When it comes to health, place matters. Where we live, work, and play influences health in both positive and negative ways. Places have a significant impact on health outcomes independent of individual-level factors [1]. Examining place-based social determinants of health, such as access to food, housing, air quality, and social support, as well as structural determinants, including the historical context, governance, policies, and norms that shape local environments, can help inform effective public health interventions [2,3].

While the concept of place can be subjective, smaller areas are often defined by census tract or zip code boundaries, and larger areas are often defined by counties, congressional districts, or states. Using these definitions, publicly accessible place-based data can be leveraged to identify factors underlying the health of a population. Prior studies have shown that neighborhood socioeconomic status, measured at the census tract-level, is associated with increased rates of heart disease, cancer, and all-cause mortality, even after individual-level socioeconomic status is taken into account [4-6]. Other studies show that zip codes in which children grow up influence their future income, educational attainment, fertility, and marriage outcomes [7]. Analyses of the impact of state Medicaid expansion status on comorbidity burden show the influence of state-level policy on health inequities [8].

Over the last decade, the availability of publicly accessible place-based data has greatly expanded [9-11]. However, the growth of data resources has been segmented and nonuniform. Many websites developed by individual governmental organizations or research groups offer domain-specific data, such as social environment, climate, health outcomes, or housing [12-15]. In addition, new or expanded place-based measures have been developed that are not yet accessible through online platforms [16,17]. For some domains (eg, rurality [18]), multiple measures exist but often are not available in one place to facilitate comparisons or assess suitability for particular applications. Furthermore, the availability of national data has become inconsistent and uncertain due to cuts to federal funding for biomedical research and shifting priorities at federal agencies [19-21]. The result is an extensive but evolving, complex, and fragile landscape of place-based measures.

UCSF Health Atlas [22] provides a single online platform for place-based data to help users explore place-based factors that influence health. It includes a curated repository of publicly available population estimates as well as expert-driven composite measures. Most data are available across all 50 states in the United States, the District of Columbia, and Puerto Rico.

The development of Health Atlas was initiated by the University of California, San Francisco (UCSF) Population Health and Health Equity program and led by the UCSF Population Health Data Initiative. The website was created in collaboration with Stamen Design, and its content was developed with expertise from the UCSF Department of Epidemiology and Biostatistics, UCSF Disparities Research Environment and Multi-omics Lab, and Plain Language Health. Health Atlas development began in 2019, and the initial California-only version was launched in April 2020 [23]. Over time, a mobile-enabled version, Census 2020 geographies, and new domains of data were added. In fall 2024, the Health Atlas team launched a national version that included national maps and data [24]. An artificial intelligence (AI)–assisted search function was integrated in January 2026.

Health Atlas supports data visualization for 7 geographic levels: census tract, zip code tabulation area, county, congressional district, core-based statistical area, public use microdata area, and state. While not all data are available at every geographic level, most data are provided at the census tract level, the smallest unit of geography in Health Atlas (representing areas of approximately 4000 residents). Health Atlas provides data based on 2010 Census and 2020 Census boundaries; variables are merged on a geographic identifier, such as the Census Tract Federal Information Processing Series code [25], and transformed into flat files based on geographic granularity. Providing multiple census geographies allows researchers to select, download, and merge data relevant to their time period of interest.

To date, Health Atlas includes data on 206 variables (also referred to as “characteristics”) across 5 topical domains: demographics, socioeconomic, neighborhood, environment, and health and health care (Table 1). Domains were modeled after the Office of Disease Prevention and Health Promotion’s social determinants of health conceptual framework, which includes domains of health care access and quality, education access and quality, social and community context, economic stability, and neighborhood and built environment. Variables were selected for inclusion based on this framework, as well as feedback from expert scholars and potential users from academic, medical, and community settings. Data sources and variables are vetted for inclusion based on their relevance to place-based social determinants of health, accessibility (eg, from a public source), national scope, availability at the geographic boundaries represented in Health Atlas, and sustainability (eg, feasibility to update over time). Included data are curated primarily from public sources including the American Community Survey (ACS) [26] (eg, demographics, socioeconomics), US Centers for Disease Control and Prevention (CDC) PLACES: Local Data for Better Health [12] (eg, health and health care), and US Environmental Protection Agency (EPA) Environmental Justice Screen [27] (eg, environmental exposures).

