Estimating the accuracy of geographical imputation

“Background The process of assigning geographic information based on a street address, commonly referred to as geocoding, is increasingly being used in health research to assess geographic clustering or associations between health outcomes and area-based socioeconomic and/or environmental characteristics [1-4].”

“Imputation techniques are well established as a general useful method for the analysis of data with missing values and are included in many popular statistical packages [23,24].”

“Previous research covering population interpolation, areal interpolation, and dasymetric mapping of socioeconomic data could be adapted to allow for more accurate geographical imputation [36,40-43].”

“To reduce the number of non-geocoded cases and thus limit potential geographic selection bias, researchers and organizations sometimes include cases geocoded to postal code centroids along with cases geocoded with the greater precision of a full street address [17-21].”

“Ratcliffe (2004), for example, using crime data, tested what minimum number of geocodes was necessary to produce a statistically accurate map [44].”

“past, researchers are becoming more aware that either strategy may lead to geographic selection bias and misclassification of case attributes [3,5-9].”

“Recent studies have shown that individual addresses not able to be geocoded to a full street address are not random and are more likely to be located in rural areas because of the disproportionate number of rural route addresses, post office box addresses, unofficial street names and streets missing from geocoding ref…”

“Background The process of assigning geographic information based on a street address, commonly referred to as geocoding, is increasingly being used in health research to assess geographic clustering or associations between health outcomes and area-based socioeconomic and/or environmental characteristics [1-4].”

“Recent studies have shown that individual addresses not able to be geocoded to a full street address are not random and are more likely to be located in rural areas because of the disproportionate number of rural route addresses, post office box addresses, unofficial street names and streets missing from geocoding ref…”

“(2005) refer to maps that incorporate geocoding bias as having 'cartographic confounding' and caution that results from spatial analysis may reflect the geographic "distribution of the available data rather than real, underlying disease patterns" [8].”

“To reduce the number of non-geocoded cases and thus limit potential geographic selection bias, researchers and organizations sometimes include cases geocoded to postal code centroids along with cases geocoded with the greater precision of a full street address [17-21].”

“In Canada and other countries with highly precise postal codes, this is typically the case, as the postal code itself offers precision to a block-face or a single large apartment building [39].”

“To reduce the number of non-geocoded cases and thus limit potential geographic selection bias, researchers and organizations sometimes include cases geocoded to postal code centroids along with cases geocoded with the greater precision of a full street address [17-21].”

“Recent studies have shown that individual addresses not able to be geocoded to a full street address are not random and are more likely to be located in rural areas because of the disproportionate number of rural route addresses, post office box addresses, unofficial street names and streets missing from geocoding ref…”

“Recent studies have shown that individual addresses not able to be geocoded to a full street address are not random and are more likely to be located in rural areas because of the disproportionate number of rural route addresses, post office box addresses, unofficial street names and streets missing from geocoding ref…”

“Recent studies have shown that individual addresses not able to be geocoded to a full street address are not random and are more likely to be located in rural areas because of the disproportionate number of rural route addresses, post office box addresses, unofficial street names and streets missing from geocoding ref…”

“past, researchers are becoming more aware that either strategy may lead to geographic selection bias and misclassification of case attributes [3,5-9].”

“To reduce the number of non-geocoded cases and thus limit potential geographic selection bias, researchers and organizations sometimes include cases geocoded to postal code centroids along with cases geocoded with the greater precision of a full street address [17-21].”

“past, researchers are becoming more aware that either strategy may lead to geographic selection bias and misclassification of case attributes [3,5-9].”

“Background The process of assigning geographic information based on a street address, commonly referred to as geocoding, is increasingly being used in health research to assess geographic clustering or associations between health outcomes and area-based socioeconomic and/or environmental characteristics [1-4].”

“Previous research covering population interpolation, areal interpolation, and dasymetric mapping of socioeconomic data could be adapted to allow for more accurate geographical imputation [36,40-43].”

“They related town boundaries to the census tract boundaries and randomly assigned tracts to cases based on the proportion of the town's female population accounted for by each census tract [25,26].”

“Recent studies have shown that individual addresses not able to be geocoded to a full street address are not random and are more likely to be located in rural areas because of the disproportionate number of rural route addresses, post office box addresses, unofficial street names and streets missing from geocoding ref…”

“Huang et al. (2007) assigned random locations within imputed census blocks for Los Angeles cancer cases used in geographic analysis of cancer survival [29].”

“Background The process of assigning geographic information based on a street address, commonly referred to as geocoding, is increasingly being used in health research to assess geographic clustering or associations between health outcomes and area-based socioeconomic and/or environmental characteristics [1-4].”

“2%) to imputed census-tracts based on the postal delivery area to avoid potential geographic selection bias resulting from excluding cases not successfully geocoded [30].”

“suggesting that the above explanation is unlikely. A related concern about this study is that the inclusion of cases (18%) that lacked exact street addresses might result in potential misclassification of CT level characteristics, especially among rural CTs. Alternative approach such as “geo-imputation” would increase…”

“%) that were geocodable to a zip code level only, which were allocated by a random assignment based imputation method (Henry and Boscoe, 2008; Hibbert et al., 2009) rather than the traditional zip code centroid method (Cayo and Talbot, 2003) because the latter would create spurious clusters of…”

“3%) using geographical imputation (Henry and Boscoe, 2008).”

“We assigned Black race to cases residing in all ZIP codes with > 30% Black residents and cases in all other distributions were considered non-Hispanic White.(6) We used 2010 census data to describe median household income, race, and ethnicity distributions by ZIP code.”

