Validation of a GIS facilities database: quantification and implications of error

Area of Science:

• Environmental health research
• Public health surveillance
• Geographic information systems (GIS) in health

Background:

• Understanding the built environment's impact on physical activity and obesity is crucial for public health.
• Accurate spatial data on physical activity facilities is essential for this research.
• Commercial databases offer a potentially efficient source of such data, but require validation.

Purpose of the Study:

• To validate a commercial database of community-level physical activity facilities.
• To assess the suitability of this database for research on physical activity facility access and health outcomes.
• To quantify errors in facility counts, attributes, and locations within the commercial database.

Main Methods:

• Compared a commercial physical activity facility database with a field census in 80 census block groups across two US communities.
• Utilized agreement statistics (concordance, kappa) to assess count and attribute accuracy.
• Calculated Euclidean distance to quantify positional error between database and field-sourced locations.

Main Results:

• Moderate agreement was found for the presence of any physical activity facility (concordance: 0.39 nonurban, 0.46 urban).
• Agreement for specific facility types ranged from poor to moderate (kappa: 0.14–0.76).
• Mean positional error was substantial (757m nonurban, 35m urban), though most facilities fell within the same ZIP code (94% nonurban, 100% urban).

Conclusions:

• The commercial database exhibits significant errors in facility presence, type, and location.
• These errors likely introduce downward bias in built environment-health association studies.
• Careful consideration of data limitations is necessary when using this database for epidemiological research.