Content uploaded by Gyula Vastag
Author content
All content in this area was uploaded by Gyula Vastag on Nov 28, 2018
Content may be subject to copyright.
This article appeared in a journal published by Elsevier. The attached
copy is furnished to the author for internal non-commercial research
and education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elsevier’s archiving and manuscript policies are
encouraged to visit:
http://www.elsevier.com/copyright
Author's personal copy
Facets of operational performance in an emergency room (ER)
Taco van der Vaart
a
, Gyula Vastag
b,n
, Jacob Wijngaard
a
a
University of Groningen, Faculty of Management and Organisation, P.O. Box 800, 9700 AV Groningen, The Netherlands
b
Corvinus University of Budapest, Faculty of Business Administration, Budapest, F +
ova
´mte
´r 13-15 (So
´ha
´z), H-1093 Budapest, Hungary
article info
Available online 29 April 2010
Keywords:
Emergency care
Services
Capacity and lead time analysis
Work-in-process inventories
Exploratory data analysis
abstract
This paper, using detailed time measurements of patients complemented by interviews with hospital
management and staff, examines three facets of an emergency room’s (ER) operational performance:
(1) effectiveness of the triage system in rationing patient treatment; (2) factors influencing ER’s
operational performance in general and the trade-offs in flow times, inventory levels (that is the
number of patients waiting in the system), and resource utilization; (3) the impacts of potential process
and staffing changes to improve the ER’s performance. Specifically, the paper discusses four proposals
for streamlining the patient flow: establishing designated tracks (‘‘fast track,’’ ‘‘diagnostic track’’),
creating a ‘‘holding’’ area for certain type of patients, introducing a protocol that would reduce the load
on physicians by allowing a registered nurse to order testing and treatment for some patients, and
potentially and in the longer term, moving from non-ER specialist physicians to ER specialists. The
paper’s findings are based on analyzing the paths and flow times of close to two thousand patients in
the emergency room of the Medical Center of Leeuwarden (MCL), The Netherlands. Using exploratory
data analysis the paper presents generalizable findings about the impacts of various factors on ER’s
lead-time performance and shows how the proposals fit with well-documented process improvement
theories.
&2010 Elsevier B.V. All rights reserved.
1. Introduction
Emergency medical care is the most visible part of the health
care system where process failures (excess time spent in waiting
rooms or misdiagnoses, for example) are widely publicized and
subjected to public scrutiny and debate. According to Reinberg
(2006), ‘‘Emergency care in the United States only deserves a C-.’’
The study, conducted by the American College of Emergency
Physicians, found that ‘‘the emergency care system is over-
crowded, provides limited access to care, and is hampered by
soaring liability costs and a poor capacity to deal with public
health or terrorist disasters.’’ The root of the problem, writes
Neergaard (2006), is that ‘‘demand for emergency care is surging,
even as the capacity for hospitals, ambulance services and other
emergency workers to provide it is dropping.’’
Dykstra (1997) describes the two major models of emergency
medicine. In the ‘‘Anglo-American Model,’’ emergency care takes
place in the hospital (emergency medicine) and pre-hospital
(emergency medical services [EMS]) settings but the emphasis is
on bringing the patients to the hospital. Emergency care in these
systems has achieved clinical, academic and professional recogni-
tion and accreditation in the United States, Canada, Australia, New
Zealand and more recently, in the United Kingdom. In the
‘‘Franco-German Model,’’ emergency medicine practiced in the
pre-hospital setting (that is, physicians and technology are sent to
the scene), with medical leadership and direction generally
provided by anesthetists. Physicians with certain clinical skills
(or experience) and after completing an emergency care course
become eligible providers within this system. These physicians
are either assigned to an advanced life support (ALS) ambulance
or travel to the scene separately with the other members of the
ALS crew (‘‘rendezvous’’ system). Basic life support (BLS) calls are
generally responded to by a team of paramedics and/or emer-
gency medical technicians [EMTs] without the on-site presence of
physicians. Inside the hospital, the ‘‘Franco-German model’’
considers emergency medicine an interdisciplinary activity that
does not require specialty status. Barring a few exceptions, the
equivalent of an emergency department does not exist. As a
result, there is no clinical or academic recognition for emergency
medicine in these countries, no career track for physicians, and
circumstances obviously restrict the resources available for
research. The Franco-German model can be found in Austria,
Finland, France, Germany, Latvia, Norway, Poland, Portugal,
Russia, Slovenia, Sweden and Switzerland (Holliman and C-evik,
2000). The Hungarian system of emergency medicine represents a
unique combination of the Franco-German and Anglo-American
models: it has ERs with medical doctors specialized in ER but,
Contents lists available at ScienceDirect
journal homepage: www.elsevier.com/locate/ijpe
Int. J. Production Economics
0925-5273/$ - see front matter &2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.ijpe.2010.04.023
n
Corresponding author.
