The study's central hypothesis is that linking detailed built environment features to longitudinal population databases improves the correlation between the built environment and health outcomes, generates new knowledge on environmental contributions to disease, and enhances understanding of how the built environment shapes long-term health outcomes. Current understanding of the relationship between the urban environment and health is limited by methodological shortcomings in environmental quantification and inconsistencies in built environment measurement. Existing studies frequently use aggregate-level data (administrative boundaries, postcodes), failing to account for individual-level exposure variations. These aggregate-level approaches suffer from modifiable area unit problems and centroid biases. The authors highlight the need for better quality spatial data at finer scales, arguing that quantifying environmental exposures at the individual level provides crucial information for improving knowledge on the impact of environmental exposures on health and well-being, directly informing spatial planning decisions. Inconsistency in the measurement and conceptualization of neighborhood characteristics across health studies is also a significant problem, leading to conflicting evidence regarding specific features and health outcomes. This emphasizes the urgent need for detailed, standardized, longitudinal measurements of the built environment in health studies. The study focuses on Bradford, a city in northern England with significant health inequalities, leveraging the 'Connected Bradford' database, a whole-population database linking various data sources over 40 years. This offers a rich dataset for investigating the links between built environment and health outcomes at a high level of detail.

The paper reviews existing literature demonstrating the relationship between various built environment characteristics and health outcomes. It highlights the associations between air pollution, noise, green space, greenness, public transport accessibility, walkability, street centrality, unhealthy food environments, and indoor dwelling qualities and various health problems. Conversely, it also notes evidence for the positive effects of some built environment features, like walkable neighborhoods, on physical activity and other positive health outcomes. The authors critique existing large-scale spatial indicators, which often focus on aggregate levels and disregard individual-level exposure variations. They emphasize inconsistencies in the measurement of neighborhood characteristics across different studies, potentially explaining contrasting findings for specific environmental features and health outcomes, such as the link between food environments and obesity or urbanicity and schizophrenia. The lack of detailed and standardized longitudinal measures of the built environment is identified as a major obstacle in establishing definitive causal links between specific features and health outcomes. The authors cite the need for high-resolution geospatial information, combined with longitudinal population datasets, to improve the value of these datasets for identifying causal relationships between built environment characteristics and health outcomes, allowing for investigations into the complex interactions between different built environment characteristics and their effects on health.

The researchers developed a methodology for constructing individual-level built environment indicators for every address in Bradford. The approach centers on measuring exposure from the individual's perspective, considering their interactions with the environment both statically (at their residence) and dynamically (during journeys). A person's address serves as the origin point for measuring the likelihood of interaction with surroundings and simulated journeys to urban features. This 'lived environment' is used to quantify exposures and aggregate them at the address level, linking them to individual-level health information from the Connected Bradford database. This differs from aggregate-level approaches by capturing granular built environment differences at the individual experience level. The methodology avoids the biases associated with postcode centroids by using a series of different metric distance radii and incorporating distance decay functions to simulate a person's lived environment. The authors acknowledge that interaction with the environment is not solely defined by proximity but also by individual behaviors, attitudes, and norms. Accessibility to urban features is quantified by both direct proximity and cumulative opportunity for exposure, modeling an individual's potential for encountering a given feature. The methodology is designed for generalizability, applicable to any data source with individual-level address information. Data linkage between the environmental indicators (EIs) and the Connected Bradford database involved five steps: deriving EIs for each Unique Property Reference Number (UPRN), obtaining pseudonymized NHS numbers and address information, georeferencing this information and linking it to UPRNs using an adapted address-matching package, pseudonymizing UPRNs and removing identifiable information, and providing Connected Bradford with a dataset of pseudonymized NHS numbers, UPRNs, and linked EIs. High-resolution geospatial data from Ordnance Survey (OS) datasets (OS MasterMap, OS MasterMap Highways, OS AddressBase, OS Points of Interest) formed the basis of the analysis, providing detailed and up-to-date information. A street network model was constructed using 2021 data to measure proximities to environmental exposures. The analysis incorporated nine different spatial relationships representing various ways individuals interact with their environment, including static locations (at home, on the street, in the urban block) and dynamic interactions (along routes to urban features). Four distance types were used: metric distance, metric distance through the street network, angular distance through the street network, and distance decay metric distance. An exponential distance decay function was used to model the decreasing importance of an urban feature with increasing distance. The study generated EIs across 11 domains: air quality (using high-resolution air pollution data from Bradford Council and Defra data for rural areas), road traffic (using Trafficmaster data to derive congestion variables), greenness and greenspace (using NDVI from Landsat 8 and OS Open Greenspace data), public transport (using NaPTAN data), walkability and land-use intensity (using a pedestrian demand model), street centrality (using space syntax metrics), built form (using OS MasterMap and EPC data), indoor qualities (using EPC data), and food environments (using OS POI data). PostgreSQL and PostGIS were used for geospatial database construction and calculation of EIs. The methodology’s longitudinal capacity (data available annually since 2007) is highlighted, allowing for the creation of longitudinal EIs.

