Table of Contents

• CLAD Hexagonal Environmental Linkage (CHEL)
• Hexagonal Aggregation
• Information on Geocoding Processes
• CHEL Dataset FAQs

About CHEL

The CLAD Hexagonal Environmental Linkage (CHEL) is a novel data asset developed by the Center for Linkage and Acquisition of Data (CLAD) to enable linkage of environmental, social, and health vulnerability indicators to participant-level records in the All of Us Research Program dataset. CHEL provides a comprehensive environmental context by aggregating over 30 geospatial indicators into hexagonal grids, harmonizing data from multiple federal sources to support environmental health and equity research.

CHEL is structured to facilitate linkage at a geographic resolution that preserves participant privacy, and is designed to align with the All of Us commitment to enabling rigorous, reproducible, and representative research. Each CHEL hexagon integrates data related to environmental burden (e.g., air and water pollution, industrial proximity), social vulnerability (e.g., housing instability, income, educational attainment), and health vulnerability (e.g., prevalence of chronic conditions like asthma or diabetes), offering a cumulative risk perspective for over 693 Hex3 geographies across the continental United States.

The inclusion of CHEL in All of Us marks a significant expansion in the types of confounding and effect-modifying factors that researchers can measure and model. It enables deeper epidemiologic insights into how environmental and social contexts intersect with genetic, clinical, and behavioral data. This is particularly crucial for understanding place-based contributors to health disparities, chronic disease burden, and longitudinal health trajectories.

Summary of data sources

CHEL includes data from the following data sources. Researchers using these elements may wish to consider the listed known limitations of each source; see the CHEL data dictionary and technical documentation for further details.

U.S. Census Bureau – American Community Survey (ACS)

• Key Indicators: Minority status, poverty, education, employment, housing characteristics, internet and insurance access, age demographics, disability, language, and mobile housing.
• Limitations: Imprecise income estimates in central city neighborhoods and lower-quality income data in very poor or very rich neighborhoods.

U.S. Environmental Protection Agency – Air Quality System (AQS)

• Key Indicators: Ozone and PM2.5 concentrations.
• Limitations: Primarily urban data; limited rural coverage.

EPA – National Air Toxics Assessment (NATA)

• Key Indicators: Diesel particulate matter, cancer risk from air toxics.
• Limitations: Outdoor data only, does not capture high concentrations, some interpolations that may conflict with observations.

TomTom MultiNet® Enterprise Dataset

• Key Indicators: Proximity to parks, roads, railways, airports.
• Limitations: Better urban than rural coverage.

EPA – Facility Registry Service (FRS)

• Key Indicators: Hazardous sites (e.g., Superfund, TRI, RMP).
• Limitations: The reported geographic coordinates for some facilities are only accurate within a radius of more than 50 meters. A 50-meter inaccuracy is small compared to the size of a Hex3 cell.

EPA – National Walkability Index

• Key Indicators: Walkability access.
• Limitations: Among children in both rural and urban settings, walkability indices show no significant association with objectively measured physical activity or self-reported walking behavior.

EPA – Watershed Index Online (WSIO)

• Key Indicators: Impaired surface water quality.
• Limitations: does not include specific contaminant concentrations or ecological data needed to assess potential human exposure within a given Hydrologic Unit Code.

CDC – PLACES Estimates

• Key Indicators: Asthma, cancer, hypertension, diabetes, mental health prevalence.
• Limitations: Florida and New Jersey data are missing for some years.

U.S. Mine Safety and Health Administration – Mine Data Retrieval System (MDRS)

• Key Indicators: Locations of coal and lead mines.
• Limitations: Focused on specific mine types; lacks broader industrial exposure data.

Two versions of the CHEL dataset are available: one for 2022 and one for 2024. However, data quality issues in the 2024 version—specifically, unexpected distributions and low variability in the PM2.5 and Ozone indicators—prevented inclusion of these air quality variables. As a result, the 2024 CHEL dataset contains fewer environmental variables than the 2022 version.

Hexagonal Aggregation

Hex3 geography details 

To protect participant privacy while enabling meaningful environmental linkage, the CLAD team selected Hex3, a hexagonal gridding system with a fixed spatial resolution. Sufficiently large geographic units are required to meet reidentification risk thresholds, particularly when linking sensitive environmental data with health records. Hex3 provides a consistent, balanced unit that supports both privacy preservation and enough detail for most spatial analyses.