E-mail addresses: j.t.van.der.vaart@rug.nl (T. van der Vaart),
gyula.vastag@uni-corvinus.hu (G. Vastag), j.wijngaard@rug.nl (J. Wijngaard).
Int. J. Production Economics 133 (2011) 201–211
Author's personal copy
unlike in the USA, their work is complemented with the
Hungarian Ambulance System with assigned medical doctors.
The Dutch EMS system, although many of its features are
similar to the various American models, takes an alternate
approach to pre-hospital health care delivery. It is a nurse-driven
triage system, both at the dispatch level and at the treatment
level. The Dutch EMS system utilizes these highly trained
personnel to effect judicial use of emergency pre-hospital
resources (Dib et al., 2006).
Each emergency department (ED) or emergency room (ER), we
use these terms as synonyms, must have treatment-related
functions that are very often accompanied with teaching and
research responsibilities. First and foremost, emergency depart-
ments have to be available to receive patients in urgent need of
treatment at any time. Considering that there is a great deal of
uncertainty regarding the timing and volume of patients’ arrival
and the amount of resources needed to treat them, resource
utilization of the ER and waiting times of the patients are the two
most critical and conflicting issues hospital emergency depart-
ments are facing. A related and important issue is that most of the
patients attending ER do not require real emergency care
(Steinbrook, 1996).
A patient’s time in the ER is either spent with using a resource
(attended by a nurse or physician and/or receiving treatment,
undergoing testing) or waiting for a resource (which might be a
release note, an administrative action). It has been well-
established that ER waiting times are the major determinant of
patient satisfaction or dissatisfaction (Holden and Smart, 1999;
McMillan et al., 1986). Increasing budget pressures, however,
limit the resources (e.g., number of beds, number of physicians
and nurses) made available to ER. Consequently, facing more and
more public scrutiny and resource limitations, hospitals have to
increase patient satisfaction while making better use of their
resources (see Chan et al., 1997, for an illustrative analysis).
The next two functions of ER are stabilization and disposition.
After treating and stabilizing the patients, they either go home or
referred to further treatment that may take place in the hospital
or in a non-hospital location immediately or at a later, scheduled
time. Additionally, many emergency departments play a key role
in medical education and research; they may serve as crucial
training grounds for residents, for example.
Because of the highly stochastic nature of emergency care, rules
were developed for situations where demand exceeds available
resources. Triage in general, and the Manchester Triage System
(MTS) in particular, was the first formalized system for managing
clinical risks (Manchester Triage Group, 1997; Derlet, 2004;
Windle and Mackway-Jones, 2003). The word ‘‘triage’’ is derived
from the French ‘‘trier’’ meaning ‘‘sort’’ and much of the credit for
the development of a method for quickly evaluating and
categorizing the wounded in the battlefield goes to Dominique
Jean Larrey, a surgeon in Napoleon’s army. Now, triage means a
brief clinical assessment that determines the time and sequence in
which patients should be seen in the ER. These decisions generally
are based on a short evaluation of the patient and an assessment of
vital signs. The patient’s overall appearance, history of illness and/
or injury and mental status also are important in the triage
decision (McMahon, 2003). The question of determining priority
scores and the reliability of these scores have been discussed in
several papers (for example, George et al., 1993; Jelinek and Little,
1996). The triage system became widely accepted in the second
half of the 20th century when organized departments with on-
duty physicians became standard. For a literature review of
emergency department triage, please see McDonald et al. (1995)
and FitzGerald et al. (2010).
There are many reported cases of triage misclassifications with
occasional dire consequences. The two illustrative cases and the
following evaluation are from Derlet (2004). These cases also
illustrate the influence and critical role of the dynamically
changing ER load (the number of patients in ER).
‘‘Case 2: A 43-year-old woman presented to the ED, complain-
ing of headache. The patient had normal vital signs except for
temperature of 101.2 1F. The ED was very busy and crowded.
Since the triage nurse had seen many patients that day whose
symptoms included headache and because this patient seemed no
worse than the others, she sent the patient to the waiting room.
Four hours later, another patient came to the triage desk stating
that the woman, who was still in the waiting room, was having a
seizure. A repeat temperature 5 h after initial presentation was
104.5 1F, and she was admitted to the hospital with a diagnosis of
meningitis.
Case 4: A 55-year-old man presented to the ED complaining of
abdominal pain. He stated that he thought his condition was
secondary to eating too much greasy fast food too rapidly. His
vital signs were blood pressure, 150/100 mmHg; pulse, 100 bpm;
respiration, 22 bpm and temperature, 98 1F. As the ED was busy,
the patient was sent to the waiting room. Two hours later, the
patient’s friend complained that he looked pale and had
increasing weakness. The patient’s friend was told, that the ED
was overcrowded. Three hours after triage, the patient collapsed
in the ED waiting room. He was brought into the ED hypotensive
and was taken to surgery, where he died of a ruptured aortic
aneurysm.
