Objectives: The objective of this study is to evaluate the impact of emergency department (ED) crowding levels on patient admission decisions and outcomes.
Methods: A retrospective study was performed based on 2-year electronic health record data from a tertiary care hospital ED in Alberta, Canada. Using modified Poisson regression models, we studied the association of patient admission decisions and 7-day revisit probability with ED crowding levels measured by: 1) the total number of patients waiting and in treatment (ED census), 2) the number of boarding patients (boarder census), and 3) the average physician workload, calculated by the total number of ED patients divided by the number of physicians on duty (physician workload census). The control variables included age, gender, treatment area, triage level, and chief complaint. A subgroup analysis was performed to evaluate the heterogeneous effects among patients of different acuity levels.
Results: Our dataset included 141,035 patient visit records after cleaning from August 2013 to July 2015. The patient admission probability was positively correlated with ED census (relative risk [RR] = 1.006, 95% confidence interval [CI] = 1.005 to 1.007) and physician workload census (RR = 1.029, 95% CI = 1.027 to 1.032), but inversely correlated with boarder census (RR = 0.991, 95% CI = 0.989 to 0.993). We further found that the 7-day revisit probability of discharged patients was positively associated with boarder census (RR = 1.009, 95% CI = 1.004 to 1.014).
Conclusions: Patient admission probability was found to be directly associated with ED census and physician workload census, but inversely associated with the boarder census. The effects of boarder census and physician workload census were stronger for patients of triage levels 3-5. Our results suggested that (i) insufficient physician staffing may lead to unnecessary patient admissions; (ii) too many boarding patients in ED leads to an increase in unsafe discharges, and as a result, an increase in 7-day revisit probability.
Authors: Huiyin Ouyang, Junyan Wang, Zhankun Sun, Eddy Lang
Huiyin Ouyang - email@example.com
Preliminary data gathering/ baseline