The sampling area covered 1237.6 square miles, encompassing 7% of the total combined land area in NH and VT. Towns were well-distributed by population size: rural, n=11; small town, n=7; mid-sized town, n=8; and urban, n=6. Nine hundred forty-three food outlets were identified through public directory listings, and 960 through field observations. After accounting for overlap, this provided a sample of 1340 unique food outlets. Twenty-seven percent were food markets and 73% were eating establishments. The number of food outlets per town ranged from 1 to 275. The majority of outlets were located in either urban (62.5%, n=837) or mid-size towns (25.7%, n=345); 5.5% (n=74) were located in small towns, and 6.3% (n=84) were in rural towns. Overall, only 36.9% of identified outlets (n=495) were accurately listed through public directories, and 5.1% (n=68) were mislocated. More than one-quarter (28.4%, n=380) of outlets were identified on public directories but not found during field observation. Thirty percent (29.6%, n=397) were not listed through public directories but were located in the field.
With the exception of big box stores, less than half of all outlet types were accurately listed on the public directories. Public directory accuracy differed significantly by outlet type (, P<0.001). Accuracy was highest for big box stores (62.5%), and eating establishments (43.5% fast food restaurants; 42% full-service restaurants). None of the farm/produce stands and only 35.7% of supermarket/grocery stores were accurately identified through public directories, thus omitting important community sources of fresh produce.
Accuracy of Public Directory Data to Identify Community Food Outlets by Outlet Type
Less than 50% of food outlets in all four town population groups were accurately identified through public directory data. Public data were significantly less accurate for low population towns (, P<0.001). Approximately three-quarters of the outlets in rural and small towns (68.6% and 77.3%, respectively) were inaccurately identified through public directories, compared to about 60% in mid-sized (62.4%) and urban (58.2%) towns.
Accuracy of Public Directory Data to Identify Community Food Outlets by Town Population
The accuracy of public directories versus field observations was only coded once during onsite town visits and so we did not measure whether these categories were miscoded. To minimize the chance of categorization errors, we extensively trained the coders during pre-testing, provided detailed town driving maps, located the public directory food outlets prior to the town visits, and used two-person teams for all townwide assessments. Google Earth and Yahoo! Yellow Pages utilized data from multiple commercial sources (e.g., InfoUSA) and thus it was expected that these data would be similar to that obtained through commercial databases. For current purposes, Google Earth had the added advantage of providing an efficient mechanism for downloading geographic coordinates data to create the townwide maps used during onsite field validation. It is possible that our results would have been different if we had utilized data from a primary commercial database. The secondary data gathered may also have differed if NH and VT government databases, such as those available within state Departments of Agriculture, had been utilized. However, this information is not geographically referenced and involves aggregating data from multiple reports, both of which would make data collection and verification more burdensome for local communities. Finally, this study was regionally based and so the findings may not be generalizable to other geographic areas.