While the outcomes of some of the above cases may not have
changed had the patients been seen directly in the ED and
immediately evaluated by a physician, these cases do illustrate
that patients’ medical conditions are constantly changing, and
that triage is an active and dynamic process. If there are long lines
to see a physician in the ED, continually reassess patients.’’
From an operations management perspective, MTS has two
interrelated functions. First, the system supports the triage nurse
in sequencing, that is, the decision on the order in which the
patients will be treated. Second, the system incorporates
performance norms through the time frames defined per triage
code. Many ERs in the USA initially used a three-level approach,
and now follow the lead of the United Kingdom, Australia and
Canada to develop and implement five-level triage classification
systems (McMahon, 2003; Cook and Jinks, 1999; McDonald et al.,
1995; Pearson et al., 1995;Considine et al., 2004; Beveridge et al.,
1999; Cooke and Jinks, 1999; Zimmerman, 2001).
This paper, using detailed time measurements of patients,
examines three facets of an emergency room’s (ER) operational
performance: (1) effectiveness of the triage system in rationing
patient treatment; (2) factors influencing ER’s operational
performance in general and the trade-offs in flow times, inventory
levels (that is the number of patients waiting in the system) and
resource utilization and (3) the impacts of potential process and
staffing changes to improve the ER’s performance.
Specifically, four proposals, rooted in and linked to the
theoretical underpinnings of Operations Management and Supply
Chain Science (Hopp, 2008), for streamlining the patient flow are
discussed:
1. Establishing designated tracks (‘‘fast track,’’ ‘‘diagnostic
track’’), like in focused factories.
2. Creating a ‘‘holding’’ area for certain type of patients; an
equivalent of positioning and establishing buffers that could
reduce interarrival variability.
3. Job design and deskilling by introducing a protocol that would
reduce the load on physicians by allowing a registered nurse to
order testing and treatment for some patients and thus could
contribute to reducing process variation.
T. van der Vaart et al. / Int. J. Production Economics 133 (2011) 201–211202
Author's personal copy
4. Potentially, and in the longer term, moving from non-ER
specialist physicians to ER specialists that, in principle, would
also reduce process variation.
The paper’s findings are based on analyzing the paths of 1095
patients between November 8 and 28, 2004 and 798 patients
between March 13 and 26, 2006 in the emergency room of the
Medical Center of Leeuwarden (MCL), The Netherlands. These
observations were complemented by interviews with hospital
management and staff that also validated the findings presented
here. Using exploratory data analysis the paper presents general-
izable findings about the impacts of various factors on ER’s
lead-time performance and shows how the proposals fit with
well-documented process improvement theories. The two main
lines of inquiry of this exploratory research are focused on the
influence of triage and the necessity for and ways of reducing
variability in an ER setting.
The paper is organized as follows. Section 2 introduces the
emergency room of MCL (Medical Center of Leeuwarden, The
Netherlands) and data collected there. Section 3 presents
the results of our analysis. Section 4 deals with suggestions for
improvement. The paper ends with conclusions and suggestions
for future research (Section 5).
2. The ER at the Medical Center of Leeuwarden (MCL)
MCL is a large, regional teaching hospital in Friesland; a
province in the north of the Netherlands. In 2005, Friesland had
about 643,000 inhabitants; Leeuwarden, its capital with about
91,000 inhabitants, is in the center of the province. MCL had 914
beds, employed over 180 medical specialists (physicians) and a
number of medical residents. The hospital handled about 26,500
admissions, administered 16,000 one-day treatments and 132,000
outpatient visits per year.
The current ER facility at MCL was recently upgraded and,
since 2001, it has been the sole provider of emergency care at the
hospital (prior 2001, there were two facilities). Since 2004, after
the development of a new information and computer system, the
emergency room uses the Manchester Triage System (MTS).
Table 1 shows the number of patients in the ER annually and
the average lead-time per patient (time between registration and
departure from ER).
Table 1
Number of ER patients and the average duration of a visit at MCL (2001–2006).
Year 2001 2002 2003 2004 2005 2006
a
ER Patients per year 19,826 19,454 19,610 19,422 20,223 6773
Average flow time per patient (min)
b
95 91 101 111 115 121
a
For January, February, March and April.
b
The time between arrival at and from ER.
Fig. 1. ER Layout of MCL.
T. van der Vaart et al. / Int. J. Production Economics 133 (2011) 201–211 203
Author's personal copy
The trend of relatively constant number of ER patients per year
with increase in lead-time per patient visit continued in 2006: in
the first four months, the average length of an ER visit was
121 min.
Fig. 1 shows the layout and the patient flow at the ER at MCL.
Patients can arrive by any of three modes of transportation:
ambulance or helicopter (in extreme emergencies), through a
hospital department (e.g., X-ray or outpatient clinic), and other
(by car or walk-in). With the exception of extreme cases, patients
are registered at the reception desk and their identifiers are
entered the computer system. Data were also collected about the
reason for the visit (ini.arr: calling the Dutch emergency number,
physician referral, patient’s initiative and other) and the potential
cause behind the visit (cause: illness with unknown cause, home
accident, mutilation, not feeling well, workplace accident,
violence, sport accident, traffic accident, disaster and other).
