587 - Are you In or Out(patient)?: A Geospatial Analysis of Subjects Enrolled in a COVID-19 Biorepository

Our team created a biorepository to investigate biomarkers associated with pediatric SARS-CoV-2 infections, Multisystem Inflammatory Syndrome of Children (MIS-C), and comparator admitted cohorts (e.g. respiratory viral infections; Kawasaki Disease). Enrollment procedures for our inpatient cohorts and the mild-covid, outpatient cohort were meaningfully different: admitted patients’ families were actively approached by research assistants (RAs) but outpatients had to call or email RAs to enroll. It is unknown how this impacted the cohort comparability of inpatient vs. outpatient groups.

Objective: Determine socio-geographic differences between children enrolled in a biorepository from outpatient and inpatient settings.

Design/Methods:

This is a cross-sectional geospatial analysis of subjects enrolled in a prospective observational cohort study. Research Assistants proactively enrolled admitted patients. In contrast, outpatient subjects, or their guardians, called the RAs. After confirming they met study inclusion criteria, outpatient subjects gave saliva samples in-person at a central site in Hartford, CT. Using ArcGIS, we geocoded subject home addresses and linked each to a pediatric-specific metric of social determinants of health: The Childhood Opportunity Index (COI). We created choropleth maps highlighting the percentage of cohort enrollment per census tract, calculated enrollment hot and cold spots using Getis-Ord Gi* fixed distance band statistic, and overlaid these findings to assess for geospatial patterns.

Results:

We enrolled 178 patients in inpatient cohorts and 108 patients in our outpatient cohort (Table 1) Inpatient subjects were more likely to be White and have private insurance. Geospatially, inpatient enrollment reflected our hospital’s typical catchment: high numbers closer to Hartford, cold spots in areas served by other hospitals (Figure 1A, C). The enrollment pattern is distinctly different for outpatient enrollees. Areas northwest and southeast of the hospital were enrollment hotspots not reflected in our inpatient enrollment (Figure 1B, D). These census tracts are a deep purple, representing high levels of outpatient enrollment and COI (Figure 1E, F).

Conclusion(s):

The sociodemographic and geospatial makeups of our inpatient and outpatient cohorts were significantly different. Children from areas of high COI enrolled at disproportionate rates relative to the inpatient enrollees for our mild-covid, outpatient cohort. We plan further analysis with geospatially weighted regression to fully quantify these associations.