysis, Investigation, Methodology, Project administration, Writing – original draft, Writing – review & editing. Rowena Bailey: Conceptualization, Formal analysis, Investigation, Methodology, Validation, Writing – review & editing. Rebecca Pedrick-Case: Conceptualization, Data curation, Methodology, Validation, Writing – review & editing. Gareth Stratton: Conceptualization, Funding acquisition, Methodology, Project administration, Supervision, Writing – review & editing. Theodora Pouliou: Conceptualization, Investigation, Writing – review & editing. Amy Mizen: Writing – review & editing. Hayley Christian: Conceptualization, Methodology, Writing – review & editing. Bryan Boruff: Conceptualization, Writing – review & editing. Ronan A. Lyons: Conceptualization, Supervision, Writing – review & editing. Rich Fry: Conceptualization, Funding acquisition, Investigation, Methodology, Supervision, Writing – original draft, Writing – review & editing. Lucy J. Griffiths: Conceptualization, Funding acquisition, Investigation, Methodology, Project administration, Supervision, Writing – original draft, Writing – review & editing.

Footnotes

Appendix 1.

Environmental characteristics by phenotype

Profile name

Rural, spacious and isolated

Suburban

Deprived and underserved

Deprived and well-served

Dense, coastal and well-connected

Number households (%)

