Medical GIS: New Uses of Mapping Technology in Public Health Peter Hayward, PhD Department of Geography SUNY College at Oneonta Invited research seminar presentation at Bassett Healthcare. Cooperstown, NY, December 15, 2009
Personal History
Personal History Education B.A., Geography: Indiana University of Pennsylvania M.A., Geography: University of California Santa Barbara PhD, Geography: University of Connecticut Current Position Assistant Professor Department of Geography SUNY College at Oneonta
Structure of Presentation Medical GIS The process of applying specific geospatial techniques in geographic information systems (GIS) to approach issues in public health
Modeling Service Areas of Health Care Providers Purpose Use GIS to model the current and potential service areas of Connecticut Children s Medical Center (CCMC) and Yale New Haven Hospital (YNHH)
Modeling Service Areas of Health Care Providers Data CT Hospital Information Management Exchange (CHIME) data on hospital admissions U.S. Census population projection data, streets data Methods Population: Model market areas based on percentage of admissions by town for CCMC versus YNHH Distance: Model CCMC market area based on timed road distance Software ArcGIS 9.3
Population Based Service Areas Primary Catch greater than 50% of all hospital admissions from that town CCMC Primary Service Area YNHH Primary Service Area
Population Based Service Areas Secondary Catch greater than 30% of all hospital admissions from that town CCMC Primary Service Area YNHH Primary Service Area CCMC Secondary Service Area YNHH Secondary Service Area
Population Based Service Areas Shared Nearly equal catchment of hospital admissions from that town CCMC Primary Service Area YNHH Primary Service Area CCMC Secondary Service Area YNHH Secondary Service Area CCMC-YNHH Service Area Split
Population Based Service Areas Pediatric Population Projections U.S. Census data projections used to determine current and future market population CCMC primary market area is projected to have a stable population CCMC secondary market area is projected to have a declining population Pediatric (Ages 0-19) Population Projection for CCMC and YNHH Service Areas, 2000-2020 Source: Connecticut State Data Center based on 2000 U.S. Census data (http://ctsdc.uconn.edu/projections-towns/townlist-css.html).
Distance Based Service Areas 10 Minutes
Distance Based Service Areas 20 Minutes
Distance Based Service Areas 30 Minutes
Distance Based Service Areas 40 Minutes
Distance Based Service Areas 50 Minutes
Distance Based Service Areas 60 Minutes
Advanced Service Area Analysis Primary Catch greater than 50% of all hospital admissions from that town 20 minute service distance CCMC Primary Service Area YNHH Primary Service Area
Advanced Service Area Analysis Questions What is unique about those towns not defined as a primary service area but within a 20 minute service distance? CCMC Primary Service Area YNHH Primary Service Area
Conclusions Overall Using GIS, hospital market areas can be analyzed and predicted in population or distance based models Other Variables to Consider Income and poverty, demographic characteristics, urban versus rural status
Structure of Presentation Medical GIS The process of applying specific geospatial techniques in geographic information systems (GIS) to approach issues in public health
Towards a Preferred District Design of Health Disparities Purpose Use GIS to map health disparities at different scales and zones in Connecticut
Mapping Minority Health Disparities Small scale maps obscure the local patterns of disparities Large scale maps may not reveal the full extent of disparities Brewer 2009, pg. 8 Chen et al. 2006, pg. 10
(Re)Districting Problem Brewer 2009, pg. 8 Chen et al. 2006, pg. 10
The Modifiable Areal Unit Problem (MAUP) The interpretation of a spatial phenomenon depends on the number and arrangement of areal units imposed on the map Zoning Effect
Towards a Preferred District Design of Health Disparities Data 1985 2004 Connecticut geocoded mortality data 485,930 White deaths vs. 35,580 Black deaths U.S. Census map files with population data (blocks counties) Methods Map disparities in mortality rates between Whites and Blacks at many scales and zones Software ArcGIS 9.3
Scale Effect TOWNS COUNTIES CENSUS TRACTS White Disparities Black Disparities BLOCK GROUPS
Zoning Effect TOWNS REDISTRICTING PLAN A COUNTIES White Disparities Black Disparities REDISTRICTING PLAN B
Conclusions Overall The locations, sizes, and significance of health disparities in Connecticut is dependent on the scales and zones used to map the problem Implication The reporting of minority health disparities within existing district designs may be based on sound measurements, but the findings may not be reliable
Structure of Presentation Medical GIS The process of applying specific geospatial techniques in geographic information systems (GIS) to approach issues in public health
Geospatial Analysis of Childhood Pedestrian Collisions Purpose Use GIS to model the spatial context of childhood pedestrian collisions in Hartford, CT
Existing Research Braddock et al. 1991, pg. 1243-1244 Problem Grouping neighborhoods takes the space out of a spatial analysis; and different areas have different underlying child populations
Geospatial Analysis of Childhood Pedestrian Collisions Data Dependent Variable: frequency of childhood pedestrian collisions (90 total) determined by Hartford census tract, 2005 2006 Independent Variables: contextual variables by census tract (poverty, education, schools, public buildings, housing characteristics, children per square mile) Methods Model statistically significant clusters and perform spatial regression analysis while accounting for underlying child population Software GeoDa
Data Capture
Identify the Locations of Collisions
Incorporate Child Population Data
Identify the Frequencies of Collisions High Frequency Moderate Frequency Low Frequency
GeoDa Mapping Taking underlying child population values into account, use GIS to determine statistically significant clusters of childhood pedestrian collisions
Spatial Clusters
Identify Correlating Factors and Formulate Hypotheses Poverty ~ Collision Frequency Parks ~ Collision Frequency Schools ~ Collision Frequency Police ~ Collision Frequency Housing ~ Collision Frequency Context of high frequency neighborhoods versus moderate and low frequency areas.
GeoDa Statistics Spatial regression analysis indicates significant associations with the number public buildings, percent female headed households, education, race, and home ownership
Conclusions Overall GIS is an effective tool to investigate the spatial context of childhood pedestrian collisions Implication GIS could be used to investigate the spatio-temporal characteristics of childhood pedestrian collisions, which has important ramifications for injury prevention
Talk Summary Medical GIS The process of applying specific geospatial techniques in geographic information systems (GIS) to approach issues in public health Other Applications Epidemiological Patient Location and Care Environment System BodyViewer Personal Goals Exploring the link between environmental factors and obesity in New York
Medical GIS: New Uses of Mapping Technology in Public Health Peter Hayward, PhD Department of Geography SUNY College at Oneonta Phone: 607-436-3398 Email: haywarpm@oneonta.edu Web: http://employees.oneonta.edu/haywarpm/