Modeling travel distance to health care using geographic information systems

1
Modeling travel distance to health care using
geographic information systems

• Anupam Goel, MD
• Wayne State University
• Detroit, MI (USA)

2
Authors background

• Initially, a research project
• Determine the distance to the closest mammogram
  center for Vermont women ages 40 and older
• Applications to other public health settings

3
Learning objectives

• Define GIS
• Recognize potential data sources for GIS projects
• Recognize some strengths and limitations of using
  GIS technology

4
Performance objectives

• Recognize variations in measuring geographical
  access
• Critically review an article using GIS

5
Geographic Information Systems (GIS) overview

• A method to visualize, manipulate, analyze, and
  display spatial data, information linked to a
  specific place
• Additional description of GIS

6
Geographic Information Systems (GIS) overview
(cont.)

• Can include spatial data from many sources
• Applications include environmental modeling,
  government or military uses, and business
  forecasting

7
Software choices

• Available GIS programs
• This presentation uses ESRI
  software, namely ArcView 3.2a
  and Network Analyst
  1.1b

8
Methods to measure access

• Distance to closest facility
• Straight-line
• Driving distance
• Number of facilities within a given distance
• Travel across political boundaries

9
Using this methodology for mammography utilization

• All relevant mammography facilities
• Assign women to representative points throughout
  the state
• Road atlas
• The shortest road distance from each group of
  women to a facility

10
Mammography facilities

• Mammography Facility Registry
• Subset of mammography facilities within Vermont
  and the surrounding counties
• Mobile mammography centers and Canadian centers
  not included

11
Estimating a womans location

• Eligible women within each Vermont ZIP code
  (Claritas, Inc.)
• Assigned these women to the ZIP code population
  centroid (Geographic Data Technologies, Inc.)

12
Estimating a womans location (cont.)
X
X
X
X
P
X
X
G
X
X
X
X
Residence
X
G
Geographic centroid
P
Population centroid
13
Road network

• Census 2000 county road networks
• All counties within Vermont (n14)
• The US counties surrounding Vermont (n10)
• Canadian roads were not included

14
Distance to closest facility

• Applied mathematics
• Graph theory
• Network optimization
• Dijkstras algorithm a method to find the
  shortest path from a node to all other nodes
  connected by a network

15
What we found

• Median distance to travel for a mammogram in
  Vermont was 11.2 km (range, 0.5-49.1 km)
• Women in the most populated ZIP codes traveled
  less for a mammogram than women in the least
  populated ZIP codes

16
Limitations

• Mammography from work instead of from home
• Mobile facilities not included
• No women surveyed for their actual driving
  distance
• Driving distance, not driving time

17
Implications of this project

• Two ways to place new facilities
• 1) Reduce the longest distances traveled (place
  new facilities in rural areas)
• 2) Reduce the average distance Vermont women
  travel (place new facilities in urban areas with
  less access to mammography)

18
Next directions

• New data sources
• Census 2000, Utilization files
• New questions
• Utilization in other areas, targeting
  interventions
• New analytic approaches
• Adjusting for covariates, spatial statistics

19
Acknowledgements

• University of Vermont
• Benjamin Littenberg, Richard G. Pinckney,
  Division of GIM
• Austin Troy, SNR
• Berta Geller and VBCSS
• National Cancer Institute funding
• 3 P30 CA22435-17S3 and 1 R03 CA101493-01