Positional accuracy of geocoding from residential postal codes versus full street addresses.

BACKGROUND: Postal codes are often the only geographic identifier available for assigning contextual or environmental information to a study population. This analysis assesses the influence of three factors-delivery mode type (mode of postal delivery), representative point type (source of latitude-longitude coordinates), and community size-on the accuracy of postal code spatial assignment. DATA AND METHODS: PCCF+ (Postal Code Conversion File Plus) was used to assign delivery mode type, representative point type and community size to each individual in the 2011 Census of Canada. A sample (n = 1,004) was randomly selected with a minimum of 90 observations for each category of those three factors. Based on the address information of individuals in the sample, measures of positional accuracy for geocoding from residential postal codes (PCCF+) versus reference locations as determined by full street addresses (Google Maps) were calculated using a geographic information system. Accuracy was measured as the distance that the geocoded position differed from the full street address.
RESULTS: Positional accuracy was related primarily to mode of postal delivery. Rural and mixed (partly urban, partly rural) modes had much higher geocoding error than did urban modes. Rural and small-town Canada and latitude and longitude based on dissemination area centroids had low accuracy, largely because of their close relationship to rural and mixed modes of delivery. DISCUSSION: The accuracy of geocoding from postal codes can vary. Geocoding imprecision may result in misclassification, depending on the spatial resolution of the environmental or contextual measures. The spatial resolution required for a study helps to identify subpopulations that should be excluded because of inadequate positional accuracy.