Background: Agreement between self-reported data and data obtained from medical records is far from perfect and few studies have analyzed the element of language when self-reported data are given in one language and this information is recorded in another language in the medical record. Our objective was to assess agreement between self-reported data and medical record data with regard to prenatal risk factors in pregnant Latina women.
Methods: We interviewed 350 Latina women at >or =20 weeks' gestation regarding alcohol use, tobacco use, use of prenatal vitamins, age, education, use of prenatal care, and medical conditions. Kappa statistic (kappa) and 95% confidence intervals (95% CIs) were used to calculate agreement between self-reported responses and medical record data. Multiple logistic regression analysis was used to evaluate effect of maternal characteristics on likelihood of disagreement.
Results: Agreement between self-reported and medical record data was generally lower for behavioral factors (alcohol kappa=0.37 and prenatal vitamin use kappa=0.09) than for medical conditions (anemia kappa=0.63, gestational diabetes kappa=0.83, and hypertension kappa=0.68). In general, maternal characteristics did not significantly predict patterns of disagreement.
Conclusions: Among pregnant Latina women, self-reported data on behavioral factors had lower agreement than self-reported data on medical conditions. Further study is needed to define the effect of other factors, such as social norms, on accuracy of self-reported data during pregnancy.