Urgent patients (arriving by ambulance, for example) are
delivered directly to ER and their data are entered later.
Registration processing is very quick (a couple minutes) and the
patient’s data are automatically transferred to the ER by
computer. The patient’s card with a bar code is printed in the
central station of the ER. In the March 2006 data, each patient, in
addition to our assigned unique code, has a seven-digit unique
code; the same code is assigned to the same person (it is not true
for November 2004), so multiple visits in the research period can
also be accounted for.
From the reception desk patients proceed to the waiting room
where they are called from by the triage nurse. In the triage, an
experienced nurse uses a computer program and a set of
questions to assign one of the five color codes of the Manchester
Triage System (blue¼non-urgent, green¼standard, yellow ¼
urgent, orange¼very urgent, red¼immediate) to the patient.
MTS (Manchester, 1997) consists of five steps: (1) Identify the
complaint voiced, and pick an appropriate flow chart. (2) Gather
and analyze information using six general key discriminators (like
life threat, pain and conscious level) to determine a level of
priority. (3) Evaluate and select alternatives, using discriminators
within the flow chart to identify the patient’s general acuity.
(4) Implementation and simplified documentation. (5) Evaluation.
In addition to the triage code, the triage nurse also assigns a
specialty code to the patient. If the ER visit was initiated by the
patient then the triage nurse decides about the code assigned.
Otherwise, if the patient was sent by a physician (e.g., family
doctor), she uses the code assigned previously.
MTS also provides limits to the waiting time before a physician
sees the patient. These norms are on display in the ER waiting
room (see Fig. 2): 4 h for blue patients, 2 h for green, 1 h for
yellow, 10 min for orange, and 0 min for red patients.
Depending on the patient’s condition and on the load in the ER,
the patient is either sent back to the waiting room or transferred
to the ER immediately for treatment. The ER, in addition to the
triage room, consists of eight standard rooms and eight specia-
lized rooms that are listed in Fig. 1. All rooms have sliding doors
and curtains to provide the level of privacy needed. In the
experience of ER personnel, patients prefer to leave the doors
open, so they do not feel isolated even when the curtain is closed.
Each room is, largely, customizable by carts of supplies and
equipment. Some rooms, however, are more specialized with
expensive equipment (trauma room with X-ray, for example) or
with specific supplies (casting, burns).
Fig. 2. Time norms of the Manchester Triage System (MTS).
Fig. 3. Schedule of nurses (number of nurses in the ER).
T. van der Vaart et al. / Int. J. Production Economics 133 (2011) 201–211204
Author's personal copy
All 26 nurses in the ER’s nurse-pool are trained in triage; in
every shift, one nurse works in the triage, the others are in the ER.
The nurses can work in one of five shifts: Night, A, C, D, and F
where the F shift is used sporadically to address high demand
situations. Fig. 3 shows the total number of nurses on duty
(including the triage nurse) by the hour of the day. It should be
noted that between 15:30 and 16:00 hours, there is a planned
half-an hour overlap between A and D shifts to discuss and hand
over patients. In this period, eight (if there is no F shift) or nine (in
case of an F shift) nurses are present, but based on our discussions
with ER personnel, we showed the effective capacity of the ER as
six and five nurses, respectively.
There are two resident physicians on duty all the time.
Typically, one of them is a surgical resident, the other one is an
internist and each treats patients with complaints related to their
specialty. The residents are supervised by attending MCL
physicians in their specialties and both the residents and the
supervisors can request consultation with other doctors from
other fields (cardiology, plastic surgery, gynecology, pediatrics,
etc.). Consequently, the flow of patients served by the internal
medicine residents and the flow served by the surgery residents
are independent to a certain degree. Before triage, there is one
queue of patients and the two groups of patients are served with
the same resources, i.e. nurses and ER cubicles. Especially at busy
times, the two lines for the two specialties become more mixed
strengthening the conclusion that there are no separate streams
for different groups of patients. If needed, the residents consult
with the attending physician and other physicians at MCL. In each
Fig. 4. Timeline (t.xxx) and related measurements (variables).
T. van der Vaart et al. / Int. J. Production Economics 133 (2011) 201–211 205
Author's personal copy
research period, 20–25 residents, supervised by more than 60
attending physicians, treated patients in the ER.
The nurses and residents are assisted by a secretary (24-h
position) who collects patient information and create patient files,
does the electronic registration, makes notes on the plan board for
residents and physicians, has contact with the reception (it is also
a 24-h post) about arriving helicopters, notifies trauma teams,
sends blood samples to the lab, provides administrative support
(patient phone, fax, etc.), keeps the valuables of patients in a safe,
takes care of the patient’s family and sets additional appoint-
ments for casting and outpatient clinic.
