Background: In case-control studies that assess associations between medication use and birth defects, detailed information on type of medication and timing of use is essential to prevent misclassification. However, data on the accuracy of recall of medication use during pregnancy are scarce.
Objective: The aim of this study was to validate a self-administered questionnaire to assess prescription medication use in the 3 months before and during pregnancy.
Methods: This validation study was embedded in Eurocat Northern Netherlands, a population-based birth defects registry that covers 10% of all births in The Netherlands. The questionnaire was validated among 560 mothers of infants with major birth defects registered from 1 January 2009 through 30 June 2010 by comparing it with a reference standard consisting of pharmacy data which were checked for compliance by maternal interviews. Sensitivity and specificity were calculated to quantify validity for any prescription medication use, groups of medications and individual medications. In addition, we determined whether maternal characteristics influenced disagreement between the questionnaire and the reference standard using logistic regression analyses.
Results: The sensitivity for any prescription medication use was 0.57, ranging between 0.07 (dermatological corticosteroids) and 0.83 (antihypertensives) for medication groups, and between 0.00 (naproxen) and 0.73 (salbutamol) for individual medications. Overall, specificity was high (0.93-1.00). Smoking during pregnancy and completing the questionnaire>2 years after delivery were associated with increased disagreement between the questionnaire for prescription medication use and the reference standard.
Conclusions: The validity of the self-administered questionnaire for prescription medication use during pregnancy was moderate to poor for most medications and disagreement differed by some maternal characteristics. As many epidemiological studies use similar questionnaires to assess medication use these studies may need additional data sources such as pharmacy records or prescription databases for medication use next to self-reported methods. Also, previous knowledge on the effect of questionnaire design should be taken into account.