24,266

14.5

29,324

17.5

39,227

23.4

53,210

31.7

21,762

13.0

CharacteristicMedian(IQR)RangeMedian(IQR)RangeMedian(IQR)RangeMedian(IQR)RangeMedian(IQR)Range
Parks & gardens0(0)0130(0)0150(2)0150(2)0154(7)015
Playing fields0(0)002(3)1130(0)003(4)1131(4)013
Play spaces0(1)071(2)0102(2)0102(2)0102(3)010
Tennis courts0(0)000(0)000(0)010(0)010(0)01
Other sports facilities0(0)0100(0)0130(1)0130(2)0131(3)013
Total greenspaces0(1)0134(3)1253(4)0279(6)12712(9)027
Bakeries0(0)020(0)040(0)060(1)062(2)06
Cafes0(0)040(0)060(1)0180(1)0185(7)018
Confectioners0(0)010(0)010(0)030(0)031(1)03
Convenience stores0(0)081(2)072(2)0192(2)0196(6)019
Fast food outlets0(0)0100(1)0101(3)0282(4)02812(11)028
Restaurants0(0)0100(0)090(1)0220(1)0224(8)022
Supermarkets0(0)030(0)030(1)040(1)041(2)04
Total food outlets0(0)0171(3)0254(5)0827(7)18232(29)382
EVI0.4(0.2)0.20.80.3(0.1)00.70.3(0.2)00.70.3(0.1)00.70.2(0.2)00.6
Frequency%Frequency%Frequency%Frequency%Frequency%
Size nearest Park or Garden
 No park or garden23,45696.7––26,99892.1––26,10266.5––34,42164.7––439620.2––
 Small1040.4––7652.6––27226.9––566110.6––492222.6––
 Medium1790.7––6632.3––463011.8––620311.7––719333.1––
 Large5272.2––8983.1––577314.7––692513.0––525124.1––
Size nearest Playing field
 No playing field24,266100.0––00.0––39,227100.0––00.0––878740.4––
 Small00.0––12,23041.7––00.0––13,18424.8––411618.9––
 Medium00.0––904430.8––00.0––18,98735.7––389317.9––
 Large00.0––805027.5––00.0––21,03939.5––496622.8––
Size nearest play space
 No play space17,66572.8––901930.8––746119.0––51569.7––372817.1––
 Small295312.2––796927.2––12,43831.7––16,76531.5––403718.6––
 Medium20408.4––709724.2––935923.9––15,96130.0––674131.0––
 Large16086.6––523917.9––996925.4––15,32828.8––725633.3––
Size nearest tennis court
 No tennis court24,266100.0––29,324100.0––38,99799.4––52,80799.2––20,70795.2––
 Small00.0––00.0––460.1––2090.4––2371.1––
 Medium00.0––00.0––130.0––1530.3––5892.7––
 Large00.0––00.0––1710.4––410.1––2291.1––
Size nearest other sports facility
 No other sports facility23,90098.5––27,53793.9––26,84968.4––31,18458.6––958044.0––
 Small1460.6––8172.8––477012.2––772314.5––472821.7––
 Medium1280.5––4251.4––33758.6––645712.1––300113.8––
 Large920.4––5451.9––423310.8––784614.7––445320.5––
Garden size (m2)
 No Garden13565.6––9113.1––7041.8––7461.4––5502.5––
 0-99.921989.1––467215.9––935223.8––13,97026.3––11,13451.2––
 100-199.9501620.7––967933.0––13,55834.6––18,03033.9––437520.1––
 200-299.9396116.3––638521.8––781319.9––10,70420.1––19228.8––
 300-399.921799.0––27019.2––28617.3––36726.9––7053.2––
 400-499.914846.1––12754.3––11803.0––13572.6––3051.4––
 500+682328.1––25078.5––19445.0––18593.5––4632.1––
 Missing12495.1––11944.1––18154.6––28725.4––230810.6––
Nearest primary school
 Hyper-local14065.8––332311.3––452711.5––903817.0––487722.4––
 Local250710.3––754625.7––11,14028.4––19,08335.9––907141.7––
 Neighbourhood288011.9––807227.5––10,52026.8––15,34528.8––490322.5––
 >900m17,44571.9––10,34435.3––12,98133.1––965518.1––281412.9––
 Missing280.1––390.1––590.2––890.2––970.4––
Average distance to coast (miles)
 <13561.5––3231.1––8912.3––10802.0––13926.4––
 119618.1––299610.2––444211.3––565210.6––407418.7––
 214956.2––18406.3––33668.6––30585.7––340815.7––
 313335.5––23608.0––26116.7––45598.6––16307.5––
 421018.7––17936.1––414610.6––37857.1––18888.7––
 510354.3––19526.7––24006.1––31595.9––20269.3––
 620178.3––22747.8––19645.0––36856.9––6513.0––
 7-10405016.7––448215.3––459111.7––577210.8––16067.4––
 11-14301212.4––373712.7––35749.1––694213.0––13146.0––
 15+690628.5––756725.8––11,24228.7––15,51829.2––377317.3––
Walkability index
 1 – Most Walkable15,70164.7––12,37842.21––740518.88––836115.71––11115.11––
 2305712.6––643721.95––11,66129.73––16,84731.66––700332.18––
 33701.52––16965.78––610915.57––985918.53––645329.65––
