A Clinical Outcomes Data Archive for a Comprehensive Fetal Diagnosis and Treatment Center

Data on near- and long-term clinical outcomes are critical for the care of all maternal-fetal patients presenting to a fetal center. This is especially important since physiologic and neurodevelopmental attributes do not manifest until later childhood when multilevel (e.g., individual, family, policy) factors have a direct influence on health outcomes. Electronic health records (EHRs) create opportunity for efficient data collection. However, documentation structures are not designed for acquisition of key attributes, and changes over time and between-clinician differences can affect resultant output. Therefore, EHR derived datasets have limited ability to accurately characterize the clinical presentation and care trajectory of patients with congenital anomalies. In addition, in most systems, the fetus lacks a digital identity and requires relinking fetal attributes documented in the maternal chart to those from the pediatric EHR. This conundrum amplifies in the setting of multiple gestation, returning maternal patients, and pregnancies with fetal demise. Moreover, current data capture systems result in incomplete abstraction of variables that may confound, mediate, or moderate critical associations. Our objective was to develop and implement a prospective data capture platform to transform EHR data into an analytic-grade database for multipurpose use.

A unified platform for longitudinal follow-up of maternal-child dyads cared for at our fetal center, named the Clinical Outcomes Data Archive (CODA), was constructed. CODA was designed using a data dictionary based on multidisciplinary and interprofessional expert input, a relational identity for each patient, fetus, and pregnancy, and a process by which EHR-sourced and chart-abstracted data are validated by a well-trained team. Descriptive analyses were performed for data acquired between July 2022 and July 2023, and a comparison of studies before and after implementation of CODA is presented.

5,394,106 data points were validated for 7,662 patients across 12 conditions. 2% of data points were found to be unreliable or undocumented. 91% of data points were sourced from the EHR. Eighty-five percent of condition-specific variables required manual chart abstraction. The study conducted with CODA was able to contribute to 18 other studies. CODA successfully merges EHR-sourced and manually abstracted documentation for longitudinal study of the maternal-child dyad.