Medical records vs. interview responses: a comparative analysis of selected variables for linked birth defect cases.

Background: Data pertaining to birth defects are subject to certain limitations depending on the collection method. This study compares the agreement of data from medical records and maternal interviews.

Methods: The medical records and maternal interviews were linked for 1017 deliveries. Prevalence, concordance and kappa coefficients were calculated for maternal gestational and non-gestational diabetes, insulin use, seizures/epilepsy, Hispanic ethnicity, and infant/fetus sex.

Results: The prevalence of non-gestational diabetes was 4.3% in the medical records and 3.4% in the maternal interviews, with 98.1% agreement. The prevalence of gestational diabetes was 7.9% in medical records and 9.2% in maternal interviews, with 94.3% agreement. Similar prevalences and high levels of agreement were observed between the two systems for infant/fetus sex and mother's Hispanic ethnicity. Although high concordance was observed for seizures/epilepsy, kappa value was moderate.

Conclusions: The availability of two distinct sources of data provides an exceptional opportunity to compare and validate both data sources. We found that the data for certain variables from maternal interviews strongly agreed with information from medical records. However, the extent of that agreement depended on the type of variable measured. Our results suggest that for some variables such as demographic variables, researchers can use either of the two data sources.