Abstract
Observational studies often utilize area-level socioeconomic status (aSES) data when individual-level (iSES) data are unavailable. However, correlations between iSES and aSES in pediatric cancer populations are not well-described. We explored correlations between aSES and iSES in childhood cancer cases co-enrolled on Children's Oncology Group registry and epidemiology protocols (n = 1379) in the United States. We created four composite measures of iSES using maternal questionnaire data for household poverty, educational attainment, occupation, and housing type for comparison with Yost and Social Vulnerability Index (SVI) socio-economic domain at the census tract (CT)- and ZIP code (ZIP)-level. CT- and ZIP- aSES were weakly to moderately correlated with iSES (ρ(SVI)=0.46, 0.36, respectively); results were similar across the four composite models (range ρ(SVI)=0.43-0.48 [CT-aSES], 0.34-0.36 [ZIP-aSES]). When stratifying by race and ethnicity, the highest correlations were among non-Hispanic Black and Hispanic cases, with CT-level correlations stronger than ZIP-level. Correlations were dramatically lower for all rural cases (ρ(SVI) range=-0.10-0.25), particularly non-Hispanic White (ρ(SVI) range=-0.15-0.23); sample size was insufficient to evaluate other subgroups. We found varying levels of correlation between aSES and iSES, dependent upon geographic resolution of aSES measure, with CT-aSES generally being more strongly correlated with iSES than ZIP-aSES, and varying by race, ethnicity, and urbanicity.