The methodology yielded over 500 environmental indicators. The paper presents visualizations of fast-food exposure using distance decay weighted counts, showcasing geographical insights into spatial distribution and exposure variation across Bradford. A synthetic individual, 'Samina', a 12-year-old British Pakistani girl, is used to illustrate the potential insights offered by the data. Samina's environmental exposure profile is detailed, including her exposure to air pollution, traffic, green spaces, public transport, walkability, and fast-food outlets. The study also presents maps of exemplary EIs for the entire Bradford metropolitan district and a specific ward, highlighting the spatial variations in exposure. A trial analysis linking fast-food exposure variables with data from the Born in Bradford (BiB) longitudinal cohort study on childhood obesity found that increasing spatial precision in fast-food outlet exposure quantification did not alter the associations with childhood obesity, challenging previous research. This highlights the complexities of measuring environmental exposures and the importance of controlling for confounding factors, such as genetic predisposition, behavior, financial situations, and deprivation, in assessing the relationship between the built environment and health. This is further emphasised by the observation that while obesity is linked to economic deprivation, the density of fast-food outlets within reach of a home address does not appear to be the primary factor.

The study's findings address the research question by demonstrating the feasibility and value of creating large-scale, individual-level environmental exposure datasets. The high-resolution spatial data at the street and building level enables the generation of guidelines and spatial planning policies for modifying the built environment to improve health outcomes. The methodology's generalizability and applicability across diverse geographical contexts and exposure types are key strengths. The ability to conduct causal inference research designs using involuntary house move information to investigate causal effects is highlighted. The study's example using the Connected Bradford database to investigate the causal effect of air pollution on respiratory diseases illustrates this potential. The possibility of applying the methodology to other longitudinal cohort studies (NCDS, BCS70, UKHLS, BHPS, MCS, Next Steps) is discussed, emphasizing the untapped potential of linking individual-level address information to environmental exposure data. The study highlights challenges related to data availability, comparability (quality, resolution, precision, completeness, classification), and measurement (definition, classification, operationalization). The study addresses these by selecting datasets with comparable global alternatives and using the earliest available data for datasets with little temporal variation. However, the use of inferred classifications for older data introduces potential comparability issues.

The paper presents a novel methodology for constructing high-resolution individual-level environmental exposure indicators. The Connected Bradford dataset serves as a unique information source, offering valuable insights for research and policy-making. The methodology’s strength lies in its generalizability and potential for nationwide and international replication, facilitating the creation of similar databases in other locations. It addresses the critical need for precise, meaningful data for research and planning in this policy domain. Future research could explore the longitudinal aspects further and expand upon the analysis of the interactions between environmental factors and health outcomes.

While the methodology is designed for broad applicability, challenges remain related to data availability and comparability, particularly for historical data. Inconsistencies in data quality, spatial resolution, precision, completeness, and classification across datasets can affect the accuracy and generalizability of the results. The study also acknowledges limitations in defining and operationalizing environmental exposure variables, as demonstrated by the trial analysis of fast-food exposure and childhood obesity which highlights the influence of multiple confounding factors beyond simple spatial proximity.