Additionally, the ER shares the following resources with MCL:
laboratory, VP-scan, CT-scan, ultrasound and additional physi-
cians (residents, attending and consulting doctors).
Patient data was obtained from the ER’s computer system to
which, during the time periods examined, a research assistant
added additional measurements. Patient lead-times were mea-
sured in detail including arrival time, triage time, start treatment,
contact moments with physician, lead times of tests conducted by
the hospital’s laboratory and departure time (admission to the
hospital or discharge). The most important departure from the
computer system was that the research assistant, with help from
the nurses, recorded the time when the patient had contact with
physicians. In the first test week, the nurses got instructions every
day, to guarantee good registration. Fig. 4 shows the timeline and
lists variables measured in the process. Additionally, we have
used several derived variables in the analysis (e.g., number of
patients in the ER at any given moment with a specific triage
code) that we will discuss later in the paper.
In November 2004, there were records for 1103 patients. From
the 1103 patients eight observations were removed because
crucial data (like triage code and triage time) were missing;
leaving us with 1095 patient records. For most measures, the
completeness of the data was very good. These observations were
complemented with 804 records collected in March of 2006; six of
them were missing crucial data, so we used 798 records.
3. Analyses and results
In this section, using exploratory data analysis we examine
three facets of MCL ER operational performance: (1) effectiveness
of the triage system in rationing patient treatment; (2) factors
influencing ER’s operational performance in general and the
trade-offs in flow times, inventory levels (that is the number of
patients waiting in the system) and resource utilization. In the
next section, we will review the planned process and staffing
changes to improve the ER’s performance.
Table 2 shows some aggregate statistics for the total time
spent in MCL ER (the time elapsed between the start of triage
(registering with the triage nurse) and being released from ER) by
triage code.
In the table, both the mean and the median of the total flow
time per patient are shown because these distributions are
significantly and positively skewed (there is an asymmetric tail
extending to the positive values). Skewness statistics are in the
neighborhood or in excess of three standard errors of skewness;
the standard errors of skewness were estimated using the formula
suggested by Tabachnick and Fidell (1996). The flow times are
also, to a lesser extent, leptokurtic (too tall or peaked); in case of
the ‘‘Green’’ and ‘‘Red’’ codes, we can speak of significant positive
kurtosis (they are leptokurtic) as the shown statistics are 3.5
times greater than the estimated standard error of kurtosis
(Tabachnik and Fidell, 1996).
Fig. 5, by showing data for individual patients, confirms that
flow times for codes 2, 3 and 4 (green, yellow and orange) are
surprisingly close to each other.
MTS standards set upper limits for patient waiting times. Fig. 6,
on the horizontal axis, shows the time elapsed between triage and
an ER nurse seeing the patient (lt.nurse). The vertical axis of the
same graph shows the time difference between triage and the
resident physician’s first meeting with the patient (lt.resid).
The conclusion seems to be that in the overwhelming majority of
the cases a nurse sees the patient within the time limit set.
However, the waiting time for doctors is much longer; often they
see patients later than required. For triage code yellow (code.tr: 3),
for example, the norm is 60 min. There are only five cases that the
nurses attended to more than an hour after triage but there are
numerous instances with doctors’ lead time longer than an hour.
However, we must use some caution in interpreting this
finding. In the dynamically changing environment of the ER
where physicians and nurses are working in close proximity of
each other, the patients’ conditions are frequently and routinely
communicated. Consequently, even though the physician may
have been ‘‘late’’ to see the patient (as the record shows), he/she
may have been aware of the patient’s condition and the treatment
administered.
We used regression tree analysis (Brieman et al., 1984; Vastag
et al., 1996) to classify the patients into more homogenous
groups; see Fig. 7. This method looks at all possible splits for all
variables included in the analysis, and it conducts searches
through all of them. The process is considerably simplified
0
1
2
3
4
5
6
Triage Codes
0 100 200 300 400 500
ER Lead Time (minutes)
Fig. 5. Distribution of patient times in MCL ER by triage codes.
Table 2
Total time spent in MCL ER by patient and by Manchester Triage System Codes
(2004; in minutes).
Statistic Triage code (color and number)
Blue (1) Green (2) Yellow (3) Orange (4) Red (5)
Number of cases 11 506 367 166 34
Minimum 0 0 1 1 9
Maximum 131 310 389 434 403
Median 10 72.5 128 117.5 137.5
Mean 32.64 85.56 134.19 127.81 140.35
Standard deviation 41.43 64.44 74.20 67.05 78.92
Coefficient of variation 1.27 0.75 0.55 0.53 0.562
Skewness 1.57 1.11 0.57 1.21 1.32
Kurtosis 2.16 0.92 0.376 2.655 2.91
T. van der Vaart et al. / Int. J. Production Economics 133 (2011) 201–211206
Author's personal copy
because the method always asks questions that have a yes or no
answer. The next step is to rank each splitting rule on the basis of
a goodness-of-split criterion. One criterion commonly used is a
measure of how well the splitting rule separates the classes
contained in the parent node. Once a best split is found, the search
process is repeated for each child node and continues recursively
until further splitting is impossible or stopped for some other
reason (e.g., the node has too few cases). We used the trimmed
mean loss function (20% of the extreme observations are removed
before computing the mean) that gave a PRE (proportional
reduction in error) value of 0.334. The cutting variables were
the specialty codes and the number of lab requests: if patients are
assigned to general surgery (specialty code¼1) or plastic surgery
(specialty code¼2) and have no lab requests (no_labo1) then, on
average, these patients (there are 395 of them) spend 69.1 min in
MCL ER. On the other hand, there were 392 patients (most of
them assigned to internal medicine) who had at least one lab test;
on average, their time in the ER was 151.9 min.
