The epidemiologic case for urban health: conceptualizing and measuring the magnitude of challenges and potential benefits

1.1 The importance of urban health

As the proportion of humanity living in cities continues to rise, the health of people living in urban areas will represent an increasingly important component of global health.¹ In 2018, the share of the world’s population residing in urban areas was 55% and is expected to increase to 60% by 2030.¹ Urbanization (the process of an area becoming more urban) is accelerating in particular in low- and middle-income countries in Asia and Africa.¹ A clear implication of global urbanization trends is that improving public health globally will require a major focus on health in urban areas.² However, the scale of current and future urban health challenges and opportunities is not well characterized and depends on assumptions about the nature of the relationships between cities and health and the scope of urban health action.

As a preliminary note on language, an urban area has been defined “as a geographical space characterized by a continuous urban settlement.”³ The phrase is commonly used in a more general sense than city, which can refer to a discrete urban area defined with administrative boundaries.² For a rigorous discussion of these terms, we recommend Chapter 1 of Urban Public Health by Diez-Roux² and Tonne et al (2021).⁴ In this article, we usually use the more general term—urban area—but occasionally use the word city as a rhetorical shorthand.

Urban areas present both challenges and opportunities for public health.^5–7 Compared with rural areas, urban areas can promote health through higher standards of living and better access to resources and services, including healthcare services.^4,8,9 Nevertheless, characteristics common to many urban environments can also harm health, including air pollution;¹⁰ noise pollution; heat exacerbated by urban heat islands;¹¹ inadequate green space; high availability of nutrient-poor, calorie-dense food; and unsafe transportation systems resulting in road traffic injuries and discouraging physical activity.^4,12 Urban areas are also vulnerable to some infectious diseases. Some, such as COVID-19,¹³ may spread rapidly in areas with high population density and connectivity.¹² Others, such as malaria, have historically been more common in rural areas,¹⁴ but climate change and landscape modifications including deforestation are changing spatial patterns of vector-borne diseases, increasing their occurrence in urban areas.¹⁵

While urban areas can share common health-relevant environmental and social features, cities are distinct, and the health of urban residents—and the impact of urban living on health—varies both between and within cities.^16,17 Unequal exposure over time to social, political, and environmental conditions gives rise to health inequities in urban populations by characteristics such as income and wealth, residential neighborhood, race and ethnicity, age, disability, gender identity, sexual orientation, or migration status.¹⁷ Ultimately, variation in health between and within cities reveals an opportunity: urban areas can be designed and managed to promote health and health equity. Furthermore, in the context of global climate change, strategies to improve health in cities can have co-benefits for climate-change mitigation and adaptation.^4,18–20 Given this opportunity, there is a need for a clearer understanding of how urban environments and the forces that change them affect public health and of the scale of the health challenges and opportunities inherent in the urban context.

1.2 The role of epidemiology

Many disciplines will contribute to this needed understanding, and epidemiologists, those who study the distribution and determinants of health in populations, are well-equipped to contribute in several ways. Some tools from epidemiology, in particular descriptive epidemiology, have an established history of informing urban health. Other more contemporary topics in epidemiology, such as methods for causal inference, have yet to realize their potential in helping to answer urban-health questions. At the same time, the discipline may be well-suited to advance urban-health knowledge in ways traditionally not under its purview, such as the surveillance of the urban physical and policy environment. In this article, we discuss how epidemiology has been and can continue to be used to advance understanding of the links between urban areas and health informed by an existing urban-health conceptual framework. We then propose actions needed to continue to advance this understanding.

2.1 A framework relating urban areas and health

Urban-health scholars have proposed various frameworks to represent and explain the links between urban environments and population health.^4,21–26 In the framework by Diez Roux presented in Chapter 3 of Urban Public Health, the links are conceptualized in four inter-related ways.²⁵ First, urban areas are contexts for health. That is, they are settings in which population health occurs. Second, urban areas are causes or determinants of health. These place-based causes occur at various levels, for example, the neighborhood (e.g., features of its built environment) or the city (e.g., housing policy). Third, and related to the second, urban areas can modify or reinforce individual-level health pathways. For example, genetic susceptibility to obesity can be modified by the neighborhood food environment.²⁷ Fourth—and encompassing each of the above—urban areas are part of complex systems that shape population health.^4,22,28

