Accuracy of Birth Certificate and Hospital Discharge Data: A Certified Nurse-Midwife and Physician Comparison

Abstract

Birth certificate and hospital discharge data are relied upon heavily for national surveillance and research on maternal health. Despite the great importance of these data sources, the recording accuracy in these datasets, comparing birth attendant type, has not been evaluated. The study objective was to assess the variation in chart documentation accuracy between certified nurse-midwives (CNMs) and physicians (MDs) for selected maternal variables using birth certificate and hospital discharge data. Data was obtained on women delivering in 10 Washington State hospitals that had both CNM and MD-attended births in 2000 (n = 2699). Using the hospital medical record as the gold standard of accuracy, the true positive rate (TPR) for selected maternal medical conditions, pregnancy complications, and intrapartum and postpartum events was calculated for CNMs and MDs using birth certificate data, hospital discharge data, and both data sources combined. The magnitude of TPRs for most recorded maternal medical conditions, pregnancy complications, and intrapatum and postpartum events was higher for CNMs than for MDs. TPRs were significantly higher in birth certificate records for pregnancy-induced hypertension, premature rupture of membranes, labor augmentation, induction of labor, and vaginal birth after cesarean (VBAC) for CNM-attended births relative to MDs. Among combined data sources, CNM TPRs were significantly higher for pregnancy-induced hypertension and premature rupture of membranes. CNMs had consistently higher accuracy of recorded maternal medical conditions, pregnancy complications, and intrapartum and postpartum events when compared to MDs for all data sources, with several being statistically significant. Our findings highlight discrepancies between CNM and MD hospital chart documentation, and suggest that epidemiologic researchers consider the issue of measurement error and birth attendant type.