Objectives: The goals of this study were to (1) identify valid variables that correlate with emergency department (ED) crowding and (2) determine a model that could be used to accurately reflect the degree of ED crowding.
Methods: A site sampling form was applied to convenience sampling of 13 community hospitals in California between April 6, 2011, and May 1, 2011. The outcome variable was average perception of crowding by the ED physician and charge nurse on a 100-mm visual analog scale. We focused on 20 candidate predictor variables that represented counts and times in the ED that were collected every 4 hours. A prediction model was developed using multivariable linear regression to determine the measures that predicted ED crowding. A parsimonious model was developed to allow for a clinical useful tool that that explained a significant amount of variability predicted by the full ED crowding model.
Results: A total of 2006 data sets were collected for each of the participating hospitals. A total of 1628 time entries for the hospitals were included in the study. Hospital EDs had censuses ranging from 18 000 to 98 000. Full evaluation was completed on 1489 data sets. Twenty variables were considered for the full model with 7 removed due to multicollinearity. The remaining 13 variables constituted the full model and explained 50.5% of the variability in the outcome variable. Five predictors were found to represent 92% of the variability represented by the full model.
Conclusions: Five variables were highly correlated with community ED crowding and could be used to model the full set of all variables in explaining ED crowding.
Copyright © 2014 Elsevier Inc. All rights reserved.