MP39: Reducing overcapacity: applying the LEAN model to length of stay in the emergency department

Abstract

Introduction: Recently there have been many studies performed on the effectiveness of implementing LEAN principals to improve wait times for emergency departments (EDs), but there have been relatively few studies on implementing these concepts on length of stay (LOS) in the ED. This research aims to explore the initial feasibility of applying the LEAN model to length-of-stay metrics in an ED by identifying areas of non-value added time for patients staying in the ED. Methods: In this project we used a sample of 10,000 ED visits at the Health Science Centre in St. John's over a 1-year period and compared patients’ LOS in the ED on four criteria: day of the week, hour of presentation, whether laboratory tests were ordered, and whether diagnostic imaging was ordered. Two sets of analyses were then performed. First a two-sided Wilcoxon rank-sum test was used to evaluate whether ordering either lab tests or diagnostic imaging affected LOS. Second a generalized linear model (GLM) was created using a 10-fold cross-validation with a LASSO operator to analyze the effect size and significance of each of the four criteria on LOS. Additionally, a post-test analysis of the GLM was performed on a second sample of 10,000 ED visits in the same 1-year period to assess its predictive power and infer the degree to which a patient's LOS is determined by the four criteria. Results: For the Wilcoxon rank-sum test there was no significant difference in LOS for patients who were ordered diagnostic imaging compared to those who were not (p = 0.6998) but there was a statistically significant decrease in LOS for patients who were ordered lab tests compared to those who were not (p = 2.696 x 10-10). When assessing the GLM there were two significant takeaways: ordering lab tests reduced LOS (95% CI = 42.953 - 68.173min reduction), and arriving at the ED on Thursday increased LOS significantly (95% CI = 6.846 – 52.002min increase). Conclusion: This preliminary analysis identified several factors that increased patients’ LOS in the ED, which would be suitable for potential LEAN interventions. The increase in LOS for both patients who are not ordered lab tests and who visit the ED on Thursday warrant further investigation to identify causal factors. Finally, while this analysis revealed several actionable criteria for improving ED LOS the relatively low predictive power of the final GLM in the post-test analysis (R2 = 0.00363) indicates there are more criteria that influence LOS for exploration in future analyses.