Development and Validation of Claims-based Algorithms for Estimating Gestational Age of Spontaneous Abortion and Termination

Sunday, August 28, 2022

8:45 AM – 9:00 AM CEST

Location: C1M2

Publication Number: 237

Background: Over 2 million pregnancies in the US annually end with spontaneous abortion (SAB, loss at < 20 weeks' gestation) or termination. We previously validated algorithms (ALG) to identify SAB and termination in healthcare utilization databases (PPVs >86%).

Objectives: To further establish valid approaches to study the impact of medical products on these outcomes using claims, our study objective was to develop and validate gestational age (GA) ALGs for SAB and termination.

Methods: We identified all beneficiaries in the Medicaid Analytic eXtract (2000-2014), 12-55 years old, with ≥1 end of pregnancy related code (i.e., live birth, mixed birth, stillbirth, SAB, termination, ectopic pregnancy, and unclassified delivery / end of pregnancy). After implementation of an ICD-9 based hierarchical pregnancy ALG, we identified distinct potential cases of SAB and termination. We required the mothers to be continuously enrolled from 20 weeks prior to the non-live birth date until 1 month thereafter, and identified the subset who received care at Mass General Brigham, a hospital network in the Boston area. We retrieved the medical records for 195 SAB and 240 terminations, which were independently reviewed by 2 clinicians who extracted the GA at the pregnancy end date (gold standard). We developed: ALG 1: assign national median GA (i.e, SAB: 9 weeks; terminations: 8 weeks); ALG 2: assign GA randomly drawn from GA distributions reported in the literature; ALG 3: use a random forest (RF) model to predict GA based on the timing of prenatal screening tests, proximate etiologies for non-livebirth outcomes, and maternal risk factors. ALG performance was assessed based on the proportion of pregnancies with estimated GA within 1 and 4 weeks of the gold standard GA, the mean squared error (MSE) and the R-squared, estimated using 10-fold cross-validation.

Results: For both outcomes, ALG 2 performed the worst and ALG 3 (RF) performed slightly better than ALG 1. For SAB, 35.4% were correctly classified within 1 week and 84.1% were correctly classified within 4 weeks of the gold standard using the RF model (MSE: 11.4, R-squared: 0.02). For termination, the proportions were 51.7% and 90.4% respectively (MSE: 10.4; R-squared: 0.36). The R-squared was low across ALGs for both outcomes indicating the prediction of the ALGs did not explain much of the variation in GA.

Conclusions: Depending on the research questions, SABs and terminations can be studied in claims data with careful implementation of the validated algorithms. However, investigators should be cognizant of the higher level of misclassification of GA for these pregnancy outcomes compared to live births and the potential implications for misclassification of the etiologically relevant time window.

Presenting Author(s)

Krista F. Huybrechts, MS, PhD, FISPE

Brigham and Women's Hospital, Harvard Medical School
Boston, Massachusetts, United States

Slides

Co-Author(s)

Yanmin Zhu, MS, PhD

Instructor in Medicine
Brigham and Women's Hospital/Harvard Medical School
Boston, Massachusetts, United States
Sonia Hernandez-Diaz, MD, MPH, FISPE

Professor of Epidemiology
Harvard School of Public Health
Boston, Massachusetts, United States
BB

Brian Bateman

BWH Div of Pharmacoepidemiology and Pharmacoeconomics
Concord, Massachusetts, United States
KG

Kathryn Gray

Division of Maternal-Fetal Medicine, Department of Obstetrics and Gynecology, Brigham and Women's Hospital, Boston, MA
Boston, Massachusetts, United States
Loreen Straub, MD MS

Research Fellow
Brigham and Women's Hospital/Harvard Medical School
Cambridge, Massachusetts, United States
LT

Lockwood Taylor

Division of Epidemiology, Office of Surveillance and Epidemiology, Center for Drug Evaluation and Research, United States Food and Drug Administration
Austin, Texas, United States
RO

Rita Ouellet-Hellstrom

Division of Epidemiology, Office of Surveillance and Epidemiology, Center for Drug Evaluation and Research, United States Food and Drug Administration, United States
YM

Yong Ma

Lead Mathematical Statistician
Office of Biostatistics, Center for Drug Evaluation and Research, U.S. Food and Drug Administration, United States
YQ

Yandong Qiang, MD, PhD

US Food and Drug Administration
Silver Spring, Maryland, United States
SV

Seanna Vine

Division of Pharmacoepidemiology and Pharmacoeconomics, Department of Medicine, Brigham and Women's Hospital and Harvard Medical School, Boston, MA, USA, United States
HM

Helen Mogun

Division of Pharmacoepidemiology and Pharmacoeconomics, Department of Medicine, Brigham and Women's Hospital and Harvard Medical School, Boston, MA, USA, United States
WH

Wei Hua, MSc, MHS, MD, PhD

Office of Surveillance and Epidemiology, Center for Drug Evaluation and Research, U.S. Food and Drug Administration, United States