(370) Validation of Maternal Co-Morbidity Diagnoses Using Differential Data Extraction Strategies Across a Large Health System

Objective: Determining maternal co-morbidity burden during delivery admissions is essential for research, public health initiatives, and quality improvement related to reducing maternal morbidity. This study compares and validates different strategies for identifying maternal co-morbidities in an integrated health system.

Study Design: This is a retrospective cohort study of all deliveries in 4 hospitals in a health system (2 community, 2 academic) in 2019. The accuracy of utilizing ICD-10 codes versus templated electronic health record (EHR) elements including clinician-entered medical history and problem list were compared for 4 randomly selected co-morbidities: ischemic heart disease, cardiomyopathy, alcohol abuse disorder, and placenta accreta. The charts of all patients with an identified co-morbidity by either ICD-10 or EHR data were reviewed and the rate of verified diagnoses were compared between data sources (ICD-10, EHR) and across hospitals.

Results: Of 14,267 deliveries, 161 patients had one of the co-morbidities identified by either ICD-10 or EHR, with differing rates by hospital (Table legend). The breakdown of each co-morbidity diagnosis by data source and rate of chart review confirmation is displayed in the Table. Utilizing EHR data significantly increased the ability to identify cases over ICD-10 codes alone, identifying almost 50% more cases of ischemic heart disease, cardiomyopathy and alcohol abuse. Cases identified by both ICD-10 and EHR data were the most accurate, and there were no significant differences in these findings by hospital.

Conclusion: There is substantial variation in the ability to identify co-morbidities based on ICD-10 vs. templated EHR data. Utilizing EHR data identifies the most cases and is superior to ICD-10 codes alone.  Cases identified by both measures are most accurate, though this methodology has limited sensitivity. These findings suggest that a hybrid strategy should be used by healthcare organizations and researchers when studying patient co-morbidities.