 4513821.17––880330.02––13,92735.5––17,66633.2––349016.04––
 5 – Least walkable––––100.03––1250.32––4770.9––370517––
House type
 Detached11,51647.5––726224.8––724818.5––51819.7––12675.8––
 Self-Contained Flat2060.8––5671.9––10832.8––17253.2––15627.2––
 Semi-Detached554322.8––10,30435.1––11,45829.2––17,96833.8––355216.3––
 Terraced373315.4––820528.0––14,28636.4––23,19943.6––13,11960.3––
 Other840.3––200.1––180.0––130.0––680.3––
 Missing318413.1––296610.1––513413.1––51249.6––219410.1––
Urban Rural classification
 City & town less sparse932838.4––17,50959.7––31,20979.6––41,73778.4––18,38584.5––
 City & town sparse2711.1––8032.7––4771.2––7701.4––6603.0––
 Town & fringe less sparse290412.0––574119.6––456111.6––838415.8––12986.0––
 Town & fringe sparse7513.1––6822.3––10902.8––17633.3––13236.1––
 Village less sparse466419.2––304710.4––8092.1––2740.5––140.1––
 Village sparse634826.2––15425.3––10812.8––2820.5––820.4––
WIMDa
 1 – Most deprived10324.3––454815.5––10,96828.0––18,36634.5––704932.4––
 2312412.9––602920.6––788720.1––12,70523.9––574026.4––
 3587324.2––538718.4––569114.5––989318.6––398618.3––
 4772531.8––679523.2––515113.1––686212.9––303714.0––
 5 – Least deprived651226.8––656522.4––953024.3––538410.1––19509.0––
WIMD Income
 1 – Most deprived255510.5––427014.6––799220.4––10,75920.2––894341.1––
 2372015.3––555719.0––902423.0––11,48821.6––459021.1––
 3504920.8––580019.8––793620.2––11,56121.7––345415.9––
 4661127.2––698923.8––736818.8––998618.8––18928.7––
 5 – Least deprived633126.1––670822.9––690717.6––941617.7––288313.2––
WIMD Employment
 1 – Most deprived11614.8––493316.8––11,25128.7––17,45632.8––567826.1––
 2244210.1––596120.3––765019.5––14,12026.5––630229.0––
 3415617.1––618921.1––692617.7––11,46821.6––476621.9––
 4771931.8––588020.1––610015.6––630811.9––307514.1––
 5 – Least deprived878836.2––636121.7––730018.6––38587.3––19418.9––
WIMD Health
 1 – Most deprived12285.1––536418.3––11,17828.5––18,07734.0––597127.4––
 222999.5––619821.1––803820.5––14,05826.4––567626.1––
 3362514.9––638221.8––707018.0––11,75622.1––519523.9––
 4831734.3––586720.0––624315.9––596811.2––283213.0––
 5 – Least deprived879736.3––551318.8––669817.1––33516.3––20889.6––
WIMD Education
 1 – Most deprived12255.0––524617.9––11,27028.7––19,05735.8––504223.2––
 2250610.3––589120.1––805820.5––12,88124.2––625728.8––
 3435117.9––650022.2––622415.9––11,06620.8––456021.0––
 4879236.2––551518.8––635116.2––613211.5––307514.1––
 5 – Least deprived739230.5––617221.0––732418.7––40747.7––282813.0––
WIMD Access
 1 – Most deprived12,46951.4––618121.1––529513.5––46008.6––12996.0––
 2282411.6––733425.0––909023.2––13,28625.0––394218.1––
 3317413.1––629921.5––804620.5––13,32325.0––497522.9––
 4292712.1––526017.9––842621.5––11,47821.6––554125.5––
 5 – Least deprived287211.8––425014.5––837021.3––10,52319.8––600527.6––
WIMD Houing
 1 – Most deprived469219.3––28899.9––442511.3––873516.4––10,66049.0––
 2413417.0––479216.3––828521.1––13,80625.9––439320.2––
 3430217.7––654722.3––843121.5––13,00224.4––317014.6––
 4441418.2––746125.4––803220.5––12,08122.7––258911.9––
 5 – Least deprived672427.7––763526.0––10,05425.6––558610.5––9504.4––
WIMD Community Safety
 1 – Most deprived15216.3––318410.9––858221.9––15,71529.5––900641.4––
 219147.9––586320.0––814020.8––14,88128.0––553425.4––
 3339114.0––631921.5––786720.1––11,55421.7––439320.2––
 4650526.8––729124.9––818320.9––800215.0––219910.1––
 5 – Least deprived10,93545.1––666722.7––645516.5––30585.7––6302.9––
WIMD Physical Environment
 1 – Most deprived255510.5––427014.6––799220.4––10,75920.2––894341.1––
 2372015.3––555719.0––902423.0––11,48821.6––459021.1––
 3504920.8––580019.8––793620.2––11,56121.7––345415.9––
 4661127.2––698923.8––736818.8––998618.8––18928.7––
 5 – Least deprived633126.1––670822.9––690717.6––941617.7––288313.2––