Historically, MCL ER tried to schedule the nurses to anticipate
the demand, the number of patients in the system. Using a
manufacturing term, patients represent work-in-process inven-
tory and they can be in two areas: either in triage (in the system
and waiting for admission to the ER) or in the ER. Using 2004 data,
Fig. 8 shows for all three weeks the number of nurses and the
number of triage and ER patients by the minute from the
midnight of the start of the week.
It is a revealing graph; it shows that the triage capacity is not a
problem, the triage processing time is short with no significant
variation and the average number of patients in the triage is
small. The problem is in the ER. In the 2004 research period, there
were, on average, 3.82 patients in MCL ER (that is taking together
the triage and ER patients) and MCL processed, on average, 51.76
patients per day. Multiplying these two numbers and using Little’s
Law (for a thorough discussion see Cachon and Terwiesch, 2006),
we get 106.27 min for the average flow time per patient. In
Table 1, we find 111 min for 2004, so our sample reasonably
closely approximates the annual statistics. Although there are
times when all 16 beds are taken in the ER, on average there are
fewer than four patients in the system and there are time periods
when the ER is less used (see the period of November 22–28)
raising the question of trade-off between availability and resource
utilization.
Putting it in a different way: a reduction in flow time cannot be
achieved without a reduction in the average number of patients in
the system (‘‘where there is WIP, there is flow time,’’ Hopp, 2008)
but the latter, inevitably, will lead to questions about resource
utilization. This seemingly trivial but often overlooked insight is
the direct consequence of Little’s Law: [Flow Time] ¼[WIP]/[Flow
Rate]. If the average utilization of 40% (that is on average 40% of
the beds are occupied) is raising eyebrows then reducing flow
time, which is one of the stated goals of ERs, inevitably will lead to
an even lower utilization level.
4. Suggestions for improvement
The four proposals considered by MCL ER management all
aimed at reducing variability and thus fit nicely with process
improvement literature. We will discuss these proposals one-by-
one although they may have the best effect if used together.
4.1. Establish designated tracks (‘‘fast track’’ and ‘‘diagnostic track’’)
The basic idea here is to create more homogenous groups of
processing times by removing the two ends of the distribution:
the patient groups that either do not require major intervention
(-‘‘fast track’’) or those that spend more time in ER but their
condition is stable (their triage code is either blue or green) and
most of their time in the ER is spent with waiting for test results
(-‘‘diagnostic track’’). As an illustration, Fig. 9, using 2004 data,
shows how much more time the non-urgent patients on the
‘‘diagnostic track’’ spend in the ER compared to others.
Establishing a diagnostic track would also reduce the arrival
variability, as these patients are not real ER patients: they can be
scheduled in advance, so they could potentially free up treatment
capacity for other patients.
The ‘‘fast track’’ approach is also known as a ‘‘see and treat’’
strategy (King et al., 2006) where patients are seen by senior
decision-making staff who immediately discharge (at least this is
the intention) a substantial number of patients. In MCL, after
visiting the ER more than 40% of patients are admitted to the
hospital. This high acuity level probably would not justify the
establishment of the ‘‘fast track’’ or the application of the ‘‘see and
treat’’ strategy.
30 80 130 180
lt.nurse
10
120
10
120
10
120
10
120
10
120
lt.resid
code.tr: 1.00
code.tr: 2.00
code.tr: 3.00
code.tr: 4.00
code.tr: 5.00
Fig. 6. Conformance to Triage Standards (2004).
T. van der Vaart et al. / Int. J. Production Economics 133 (2011) 201–211 207
Author's personal copy
4.2. Creating a holding area for certain type of patients
This proposal, very much like the bar in a Benihana restaurant,
would reduce arrival variability by buffering non-urgent, schedul-
able patients (like the ones on the diagnostic track, for example)
and thus achieving higher utilization on some critical resources.
4.3. New protocol that would allow registered nurses to order testing
and treatment for some patients
Using a protocol for easy, standard cases would contribute to
reducing the variation in processing times and at the same time
would reduce the load on the doctors. This suggestion also fits
well with the notion of deskilling where complex work processes
are broken into smaller, simpler and unskilled tasks.
