A Framework for Automated and Reproducible Geomarker Curation and Computation at Scale

Erika Rasnick, Andrew Vancil, Cole Brokamp


Division of Biostatistics and Epidemiology, Cincinnati Children’s Hospital Medical Center
College of Medicine, University of Cincinnati

Geocoding and Geomarker Assessment

  • The role of the environment in health has been known for decades, but the data and tools needed to characterize the relationship at a population level have only recently become available.

  • Geocoding: Converting a string of text into latitude and longitude coordinates

place (+ time) → estimated past “exposures”

  • Geomarker: Similar to biomarkers in precision health and defined as any objective, contextual, or geographic measure that influences or predicts the incidence of outcome or disease

    • community deprivation, crime, cultural isolation indices
    • census, surveys
    • airborne pollutants
    • landcover classification (Figure 1a), greenspace (Figure 1b)
    • access to healthcare (Figure 1c), healthy food, pharmacies, transportation
    • noise, heat, walkability, parks, proximity to roads/traffic (Figure 1d)
Geomarker Examples

Figure 1: Geomarker Examples

Privacy and Reproducibility in Health Studies

  • Confidentiality of patients and research subjects must be safeguarded (HIPAA privacy rule, HITECH Act of 2009, and Federal Policy for the Protection of Human Subjects)
  • Address (and equivalent geocode) is considered Protected Health Information (PHI)
  • Presents challenge when integrating geocoding and geomarkers into research studies and clinical applications
    • Most existing IRB protocols and institutional data use agreements prevent transmission of PHI over the internet (e.g. Google, ArcGIS Online, Texas A&M Geocoder)
    • IRB approvals are lengthy and may not even be possible depending on original consent forms
    • Different methods introduces differential error, resulting in biased associations between geomarkers and health outcomes
    • Commercial options are cost prohibitive and are not designed for operations at scale
    • Closed source geocoder prevents transparency and reproducibility

DeGAUSS


DEcentralized Geomarker Assessment for mUlti Site Studies
DeGAUSS Workflow

Figure 2: DeGAUSS Workflow

  • Decentralized approach for geocoding and deriving community and individual level environmental characteristics
  • Reproducible and standardized geomarker assessment while preserving the privacy of PHI (never exposed to the internet)
  • Software based on a container framework that reads and writes CSV files
  • Does not require extensive computational resources or extensive knowledge of command line and/or geospatial computing
image description
degauss/geocoder batch geocoding
degauss/cchmc_batch_geocoder geocoding, census tract, deprivation index
degauss/census_block_group census block group FIPS
degauss/roads proximity and length of major roads
degauss/greenspace enhanced vegetation index
degauss/geocoder_slim API returning geocoded text string as JSON
degauss/PEPR_drivetime distance and drive time to clinical care sites
degauss/schwartz_grid_lookup schwartz grid for spatiotemporal pollutant models
degauss/schwartz daily PM2.5, NO2, and O3 concentrations

Applications

  • DeGAUSS is open source and can be customized for applications specific to different studies
  • Limited use agreement or deidentified dataset (other healthcare systems, government agencies)
  • Sharing exposure assessment models
  • Electronic health record and registry integration

Development

The setup shown below utilizes Docker and R to build DeGAUSS software that is self-contained and does not require passing PHI to the internet, but a similar workflow can be employed using other programming languages and containerization software.

DeGAUSS Development

Figure 3: DeGAUSS Development

  • R script: code that performs geospatial task
  • Dockerfile: instructions to set up the container including installation of operating system, system dependencies, required geospatial datasets, R, and required R packages
  • renv lockfile: documents sources and versions of R packages required to run the R code

Summary


DeGAUSS is a secure, decentralized, and reproducible approach to solving geospatial privacy issues in multi-site studies, but it also excels as a vehicle for making spatiotemporal exposure assessment prediction models findable, accessible, interoperable, and reusable (FAIR).

Contact

  • Free and open source development
  • Ongoing user feedback

https://degauss.org

@degauss-org

@erika_rasnick

References


Cole Brokamp, Chris Wolfe, Todd Lingren, John Harley, Patrick Ryan. Decentralized and Reproducible Geocoding and Characterization of Community and Environmental Exposures for Multi-Site Studies. Journal of American Medical Informatics Association. 25(3). 309-314. 2018. Download.

Cole Brokamp. DeGAUSS: Decentralized Geomarker Assessment for Multi-Site Studies. Journal of Open Source Software. 2018. Download.