Appendix 2.

Latent class analysis – model fit statistics

The model with nine classes failed to converge. The model with eight classes had the lowest AIC value, however the six, seven and eight class models all had classes that contained less than 10% of the sample. The five class model was selected as the best fitting model.

Classnll(null)ll(model)dfAICBIC
threeclass167,789.197383810739478903948963
fourclass167,789.192657314338534323854866
fiveclass167,789.185756417837154853717270
sixclass167,789.181853421536374983639654
sevenclass167,789.176053925135215813524098
eightclass167,789.173944128734794563482335

Appendix A. Supplementary data

The following are the Supplementary data to this article:

Multimedia component 1

Multimedia component 2

Data availability

The authors do not have permission to share data.

References

  1. AddressBase premium data products OS n.d. https://www.ordnancesurvey.co.uk/products/addressbase-premium (accessed July 2, 2025).
  2. Beynon C., Pashayan N., Fisher E., Hargreaves D.S., Bailey L., Raine R. A cross-sectional study using the childhood measurement programme for Wales to examine population-level risk factors associated with childhood obesity. Public Health Nutr. 2021;24:3428–3436. doi: 10.1017/S1368980020001913. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. 2011 census geographies – Office for national statistics n.d.https://www.ons.gov.uk/methodology/geography/ukgeographies/censusgeographies/2011censusgeographies (accessed August 4, 2025).
  4. Daniels K.M., Schinasi L.H., Auchincloss A.H., Forrest C.B., Diez Roux A.V. The built and social neighborhood environment and child obesity: a systematic review of longitudinal studies. Prev. Med. 2021;153 doi: 10.1016/J.YPMED.2021.106790. [DOI] [PubMed] [Google Scholar]
  5. Daniels K.M., Schinasi L.H., Auchincloss A.H., Forrest C.B., Diez Roux A.V. The built and social neighborhood environment and child obesity: a systematic review of longitudinal studies. Prev. Med. 2021;153 doi: 10.1016/J.YPMED.2021.106790. [DOI] [PubMed] [Google Scholar]
  6. Datar A., Nicosia N., Shier V. Parent perceptions of neighborhood safety and children’s physical activity, sedentary behavior, and obesity: evidence from a national longitudinal study. Am. J. Epidemiol. 2013;177:1065–1073. doi: 10.1093/AJE/KWS353. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Davies J., Bailey R., Mizen A., Pouliou T., Fry R., Pedrick-Case R., et al. Residential mobility amongst children and young people in Wales: a longitudinal study using linked administrative records. Int J Popul Data Sci. 2024;9 doi: 10.23889/IJPDS.V9I1.2398. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Garrett J.K., Rowney F.M., White M.P., Lovell R., Fry R.J., Akbari A., et al. Visiting nature is associated with lower socioeconomic inequalities in well-being in Wales. Sci. Rep. 2023;13:9684. doi: 10.1038/s41598-023-35427-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. GBD 2021 Adolescent BMI Collaborators Global, regional, and national prevalence of child and adolescent overweight and obesity, 1990-2021, with forecasts to 2050: a forecasting study for the global burden of disease study 2021. Lancet. 2025;405:785–812. doi: 10.1016/S0140-6736(25)00397-6/ATTACHMENT/8C915013-E8D3-441F-9B2A-9B438E91B229/MMC2.PDF. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Giles-Corti B., Vernez-Moudon A., Reis R., Turrell G., Dannenberg A.L., Badland H., et al. City planning and population health: a global challenge. Lancet. 2016;388:2912–2924. doi: 10.1016/S0140-6736(16)30066-6/ASSET/9707CF2C-54B7-4866-A924-1143B2620FE7/MAIN.ASSETS/GR1.JPG. [DOI] [PubMed] [Google Scholar]
