Variations in the accuracy of obstetric procedures and diagnoses on birth records in Washington State, 1989.

The authors abstracted a sample of 7,536 hospital medical records to validate the accuracy of the coding of obstetric information on 1) birth certificates, 2) a statewide computerized hospital discharge abstract data system, and 3) a linked file merging birth certificates and the hospital abstract data for Washington State deliveries occurring in 1989. Measures of accuracy of coding of delivery method and obstetric procedures varied greatly among the 23 hospitals that participated in the study. Computerized hospital discharge data were generally more complete and accurate than were birth certificate data. The linked file was more likely to identify obstetric procedures than was either source alone. For example, only 84.1% of cesarean deliveries noted in the hospital charts were identified on birth certificates (range among hospitals, 37-100%). Using the linked file, the authors identified 99.8% of cesarean deliveries (range, 97-100%). Linked birth certificate-hospital abstract files may become an excellent source of data for epidemiologic and health care studies; however, further training of medical record personnel and standardization of coding are needed to improve the quality of computerized data on obstetric events.