Measuring the magnitude of urban-health challenges through each of these lenses has value for researchers, policymakers, and advocates. For instance, viewing urban areas as contexts for health allows us to describe and compare health outcomes across different urban environments, highlighting disparities and identifying areas to prioritize for intervention. Understanding urban areas as determinants and modifiers of health enables us to investigate the causal impacts of specific urban characteristics and interventions on health outcomes and to inform action to change those characteristics. Finally, adopting a complex systems perspective allows us to appreciate the interconnectedness of factors influencing urban health, informing the development of holistic interventions at the structural level that can be mutually beneficial for many domains of population health, ecosystems, and climate resilience.²⁹

2.2 Urban areas as contexts for health

2.2.1 Describing health-related states and behaviors in urban populations

The most conceptually straightforward link between urban areas and health is the occurrence of diseases, health-related states, and behaviors among individuals living in urban areas. Measuring these outcomes in urban areas is important for planning and resource allocation, monitoring progress on health targets over time, comparing the distribution of health between groups, generating hypotheses, and identifying opportunities for intervention. This task has traditionally been undertaken by public health surveillance systems (such as vital statistics registries and large-scale cross-sectional surveys), cohort studies, and other tools of descriptive epidemiology. For example, national registries have been used to compare cancer incidence between urban and rural areas over time in the United States.^30,31 Vital registration data can be standardized and harmonized to allow for cross-city comparisons, such as in the SALURBAL study that described within-country heterogeneity in life expectancy across Latin American cities.¹⁶

A barrier to descriptive epidemiology in urban areas is that information is commonly collected to be representative of an administrative unit above the city (for example, of a state, province, or country). For example, the Global Burden of Disease project, which comprehensively describes the occurrence, mortality attributable to, and disability attributable to hundreds of diseases, injuries, and risk factors, presents information at the country level.³² Such comprehensive descriptive epidemiology does not exist for urban areas. Assuming 55% of the world’s population lives in urban areas¹ and that there were 55.4 million deaths in 2019,³³ then a crude calculation assuming the number of deaths in urban areas is proportional to their share of the population suggests about 30.5 million deaths (0.55*55.4 million) occurred in urban areas in 2019. Currently, only crude calculations such as this are available, and global estimates have not accounted for urban-versus-rural differences in mortality for reasons such as urban-rural differences in age structure or exposure to risks. The lack of urban-specific estimates of the health burden or of the scale of the health benefits that might result from relevant interventions poses challenges for allocating appropriate financial, institutional, and human resources to urban health action.

While more work is needed to comprehensively describe the urban burden of ill health, existing information can serve as a useful guide. When direct measurement of health outcomes (e.g., through surveys) in cities is not feasible, two approaches hold promise for advancing descriptive epidemiology within and across urban areas: health-impact modeling and small-area estimation. Health-impact modeling is often used to project the health impacts of alternative scenarios. It can also be used to estimate the burden of dimensions of health attributable to the status-quo scenario. For example, health-impact modeling was used to estimate that exposure to air pollution among residents of 13,160 urban areas globally resulted in 1.8 million excess deaths in 2019.¹⁰ While valuable, health-impact modeling studies can rely on strong assumptions, for example that health effects estimated in one location can apply to the area of interest. As we discuss further below, health effects can be context dependent.²⁵ The validity of health-impact assessment may benefit from more explicitly considering the extent to which effects are transportable from one context to another.³⁴

As health outcomes can vary considerably between and within urban areas,¹⁷ methods aiming to estimate health outcomes at a fine geographic scale are important. Small-area estimation (SAE) is a suite of such methods. SAE methods estimate outcome measures (e.g., disease prevalence) at a unit of geography beneath the level where the measure is known (e.g., from a U.S. state to a county).³⁵ SAE methods include, for example, non-parametric methods that use measured associations between the outcome and demographic characteristics at the larger geographic unit (e.g., U.S. states) and apply those associations to smaller units where the outcome is not measured (such as counties), spatial data smoothing,³⁶ regression techniques using area-specific data as predictors,³⁵ and an array of Bayesian techniques.³⁷

