Perinatal data reliability in a large teaching obstetric unit

Summary. In this inter-rater agreement study of antenatal and neonatal variables collected in a large teaching obstetric unit, information routinely collected by hospital staff was compared with that collected by a specially trained physician and a social worker. Agreement between the two sources of data was evaluated using kappa statistics and intraclass correlation coefficients. Excellent agreement was observed for some variables such as maternal and newborn anthropometric measures, and previous birthweight, but there was poor agreement for others such as indicators of physical activity, work during pregnancy and blood pressure measures. Some of the limitations are due to problems in phrasing questions, patients' recall, interviewer bias and abstracting data. We recommend that epidemiological studies should always include a reliability component, proper standardization of personnel and instruments and include, when published, validity data and examples of questions used.