(154) Utilization and Timing of Z3A Codes in Pregnancy to Estimate Gestational Age Using Administrative Claims

Publication Number: 1199

Background: Gestational age (GA) is a critical variable when studying the effects of treatments taken at specific timepoints during pregnancy. Although ICD-10 Z3A.XX codes (introduced in the US in 2015) may be used to capture week-specific GA in claims data, the validity of this measurement of GA is uncertain. For example, multiple Z3A codes over the same pregnancy may conflict in estimated GA. An improved understanding of the application of Z3A codes is important to inform GA algorithm development.

Objectives: To characterize the utilization and relative timing of Z3A gestational age ICD-10 codes in pregnancies ending in live births (LB) recorded in administrative claims data.

Methods: We used IBM® MarketScan® Commercial and Medicare Supplemental Claims data to identify a cohort of female patients aged 15-49 years with a claim for LB during 10/2016-2/2020 in the US, selecting the earliest qualifying LB for patients with multiple. Patients were required to have 365 days of continuous enrollment and no claims for any pregnancy outcome in the 45 weeks (w) prior to LB. We described demographic and pregnancy characteristics overall and among those with at least one Z3A code in the 42w prior to/including LB. Those with at least one code were further described with respect to Z3A code utilization, and we characterized the range of last menstrual period (LMP) dates estimated from these codes over pregnancy.

Results: Among 378,998 patients identified with a qualifying LB, mean (SD) age at delivery was 30.8 (5.2) years and 32% of deliveries were via cesarean section. 374,617 patients (99%) had at least one Z3A code recorded. Patient demographic and pregnancy characteristics were similar between this group and the full cohort. Among the patients with at least one Z3A code, patients had a mean (SD) 5.1 (3.5) unique codes recorded, and 95% had a code on the date of delivery. Codes corresponding to the first/second/third trimester were recorded for 56/68/99% of patients. The most common first/second/third trimester GA codes recorded were 12w (18%)/20w (24%)/39w (48%). Among 319,442 patients with ≥2 unique GA codes, the median [IQR] discrepancy between earliest and latest estimated LMP was 6 [3-10] days (90% within 26 days).

Conclusions: Nearly all patients with LB had at least one GA-specific code recorded during pregnancy or at delivery. Most patients had multiple GA estimates over pregnancy with a high degree of agreement between estimates; for most patients, earliest and latest estimated LMP were within a week apart. Future work should expand to other pregnancy outcomes and validate LMP estimates against a gold standard to inform algorithm development.