Background: Studies of drug/vaccine safety during pregnancy are needed. Administrative databases offer the potential to collect data on large populations of pregnant women. However, some challenges of working with these data include identifying pregnancy outcomes, linking mothers and babies, and determining gestational age.
Objective: To describe and evaluate a CPT-based algorithm developed to determine gestational age when EHR or birth certificate data are not available. This algorithm complements an existing algorithm (PEA) used in VSD that identifies pregnancy outcomes and assigns fixed gestational age in the absence of additional information.
Methods: The algorithm was developed and validated in two samples of the HP VSD population from 2000 to 2009 (n-developing sample=2200, n-validation sample=1000) containing data from the state birth certificate file. Inclusion of CPT pregnancy codes were based on distribution properties and expert opinion. The algorithm consisted of hierarchical rules based on CPT codes to determine periods of exposure and assign gestational age. Validation included performance measures as compared to gestational age reported in birth certificates for early delivery defined by <37 weeks, and vaccination trimester.
Results: The algorithm consisted of five events which are standard in prenatal care: nuchal screen, triple screen, GBS culture, ultrasound, and serology for rubella Ab and Hep BS Ag . Eighty percent of pregnant women had at least one of these tests during their pregnancy. Sensitivity and positive predictive value of the algorithm for identifying early delivery was 71%, and 42%. Percent of women assigned to be vaccinated in the first trimester was 16.9% according to birth certificate and 17% with the CPT-based algorithm.
Conclusion: When EHR or birth certificate data are not available, a fixed gestational age is assigned based on birth outcome. Our CPT-based algorithm was accurate in assigning trimester of exposure for vaccination, though predictive value for early delivery was low.