A distributed geospatial approach to describe community characteristics for multisite studies

Abstract

Understanding place-based contributors to health requires geographically and culturally diverse study populations, but sharing location data is a significant challenge to multisite studies. Here, we describe a standardized and reproducible method to perform geospatial analyses for multisite studies. Using census tract-level information, we created software for geocoding and geospatial data linkage that was distributed to a consortium of birth cohorts located throughout the USA. Individual sites performed geospatial linkages and returned tract-level information for 8810 children to a central site for analyses. Our generalizable approach demonstrates the feasibility of geospatial analyses across study sites to promote collaborative translational research.

Childhood Asthma Incidence, Early and Persistent Wheeze, and Neighborhood Socioeconomic Factors in the ECHO/CREW Consortium

“…Similar variability has been previously shown in the census factors across all cohorts. 9 eFigure 2 in Supplement 1 shows the distribution of child's race and ethnicity by categories of neighborhood characteristics. In tracts with low proportions of the population below the poverty level, 83% of the children were White; in tracts with high proportions of the population below the poverty level, 49% of the children were Black and 35% Hispanic, while only 13% were White.…”

“…decades with attention to longitudinal comparability of income, poverty, and geographic boundaries. 9 Our De-GAUSS approach enabled cohort data managers to assign census information to birth record addresses in a decentralized manner, preserving protections for the participants.…”

“…Briefly, US Census data including population density (population per kilometer squared), median household income, percentage below poverty level (percentage of individuals in tract whose income is below the federal poverty level), percentage of low-income households (income <$7500, $12 500, $20 000, and $25 000 for census years 1980, 1990, 2000, and 2010, respectively), percentage without high school diploma, percentage occupied housing, median gross rent, median housing value, percentage with a female head of household, percentage with a single-parent head of household, and percentage unemployed. Boundary files for the years 1980, 1990, 2000, and 2010 were provided along with R code in a Docker container to each cohort . Cohort data managers merged census data to participant birth address using census year nearest to the birth year and then removed identifiable information.…”

“…We ascertained census tract-level information for all CREW participants at their birth address using our previously described Decentralized Geomarker Assessment for Multi-site Studies (DeGAUSS) approach. 9 Briefly, US Census data including population density (population per kilometer squared), median household income, percentage below poverty level (percentage of individuals in tract whose income is below the federal poverty level), percentage of low-income households (income <$7500, $12 500, $20 000, and $25 000 for census years 1980, 1990, 2000, and 2010, respectively), percentage without high school diploma, percentage occupied housing, median gross rent, median housing value, percentage with a female head of household, percentage with a single-parent head of household, and percentage unemployed. Boundary files for the years 1980, 1990, 2000, and 2010 were provided along with R code in a Docker container to each cohort.…”

“…Boundary files for the years 1980, 1990, 2000, and 2010 were provided along with R code in a Docker container to each cohort. 9 Cohort data managers merged census data to participant birth address using census year nearest to the birth year and then removed identifiable information.…”