Figure 1 illustrates the relative spatial scale of Hex3 geographies compared to other common geographic units, including smaller hexagonal grids (e.g., Hex5, Hex7), ZIP codes, and census tracts. Hex3 maintains a consistent area of approximately 12,300 km², providing a balanced unit for spatial linkage that supports both data utility and privacy. Alternatives such as ZIP codes and census tracts vary widely in size and shape depending on population density or postal routing needs, often leading to inconsistent spatial resolution across regions. Hex-based systems offer geometric regularity, making them ideal for environmental data integration and comparative research.

Overall, Hex3 geographies offer a methodologically sound and privacy-aware approach to integrating environmental context into participant-level health data, making them ideal for longitudinal, multilevel, and place-based health research.

Information on Geocoding Processes

Source of participant location

All addresses used in this project originate from participant-provided address history and are linked to participant consent records, update events, and specimen records. Due to privacy policies, the use of commercial address verification software is restricted, which necessitated the development of custom internal processes for address cleaning and validation. As part of this process, quality flags are applied to each component of an address in accordance with USPS Publication 28 postal address formatting standards. These flags help assess formatting consistency and enhance the accuracy of downstream geocoding efforts.

Geocoding pipeline

To support linkage with CHEL variables, cleaned addresses were geocoded using both DeGAUSS and PostGIS. DeGAUSS leverages U.S. Census Bureau TIGER files, while PostGIS is an open-source spatial extension for PostgreSQL that also utilizes TIGER data.

The workflow included initial data cleaning and preprocessing to identify and flag formatting issues, followed by deduplication and generation of geocodes for addresses that passed conformance checks. A second geocoding run was performed for addresses that failed the first run or required correction based on post-hoc quality flags.

Each address was evaluated for spatial and temporal consistency, and final geocodes were reconciled by person, location, and time. Geocoding results from DeGAUSS and PostGIS were compared to assess agreement, and additional evaluations were conducted at the ZIP3 and state level to confirm geographic accuracy before delivery to the Environmental Data Linkage Team.

CHEL Frequently Asked Questions (FAQs)

1. What is the CLAD Hexagonal Environmental Linkage (CHEL)?

CHEL is a geospatial dataset that links environmental and social determinants of health to participant records using hexagonal geographic units (Hex3, ~12,300 km²). It is designed to integrate with All of Us data for research on place-based health influences.

2. Why was Hex3 used instead of census tracts or ZIP codes?

Hex3 provides a consistent spatial resolution across urban and rural areas, which supports uniform analysis and improves spatial modeling—an important consideration for environmental exposures that are not inherently tied to administrative boundaries.  Larger units were also required to meet participant privacy requirements.

3. What are the sources of the environmental data?

CHEL incorporates variables from federal sources, including the U.S. Census Bureau, EPA, CDC, and others. These include measures of environmental burden (e.g., pollution, land use), social vulnerability, climate burden, and health vulnerability.

4. How were addresses geocoded?

Addresses from the Data and Research Center (DRC) address history were geocoded using two tools: DeGAUSS (Docker-based, TIGER file-enabled) and PostGIS (open-source spatial database extension), both leveraging U.S. Census TIGER data for consistency and accuracy.

5. What versions of CHEL are available?

There are currently two versions: CHEL 2022 and CHEL 2024. Each version reflects the best available data for its respective time period.

6. Why do the 2022 and 2024 datasets differ in environmental variables?

Due to data quality concerns in the 2024 update—specifically unexpected distributions and low variability in PM2.5 and Ozone measures—these air quality indicators were excluded. Therefore, fewer environmental variables are present in the 2024 dataset.

7. How should I interpret missing environmental variables in the 2024 dataset?

Missing variables reflect intentional exclusion based on quality control criteria. Researchers are encouraged to use the 2022 dataset for analyses requiring air quality indicators.

8. Can I compare CHEL variables across versions?

Caution is advised when comparing across versions due to potential changes in variable availability, measurement methods, and geographic data updates. Consult the CHEL data dictionary and version documentation for details.

9. How is participant privacy maintained in this dataset?

CHEL uses spatial generalization (Hex3 units) and omits direct identifiers. Additionally, CHEL applies deterministic geographic jittering to protect participants located in sparsely populated Hex3–ZIP3 intersections. When the estimated population is below 20,000, the Hex3 is reassigned to an eligible Hex3 within the same ZIP3 or censored if no eligible Hex3 is available.