Long used to estimate the prevalence of health-related behaviors at the U.S. county level,^38–40 SAE methods have recently been used to estimate life expectancy of residents of U.S. census tracts (a within-city unit).⁴¹ Results suggest there is considerable variation in life expectancy between census tracts, underscoring that health varies within cities.⁴² CDC PLACES is another large-scale application of SAE methods in the U.S. context, estimating various chronic-disease-related measures at the sub-county level.⁴³ An advantage of SAE methods is that they borrow information from the higher level, allowing investigations of small areas (such as an urban neighborhood) to use fewer resources than they might require if the higher-level information were not available. As with most any method, estimation relies on assumptions, for example, that the association between demographics and the outcome at the larger area holds at the smaller area. In the urban-health context, SAE methods may face particular challenges for key populations such as migrants or those who live in slums, groups whose health is difficult to measure.^44,45 A small-area estimate aiming to estimate the health distribution of refugee settlements may be uninformative, for example, if it uses as key inputs the health distribution of surrounding areas, which may not be a strong predictor of the health of a group of individuals who recently arrived to that area. Finally, caution is recommended when using small-area estimation to monitor changes over time,⁴³ as the methods may involve inputs with a coarse temporal resolution and because the model inputs may assume time-invariant predictor-outcome associations.

2.2.2 Describing the health-relevant context of urban areas

Descriptive epidemiology conventionally measures health outcomes (such as cancer incidence and death) and individual-level risk factors (such as smoking and diet). As knowledge about the environmental impacts on health continues to grow, an increasingly important task for descriptive epidemiology, together with other disciplines, will be to better characterize the multi-level social and physical environments in urban areas that give rise to health.^24,46 As an example addressing this need, a research coalition developed a dataset covering 371 cities in Latin America with various consistently defined indicators of the social environment across economic, social, housing, governmental, institutional, and organizational domains.⁴⁶ They noted several challenges in developing this dataset including variation in data availability, quality, and spatial and temporal resolution and extent.

The surveillance of physical environmental conditions relevant for health has advanced considerably as data have become available from remote sensing (i.e., satellite imagery),⁴⁷ street-level imagery (such as Google Street View),⁴⁸ and open-source mapping software (such as OpenStreetMap).⁴⁹ For example, remote sensing has been used to measure urban green space,⁵⁰ street-view imagery to assess walkability and disorder of neighborhoods,^48,51 and OpenStreetMap to assess building quality and attributes of the roadway network.^49,52,53 One large-scale effort using some of these tools (OpenStreetMap, notably) is the creation of a multi-city dataset of livability indicators across several Australian cities.^54,55 This dataset of Australian cities is impressive for its rich information about the physical urban environment at high (within-city) spatial resolution and the comparability of its measures between cities, as constituent cities are each measured using the same methods.

As illustrated by the creation of such large-scale multi-city datasets of the built environment, modern sources of street-level imagery and mapping software have eased the objective and harmonious assessment of the physical environment at scale. Challenges remain with their completeness, accuracy,⁵³ and ongoing availability and legal use in urban health-related research.⁵⁶

Related to the need for a consistent dataset of indicators of the social and built environment across and within cities is the need to systematically inventory laws, policies, interventions, projects, and other actions that could affect these indicators. As Diez Roux writes, “Cities are acting,”² but, as a whole, the public health community lacks a robust system for collecting data on these actions. The field of policy surveillance is seeking to address this need.⁵⁷ Stemming from the tradition of public-health surveillance, policy surveillance is “the systematic tracking of policies of public importance”, as defined in Chapter 14 of the Urban Public Health text.⁵⁸ Well-executed, policy surveillance provides accessible quantitative data over place and time on attributes of policies relevant for public health. This information can be used to monitor trends in policies known to affect public health and to evaluate population-health impacts of policies.^57,58 Noteworthy policy-surveillance resources relevant for urban health in the U.S. context include LawAtlas.org⁵⁹ and CityHealth.org,⁶⁰ projects maintained by the Temple University Center for Public Health Law Research Policy Surveillance Program and collaborating organizations. In addition to the examples mentioned in Chapter 14,⁵⁸ the CDC’s Community-Based Survey of Supports for Healthy Eating and Active Living has conducted nationally representative policy surveillance of municipalities on topics such as physical activity⁶¹ and traffic safety.⁶² At the global level, the C40 Cities, a network of cities focused on climate-change action, have documented actions cities are implementing to adapt to climate change,⁶³ many of which likely have population-health co-benefits.⁶⁴

These resources provide a good foundation for urban-health policy surveillance. As compared with national or state-level policies, local-level policies are often more difficult to track systematically (owing to the complex layers of governance in municipalities, inconsistency across cities in maintenance of websites and public-facing documents, and the sheer number of municipalities as compared with other administrative units), and gathering historical policy-related information can be even more challenging.

