(245) Spontaneous Abortions Case Definition and Gestational Age Estimation Methodological Differences

Publication Number: 919

Background: For pregnancy safety and population-based studies, administrative health claims are indispensable tools for studying pregnancy and infant outcomes. However, the accurate ascertainment of pregnancy outcomes and estimating the start date of pregnancy are crucial to avoid misclassification of any drug exposure. Several pregnancy algorithms that utilize administrative health claims are available; however, their performance varies based on the pregnancy outcome, live birth vs. early pregnancy loss

Objectives: To confirm cases identified by the spontaneous abortion (SAB) identification algorithm using the HealthCore (HC) internal algorithm and the Instant Health Data Patient Profile (IHD-PP) using the HealthCore Integrated Research DatabaseSM (HIRD).

Methods: The Sentinel Cohort Identification and Descriptive Analysis tools (FDA, 2018) identified eligible pregnancies and women of childbearing age (10-49 years) between November 2015 to December 2020. Pregnancies ending in a spontaneous abortion were identified via developed and validated algorithms. Since the SAB identification algorithm could not estimate the start date of pregnancy, i.e., the last menstrual period (LMP), a harmonized gestational age (GA) of 140 days was used for all cases. The HC internal pregnancy algorithm and the IHD-PP were utilized to validate these cases. A lookup period of (-140 +7 days) was used to search for all codes ≥ 8 weeks and < 20 weeks. Furthermore, the HC algorithm considers the difference in days between the reported gestational week, i.e., Z3A code, and the event date.

Results: Using the IHD-PP, 85% of SAB cases identified by the SAB identification algorithm were confirmed by the HC algorithm as actual SAB cases, while 15% were corrected as elective termination cases. In addition, the HC algorithm recalculated the GA among all the SAB cases using all available codes. For 90% of the SAB cases, GA was inferred as 88 GA days instead of the harmonized 140 days. While for 10% of the SAB cases, reported Z3A codes were allocated, thus their GA was adjusted based on the Z3A codes, with half of the cases corrected to 60 GA days and the other half corrected at 130 GA days.

Conclusions: The performance of any pregnancy algorithm used to identify early pregnancy loss, such as spontaneous abortion, relies on the accuracy and completeness of the diagnosis coding within the administrative claims database. Moreover, the complexity associated with identifying spontaneous abortion muddles the estimation of GA and limits the assessment of the start of pregnancy date. All of which could lead to a misclassification of exposures and outcomes that may have occurred.