ABSTRACT
Study Objectives: Emergency department (ED) chief complaint data has several potential applications, including quality measurement, syndromic surveillance, operations, research, and education. However, there are no consistent methods to categorize ED chief complaints or evaluate their association with other ED outcomes, which has limited the utility of this type of data. To advance chief complaint data standardization, we report the initial results of a novel national ED chief complaint dataset under development in the Veteran Health Administration (VA). We identified common presenting ED chief complaints, characterized their associations with an ED discharge diagnosis of an emergent condition, and related admission rates. Method(s): This was a retrospective observational study of VA ED visits in FYs 2018-2020. A Natural Language Processing (NLP) program based on cTAKES, an Apache open source project, was applied to the original text of VA ED chief complaints. Results were mapped to Concept Unique Identifiers (CUI) in the Unified Medical Language System (UMLS). Multiple concepts could be identified from a chief complaint text entry. ED discharge diagnoses were defined by ICD-10 codes. Emergent diagnoses were selected based on a previously established list of codes for Emergency Care Sensitive Conditions (ECSCs), which are acute illnesses that require timely, quality emergency care to improve morbidity and mortality. Result(s): A total of 5,898,684 VA ED visits were identified with at least one ED chief complaint and a discharge diagnosis. 59% of visits had 1 chief complaint concept, 26% of visits had 2 concepts, 10% had 3 concepts, 4% had 4 concepts, and 1% had 5 concepts. The 10 most common chief complaints, associated rates of an ECSC discharge diagnosis, and respective admission rates for both ECSC and non-ECSC ED visits are depicted in Table 1. Among the most common chief complaints, dyspnea had a majority of ED visits with an ECSC diagnosis, likely due to the COVID-19 pandemic. Otherwise, rates of ECSC visits varied from 24% (coughing) to 5% (back pain). However, admission rates for ECSC visits ranged from 67% (abdominal pain) to 15% (pharmaceutical preparations). Conclusion(s): To our knowledge, this national ED chief complaint dataset is the largest in the country, and representative of a diverse patient population (ie by age, region, rurality). Initial work has highlighted areas for refinement of this dataset. Further work is ongoing to examine combinations of chief complaints to better predict ECSC diagnosis and admission rate given the variation in initial findings. Additionally, ongoing work to improve context detection and reduce mapping errors is underway, and will improve utility in multiple applications. [Formula presented] No, authors do not have interests to disclose Copyright © 2022