On the wrong side of the tracts? Evaluating the accuracy of geocoding in public health research.

This study sought to determine the accuracy of geocoding for public health databases. A test file of 70 addresses, 50 of which involved errors, was generated, and the file was geocoded to the census tract and block group levels by 4 commercial geocoding firms. Also, the "real world" accuracy of the best-performing firm was evaluated. Accuracy rates in regard to geocoding of the test file ranged from 44% (95% confidence interval [CI] = 32%, 56%) to 84% (95% CI = 73%, 92%). The geocoding firm identified as having the best accuracy rate correctly geocoded 96% of the addresses obtained from the public health databases. Public health studies involving geocoded databases should evaluate and report on methods used to verify accuracy.