The computerized perinatal database: are the data reliable?

The purpose of this study is to examine the correctness of the clinical data from the computerized perinatal database (PC-Log) at a Mayo Health System hospital. This computerized database is used for electronic transmission of birth certificates in Wisconsin. The paper medical record is chosen for the comparison. Random selection of 99 charts from a total of 893 births at a tertiary perinatal center during 1995. Of 310 fields in the database, 32 variables were compared to a hand abstraction of the paper medical record. PC-Log had 100% positive-predictive value (PPV) for eclampsia, prolonged rupture of membranes, pre-existing diabetes, cesarean section, and transports. The sensitivity, specificity, and PPV for other variables (abortion, congenital anomalies, gestational diabetes, maternal hypertension, and maternal employment) showed moderate to high agreement, but was poor for maternal ethanol use during pregnancy. Compared to hand abstraction, PC-Log had no recorded cases of substance abuse, antenatal steroids, hyaline membrane disease, circumcision, maternal and infant length of stay. Means for birth weight 5 minute Apgar scores did not differ, and the correlations were r = 0.982 and r = 0.960. The PC-Log showed good agreement for many but not all the variables of clinical interest.