1. Adding Value to Registries through Geospatial Big Data Fusion Geospatial Health Context Big Table Facilitating Geospatial Analysis in Health Research
Tim Haithcoat & Chi-Ren Shyu
University of Missouri Informatics Institute
June 13, 2019

3. THE GOAL
Develop robust processes for health researchers and practitioners to more easily incorporate spatially integrated health, social, cultural, access, infrastructure, and environmental parameters/factors and spatial context in their research using scalable geospatially enabled databases, analytics, and visualizations.

4. Unique Infrastructure
Typical Relational DB
Typical Geospatial DB
Talk about current state-of-the-art and shortcomings of raster and vector approaches
Why designed this way. Column based – why not row based?
Why did we pick this method and design;

5. Tessellation over Census blocks
Block centroids = 343,565 points

6. Thiessen Proximal Polygons
Tessellation with Census Centroids
Thiessen Proximal Polygons

7. Extent of the Data Table
Defined a point file with 318 million points for contiguous 48 states.
How many columns (attributes)? Projection  10,000+
How many data sets? US Data.gov – Federal GIS > 1,000
What is the size of the table? 1.5 Gb/attribute Growth Projection  90 Tb
Using Spark big data ecosystem
Australian Cancer Atlas
Determined Main Common Keys
Census Geography
Zip Code
Watershed
Etc.
Created point summary counts for all geographies to use for analytics
How deal with various resolutions and scale across database
How & why set up regions and to speed query and retrieval
How mapped? KD Tree: how make it work and make it efficient

8. Establishing Context Inter-layer Distance measures
Coded 1st & 2nd Order Relationships

9. Registry Data Loading
Registry Data
Records

10. Leveraging Geospatial in Registries
Geocoding of Registry
Attach an X,Y coordinate to each record with associated confidence (strongest)
Attach a primary key(s) (i.e. Census ID, Zip Code Tabulation Area) based on geocode of address to create ‘easy’ linkage to associated data when needed.
Use geocoded location to determine association with a primary key to move attributes of interest directly to the registry record.
Determine what information, and at what geographic summarization level, registry data gets shared

11. Using the Big Data Table
Geospatial Health Context Big Table
Data Required
Socio-Economic
Demographic
Infrastructure
Environmental
Cultural
Derived
Physical
Modeled
LIFESTYLE
50%
HEALTH
CARE
25%
BIOLOGY
15%
ENVIRON
10%
User Data
Address
Zip Code
Tract
County
Inquiry Type
Exploratory
Simple Question
Complex Question
Complex Question w Temporal
Aggregation Unit
Zip Code Tract
Block Group County
Watershed School Dist
Health Service Area

13. Choose an Issue
Right-Sizing Care: Over the next decade, the aging American population is expected to place increased demands on the U.S. healthcare system. For older Americans, a review of medical records, found that 38% of doctor visits, including 27% of Emergency Room (E.R.) visits could have been replaced with telemedicine.
Effort Required Census data tables (2 hrs) Census geography (1 hr) Hospital types (2 hrs) Road network zones (time and/or distance) (1 week) Broadband type (2 hrs)
Query Elements
Age > 60 years
Gender
Hospital Service Area
Broadband Service
The Data Needed
Census age & gender
Hospital locations
Attributed road network
Broadband attributes
Census geography
Select & retrieval
MAUP can be addressed simply
Key geographies

14. GeoHCBT: A case study of Leukemia

15. Example Complex Questions
What factors in different demographic groups or locations discourage people from cancer treatment?
How can we update our healthcare delivery strategy based on availability of medical services with relation to cancer risk based on population growth, ageing, and cancer type?
Can we identify any new relationships between cancer occurrence and environmental, socio-cultural, infrastructural, or other data to explore or generate new hypotheses?
What is the magnitude of population cancer disparities in an area, where are they located, and what factors might be creating these ‘hot spots’?

16. Relevance The Geospatial Health Context Big Table provides:
Cancer Researchers an integrated big data repository to:
Search - Enable stronger research designs (i.e. develop sampling / surveillance approached).
Explore - Understand spatial interaction of a multitude of attributes.
Ability to add contextual information based on neighborhood
Decision Makers with a new tool to evaluate policy implications and focus on areas / populations affected.
Public Health Professionals an ability to identify, mitigate, and potentially prevent health disparities in cancer incidence.

17. Acknowledgments Looking for research collaborations:
Collaborators:
Chi-Ren Shyu, PhD Richard D. Hammer, M.D.
Tim Matisziw, PhD Iris Zachary, PhD
Eileen Avery, PhD Kelly Bowers, D.O.
Mirna Becevic, PhD
This work is supported by the NIH BD2K T32 Training grant (5T32LM )
The Big Data ecosystem is supported by the NSF CNS
Looking for research collaborations:
Contact: