The Impact of Leveraging an Emergency Department Observation Unit for Pandemic Response on Observation Outcomes: A Retrospective Observational Difference-In-Differences Study.

Objectives: It is unknown what the impact of leveraging an emergency department observation unit (EDOU) for alternative purposes during a public health emergency might be. The study objective was to determine the impact of shifting observation patients from an EDOU to inpatient beds on hospital observation outcomes-length of stay (LOS), total cost, and inpatient admission rate. Methods: This is a retrospective observational difference-in-differences study across 4 hospitals within a large academic health system in Atlanta, Georgia. All had type 1 EDOUs managed by the department of emergency medicine. At the onset of the pandemic, the intervention hospital converted the EDOU to a COVID unit, displacing all observation patients to inpatient beds (April 2020). The remaining 3 hospitals maintained operations and served as controls. Data were abstracted from hospital clinical and financial observation databases. Comparisons were made between the preintervention period (August 2019 to March 2020) and the intervention period (April 2020), and between intervention and control hospitals. The primary outcomes included: average per patient cost, total (observation + inpatient) LOS, observation LOS, and inpatient admission rate. Outcomes were evaluated using mixed-effects gamma regressions and logistic regressions. Analyses were adjusted for age, sex, triage level, diagnosis, and procedural International Classification of Diseases codes. We present mean ratios (MRs) for continuous outcomes, odds ratios for binary outcomes, and 95% CIs. The difference-in-differences effect was tested using a time-by-site interaction term. Results: Over the study period there were 22,263 observation patients (3080 intervention patients, 19,183 control patients). Average age was 58 (43 to 72), and 56% were female. Five most common Clinical Classification System (CCS) conditions were chest pain (10.8%), hypertension (5.4%), fluid electrolyte disorders (3.7%), syncope (2.5%), and abdominal pain (2.4%). The difference-in-differences interaction was significant for all outcomes. Total cost for the control hospitals remained stable ($3876.6 vs $3997.1; MR =1.03, 95% CI: 0.99-1.08) but increased from $3449.1 to $4645.5 (MR = 1.35, 95% CI: 1.20-1.52) for the intervention hospital. For total LOS, control hospitals saw a significant decrease in LOS (36.4 h vs 34.3 h, MR = 0.94, 95% CI: 0.90-0.99); the intervention hospital saw a nonsignificant increase (32.1 h vs 36.6 h, MR = 1.14, 95% CI: 1.00-1.30). For only observation LOS, control hospitals saw a significant decrease (23.8 h vs 19.8 h; MR = 0.83, 95% CI: 0.79-0.87), whereas the intervention hospital saw a significant increase (21.3 h vs 26.1 h; MR = 1.22, 95% CI: 1.09-1.38). Finally, for admission rates, control hospitals remained stable (16.4% vs 16.8%; odds ratio of 1.02, 95% CI: 0.88-1.19), whereas the intervention hospital saw an increase from 13.3% to 21.0% (odds ratio of 1.74, 95% CI: 1.16-2.61). The findings were similar when studying only discharged patients. Conclusion: Displacing EDOU observation patients to inpatient beds to accommodate disaster or surge patients is associated with increases in observation patient cost, LOS, inpatient bed use, and inpatient admission rate.