The quality of routinely collected maternity data

OBJECTIVE: To assess the validity of clinical information held on a regional maternity database, the St Mary's Maternity Information System (SMMIS). DESIGN: A retrospective review of 892 maternity case notes and matched SMMIS records, by a midwife trained in clinical coding techniques. SETTING: Three maternity units in the North West Thames Region. MAIN OUTCOME MEASURES: Percentage agreement for 17 directly recorded SMMIS data items and equivalent data abstracted from the notes. Frequencies of diagnosis codes abstracted from case notes, as compared with those generated by SMMIS on the basis of directly recorded data. RESULTS: A generally high level of agreement was observed between the abstracts of the notes and the SMMIS records. Of the 17 data items examined, 10 showed 95% agreement or better, and all but two exceeded 80% agreement. Little difference was found between the levels of agreement observed at the three sites. A greater number and range of diagnosis codes were abstracted from the notes than were generated by SMMIS. CONCLUSIONS: The directly recorded clinical data held on the SMMIS regional database is largely accurate and consistently recorded across a variety of units. The database can therefore be considered a valuable resource for the comparative audit of maternity practice. The SMMIS technique for deriving, on a semi-automatic basis, diagnosis codes from the directly recorded fields, appears to work moderately well. We suggest that the direct method of data collection used in SMMIS could provide a model for other specialties in the National Health Service.