  11. Guseman E.H., Sisson S.B., Whipps J., Howe C.A., Byra M.M., Silver L.E. Neighborhood and family characteristics associated with adiposity and physical activity engagement among preschoolers in a small rural community. Int. J. Environ. Res. Publ. Health. 2022;19 doi: 10.3390/IJERPH192113964. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Hirsch J.A., Meyer K.A., Peterson M., Rodriguez D.A., Song Y., Peng K., et al. Obtaining longitudinal built environment data retrospectively across 25 years in four US cities. Front. Public Health. 2016;4 doi: 10.3389/FPUBH.2016.00065. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Hirsch J.A., Meyer K.A., Peterson M., Rodriguez D.A., Song Y., Peng K., et al. Obtaining longitudinal built environment data retrospectively across 25 years in four US cities. Front. Public Health. 2016;4 doi: 10.3389/FPUBH.2016.00065. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Huete A., Didan K., Miura T., Rodriguez E.P., Gao X., Ferreira L.G. Overview of the radiometric and biophysical performance of the MODIS vegetation indices. Remote Sens. Environ. 2002;83:195–213. doi: 10.1016/S0034-4257(02)00096-2. [DOI] [Google Scholar]
  15. Jenkin G L., Pearson A L., Bentham G., Day P., Kingham S. Neighbourhood influences on children’s weight-related behaviours and body mass index. AIMS Public Health. 2015;2:501–515. doi: 10.3934/publichealth.2015.3.501. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Jia P., Cao X., Yang H., Dai S., He P., Huang G., et al. Green space access in the neighbourhood and childhood obesity. Obes. Rev. 2020 doi: 10.1111/obr.13100. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Jia P., Zou Y., Wu Z., Zhang D., Wu T., Smith M., et al. Street connectivity, physical activity, and childhood obesity: a systematic review and meta‐analysis. Obes. Rev. 2021;22 doi: 10.1111/obr.12943. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Jones K.H., Ford D.V., Jones C., Dsilva R., Thompson S., Brooks C.J., et al. A case study of the secure anonymous information linkage (SAIL) gateway: a privacy-protecting remote access system for health-related research and evaluation. J. Biomed. Inf. 2014;50:196–204. doi: 10.1016/j.jbi.2014.01.003. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Jones K.H., Ford D.V., Thompson S., Lyons R. A profile of the SAIL databank on the UK secure research platform. Int J Popul Data Sci. 2019;4 doi: 10.23889/ijpds.v4i2.1134. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Kegler M.C., Escoffery C., Alcantara I., Ballard D., Glanz K. A qualitative examination of home and neighborhood environments for obesity prevention in rural adults. Int. J. Behav. Nutr. Phys. Activ. 2009;5 doi: 10.1186/1479-5868-5-65. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Ling J., Chen S., Zahry N.R., Kao T.S.A. Economic burden of childhood overweight and obesity: a systematic review and meta-analysis. Obes. Rev. 2023;24 doi: 10.1111/OBR.13535;PAGE:STRING:ARTICLE/CHAPTER. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Lyons R.A., Jones K.H., John G., Brooks C.J., Verplancke J.-P., Ford D.V., et al. The SAIL databank: linking multiple health and social care datasets. BMC Med. Inf. Decis. Making. 2009;9:3. doi: 10.1186/1472-6947-9-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Mah S.M., Dasgupta K., Akbari A., Ross N.A., Fry R. An international comparative study of active living environments and hospitalization for Wales and Canada. SSM Popul. Health. 2022;18 doi: 10.1016/J.SSMPH.2022.101048. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Maitland C., Stratton G., Foster S., Braham R., Rosenberg M. A place for play? The influence of the home physical environment on children’s physical activity and sedentary behaviour. Int. J. Behav. Nutr. Phys. Activ. 2013;10:1–21. doi: 10.1186/1479-5868-10-99/TABLES/4. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Malacarne D., Handakas E., Robinson O., Pineda E., Saez M., Chatzi L., et al. The built environment as determinant of childhood obesity: a systematic literature review. Obes. Rev. 2022;23 doi: 10.1111/obr.13385. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Mayor S. Children living near fast food outlets in England are more likely to be overweight, study shows. BMJ. 2014;348 doi: 10.1136/BMJ.G1557. [DOI] [PubMed] [Google Scholar]
