Time Series Analysis of Emergency Department Length of Stay per 8-Hour Shift

Niels K. Rathlev, MD; Dan Obendorfer, MBA; Laura F. White, PhD; Casey Rebholz, MPH; Brendan Magauran, MD, MBA; Willie Baker, MD; Andrew Ulrich, MD; Linda Fisher, RN; Jonathan Olshaker, MD

Disclosures

Western J Emerg Med. 2012;13(2):163-168. 

In This Article

Abstract and Introduction

Abstract

Introduction: The mean emergency department (ED) length of stay (LOS) is considered a measure of crowding. This paper measures the association between LOS and factors that potentially contribute to LOS measured over consecutive shifts in the ED: shift 1 (7:00 AM to 3:00 PM), shift 2 (3:00 PM to 11:00 PM), and shift 3 (11:00 PM to 7:00 AM).
Methods: Setting: University, inner-city teaching hospital. Patients: 91,643 adult ED patients between October 12, 2005 and April 30, 2007. Design: For each shift, we measured the numbers of (1) ED nurses on duty, (2) discharges, (3) discharges on the previous shift, (4) resuscitation cases, (5) admissions, (6) intensive care unit (ICU) admissions, and (7) LOS on the previous shift. For each 24- hour period, we measured the (1) number of elective surgical admissions and (2) hospital occupancy. We used autoregressive integrated moving average time series analysis to retrospectively measure the association between LOS and the covariates.
Results: For all 3 shifts, LOS in minutes increased by 1.08 (95% confidence interval 0.68, 1.50) for every additional 1% increase in hospital occupancy. For every additional admission from the ED, LOS in minutes increased by 3.88 (2.81, 4.95) on shift 1, 2.88 (1.54, 3.14) on shift 2, and 4.91 (2.29, 7.53) on shift 3. LOS in minutes increased 14.27 (2.01, 26.52) when 3 or more patients were admitted to the ICU on shift 1. The numbers of nurses, ED discharges on the previous shift, resuscitation cases, and elective surgical admissions were not associated with LOS on any shift.
Conclusion: Key factors associated with LOS include hospital occupancy and the number of hospital admissions that originate in the ED. This particularly applies to ED patients who are admitted to the ICU.

Introduction

Emergency department (ED) mean length of stay (LOS) per patient measured from the patient's arrival to departure has been promoted as a surrogate indicator of crowding in the absence of a standard or universal definition. It is also frequently considered a key process indicator for performance improvement and clinical and operational efficiency.[1] From the patient's perspective, total LOS and long wait times to see a physician or for test results create the sense that the ED is busy and crowded; however, a major limitation of LOS as an indicator of crowding is the retrospective nature of the measure; it cannot readily be used to manage flow real time.

It is important that variables associated with crowding are identified for optimal management of flow in the ED. It is of particular interest to define variables that are amenable to administrative change and redesign in order to reduce variability in performance.[2] Asplin et al developed an inputthroughput- output conceptual model of ED crowding which forms the basis for our selection of factors for study.[3] Input factors, such as the number of unscheduled ambulatory care ED visits and emergency medical services transports, cannot be controlled except by diverting ambulances away from the ED. These factors were therefore not included as covariates in the current analysis. Throughput relates to factors that are influenced by flow processes in the emergency, laboratory, and radiology departments from the time of triage to the time of disposition. Nursing staffing was included as a throughput factor in this study, as it can be controlled by administrative design. Mandatory ED nursing staffing ratios of 1 nurse for every 4 patients has led to a reduction in wait time to ED bed and ED bed to departure in the State of California.[5] While staffing patterns would appear to be an important throughput factor, previous studies have not found attending physician staffing to be a statistically significant factor influencing LOS.[4] Consequently, attending physician staffing was not included in the present analysis. Output factors relate to the disposition of patients to home, chronic care facilities, or inpatient beds. Intense competition for the latter may occur especially on weekdays, when patients are frequently admitted to an inpatient bed either before or after elective surgical procedures; this is an example of artificial variability that is amenable to change on an institutional level. Similarly, the expediency of the departure process from the ED is an important output factor and the number of discharged patients per shift (ED discharges) leaving for home, an inpatient bed, or for another facility was therefore incorporated as a covariate in the study.

Studies have documented statistically significant associations between LOS measured over 24-hour periods and hospital occupancy, the number of ED admissions, and the number of elective surgical cases that were admitted directly to an inpatient bed while bypassing the ED.[4] As recommended by other investigators, our goal was to measure the associations between throughput and output factors and LOS in more discreet time periods than previously reported.[6,7] Since ED crowding and volume vary greatly during a given 24-hour period, we measured the independent variables during 3 separate 8-hour shifts per 24-hour period, when possible. Our purpose was to discern which covariates of interest were associated with LOS and, when relevant, whether this relationship was present during all shifts or only specific 8-hour shifts.

processing....