Emergency department crowding and hospital patient flow : influential factors and evidence-Informed solutions

Abstract

Background: Internationally, one of the biggest challenges in emergency departments is crowding – when demand for emergency care exceeds its capacity in resources and timeliness. Crowding is associated with increased morbidity, mortality, cost and decreased patient and health-care worker satisfaction. Consequently, governments in the United Kingdom, Australia and some Canadian provinces have implemented time targets for emergency department length-of-stay, but have had difficulty achieving them. Although there is much literature on etiology and solutions for emergency department crowding, there is a lack of evidence-informed policy and cost-effectiveness analyses on solutions for reaching targets. Which are the most appropriate interventions for the individual hospital? What factors associated with failing targets should the hospital prioritize?

Objectives: The objectives of this thesis were to find factors strongly associated with failing to meet emergency department (ED) length-of-stay (LOS) targets and rigorously evaluate ED crowding solutions. The first two objectives were to determine the effectiveness of a supplementary physician-nurse team at triage (MDRNSTAT) on EDLOS, quality of care and its cost-effectiveness from the hospital and patient perspective. The third objective was to determine predictors of target time failure for discharged high acuity patients and intensive care unit (ICU) admissions at Sunnybrook Health Sciences Center (Sunnybrook), an academic tertiary-level hospital in Ontario, Canada. Finally, we compared performance and factors predicting failure of government time targets between 2012 and 2013 at Sunnybrook and between Sunnybrook and Austin Health (Austin), an Australian academic tertiary level hospital.

Methods: Study I was a pragmatic cluster randomized trial comparing shifts with and without the MDRNSTAT. The primary outcome was emergency department length-of-stay (EDLOS) for non-consulted discharges. Secondary outcomes included EDLOS for patients initially seen by the emergency department, and subsequently consulted and admitted, patients reaching government-mandated thresholds, time to initial physician assessment, left-without being seen rate, time to investigation, and measurement of harm. Study II was a cost-effectiveness evaluation of the MDRNSTAT.

Study III was a retrospective, observational study of 2012 Sunnybrook Hospital (Canada) emergency department data using multivariable logistic regression. The main outcome measure was failure to reach government EDLOS targets for high acuity discharges and ICU emergency admissions. Study IV was a retrospective, observational study of 2012, 2013 Sunnybrook Hospital (Canada) and 2012 Austin Health (Australia) administrative data using descriptive statistics and multivariable logistic regression. The main outcome measure was reaching ED time targets by subgroup: admissions, low and high acuity discharges. Secondary outcomes for Study III and IV were predicting failure of government targets and a select group of hospital factors.

Results: For Study I, the MDRNSTAT decreased discharged, non-consulted, high acuity patients EDLOS by 34 minutes [CI: 16 to 52]. For discharged, non-consulted, low acuity patients, EDLOS decreased by 52 minutes [CI: 38 to 65]. Physician initial assessment duration (PIAD) decreased by 53 minutes [CI: 48 to 57]. The MDRNSTAT-associated shifts’ leftwithout- being-seen rate was 1.5% versus 2.2% for the control (p=0.06). No patients returned to the emergency department after being discharged by the MDRNSTAT at triage. From Study II, the added cost of the MDRNSTAT was $3,597.27 [$1729.47 to ∞] per additional patient-seen, $75.37 [$67.99-$105.30] per physician-initial-assessment hour saved and $112.99 [$74.68 – $251.43] per EDLOS hour saved. From the hospital perspective, the costbenefit ratio was 38.63 [18.96 to ∞] and net present value of -$447,996 [-$435,646 to - $459,900]. For patients, the cost-benefit ratio for satisfaction was 2.8 [2.3-4.6].

For Study III, factors predicting EDLOS target failure for Sunnybrook’s discharged high acuity patients were: having PIAD>2hrs (OR 5.63 [5.22-6.06]), consultation request (OR 10.23 [9.38-11.14]), a MRI (OR 19.33 [12.94-28.87]), CT (OR 4.24 [3.92-4.59]), or US (OR 3.47 [3.13-3.83]). For ICU admissions, factors predicting EDLOS target failure were: bed request duration (BRD)>6hrs (OR 364.27 [43.20-3071.30]) and access block (AB)>1hr (OR 217.27 [30.62-1541.63]). For discharged low acuity patients, factors predicting failure for the 4hr target were: PIAD> 2hrs (OR 15.80 [13.35-18.71]), consultation (OR 20.98 [14.10- 31.22]), TnI (OR 13.37 [6.30-28.37]), MRI (OR 31.68 [6.03-166.54]), or CT (OR 16.48 [10.07-26.98]). Study IV found that the Australian hospital, Austin Health, succeeded for all targets except for low acuity discharges. Sunnybrook failed all time targets. For low acuity discharges, Austin factors for failing government targets were PIAD>2 hrs (OR 11.62 [10.40- 12.99]), consultation (OR 6.99 [5.83-8.38]) and CT (OR 7.16 [5.19-8.66]). For high acuity discharges, Austin factors were evening shift (OR 4.09 [3.40-4.93]), consultation (OR 8.82 [7.62-10.21) and MRI (OR 8.16 [3.07-21.70]). For admissions, Austin factors were AB>1hr (OR 57.35 [39.31-83.67]) and BRD>6hrs (OR 46.07 [33.23-63.88]). Comparing 2012 to 2013 at Sunnybrook, the factors for failing targets remained similar for admissions, low and high acuity discharges.

Conclusions: The MDRNSTAT reduced delays and left-without-being-seen rate without increased return visits or jeopardizing urgent care of severely ill patients; however, it was not a cost-effective daytime strategy at Sunnybrook. The MDRNSTAT would be more feasible during time periods with higher access block, such as the afternoon to late evening. Sunnybrook factors predicting failure of government targets for high acuity discharges and ICU admissions were hospital-controlled. The Australian hospital out-performed the Canadian hospital on government time targets. Factors predicting failure of government targets remained consistent over time in the same hospital but were different between hospitals. Irrespective of time and location, factors most associated with target failure were hospital-controlled. Therefore, hospitals should individualize their approach to shortening EDLOS by analyzing its patient population and resource demands.

Study I Trial registration number: NCT00991471 ClinicalTrials.gov