Revisits After Emergency Department Discharge for Conditions with High Disposition-Decision Variability at Hospitals with High and Low Discharge Rates

Abstract

Introduction: The first proposed emergency care alternative payment model seeks to reduce avoidable admissions from the emergency department (ED), but this initiative may increase risk of adverse events after discharge. Our study objective was to describe variation in ED discharge rates and determine whether higher discharge rates were associated with more ED revisits. Methods: Using all-payer inpatient and ED administrative data from the California Office of Statewide Health Planning and Development (OSHPD) 2017 database, we performed a retrospective cohort study of hospital-level ED discharge rates and ED revisits using conditions that have been previously described as having variability in discharge rates: abdominal pain; altered mental status; chest pain; chronic obstructive pulmonary disease exacerbation; skin and soft tissue infection; syncope; and urinary tract infection. We categorized hospitals into quartiles for each condition based on a covariate-adjusted discharge rate and compared the rate of ED revisits between hospitals in the highest and lowest quartiles. Results: We found a greater than 10% difference in the between-quartile median adjusted discharge rate for each condition except for abdominal pain. There was no significant association between adjusted discharge rates and ED revisits. Altered mental status had the highest revisit rate, at 34% for hospitals in the quartile with the lowest and 30% in hospitals with the highest adjusted discharge rate, although this was not statistically significant. Syncope had the lowest rate of revisits at 16% for hospitals in both the lowest and highest adjusted discharge rate quartiles. Conclusion: Our findings suggest that there may be opportunity to increase ED discharges for certain conditions without resulting in higher rates of ED revisits, which may be a surrogate for adverse events after discharge.