The data are selected based on the most recent version available for a given geographic boundary. For example, the 2015‐2019 ACS variables (eg, for race/ethnicity) are available for Census 2010 geographies because they are the most recent variables available for that geography; the Census 2020 boundary is currently associated with the 2019‐2023 ACS variables. Health Atlas aims to update data annually following the release of new data from the Census and to search for additional updates to other variables at the same time. Health Atlas does not maintain longitudinal data, so when new data are released at an existing geographic boundary, it replaces the old version.

Full descriptions, data sources, available years, and geographies for each variable are provided on the Health Atlas and in Multimedia Appendices 1 and 2.

Data are partitioned by geographic level, census year, and variable and stored in Amazon Web Services S3 buckets, a public cloud storage resource [51]. This type of data organization (commonly called “hive partitioning”) is performed by using the DuckDB [52] and Polars [53] software packages. Storing the data statically reduces dependence on the server, making the platform more resilient.

Maps are generated using Mapbox [54] maps, which include data from OpenStreetMap [55]. Data descriptions are updated in Airtable [56], a cloud-based data storage platform, and mapped to the public-facing website. Metadata about each variable are loaded from Airtable into Health Atlas. When a user selects a variable and geography, the platform pulls only the relevant data from the S3 bucket and can load the values for the selection independently.

We aimed to meet Web Content Accessibility Guidelines level AA compliance for the interface with high text-to-background contrast ratios, minimum text sizes, and minimal decorative text. Maps and data visualizations are not strictly accounted for under Web Content Accessibility Guidelines, but we tested common points of concern, such as colorblind-friendly color palettes and figure-to-background ratios, to improve legibility during the design process.

An AI-assisted search function is integrated into Health Atlas to help users select their data of interest more efficiently. Suggestions are provided based on natural language processing via large language models and are limited to variables available in the Health Atlas.

Health Atlas is a publicly available resource with over 200 place-based variables that can support a wide array of users interested in public health. Researchers can download data for analysis and maps for presentations or reports. Policymakers can gain insights into custom regions of interest. Community members can visualize local health and environmental factors for advocacy. Educators and learners can use mapping features to explore a variety of topics. By including national-level data, Health Atlas expands opportunities for collaboration by facilitating multistate comparisons and partnerships. Furthermore, it provides access to data that have not been readily available previously, such as segregation and ethnic enclave measures.

Health Atlas is aligned with contemporary conceptual frameworks that emphasize that health inequities largely stem from structural and social determinants of health across multiple levels of influence, including place [1,59]. Data hosted on Health Atlas, therefore, include measures of structural and social determinants of health across domains, such as the built and social environment, environmental and climate stressors, and health care access and use.

Health Atlas provides place-based variables across multiple domains free of cost to the user, curated from publicly available sources and/or research groups that collectively have not been previously collated or easily accessed (Table 1) [16,17,32]. It also provides comprehensive data downloads and data visualizations across domains and geographic units (eg, census tract, county, state). Providing granular levels of data, such as census tracts, is key to meaningfully capturing neighborhood-level exposures. Providing data at multiple levels, including nonnested geographic units such as congressional districts and public use microdata areas, provides users the flexibility to use the data for different applications such as advocacy, policy, and administration. Other resources, such as PolicyMap [60] and SocialExplorer [61], include a wider set of similar data but require institutional licenses for access to the data, whereas Health Atlas data are publicly available. Tools such as ArcGIS [62] can also provide maps with more sophisticated features, including spatial analysis or the ability to layer study data in a secure environment; however, access to these tools typically requires licenses. While some licensed access additionally allows for a platform in which users can upload and incorporate their own data into maps, Health Atlas allows users to download a customizable dataset to append to their data.

