Reliability of birth certificate data: a multi-hospital comparison to medical records information.

OBJECTIVE: To examine the reliability of birth certificate data and determine if reliability differs between teaching and nonteaching hospitals. METHODS: We compared information from birth certificates and medical records in 33,616 women admitted for labor and delivery in 1993-95 to 20 hospitals in Northeast Ohio. Analyses determined the agreement for 36 common data elements, and the sensitivity, specificity, and positive and negative predictive values of birth certificate data, using medical record data as a "gold standard." RESULTS: Sensitivity and positive predictive value varied widely (9-100% and 2-100%, respectively), as did agreement, which was "almost perfect" for measures of prior obstetrical history, delivery type, and infant Apgar score (K = 0.854-0.969) and "substantial" for several other variables (e.g., tobacco use (K = 0.766), gestational age (K = 0.726), prenatal care (K = 0.671)). However, agreement was only "slight" to "moderate" for most maternal risk factors and comorbidities (K = 0.085-0.545) and for several complications of pregnancy and/or labor and delivery (K = 0.285-0.734). Overall agreement was similar in teaching (mean K = 0.51) and nonteaching (K = 0.52) hospitals. Although agreement in teaching and nonteaching hospitals varied for some variables, no systematic differences were seen across types of variables. CONCLUSIONS: Our findings indicate that the reliability of birth certificate data vary for specific elements. Researchers and health policymakers need to be cognizant of the potential limitations of specific data elements.