(233) Algorithms to identify major congenital malformations in routinely collected healthcare databases: A systematic literature review

Background: Most of the evidence on the safety of medication use in pregnancy is derived from post-marketing observational studies using real-world data (RWD). While major congenital malformations (MCM) are commonly a primary outcome of interest, identifying cases in RWD can be challenging due to complex billing or coding practices, inconsistent availability of linked maternal and child records, and heterogeneous definitions and coding systems.

Objectives: To identify and summarize algorithms used to identify MCMs in RWD sources in the United States, Canada, and Europe by conducting a systematic literature review.

Methods: We developed a search strategy to identify studies containing algorithms for MCMs from January 1, 2010 to March 17, 2023. Search terms included keywords to identify studies with each of the following: MCMs as an outcome, routinely collected healthcare data, epidemiologic designs likely to incorporate algorithms, and conducted among pregnant individuals and/or infants. To determine study eligibility and inclusion, we followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. First, two independent authors reviewed titles and abstracts followed by full-texts. For studies meeting inclusion criteria, data extraction was conducted using a standardized data collection form.

Results: Among the initially identified 1922 studies, 791 were selected for full-text review. Of these, 70.4% were excluded, leaving 234 studies. Algorithm design varied by data source type, with claims-based studies typically including more complex logic (e.g., requiring multiple codes in specific time windows for mother and/or infant) compared with studies based on national registers (e.g., requiring a single code any time during the infant’s first year of life). Over half (54.7%) of the included studies were from Europe, predominantly from Nordic countries using national register data (n=105, 82.0%). Studies using claims (17.1%) or hospital discharge data (18.4%) were also common. The majority of studies (71.8%) provided algorithms for MCMs overall, while the remainder focused on specific malformations or anatomical systems. Though there was heterogeneity in the timing of MCM assessment, most studies (57.3%) collected MCMs through the infant’s first year of life. There were 24 (10.3%) validation studies, 70.8% based on claims and/or electronic health record data only. Most had positive predictive values greater than 70%, though variation existed depending on MCM type or anatomical site.

Conclusions: We provide the first comprehensive systematic literature review of algorithms used to identify MCM in RWD, aiding researchers in their ability to generate reliable evidence in pregnancy safety pharmacoepidemiology.