Should aggregate US Census data be used as a proxy for individual household income in a birth defects registry?

BACKGROUND

Birth Defects Registries do not have access to income data and low household income is associated with adverse pregnancy outcomes of stillbirth, preterm birth, and birth defects. We compared 1999 income data from the National Birth Defects Prevention Study (NBDPS) with 2000 Census block group income data for the residence location of these same mothers.

METHODS

We geocoded 339 case mothers and 121 control mothers and assessed household income among case and control mothers by using NBDPS and census block group data. Correlation and concordance were assessed between the 2 data sources' household income data.

RESULTS

The household income distribution was similar between case and control mothers within each data source. Both case and control mothers in the NBDPS's lowest household income category (<$20,000/year) reported lower income than was documented in their census block group's median household income (p-value<0.0001). A weak correlation was identified between NBDPS interview and census block group income data (control mothers, rs=0.53; case mothers, rs=0.32). There was also poor to fair concordance between the 2 data sources (control mothers, kw=0.28; 95% CI=0.19-0.37 and case mothers, kw=0.18; 95% CI=0.13-0.24).

CONCLUSION

These findings demonstrate dissimilar household incomes between NBDPS and census block group data. Caution should be used if block-level data is used as a proxy for individual-level household incomes in population-based birth defects surveillance and research.