Characterising the Obesogenic Built Environment using Composite Indicators: A moonshot or a paper moon?

Obesity is a chronic, relapsing disease with multiple comorbidities affecting over two billion people globally. Beyond individual causes, the built environment may influence diet, sedentary behaviour, physical activity, and transport choices, making it a key target for population-level prevention of obesity. Objective measures of obesogenic built environments are essential, but studies often rely on single-exposure models that fail to capture complex interactions. This thesis explores methods for developing composite indicators to characterise obesogenic built environments and examines their associations with health behaviours and obesity-related outcomes. Chapter 2 presents an umbrella review of 32 systematic reviews on built environment factors and their links to obesity in adults. Findings suggest that urbanisation is associated with lower weight in high-income countries but higher weight in low-income countries. Socioeconomic status plays a role as both a determinant and confounder. Some evidence links fast food availability, urban sprawl, and reduced land use mix to higher weight in high-income settings, but overall, food environment associations were less consistent than those of the physical activity environment. To characterise the physical activity environment, we developed two spatial indices: walkability and drivability These indices measure land use compactness (population density, land use mix), walking infrastructure (intersections, sidewalks), driving infrastructure (highways, parking supply), public transport, daily destinations (shopping, jobs), and green space. Chapters 3, 4, and 5 used the 2017 Dutch National Travel Survey to validate and improve these indices. In Chapter 3, a 10% increase in residential walkability was associated with 8.5 additional minutes of daily walking (n=16,055). Chapter 4 found that a 10% increase in workplace walkability increased the odds of commuting on foot by 32% (n=6,769), with stronger effects when residential and workplace walkability were combined. Chapter 5 refined the drivability index by incorporating urbanisation as an interaction term. The final index predicted driving more accurately in urban (AUC=0.64) than in rural areas (AUC=0.54). A 10% increase in residential drivability was associated with a 31% higher likelihood of driving. Both indices have implications for urban planning and public health policies. In Chapter 6, we combined 12 physical activity environment components (walkability, drivability, sports facilities, and bike path density) with five food environment components (density of fast food outlets, supermarkets, restaurants, takeaways, and local food stores) to create a theory-based obesogenic environment index for the Netherlands. The index showed regional differences but no clear socioeconomic gradient in obesogenicity. Chapter 7 linked the index to residential addresses in the 2016 Dutch Public Health Monitor (n=255,187). Surprisingly, the index was inversely associated with BMI, largely driven by the food environment component. In Chapter 8, we adapted the Dutch obesogenic environment model to Singapore. From 11 spatial measures, four key variables (subway access, park density, residential density, and general food access) were identified as significant predictors of overweight risk in a multiethnic adult cohort (n=18,919). Combined together, this data-driven index was more strongly associated with obesity-related outcomes (e.g., waist-to-hip ratio, cholesterol levels) than a theory-based version. Overall, this thesis confirms that the physical activity environment is more consistently linked to obesity-related outcomes than the food environment. Associations between indices and health outcomes were influenced by urbanisation and socioeconomic context. Methodologically, this research presents different approaches for constructing composite indicators, each with trade-offs. While no single method is universally optimal, choices should align with research goals, user needs, and data availability. The developed walkability, drivability, and obesogenic environment indices highlight neighborhood disparities in environmental characteristics relevant to urban and health planning and could be aggregated for regional assessments. Future research should further integrate composite indicators, geospatial analysis, and epidemiological methods to refine these measures and advance the field