Keran Moll, Hui Lee Wong, Kathryn Fingar, Shayan Hobbi, Minya Sheng, Timothy Burrell, Bethany Baer, Steven Immel, Nasim Lari, Linda Eckert, Flor Munoz, William Rosenfeld, Joyce Obidi, Kristin Sepulveda, Azadeh Shoaibi and Steven Anderson
Background: Identification of pregnancy outcomes and estimation of gestational age (GA) are critical in maternal vaccine safety studies. Validated pregnancy outcome algorithms based on the International Classification of Diseases, Tenth Revision, Clinical Modification/Procedure Coding System (ICD-10-CM/PCS) have yet to be published.
Objectives: To develop ICD-10-CM/PCS-based pregnancy outcome algorithms in claims databases using linked electronic medical records (EMRs) adjudicated by physicians as the reference method.
Methods: Pregnancy outcomes (live birth, spontaneous abortion [SA], and stillbirth [SB]) were identified using ICD-10-CM/PCS and service codes on claims for women aged 12-55 years in the 2016-2018 IBM® MarketScan® Explorys® Claims-EMR Data Set (CED). Multiple outcomes for the same woman were distinguished based on a hierarchy and spacing requirements. If multiple estimates of GA were present, a pregnancy start date was determined using estimates with the highest clinical accuracy (i.e., fertility procedures, GA on diagnosis, and prenatal scans). Among pregnancies with available structured obstetric EMR data, 100 full-term and preterm live births (FLBs/PLBs), 100 SAs, and 24 SBs were drawn as validation samples for which EMR charts were populated. Guided by Global Alignment of Immunization safety Assessment in pregnancy (GAIA) case definitions, physicians adjudicated cases and assigned a GAIA level of certainty to each one. Percentage positive agreement (PPA) was calculated, comparing results from the algorithms and adjudication.
Results: The claims-based algorithms identified 34,204 pregnancy episodes and their outcomes in CED (26,825 FLBs, 3,776 PLBs, 3,452 SAs, 151 SBs) and estimated GA at outcome date. Obstetric EMRs were available for 10% (3,390: 2,736 FLBs, 363 PLBs, 267 SAs, 24 SBs) of the episodes, which resembled the overall CED pregnancy population on demographic and clinical characteristics (most standardized mean differences <0.2). Of sampled episodes, physicians adjudicated the outcome and GA at outcome with GAIA levels of certainty of 1-3 for 93 PLBs, 92 FLBs, 75 SAs, and 24 SBs. Among these episodes, the PPA on the outcome was 98% for FLBs, 62% for PLBs, 100% for SAs, and 71% for SBs. The PPA on GA within 7 and 14 days, respectively, was 86% and 99% for FLBs, 82% and 93% for PLBs, 61% and 81% for SAs, and 67% and 79% for SBs.
Conclusions: ICD-10-CM/PCS-based pregnancy outcome algorithms were developed and validated with a high level of agreement with physician-adjudicated results from EMR review. This study demonstrates the application of GAIA case definitions to retrospectively collected EMR data.