A multi-stage approach to maximizing geocoding success in a large population-based cohort study through automated and...

Area of Science:

• Epidemiology
• Geographic Information Systems (GIS)
• Public Health

Background:

• Large-scale prospective cohort studies require accurate spatial data for geographic analyses.
• Geocoding participant addresses is crucial for linking health outcomes to environmental exposures.
• Previous geocoding methods often struggle with high match rates in extensive, geographically diverse populations.

Purpose of the Study:

• To develop and implement a multi-stage geocoding protocol to maximize address-level geographic coordinate assignment for a large prospective cohort study.
• To improve the efficiency and accuracy of geocoding for spatial analyses in epidemiological research.
• To assess factors influencing geocoding success rates in a large, multi-state cohort.

Main Methods:

• A multi-stage geocoding protocol involving address parsing, cleaning, standardization, and a combination of automated and interactive tools was developed.
• Multiple base maps were utilized to enhance automated geocoding accuracy.
• Statistical analyses, including odds ratios and confidence intervals, were used to evaluate the influence of population density and data collection methods on geocoding outcomes.

Main Results:

• The geocoding protocol increased the non-Post Office (PO) Box match rate from 74.5% to 97.6%.
• Overall, 99.96% of participant addresses were geocoded, with only 5.2% resolved to the ZIP code centroid level.
• Low population density (<44 persons/km2) and in-person interviews significantly increased the likelihood of requiring resource-intensive interactive geocoding.

Conclusions:

• A comprehensive, multi-stage geocoding protocol can achieve high address-level accuracy in large-scale, geographically dispersed cohort studies.
• Population density and the method of address ascertainment are critical factors influencing automated geocoding success.
• The developed protocol significantly enhances the feasibility of spatial analyses in large epidemiological studies.