Methodological considerations in the design of an obstetric database abstracted from medical records.

The quality of maternal information contained within contemporary obstetric notes was investigated by abstracting data from the medical records of multiparous women who were admitted to a major city hospital in the South Thames Region. Potential sources of error were identified by comparing information recorded in different sections of each obstetric notes and within the obstetric notes of consecutive pregnancies. The format of the obstetric notes largely determined which variables were recorded and, to some extent, the accuracy of information collected. However, the quality of the data ultimately depended upon whether each variable was self-reported or directly measured. Self-reported variables were subject to selective omission and subjective bias, while measured variables were susceptible to inaccurate equipment and poor measurement practice. By interviewing a sample of midwives currently involved in antenatal care at the Trust it was possible to confirm that extensive variation in measurement and recording procedures routinely occurred.