4.4. Moving from non-ER specialist physicians to ER specialists
Currently, there is a significant variation in processing times
by specialty (internists, on average, take more time than surgeons;
see Fig. 10) and by experience; inexperienced residents, on
average, are associated with longer times in the ER than
experienced ones.
CODE_SP<3.000
NO_LAB<1.000 NO_LAB<1.000
Mean=114.182
SD=71.603
N=1025
Mean=93.476
SD=67.527
N=546
CODE_SP<3.000
Mean=137.785
SD=68.820
N=479
Mean=69.091
SD=44.279
N=395
NO_LAB<1.000
Mean=157.265
SD=75.865
N=151
Mean=76.057
SD=52.869
N=87
NO_LAB<1.000
Mean=151.485
SD=64.337
N=392
Fig. 7. Regression tree of lead time determinants.
T. van der Vaart et al. / Int. J. Production Economics 133 (2011) 201–211208
Author's personal copy
Teaching and employing an ER specialist, like in the U.S.A., would
mean greater standardization in procedures and approaches and
would contribute to reduced variability in processing times.
These proposals individually, or taken together, would create
more homogeneous patient subgroups and reduce both arrival
and process variability in the ER.
5. Conclusions
In this paper, we outlined both production control (see
for example Malakooti et al., 2004) type of improvements
(Proposals 1 and 2) and ideas more related to organization
and technology (Proposals 3 and 4). Our focus has been on the
Number of Nurses (Nov 8-14)
0 4000 8000 12000
Time (minutes)
1
3
5
7
NURSES
Number of Nurses (Nov 15-21)
0 4000 8000 12000
Time (minutes)
1
3
5
7
NURSES
Number of Nurses (Nov 22-28)
0 4000 8000 12000
Time (minutes)
1
3
5
7
NURSES
Triage Load (Nov 8-14)
0 4000 8000 12000
Time (minutes)
0
2
4
6
Number of Patients
Triage Load (Nov 15-21)
0 4000 8000 12000
Time (minutes)
0.0
0.7
1.4
2.1
2.8
3.5
Patients
Triage Load (Nov 22-28)
0 4000 8000 12000
Time (minutes)
0
1
2
3
4
5
6
7
Patients
ER Load (Nov 8-15)
0 4000 8000 12000
Time (minutes)
0
5
10
15
20
Patients
ER Load (Nov 15-21)
0 4000 8000 12000
Time (minutes)
0
5
10
15
Patients
ER Load (Nov 22-28)
0 4000 8000 12000
Time (minutes)
0
4
8
12
16
Patients
Fig. 8. Patient load and nurse capacity in 2004.
0 100 200 300 400 500
Time between release and arrival (minutes)
path.dia: Diagnostic
path.dia: Other
Fig. 9. Lead time of potential ‘‘diagnostic track’’ patients (2004).
T. van der Vaart et al. / Int. J. Production Economics 133 (2011) 201–211 209
Author's personal copy
former group, proposals that are directly related process im-
provement.
The problem with triage, as a prioritization tool, is that urgent
patients will always override the less urgent ones (even though
there are time limits for all triage classes) and, therefore, the
average flow time will be approximately the same in all triage
categories. As we found out, in predicting total flow times,
specialty codes (surgery versus internal medicine, for example)
and the number of lab requests were the key variables and they
are not the control of an operations manager.
The diagnostic path, however, can free up capacity for the real
urgent patients. The fact that the visits by these patients, to some
extent, can be planned (within 24 h) means that it is possible to
smooth the demand, which can have a positive effect on lead
times during the busy hours.
King et al. (2006), in an environment similar to MCL,
introduced a simple modification to the triage-based ER. In the
first step, the triage nurse assigns the patient into one of two
groups: (1) dischargeable or (2) admissible. This classification is
based on her/his judgment of the patient’s predicted outcome: the
patient will either be discharged after the examination or be
admitted to the hospital. The emphasis here is not so much on the
accuracy of the prediction (triage nurses were accurate
about 80% of the time) rather it is on the speed of the
classification. In the dischargeable group, the priority rule is FIFO
but patients could be re-classified to the other group. Triage is
used only in the admissible group but even there the priority rule
for the less serious categories can be changed to FIFO. These
simple changes led to significantly shorter flow times in all
patient groups and, consequently, the number of patients in the
ER was also reduced.
While the ER system reported here is quite general and
can be found in many countries, the study’s findings should
further be validated with data from other ER models using
approaches of causal modeling (e.g., Bayesian networks) and/or
simulation.
References
Beveridge, R., Ducharme, J., Janes, L., Beaulieu, S., Walter, S., 1999. Reliability of the
Canadian emergency department triage and acuity scale: inter-rater agree-
ment. Annals of Emergency Medicine 34 (2), 155–159.
Brieman, L., Friedman, J., Olshen, R., Stone, C., 1984. Classification and Regression
Trees. Wadsworth, Pacific Grove, CA.
Cachon, G., Terwiesch, C., 2006. Matching Supply with Demand: An Introduction to
Operations Management. McGraw Hill/Irwin, New York, NY.
