Statistician, postdoctoral research fellow Leibniz Institute for Prevention Research and Epidemiology - BIPS, Germany
Background: In claims data, pregnancies are usually identified by searching for outcome-specific diagnostic and procedure codes. Pregnancies without a recorded outcome—e.g., ongoing pregnancies, spontaneous abortions not requiring medical treatment or induced abortions without a medical indication—are missed by this approach. In German claims data, there may be options to capture these pregnancies and estimate their onset, as the expected delivery date (EDD) is usually coded early in pregnancy.
Objectives: To identify pregnancies without a recorded outcome in German claims data and estimate their beginning and end.
Methods: Using data from the German Pharmacoepidemiological Research Database (GePaRD) from 2004-2020, we searched in girls and women of childbearing age (12-50 years) for EDDs that did not belong to a pregnancy already identified by a previously developed outcome algorithm, which uses outcome-specific diagnostic and procedure codes. If the EDD was later than the end of data availability or the end of insurance of the respective woman, the pregnancy was classified as “ongoing” and the beginning of the pregnancy was set to EDD-280 days (same as for pregnancies identified by the outcome algorithm. For all other EDDs, we searched for outpatient care codes, non-outcome-specific inpatient diagnoses, and in- and outpatient procedures codes indicating a pregnancy (“pregnancy markers”) The date of the code closest to the EDD was used to estimate the end of the pregnancy and the beginning was set to EDD-280 days. EDDs without a pregnancy marker were disregarded.
Results: Overall, we identified 190,127 ongoing pregnancies and 294,759 pregnancies without a recorded outcome (as compared to 2,410,840 pregnancies identified by the outcome algorithm). For 79% of the pregnancies that were classified as ongoing on 31.12.2019, an outcome could be identified once 2020 data were included. In the pregnancies without a recorded outcome not classified as ongoing, we identified 3,670,499 pregnancy markers (per pregnancy median 10 markers), thereof 2.8% inpatient diagnoses, 0.03% in- or outpatient procedures codes, and 97.2% outpatient care codes. Timing of pregnancy markers was plausible, e.g., blood tests supposed to be performed early in pregnancy were in median coded 6.9 weeks and in 95% up to 12.4 weeks after the onset of pregnancy and pregnancies lengths matched in more than 90% with the coded duration (O.09).
Conclusions: Our study suggests that ongoing pregnancies and pregnancies without a recorded outcome can be identified and their onset and end plausibly be estimated in German claims data.