The ability to gain insights from data is key and should not be limited to those with advanced training. Health Atlas provides easy-to-access maps, histograms, scatterplots, and summary plots to allow users to better understand the distribution of the data. Data descriptions and links to original sources are included for each variable. Through partnership development, Health Atlas makes new measures available to the public (eg, segregation, climate). For these measures, more detailed descriptions of how variables were created are provided, with citations to relevant published articles.

Health Atlas has remained nimble in response to user input, new partnerships, and changing external circumstances. The site has evolved to include tags for filtering variables related to selected topics, additional variable domains and categories, and a function to easily export map images. The Health Atlas scientific team also has back-end capabilities to update data and data descriptions as needed.

Since Health Atlas was first released in April 2020, it has been used for a variety of public health applications. During the COVID-19 pandemic, Health Atlas allowed users to visualize COVID-19 data alongside place-based data to make evidence-based decisions for organizations and individuals [63]. It also helped researchers and policymakers better understand how different communities were being impacted by the pandemic, and how social factors, such as neighborhood-level English proficiency and disability, contributed to disparities in COVID-19 outcomes [64,65]. The University of California Davis Comprehensive Cancer Center is currently using Health Atlas to identify areas with low breast cancer screening to plan locations for their mobile mammogram bus program (J. Dang, personal communication, October 30, 2025).

Health Atlas has also been used to link place-based data to geocoded patient addresses from individual-level datasets or electronic health records, providing insight into multilevel determinants of health. At UCSF, a project linking nSES from Health Atlas to electronic health records data found that older adults living in low SES neighborhoods had lower odds of advance care planning [66], a process by which people communicate their preferences for future medical care. Another project used nSES from Health Atlas to adjust for confounding at the neighborhood level and provide a clearer picture of the direct benefits of a digital health intervention among patients participating in cardiac rehabilitation [67]. Additional research used geospatial methods to plot hotspots of patients with serious mental illness and characterize neighborhoods based on racial or ethnic composition and socioeconomic status using data accessed through Health Atlas [68]. These examples of multilevel research studies demonstrate the intersection of individual and place-based factors and provide context and nuance for health professionals and policymakers to better address structural and social determinants of health.

While Health Atlas can help to quickly visualize geographic differences within the framework of an ecological study, users should avoid attributing place-based factors or associations to individuals residing within those places (ie, ecological fallacy) [69]. The interpretation of descriptive results should include consideration of the time periods over which measures were derived (source years) and possible confounding factors. Moreover, place-based data cannot replace local expertise on how residents interact with places and how these various attributes influence their health.

Data available through Health Atlas are pulled primarily from publicly available sources. If the reliability and availability of emerging data change, the tool may not accurately represent the population and the places where they reside. In some cases, data were not available for certain areas or years (eg, PLACES Health-Related Social Needs, which were not derived for all states by the CDC) not relevant for certain regions (eg, ethnic enclaves, which were only generated for states with sufficient representation of Asian American or Hispanic residents), or were not generated for all states due to the complexity of the data (eg, modeled climate data such as heat). Furthermore, the use of geographic boundaries such as census tracts can limit data presentation and analysis for Health Atlas [59]. While the availability of multiple geographic units allows for flexibility, the results of an analysis may be dependent on the size, shape, or orientation of the chosen geographic unit (ie, Modifiable Areal Unit Problem) [70].

Health Atlas is a scientifically robust data repository that represents an important contribution to the landscape of place-based data available to the public. We envision that the national Health Atlas will contribute to evidence-based community-based initiatives, impactful health equity research, and effective public health programs. Future directions for Health Atlas include further integration of AI, improved accessibility, incorporation of longitudinal data, and automated reports.