P119: Emergency department census is useful as a real-time measure of crowding

Abstract

Introduction: Crowding is associated with poor patient outcomes in emergency departments (ED). Measures of crowding are often complex and resource-intensive to score and use in real-time. We evaluated single easily obtained variables to establish the presence of crowding compared to more complex crowding scores. Methods: Serial observations of patient flow were recorded in a tertiary Canadian ED. Single variables were evaluated including total number of patients in the ED (census), in beds, in the waiting room, in the treatment area waiting to be assessed, and total inpatient admissions. These were compared with Crowding scores (NEDOCS, EDWIN, ICMED, three regional hospital modifications of NEDOCS) as predictors of crowding. Predictive validity was compared to the reference standard of physician perception of crowding, using receiver operator curve analysis. Results: 144 of 169 potential events were recorded over 2 weeks. Crowding was present in 63.9% of the events. ED census (total number of patients in the ED) was strongly correlated with crowding (AUC = 0.82 with 95% CI = 0.76 - 0.89) and its performance was similar to that of NEDOCS (AUC = 0.80 with 95% CI = 0.76 - 0.90) and a more complex local modification of NEDOCS, the S-SAT (AUC = 0.83, 95% CI = 0.74 - 0.89). Conclusion: The single indicator, ED census was as predictive for the presence of crowding as more complex crowding scores. A two-stage approach to crowding intervention is proposed that first identifies crowding with a real-time ED census statistic followed by investigation of precipitating and modifiable factors. Real time signalling may permit more standardized and effective approaches to manage ED flow.

Association of emergency department crowding with inpatient outcomes

“…Our data also did not permit us to directly measure ED crowding; we instead followed the previous literature 49,50 using ED census as a proxy. While this may introduce measurement error, we do not think that it biases results since ED census has been shown to closely approximate ED crowding 49,53 . Furthermore, this error is likely minimal because our ED census measure is based on a hospital‐level percentile, rather than mean, and we explicitly controlled for key factors that may influence ED crowding, including seasonality, day of the week, and hospital fixed effects, which account for any time‐invariant hospitals characteristics that influence ED crowding.…”

“…Although more precise measures of ED crowding exist, 51,52 these measures require the information we lack, such as triage category. However, ED census approximates these more precise measures 49,53 . We were not able to measure ED length of stay, so we assumed that all patients in the ED were discharged, admitted, or transferred on the same day.…”

“…However, ED census approximates these more precise measures. 49,53 We were not able to measure ED length of stay, so we assumed that all patients in the ED were discharged, admitted, or transferred on the same day. Supporting this assumption is research suggesting that approximately 3% of ED discharges take longer than 24 h, 54 and our own analyses suggest approximately 93% of ED visits are discharged, admitted, or transferred the same day (Appendix 2).…”