An empirical study of the behavior of patients who leave the emergency department without being seen

• Author: Jennifer Wiler,Nicole DeHoratius,Ehsan Bolandifar,Tava Olsen
• Published date: 01 July 2019
• Date: 01 July 2019
• DOI: http://doi.org/10.1002/joom.1030

RESEARCH ARTICLE

An empirical study of the behavior of patients who leave the

emergency department without being seen

Ehsan Bolandifar

1

| Nicole DeHoratius

2

| Tava Olsen

3

| Jennifer Wiler

4

1

The Chinese University of Hong Kong,

Hong Kong

2

Booth School of Business, University of

Chicago, Chicago, Illinois

3

University of Auckland, Auckland, New

Zealand

4

University of Colorado, Aurora, Colorado

Correspondence

Ehsan Bolandifar, The Chinese University

of Hong Kong, No.12 Nam Cheung Street,

Cheng Yu Tung Building, Room 922, Hong

Kong, Hong Kong.

Email: ehsan@cuhk.edu.hk

Handling Editor: Anita Tucker

Abstract

Queue abandonment has a significant impact on system performance. However,

the key drivers for abandonment, particularly in observable systems, are not well

understood. To better inform our understanding of abandonment behavior, we

study the effect of three operational drivers of abandonment from a hospital emer-

gency department (ED), namely, waiting time, queue length, and observed service

rate. We confirm that all three factors affect a patient's propensity for leaving the

waiting area without being seen by a physician (LWBS), that is, abandoning the

queue. Further, these factors interact with each other in a nonlinear fashion. Both

ED crowding and observed service rate influence a patient's perception of waiting

time. Moreover, patients are not homogenous in their abandonment response, and

we observe behavior that is distinct for patients with severe conditions. Specifi-

cally, patients who report to a congested ED with more severe conditions are more

inclined to abandon the ED early in the process compared to patients with less

severe conditions. Further, we observe that patients with severe conditions who

elect to remain in the crowded ED exhibit less sensitivity to waiting time and

observed service rate than other patient types. We discuss the implications of this

observed abandonment behavior on ED management.

KEYWORDS

abandonment, emergency department crowding, empirical study, health-care operations, left without

being seen

1|INTRODUCTION

Patients who leave the emergency department (ED) without

being seen by a physician risk adverse health consequences

due to treatment delays (e.g., Hunt, Weber, & Shoastack,

2006). Approximately 5 % of patients who leave without

being seen present to a hospital within 14 days of their aban-

donment with more severe conditions (Baker, Stevens, &

Brook, 1991). The rate of patient ED abandonment varies

substantially in the United States, ranging from 1 % to

20.3% of patient arrivals to the ED (Hsia et al., 2011). The

factors that drive patients to leave the ED without being seen

are not well understood and reflect a broader lack of under-

standing of queue abandonment, particularly within observ-

able systems.

Although Lu, Olivares, Musalem, and Achilkrut (2013)

and Batt and Terwiesch (2015) identify several factors that

affect customers' abandonment behavior in physical queues

(considering balking and reneging behavior, respectively),

most of the literature on abandonment focuses on “invisible”

queues, such as call centers (e.g., Gans, Koole, &

Mandelbaum, 2003). This literature assumes that abandon-

ment is primarily driven by queue waiting time. However, in

physical queues, the psychology of queueing suggests that

Received: 25 March 2018 Revised: 26 March 2019 Accepted: 18 April 2019

DOI: 10.1002/joom.1030

430 © 2019 Association for Supply Chain Management, Inc. J Oper Manag. 2019;65:430–446.wileyonlinelibrary.com/journal/joom

abandonment behavior may be driven by more than waiting

time alone (e.g., Maister, 1985). Berry, Seiders, and Grewal

(2002) provide a survey of papers on queueing psychology

(or “service convenience”as they refer to it), and Bitran,

Ferrer, and Oliveira (2008) incorporate the psychology of

waiting into the design of the queue.

We study the operational factors that affect abandonment

behavior in a physical queue; namely, the waiting area of the

ED, using data collected from the electronic patient tracking

system of a large, adult only, urban ED. We examine aban-

donment behavior—LWBS—as a function of ED crowding

(i.e., queue length) and observed service rate, in addition to

waiting time. Batt and Terwiesch (2015) partially character-

ize the effect of these operational factors on a patient's

LWBS probability. We extend their findings to show that

the impact of waiting time on abandonment depends on the

observed service rates and provide further details on the sen-

sitivity of patients to waiting times at different level of ED

crowding.

