Event Title

Validation of a Commercial Geographical Information Systems Database of Walking and Bicycling Destinations

Description

Background: Recent interdisciplinary studies in public health, transportation, and urban planning have shown that stores and other destinations such as banks, post offices, and physical activity facilities within close proximity to residences are positively related to recreational and transportation physical activity. The built environment has been measured several different ways, including conducting field audits and by surveying individuals’ perceptions of their neighborhood. Increasingly researchers are also using geographic information systems (GIS) software and commercially available data sources to create objective measures of the built environment. The advantages of commercial data are that they are relatively easy to access and are regularly updated. Despite these advantages it is important to assess the validity of these databases for developing measures of accessibility and density of neighborhood destinations. Two recent studies have investigated the validity of GIS databases of physical activity facilities and food stores, but to our knowledge less research has been conducted to validate a broader range of facilities that may serve as important walking and bicycling destinations.

Objective: The objective was to assess the validity of a commercially-available GIS database of facilities that may serve as walking and bicycling destinations for adults.

Methods: Researchers conducted field audits to verify the presence of 402 facilities contained in a commercial database. A list of North American Industrial Classification System codes was reviewed to identify the types of commercial facilities in the database which could serve as walking or bicycling destinations for adults. These were further categorized into five domains; food and drink (n=139), social or cultural organizations (n=115), retail establishments (n=101), services (n=28), and physical activity resources (n=19). Two high, medium, and low population density tracts in both Hartford County, Connecticut and Tippecanoe County, Indiana were selected for the analysis (12 tracts in total). Three levels of agreement were defined; 1) facilities in the database were considered to be an “exact match” if they were located on the same street segment and had the same proprietary name, 2) “close to exact match” if the facility was located on the street segment and was of the same domain, but with a different proprietary name, and 3) an “adjacent street segment match” if the facility was found to be located on an adjacent street segment. The percentages of facilities in the database that were located in the field were calculated overall, and by county, population density, and domain. Chi-square analyses were used to examine differences in match rates by county, population density, and type of facility.

Results: Overall, among the 402 facilities examined, field audits identified 67.7% were an exact match. When the ‘close to exact matches’ were included the percentage matched increased to 76.9%, and with the addition of adjacent street segments it increased to 85.8%. Percent agreement for exact matches was higher in Tippecanoe County than Hartford County (71.5% vs. 63.9%). However when all three levels of matches were included the percent agreements for the two counties were more similar (86.5% vs. 85.1%). Overall, match rates were higher in high population density census tracts than in low population density tracts (71.0% vs. 60.6%). Among the five facility domains, the exact match rates were 64.0% for food and drink establishments, 64.3% for services, 67.3% for retail establishments, 70.4% for social and cultural organizations, and 84.2% for physical activity facilities. Overall, chi-square analyses did not show statistically significant differences in match rates by county, population density, or by domain.

Conclusions: The results of this validation study demonstrated moderate to good accuracy of the commercial GIS database with more than two-thirds of the facilities correctly located in the field overall. The estimates generated in this study were similar to those in two previous validation studies of physical activity facilities and food stores which found agreement was 71%-73%. The findings in this study suggest that the commercially available GIS database provided a valid alternative to conducting extensive field audits or resident surveys.

Start Date

11-2009

Document Type

Presentation

Keywords

facilities, validation, physical activity, built environment

Session List

Poster

Share

COinS
 
Nov 1st, 12:00 AM Jan 1st, 12:00 AM

Validation of a Commercial Geographical Information Systems Database of Walking and Bicycling Destinations

Background: Recent interdisciplinary studies in public health, transportation, and urban planning have shown that stores and other destinations such as banks, post offices, and physical activity facilities within close proximity to residences are positively related to recreational and transportation physical activity. The built environment has been measured several different ways, including conducting field audits and by surveying individuals’ perceptions of their neighborhood. Increasingly researchers are also using geographic information systems (GIS) software and commercially available data sources to create objective measures of the built environment. The advantages of commercial data are that they are relatively easy to access and are regularly updated. Despite these advantages it is important to assess the validity of these databases for developing measures of accessibility and density of neighborhood destinations. Two recent studies have investigated the validity of GIS databases of physical activity facilities and food stores, but to our knowledge less research has been conducted to validate a broader range of facilities that may serve as important walking and bicycling destinations.

Objective: The objective was to assess the validity of a commercially-available GIS database of facilities that may serve as walking and bicycling destinations for adults.

Methods: Researchers conducted field audits to verify the presence of 402 facilities contained in a commercial database. A list of North American Industrial Classification System codes was reviewed to identify the types of commercial facilities in the database which could serve as walking or bicycling destinations for adults. These were further categorized into five domains; food and drink (n=139), social or cultural organizations (n=115), retail establishments (n=101), services (n=28), and physical activity resources (n=19). Two high, medium, and low population density tracts in both Hartford County, Connecticut and Tippecanoe County, Indiana were selected for the analysis (12 tracts in total). Three levels of agreement were defined; 1) facilities in the database were considered to be an “exact match” if they were located on the same street segment and had the same proprietary name, 2) “close to exact match” if the facility was located on the street segment and was of the same domain, but with a different proprietary name, and 3) an “adjacent street segment match” if the facility was found to be located on an adjacent street segment. The percentages of facilities in the database that were located in the field were calculated overall, and by county, population density, and domain. Chi-square analyses were used to examine differences in match rates by county, population density, and type of facility.

Results: Overall, among the 402 facilities examined, field audits identified 67.7% were an exact match. When the ‘close to exact matches’ were included the percentage matched increased to 76.9%, and with the addition of adjacent street segments it increased to 85.8%. Percent agreement for exact matches was higher in Tippecanoe County than Hartford County (71.5% vs. 63.9%). However when all three levels of matches were included the percent agreements for the two counties were more similar (86.5% vs. 85.1%). Overall, match rates were higher in high population density census tracts than in low population density tracts (71.0% vs. 60.6%). Among the five facility domains, the exact match rates were 64.0% for food and drink establishments, 64.3% for services, 67.3% for retail establishments, 70.4% for social and cultural organizations, and 84.2% for physical activity facilities. Overall, chi-square analyses did not show statistically significant differences in match rates by county, population density, or by domain.

Conclusions: The results of this validation study demonstrated moderate to good accuracy of the commercial GIS database with more than two-thirds of the facilities correctly located in the field overall. The estimates generated in this study were similar to those in two previous validation studies of physical activity facilities and food stores which found agreement was 71%-73%. The findings in this study suggest that the commercially available GIS database provided a valid alternative to conducting extensive field audits or resident surveys.