(920) Does Utilizing EHR in Addition to Administrative Data Better Identify Risk Factors for Maternal Morbidity?

Objective: Reliable methods to ascertain severe maternal morbidity (SMM) patterns from electronic health records (EHR) are necessary to accurately capture data on this important outcome and quality metric. This study investigates the role of utilizing EHR data to enhance billing code identifiers of maternal co-morbidities.

Study Design: This retrospective cohort study includes delivery admissions from four hospitals within a health system over a 1-year period (2019). Co-morbidity components of a validated index (OB-CMI) were identified by established ICD-10 codes as well as clinician-entered medical history and problem lists templated in the EHR. Concordance of individual co-morbidity diagnoses based on ICD-10 codes and EHR data was examined, as was the ability of the OB-CMI score to predict SMM using receiver-operator-characteristic (ROC) curves based on different co-morbidity identification strategies.

Results: 14,267 deliveries were included with co-morbidity frequency by data source demonstrated in the Table. There is heterogeneity across conditions in terms of predominant source of identification (12.0-100% for ICD-10 codes, 8.0-94.6% for EHR data) as well as in cases identified by both methods (4.8-57.6%). The ROC curve demonstrated area-under-the-curve (AUC) for OB-CMI score defined based on ICD-10 codes alone of 0.8055, and in combination with EHR data of 0.8112 (p = 0.43) [Figure 1a]. This was also assessed by individual hospital to evaluate potential impact of hospital-specific documentation and coding patterns, with similar findings [Figure 1b].

Conclusion:

While identification of maternal co-morbidities can be expanded using EHR data, there was no difference in risk prediction for SMM with an OB-CMI score utilizing EHR in addition to administrative data. These findings support the validity of billing codes to capture risk for SMM in hospital-specific and population health research.