Chan, L., Reilly, K.M., Salluzzo, R.F., 1997. Variables that affect patient throughput
times in an academic emergency department. American Journal of Medical
Quality 12 (4), 183–186.
Considine, J., LeVasseur, S.A., Villanueva, E., 2004. The Australian triage scale:
examining emergency department nurses’ performance using computer and
paper scenarios. Annals of Emergency Medicine 44 (5), 516–523.
Cooke, M.W., Jinks, S., 1999. Does the Manchester triage system detect the
critically ill? Journal of Accident and Emergency Medicine 16 179–181.
Derlet, R., 2004. Triage. Accessed on June 15, 2006 at /http://www.emedicine.
com/emerg/topic670.htmS.
Dib, J.E., Naderi, S., Sheridan, I.A., Alagappan, K., 2006. Analysis and applicability of
the Dutch EMS system into countries developing EMS systems. The Journal of
Emergency Medicine 30 (1), 111–115.
Dykstra, E.H., 1997. International models for the practice of emergency care. The
American Journal of Emergency Medicine 15 (2), 208–209.
FitzGerald, G., Jelinek, G.A., Scott, D., Gerdtz, M.F., 2010. Emergency department
triage revisited. Emergency Medicine Journal 27, 86–92.
George, S., Read, S., et al., 1993. Differences in priorities assigned to patients by
triage nurses and by consultant physician in accident and emergency
departments. Journal of Epidemiology and Community Health 47, 312–315.
Holden, D., Smart, S., 1999. Adding value to the patient experience in emergency
medicine: What features of the emergency department visit are the most
important to patients? Emergency Medicine Journal 11 3–8.
Holliman, J., C-evik, A.A., 2000. International emergency medicine. Turkish Journal
of Trauma and Emergency Surgery 6 (1), 7–13.
Hopp, W.J., 2008. Supply Chain Science. McGraw-Hill/Irwin, New York, NY.
Jelinek, G.A., Little, M., 1996. Inter-rater reliability of the national triage
scale over 11,500 simulated occasions of triage. Emergency Medicine 8,
226–230.
King, D.L., Ben-Tovim, D.I., Barsham, J., 2006. Redesigning emergency department
patient flows: application of lean thinking to health care. Emergency Medicine
18, 391–397.
Manchester Triage Group, 1997. In: Mackway-Jones, K. (Ed.), Emergency Triage.
BMJ Publishing Group, London.
Malakooti, B., Malakooti, N.R., Yang, Z., 2004. Integrated group technology, cell
formation, process planning, and production planning with application
0 100 200 300 400
lt.er
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
code.tr
agg.res: 1.00
agg.res: 2.00
agg.res: 3.00
agg.res: 4.00
Fig. 10. Time spent in MCL ER by specialty (2004). (agg.res: 2¼Internal Medicine, agg.res: 1 ¼Surgery).
T. van der Vaart et al. / Int. J. Production Economics 133 (2011) 201–211210
Author's personal copy
to the emergency room. International Journal of Production Research 42 (9),
1769–1786.
McDonald, L., Butterworth, T., et al., 1995. Triage: a literature review 1985-1993.
Accident and Emergency Nursing 3 (4), 201–207.
McMahon, M.M., 2003. ED triage. The American Journal of Nursing 103 (3), 61–63.
McMillan, J.R., Younger, M.S., DeWine, L.C., 1986. Satisfaction with hospital
emergency departments as a function of patient triage. Health Care Manage-
ment Review 11, 21–27.
Neergaard, L., 2006. Probe says U.S. emergency care in trouble. Accessed on June
15, 2006 at /http://news.yahoo.com/s/ap/emergency_careS.
Pearson, S.D., Goldman, L., et al., 1995. Triage decisions for emergency department
patients with chest pain: do physicians’ risk attitudes make the difference?
Journal of General Internal Medicine 10 (10) 557–564.
Reinberg, S., 2006. U.S. Emergency Care Not Making the Grade: Report. Accessed
on June 15, 2006 at /http://health.yahoo.com/news/142892S.
Steinbrook, R., 1996. The role of the emergency department. The New England
Journal of Medicine 334 (10), 657–658.
Tabachnick, B.G., Fidell, L.S., 1996. Using Multivariate Statistics 3rd edition Harper
Collins, New York.
Vastag, G., Rondinelli, D.A., Kerekes, S., 1996. Evaluation of corporate environ-
mental management approaches: a framework and application. International
Journal of Production Economics 43, 193–211.
Windle, J., Mackway-Jones, K., 2003. Don’t throw triage out with the bathwater.
Emergency Medical Journal 20, 119–120.
Zimmerman, P.G., 2001. The case for a universal, valid, reliable 5-tier triage acuity scale
for US emergency departments. Journal of Emergency Nursing 27 (3), 246–254.
T. van der Vaart et al. / Int. J. Production Economics 133 (2011) 201–211 211
