![[ Visit Client Website ]](images/banner.jpg)
BACKGROUND: Birth registry datasets geocoded with latitude and longitude information have been widely used in epidemiologic studies. However, the validity of the results observed in some studies is questionable given the spatial uncertainty present in geocoded data. Incorrect mailing address information can result in geocoded uncertainty and may introduce errors and missing sources for researchers trying to investigate with mailed surveys. This is especially true if the population with spatially uncertain addresses and unreachable by mail differs significantly from the population with addresses that can be geocoded. We will explore the characteristics of pregnant women using geocoded Birth Registry Data, comparing the maternal behaviors and pregnancy outcomes between mothers with spatially certain and spatially uncertain addresses.
METHODS: Mapmarker USA Version 27 software was used to geocode addresses in 2013 Florida Birth Registry Data. Addresses that were unable to be geocoded were identified as having spatial uncertainty. Analyses were conducted using SAS 9.3. Data from mothers with mailing addresses outside of Florida or from births involving adoption were excluded from this analysis. Comparisons were made using logistic regressions to determine whether or not spatial uncertainty predicting the likelihood of various maternal behaviors or pregnancy outcomes.
RESULTS: Approximately 1.8% of maternal mailing address data shows spatial uncertainty in the geocoded 2013 Florida Birth Registry Data. Mothers who had an uncertain address in the birth registry were most likely to be unmarried, non-Hispanic Black or Hispanic, and younger than 25 years old. Additionally, they more often report having less than a high school education, using Medicaid or having no insurance for delivery, or having participated in WIC. Compared to mothers with certain addresses, mothers registered with uncertain addresses were more likely to smoke during pregnancy (OR=1.2), drink during pregnancy (OR=1.5), have low birth weight babies (OR=1.2), and give birth preterm (OR=1.1). Mothers registered with uncertain addresses were less likely to have received adequate prenatal care (OR=0.8) or to breastfeed their babies (OR=0.8). All results are statistically significant (p<0.05).
CONCLUSIONS: Very few studies have been conducted addressing differences in population characteristics for those with uncertain geocoded data. Spatial uncertainty in geocoded data has the potential to affect the validity of population-based epidemiology studies. These findings provide important information to survey designers, data collectors, and epidemiology researchers for understanding the characteristics and activity patterns of pregnant women who might be unreachable in their potential studies.