  27. Miller L.J., Joyce S., Carter S., Yun G. Associations between childhood obesity and the availability of food outlets in the local environment: a retrospective cross-sectional study. Am. J. Health Promot. 2014;28 doi: 10.4278/AJHP.130214-QUAN-70. [DOI] [PubMed] [Google Scholar]
  28. Mizen A., Rodgers S., Fry R., Lyons R. Linking environment and health data to investigate the association between access to unhealthy food and child BMI. Int J Popul Data Sci. 2018;3 doi: 10.23889/ijpds.v3i4.906. [DOI] [Google Scholar]
  29. Mizen A., Song J., Fry R., Akbari A., Berridge D., Parker S.C., et al. Longitudinal access and exposure to green-blue spaces and individual-level mental health and well-being: protocol for a longitudinal, population-wide record-linked natural experiment. BMJ Open. 2019;9 doi: 10.1136/bmjopen-2018-027289. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Mizen A., Thompson D.A., Watkins A., Akbari A., Garrett J.K., Geary R., et al. The use of enhanced vegetation index for assessing access to different types of green space in epidemiological studies. J. Expo. Sci. Environ. Epidemiol. 2024;34:753–760. doi: 10.1038/s41370-024-00650-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Mizen A., Thompson D.A., Watkins A., Akbari A., Garrett J.K., Geary R., et al. The use of enhanced vegetation index for assessing access to different types of green space in epidemiological studies. J. Expo. Sci. Environ. Epidemiol. 2024;34:753–760. doi: 10.1038/S41370-024-00650-5;KWRD=MEDICINE. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Nordbø E.C.A., Raanaas R.K., Nordh H., Aamodt G. Neighborhood green spaces, facilities and population density as predictors of activity participation among 8-year-olds: a cross-sectional GIS study based on the Norwegian mother and child cohort study. BMC Public Health. 2019;19:1426. doi: 10.1186/s12889-019-7795-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Obesity and overweight n.d. https://www.who.int/news-room/fact-sheets/detail/obesity-and-overweight (accessed June 18, 2025).
  34. Office for National Statistics . 2013. Office for National Statistics 2011 rural/urban Classification. [Google Scholar]
  35. OS MasterMap topography layer, great Britain’s landscape vector map data n.d. https://www.ordnancesurvey.co.uk/business-government/products/mastermap-topography (accessed April 21, 2021).
  36. OS open greenspace data, products OS n.d. https://www.ordnancesurvey.co.uk/products/os-open-greenspace#get (accessed August 4, 2025).
  37. Poole R., Moon G. What is the association between healthy weight in 4–5-year-old children and spatial access to purposefully constructed play areas? Health Place. 2017;46:101–106. doi: 10.1016/j.healthplace.2017.05.012. [DOI] [PubMed] [Google Scholar]
  38. QGIS manual. i.Vi – GRASS 2020. https://grass.osgeo.org/grass78/manuals/i.vi.html. (accessed August 22, 2024).
  39. Reimers A.K., Knapp G. Playground usage and physical activity levels of children based on playground spatial features. J. Publ. Health. 2017;25:661–669. doi: 10.1007/S10389-017-0828-X/TABLES/5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Robinson T., Boruff B., Duncan J., Murray K., Schipperijn J., Nathan A., et al. Understanding variations in the built environment over time to inform longitudinal studies of young children’s physical activity behaviour – the BEACHES project. Health Place. 2024;90 doi: 10.1016/J.HEALTHPLACE.2024.103345. [DOI] [PubMed] [Google Scholar]
  41. Rodgers S.E., Demmler J.C., Dsilva R., Lyons R.A. Protecting health data privacy while using residence-based environment and demographic data. Health Place. 2012;18:209–217. doi: 10.1016/J.HEALTHPLACE.2011.09.006. [DOI] [PubMed] [Google Scholar]