“Population or socioeconomic status–weighted geoimputation may improve the accuracy of the travel impedance estimates (Henry and Boscoe, 2008).”

“Finally, it can facilitate and improve measurement-error model methods and imputation methods for adjusting spatial statistical analyses for geocoding inaccuracy [24-26].”

“Henry and Boscoe [17] saw a similar accuracy of 25.9% using total population as a weighting mechanism.”

“…increasing in accuracy and parcel files are becoming more available every day, many recent reports continue to find high levels of non-street level matches (Henry and Boscoe 2008, Hibbert et al. 2009, Zandbergen 2008a) for which a USPS ZIP code or city would typically be the next selection.”

“To obtain cancer data with a high resolution and reduce spatial aggregation error, Henry and Boscoe (2008) used demographically-based geo-imputation to assign cancer cases from zip codes to census tracts [15].”

“All geocoded cases were assigned a 2010 census tract, and the remaining cases not geocoded ( n = 5) were assigned a census tract based on their zip code using geographic imputation methods described previously by Henry and Boscoe (42).”

“…been shown to be dependent on community characteristics such as population density as well as age, race/ethnicity, and socioeconomic status at both the community level (Henry and Boscoe 2008, Oliver et al. 2005) and the individual level (Gregorio et al. 1999), thereby indicating potential for bias.”

“Since then, a large amount of research has focused on georeferencing domestic disease occurrences [1,11,72-74].”

“Coarse locational data may be combined with the observed point-level data to make valid statistical inferences in the presence of geographic bias via a coarsened-data maximum likelihood estimation procedure [36], or it may be used in concert with demographic information to impute a surrogate point location for each ad…”

“Versions of these strategies were originally developed and applied by Klassen et al. 2004 [11] and are similar to those presented by Boscoe 2008 [25] and Henry and Boscoe 2008 [33], see also Zimmerman 2008 [34] for related discussions.”

“Henry and Boscoe (2008) cite eight studies that used the demographic characteristics of finer level census geographies to assign geographic locations to cases for which only their larger geographic areas were known [28].”

“The 15% (300) of addresses that were geocodable to zip code level only were allocated to Census tract by a random assignment based imputation method and process which was weighted by demographic information (Henry & Boscoe, 2008; Hibbert et al., 2009).”

“…national and international information infrastructures and determining geographical location features (Boqiu et al., 2010; Coetzee and Bishop, 2009; Coetzee and Cooper, 2007; Coetzee et al., 2008; Geymen et al., 2008; Henry and Boscoe, 2008; Lind, 2008; Zandbergen, 2008; Zandbergen, 2010).”

“Such imputation or geocoding at a coarse spatial resolution can introduce geographic bias/positional errors in the exposure classification (Henry and Boscoe, 2008; Zandbergen, 2009).”

“[10, 19]).”

“Geographical imputation (Henry & Boscoe, 2008; Sheehan et al., 2004) was used to allocate each of the 24,100 cases to a census tract based on the relative proportion of the ZIP code’s male population accounted for by each census tract for the decade when the case was diagnosed.”

“…Federation and the Länder (2016) Figure 5.3: Spatial Linking by Location with Immigrant Rates Data This method of spatial linking yet is prone to inaccuracy errors for the following reasons: First, geospatial data can be inaccurate through measurement errors or missing data (Henry & Boscoe, 2008).”

“For the remaining records, we used the ZIP code or town or county along with race, ethnicity, and age to impute the 2010 census tract following the method of Henry and Boscoe [20].”

“of cancer rates estimated via areal interpolation [32, 34].”

“The predictions are based on a Bayesian space–time hierarchical model that combines (1) observed PM2.”

“Since street blocks can support multiple complexes, the opportunity for overlap between centroids when precision is broadened is quite high and explains the significant drop in unique coordinates.(178,180) Conversely, in rural communities where individual postal codes span much larger geographic areas, the opportunity…”

“These coarser locations may be combined with point-level and/or demographic data to make valid inferences for intensity or risk in the presence of geographic bias via either (a) coarseneddata maximum likelihood estimation procedures (Zimmerman, 2008b; Zimmerman and Fang, 2012), or (b) imputation of a surrogate point l…”

“Various methods, such as multiple imputation (MI) 9,10 , geo-imputation 11,12 , dasymetric mapping 13–16 , geo-statistical interpolation 17 , and machine learning-based approaches 18,19 , have been employed for missing data handling and data downscaling.”

“Switching to using population weighted centroids may improve the performance on some tasks (Henry & Boscoe, 2008).”

“Selecting a random point as opposed to a centroid avoids generating artificial clusters.(5) The same procedures were repeated for the AI population using the additional Census information to identify the proportion of AI in each census tract from the 2010 Tiger Maps.”

“These methods assume that the ‘consensus’ of multiple platforms is closer to the ground truth, effectively mitigating the influence of extreme outliers (Henry and Boscoe 2008; Yin, Ma, and Goldberg 2019).”

“Applications on Census data and health studies can be found in [7–9] and references therein.”

“Such geo-imputation is common in the health sciences literature when developing models in the absence of data describing individuals (Curriero et al., 2010; Henry & Boscoe, 2008; Klassen & Platz, 2006).”

“…infrastructure correlates with lower immunization coverage and contributes to the rise of diseases like malaria, polio, and measles, compounded by absent or unstable governments and poor coordination among health agencies, as seen during military conflicts in Afghanistan [19] and Ukraine [20].”