Background: The maintenance of health registers has become routine. The main prerequisite for their use is that registers be complete and that their contents correspond to reality.
Methods: Data on all primiparous women who gave birth between 1987 and 1989 (N=73009) and on their second (N=55388) and third births (N=22904) in the 1987-1998 period were retrieved from the Finnish Medical Birth Register (MBR). The consistency of the MBR data on reproductive history and on previous Caesarean section was investigated by comparing the records on subsequent births.
Main results: In total 98.5% of the information on reproductive history corresponded with the previous data in the MBR. Data quality decreased over time and with increasing parity. There were problems with the registration of rare cases, e.g. several extrauterine pregnancies or stillbirths. The quality deteriorated in the late 1990s, because no data on previous induced abortions and extrauterine pregnancies were collected between 1991 and 1995. The quality of data on previous Caesarean section was poor in 1987-1990, a period when the data were collected by using ICD-9 codes, but the quality improved after the introduction of a check-box format in 1991.
Conclusions: Changes in question formats may change the quality of register data significantly. Check-boxes seem to improve quality compared to open-ended questions. The data on reproductive history and previous Caesarean sections could be combined routinely to improve the quality of the MBR.