Validation of Claims‐Based Algorithms for Identifying Congenital Urinary Tract, Genital, Gastrointestinal, and Musculoskeletal Malformations

ABSTRACT Aim Congenital malformations are important outcomes when evaluating medication safety in pregnancy. However, the accuracy of claims‐based algorithms for identifying organ‐specific malformations remains understudied. We validated algorithms for four malformation groups: urinary tract, genital, gastrointestinal, and musculoskeletal. Materials and Methods Using the Mass General Brigham (MGB, 2007–2020) and Stanford Medicine (2016–2023) databases, we identified infants with potential malformations of interest based on diagnosis and procedure codes within 90 days of birth. A total of 150 cases were sampled for each malformation group. Positive predictive values (PPV) were estimated to quantify the validity of the algorithms, separately by site and by International Classification of Diseases (ICD) coding system. Results For MGB ICD‐9, MGB ICD‐10, and Stanford ICD‐10 algorithms, respectively, PPVs were 100%, 98.0%, and 95.8% for urinary tract malformations; 98.0%, 95.9%, and 93.8% for genital malformations; 76.0%, 78.0%, and 76.0% for gastrointestinal malformations; and 85.4%, 84.0%, and 62.0% for musculoskeletal malformations. False positives were primarily attributed to suspected malformations that were later ruled out and to broader or inaccurate coding for diagnoses that were not major malformations. Conclusions Claims‐based algorithms demonstrated high PPVs for urinary tract and genital malformations and moderate for gastrointestinal malformations, but there was variation in musculoskeletal PPVs between the institutions with different patient populations.