Beyond the surveillance of codified policy, challenges are perhaps even more formidable in tracking sub-city actions that might not be viewed as policy per se but could nevertheless affect public health. Examples include the conversion of a vacant lot to housing, the construction of a neighborhood greenway, or the closing of a coal power plant. Epidemiologists, with expertise in measurement across various scales of time and place, data-management skills, and a high value placed on methodological rigor, are well-suited to contribute to the surveillance of policies, interventions, and other population-level actions that could affect public health. The field of legal epidemiology provides a direct avenue for epidemiologists to collaborate with lawyers and other policy experts to contribute to policy surveillance, specifically.⁶⁵ And existing global collaborations between urban planners, data scientists, and public-health professionals provides guiding examples of tracking other sub-city interventions.^49,55 Critical data include where and when the action occurred, the geographic scale of the action, and, as applicable, the level of governance responsible.⁶⁶ For evaluation purposes, documenting when the intervention occurred is especially important so that methods considering intervention timing (described in the next section) can be applied.

3.1 Summary of needed understanding

The reviewed conceptual model considers various ways to understand—and even define—urban health. Within each conceptual frame, more work is needed to advance understanding of urban health and support action to improve the health of urban populations. For example, it is unknown how much ill health occurs in urban populations. In this article, we crudely estimated that about 30-35 million deaths occur annually in urban areas, but this estimate relies on strong assumptions and ought to be improved upon. More descriptive epidemiology is needed to characterize urban health with both fine (within-city) resolution and at a large (global) scale to better understand differences in health between and within urban areas. There is a need for improved measurements of both the health of urban populations and of the environments that give rise to health. Creating databases at the multi-city scale containing high-resolution indicators of the social and built environments and related policies, interventions, and other actions should be encouraged. Noted examples characterizing indicators and policies in Latin America⁴⁶ and Australia⁵⁵ may serve as a guide.

There is also a need to understand what works to improve urban health. Decades of research have shown links between social and physical urban environments and health, but research evaluating the impacts of actual actions (e.g., policies and interventions) that may affect urban health and health equity remains comparatively scarce. Monitoring actions between and within cities would enable such evaluations. Finally, the argument for considering the health effects of urban living as part of a complex system is persuasive,⁸⁵ but our tools for quantitatively understanding these system-level processes remain under-developed in the public-health literature, in particular for the evaluation of urban-health interventions.⁸⁹ That many of these systems modeling methods require information-rich empirical evidence also underscores the need for traditional epidemiologic research to inform their components. While it will take a multi-disciplinary effort to address these gaps in understanding, the discipline of epidemiology, specifically, can inform urban-health understanding by adopting a broad view of descriptive epidemiology that includes surveillance of the urban physical and policy environments, continuing to evaluate group-level interventions between and within urban areas, and advancing research on systems thinking in urban health issues.

3.2 Actions needed to support gaps in understanding

To support this needed understanding, we recommend several actions that can be taken by stakeholders in national governments, local governments, not-for-profit organizations, coalitions, and academia. National governments could increase funding for fine-scale urban health surveys and surveillance systems and develop dashboards displaying resulting measures. U.S.-based examples of such tools include New York University’s City Health Dashboard^92,93 and the University of Wisconsin’s County Health Rankings.⁹⁴ As individual countries improve urban-health surveillance, the magnitude of urban-health challenges at the global or regional scale may become more apparent. Public health practitioners and other officials in local governments could similarly collect small-area health data and, ideally, present it in standardized formats on open-data portals.⁹³ Local governments are especially well-suited to conduct policy surveillance on actions that could affect urban health, as they are often responsible for their on-the-ground implementation. If data collection on all relevant interventions and actions is not feasible for government officials, they could partner with academic researchers, not-for-profit organizations, or global coalitions such as C40.⁹⁵ The evaluation of policies and interventions that could affect urban health at the local level is another area well-served by cross-sectoral collaboration. This collaborative evaluation research is particularly important in sectors not traditionally viewed under the public-health umbrella ranging from but not limited to transportation systems, housing policy, taxation policy, and energy systems. Throughout urban health surveillance and evaluation, care should be taken to collect data to represent marginalized groups and evaluate impacts on equity-related outcomes.