  42. Rodgers S.E., Lyons R.A., Dsilva R., Jones K.H., Brooks C.J., Ford D.V., et al. Residential anonymous linking fields (RALFs): a novel information infrastructure to study the interaction between the environment and individuals’ health. J. Public Health. 2009;31:582–588. doi: 10.1093/PUBMED/FDP041. [DOI] [PubMed] [Google Scholar]
  43. Sallis J.F., Floyd M.F., Rodríguez D.A., Saelens B.E. Role of built environments in physical activity, obesity, and cardiovascular disease. Circulation. 2012;125:729–737. doi: 10.1161/CIRCULATIONAHA.110.969022. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Schalkwijk A.A.H., Van Der Zwaard B.C., Nijpels G., Elders P.J.M., Platt L. The impact of greenspace and condition of the neighbourhood on child overweight. Eur. J. Publ. Health. 2018;28:88–94. doi: 10.1093/EURPUB/CKX037. [DOI] [PubMed] [Google Scholar]
  45. Schalkwijk A.A.H., Van Der Zwaard B.C., Nijpels G., Elders P.J.M., Platt L. The impact of greenspace and condition of the neighbourhood on child overweight. Eur. J. Publ. Health. 2018;28:88–94. doi: 10.1093/EURPUB/CKX037. [DOI] [PubMed] [Google Scholar]
  46. Singh G.K., Siahpush M., Kogan M.D. Neighborhood socioeconomic conditions, built environments, and childhood obesity. Health Aff. 2010;29:503–512. doi: 10.1377/HLTHAFF.2009.0730;WGROUP:STRING:PUBLICATION. [DOI] [PubMed] [Google Scholar]
  47. Spence J.C., Cutumisu N., Edwards J., Evans J. Influence of neighbourhood design and access to facilities on overweight among preschool children. Int. J. Pediatr. Obes. 2008;3:109–116. doi: 10.1080/17477160701875007. [DOI] [PubMed] [Google Scholar]
  48. Terrón-Pérez M., Molina-García J., Martínez-Bello V.E., Queralt A. Relationship between the physical environment and physical activity levels in preschool children: a systematic review. Curr. Environ. Health Rep. 2021;8(2):177–195. doi: 10.1007/S40572-021-00318-4. 2021;8. [DOI] [PubMed] [Google Scholar]
  49. The Child Measurement Programme for Wales Publication details Title . 2017. Child Measurement Programme for Wales: History, Legislative Framework and Technical Aspects. [Google Scholar]
  50. Thompson D.A., Geary R.S., Rowney F.M., Fry R., Watkins A., Wheeler B.W., et al. Cohort profile: the green and blue spaces (GBS) and mental health in Wales e-cohort. Int. J. Epidemiol. 2022;51:e285–e294. doi: 10.1093/ije/dyac080. [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. tmap Tennekes M. Thematic maps in R. J. Stat. Software. 2018;84:1–39. doi: 10.18637/JSS.V084.I06. [DOI] [Google Scholar]
  52. Urie Bronfenbrenner. Harvard University Press; Cambridge, MA: 1979. Ecology of Human Development : Experiments by Nature and Design. [Google Scholar]
  53. Van Der Horst K., Oenema A., Ferreira I., Wendel-Vos W., Giskes K., Van Lenthe F., et al. A systematic review of environmental correlates of obesity-related dietary behaviors in youth. Health Educ. Res. 2007;22:203–226. doi: 10.1093/HER/CYL069. [DOI] [PubMed] [Google Scholar]
  54. Van Der Zwaard B.C., Schalkwijk A.A.H., Elders P.J.M., Platt L., Nijpels G. Does environment influence childhood BMI? A longitudinal analysis of children aged 3–11. J. Epidemiol. Community Health. 2018;72:1110–1116. doi: 10.1136/JECH-2018-210701. [DOI] [PubMed] [Google Scholar]
  55. Welsh Government. Schools in Wales DataMapWales n.d. https://datamap.gov.wales/maps/schools-in-wales/(accessed March 27, 2024).
  56. Yang S., Chen X., Wang L., Wu T., Fei T., Xiao Q., et al. Walkability indices and childhood obesity: a review of epidemiologic evidence. Obes. Rev. 2021;22 doi: 10.1111/obr.13096. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Multimedia component 1

Multimedia component 2

Data Availability Statement

The authors do not have permission to share data.

Connect

Upvote0PointsDownvote

0 Votes: 0 Upvotes, 0 Downvotes (0 Points)

Leave a reply

Related Posts

Previous Post

Next Post

Loading Next Post...
Follow
Search
Popular Now
Show More
Scroll to Top
Loading

Signing-in 3 seconds...

Signing-up 3 seconds...

css.php