Use of an automated database to evaluate markers for early detection of pregnancy

Abstract

The objective of this study was to develop and validate algorithms to detect pregnancies from the time of first clinical recognition by using Kaiser Permanente automated databases from Portland, Oregon. In 1993--1994, the authors evaluated these databases retrospectively to identify markers indicative of initial clinical detection of pregnancy and pregnancy outcomes. Pregnancy markers were found for 99% of the women for whom pregnancy outcomes were included in the automated databases, and pregnancy outcomes were identified for 77% of the women for whom there were pregnancy markers. The earliest marker most predictive of a pregnancy outcome was a positive human chorionic gonadotropin test; least predictive was an obstetric outpatient visit. Medical record review indicated that in a sample of women with pregnancy markers in the database, an estimated 6% of pregnancy outcomes (primarily early fetal deaths and elective terminations) were lost. Pregnancies were first captured in automated databases 6--8 weeks after the last menstrual period, and a combination of a positive human chorionic gonadotropin test and an outpatient obstetric visit was the most sensitive and specific early marker of pregnancy. When combined with automated pharmacy records, these databases may be valuable tools for evaluating prescription drug effects on all major outcomes of clinically recognized pregnancies.