Impact of Major Congenital Malformation Algorithms on Their Prevalence in Large Population-Based Mother-Child Cohorts

Major congenital malformations (MCM) affect 2%-6% of pregnancies globally. Identifying MCM using real-world data is essential, but various definitions exist with differing performances and case ascertainment criteria, limiting collaborations.

To compare published definitions and algorithms to identify MCM using data from the Québec Pregnancy Cohort (QPC).

We conducted a comparative study of 10 algorithms for identifying MCM using QPC, a pre-established population-based cohort of pregnancies covered by Québec's public drug insurance plan between 1998 and 2015. We included infants from singleton pregnancies (live or stillbirth) who met the minimum postnatal follow-up duration required by the applied algorithm. Infants were excluded if they had an isolated major chromosomal congenital malformation or a gestational age ≤ 20 weeks. Algorithms varied by time window of detection (28 days, 6 months, 1 year), data source considered (inpatient, outpatient), diagnostic codes (ICD-9, ICD-10) and procedural codes requirements. The prevalence of global MCM and organ-specific malformation was calculated and compared to North American prevalence.

Among 233,338 infants meeting inclusion criteria, global MCM prevalence ranged from 2.9% to 9.0%. Across all algorithms, the most prevalent malformations were musculoskeletal and circulatory anomalies, with prevalence ranging from 0.9% to 3.3% and from 0.4% to 2.4%, respectively. Using ≥ 1 diagnostic code in the first year of life overestimated the prevalence, whereas using ≥ 1 inpatient or ≥ 2 outpatient codes for the same organ system on different days yielded a prevalence of 5.0%, which is close to what is expected compared to international prevalences. Considering this algorithm appears well-suited to the QPC setting, where inpatient diagnostic codes have been shown to be more reliable.

Algorithm choice substantially impacts MCM prevalence estimates. Transparent definitions enhance reproducibility and cross-study comparability in pharmacoepidemiologic research.