Mandelbaum, Sakov, and Zeltyn (2000) and Mandelbaum

and Zeltyn (2013) demonstrate that queueing behavior is not

homogenous across all groups. They highlight empirically

that the willingness of customers to wait differs depending on

the extent of service required. For ED managers, it is of

utmost importance to know whether a patient's LWBS behav-

ior (i.e., the propensity to abandon) depends on the extent of

service required because this can have important implications

for ED design. We will show that acuity level (which is a

measure of the extent of service required) does matter in char-

acterizing abandonment behavior, but that the relationship is

not straightforward.

Our study contributes to the literature on queueing and

abandonment behavior in several ways. We provide evi-

dence that service rate has a significant impact on abandon-

ment behavior. To the best of our knowledge, please refer to

Lu et al. (2013), the impact of observed service rate is here-

tofore unexplored in the empirical literature. Moreover, our

findings contribute to the growing literature on process

transparency (e.g., Buell, Kim, & Tsay, 2017; Buell &

Norton, 2011) by characterizing the effect of observed ser-

vice rates and crowding levels on patients' LWBS behavior.

We examine the interaction effects between waiting time,

service rate, and ED crowding and provide new insights into

LWBS behavior. We observe and characterize the heteroge-

neous abandonment behavior among different patient types.

Finally, we summarize the implications of our findings for

ED management, with a particular focus on FastTrack

scheduling.

The rest of this article is organized as follows. In

Section 2, we review the related literature and present our

hypotheses on patient behavior. We detail the setting for our

empirical study and the data gathered in Section 3. Section 4

presents our econometric specifications and the results of

our hypothesis tests. We elaborate on our contribution to

theory in Section 5. Section 6 provides concluding remarks

including the implications of our findings, their potential

limitations, and avenues for future research.

2|RELATED LITERATURE AND

HYPOTHESES DEVELOPMENT

Queueing systems with impatient customers represent a wide

range of service systems in which customers may abandon

the system if they must wait to get served. The bulk of the

literature on abandonment assumes that customers are endo-

wed with a given patience distribution and once a customer's

wait exceeds their patience, the customer abandons the sys-

tem. The literature on such systems is large so here we focus

on the literature that is most relevant to the development of

our hypotheses and refer readers to existing survey articles

that provide a more exhaustive review of different research

streams (e.g., Mandelbaum & Zeltyn, 2013).

Instead of a patience distribution, another approach to

model abandonment from queues is the equilibrium analysis

of reneging by rational customers, where abandonment deci-

sions are made based on customers' utility functions (see

Hassin & Haviv, 2003, for an overview and Shimkin &

Mandelbaum, 2004, for a canonical model). A related

approach is taken by Aks¸in, Ata, Etemadi, and Su (2013)

who study a dynamic decision model of abandonment for

rational customers. Each customer solves an optimal stop-

ping time problem to decide whether to wait or abandon the

queue. In their model, they do not apply queueing theory to

derive the equilibrium waiting time distribution but instead

they use structural estimation to deduce this distribution

from their data. Unlike Aks¸in et al. (2013), we are dealing

with a semiobservable queue where patients can observe ED

crowding and the rate of service, which are naturally absent

in their call-center focused paper.

The medical literature acknowledges that waiting time is

important for LWBS decisions. For example, Bindman,

Grumbach, Keane, Rauch, and Luce (1991) report that, in

their studied institution, 86% of patients who left without

being seen by a physician left because of the long waiting

times. Waiting time is important, in part, because a patient's

medical condition may improve or deteriorate over time. For

example, Goldman, Macpherson, Schuh, Mulligan, and Pirie

(2005) report that 37% of LWBS patients left because their

symptoms had resolved, or they started to feel better. On the

other extreme, Baker et al. (1991) found that 53% of the

LWBS patients left because they felt too ill to wait. This

group of patients who left because they are too ill to wait

represents a high-risk group for serious adverse effects that

any delay in their treatment may have on their health.

BOLANDIFAR ET AL.431