Perinatal & Pediatric
The Validity of Diagnostic Codes for Severe Maternal Morbidity from Electronic Medical Record Data Kelli Ryckman* Kelli Ryckman Eva Sileo Claire Carlson Stephanie Radke
Background: The rates of severe maternal morbidity (SMM) have been increasing in the United States. Use of electronic health record (EHR) data for research on and surveillance of SMM is important for identifying temporal trends and monitoring quality improvement initiatives. There has been an absence of literature evaluating the validity of the International Classification of Diseases, Tenth Revision (ICD-10) coding system in EHR data for accurately identifying cases of SMM.
Objective: To determine the validity of ICD-10 codes for non-transfusion related SMM.
Study Design: This retrospective cohort study using EHR data for deliveries occurring at a single midwestern tertiary care unit between July 1st 2016 and June 30th 2019. A total of 6,456 deliveries were evaluated for the presence of ICD-10 diagnostic codes for one or more of sixteen preselected clinical indicators of non-transfusion related SMM. Two trained primary reviewers reviewed the discharge summaries and clinical notes for the 226 deliveries identified as having an ICD-10 code for one or more clinical indicator for SMM. The percentage of correct codes for each of the 16 preselected clinical SMM indicators was calculated yielding a positive predictive value (PPV) and 99% confidence interval (CI).
Results: The overall number of correctly assigned ICD-10 codes was 149 out of 309 (PPV: 48.2%; 99% CI: 41.0%-55.5%). The most prevalent (>10 occurrences) clinical indicators of SMM with the lowest <40% PVV were acute renal failure, disseminated intravascular coagulation and puerperal cerebrovascular disorders. When removing these 3 clinical indicators the number of correctly assigned ICD-10 codes was 123 out of 176 (PPV: 70.3%, 99% CI: 60.8%-78.3%).
Conclusions: In general, care should be taken when using ICD-10 codes to classify cases of SMM, particularly for select conditions with low PPV’s. This data can inform how to develop more nuanced and accurate ways for identifying cases of SMM in EHR data.