Testing the Sensitivity of the New York University Emergency Department Algorithm to Detect Changes in Emergency Department Usage Patterns

Background/Purpose: The Emergency Department Algorithm (EDA) is a tool developed by New York University Center for Health and Public Service Research that uses administrative discharge data to categorize emergency department (ED) visits by need for emergent care. Our analysis explains why various EDA users have observed EDA-generated outputs that do not change significantly with significant changes in patient populations and/or conditions in the associated healthcare system. Methods: Mathematical simulation was used to estimate EDA outcomes across all possible (hypothetical) patient populations; sensitivity analysis was used to demonstrate the changes in EDA outcomes for changes in a real-world baseline patient population. The baseline population consists of all ED cases from a community hospital's administrative database that did not result in inpatient admission and for which discharge ICD-9s (the input into the EDA) were available (n = 276,862). Results/Outcomes: The vast majority of possible EDA outputs do not differ significantly from one another. Thus, the vast majority of possible EDA outputs from the algorithm are so concentrated that they do not significantly differ even with significant changes to inputs to the algorithm. Using an actual patient population, the most extreme hypothetical increases in high-probability non-emergent, primary care treatable, or preventable visits would produce only a 9% decrease in EDA-estimated emergent cases. Conclusions: The EDA is being used nationally to assess ED usage; however, it is insufficiently sensitive to changes in ED use patterns to be of use to policymakers or hospital administrators in addressing possibly unnecessary ED use.