LO52: Combination of easily measurable real time variables to predict ED crowding. (15th May 2017)

• Record Type: Journal Article
• Authors: Clouston, R.V.; Howlett, M.; Stoica, G.; Fraser, J.; Atkinson, P.R.
• Is Part Of: CJEM. Volume 19(2017:May)Supplement 1
• Page Start: S45
• Page End: S46
• Publication Date: 2017-05-15
• Subjects: emergency department, -- crowding, -- overcapacity; Emergency Treatment -- Periodicals; Emergency Medicine -- Periodicals; Emergency medical services -- Canada -- Periodicals; Medical emergencies -- Canada -- Periodicals; Emergency medical services; Medical emergencies; Canada; Periodicals; 616.02505
• DOI: 10.1017/cem.2017.114
• Languages: English
• ISSNs: 1481-8035

Abstract

Abstract : Introduction: Almost every domain of quality is reduced in crowded emergency departments (ED), with significant challenges around the definition, measurement and interventions for ED crowding. We wished to determine if a combination of 3 easily measurable variables could perform as well as standard tools (NEDOCS score and a NEDOCS-derived LOCAL tool) in predicting ED crowding at a tertiary hospital with 57, 000 visits per year.Methods: Over a 2-week period, we recorded ED crowding predictor variables and calculated NEDOCS and LOCAL scores. These were compared every 2 hours to a reference standard Physician Visual Analog Scale (range 0 to 10) impression of crowding to determine if any combination of variables outperformed NEDOCS and LOCAL (crowded=5 or greater). Five numeric variables performed well under univariate analysis: i) Total ED Patients; ii) Patients in ED beds + Waiting Room; iii) Boarded Patients; iv) Waiting Room Patients; v) Patients in beds To Be Seen. These underwent multivariate, log regression with stratification and bootstrapping to account for incomplete data and seasonal and daily effect.Results: 143 out of a possible 168 observations were completed. Two different combinations of 3 variables outperformed NEDOCS and LOCAL. The most powerful combination was: Boarded Patients; plus Waiting Room Patients; plus Patients in beds To Be Seen, with Sensitivity 81% and Specificity 76% (r=0.844, β=0.712, p<0.0001, strong positive correlation). This compared favourably with NEDOCS and LOCAL, each with Sensitivity 71% and Specificity 64%[PA1] (r=0.545 and r=0.640 respectively). We will also present a sensitivity and specificity analysis of all combinations of predictor variables, using various reference standard cut-offs for crowding.Conclusion: A combination of 3 easily measurable ED variables (Boarded Patients; plus Waiting Room Patients; plus Patients in beds To Be Seen) performed better than the validated NEDOCS tool and a NEDOCS-derived LOCAL score at predicting ED crowding. Work is on going to design a simple tool that can predict crowding in real time and facilitate early interventions. Correlation with ED system and clinical outcomes should be studied in different ED environments.