Hospital Emergency Department (ED) Simulation for Resource Management

Difficulty Level Full

Suggested Language C++/Java
Maximum Team Size 4
Skill Set C++/Java, GUI, graphics

1. Introduction In addition, you learn analysis of simulation results to

The Emergency Department (ED) is an integral part of evaluate ED performance efficiency. The performance of
hospitals. Admissions from the ED account for a ED can be measured by one or more of the following
significant proportion for a hospital’s activity. Ensuring a performance indicators: waiting time; bed utilization and
timely and efficient flow of patients through the ED is access block.
crucial for optimizing patient care. In recent years, ED
overcrowding and its impact on patient flow has become a 2. Problem Description
major issue facing the health sector. Figure 1 shows the ED patient flow. Specific elements
Simulation is rapidly becoming a tool of choice along the path that patients follow include some or all of
when examining hospital systems due to its capacity to the following stages: arrival; triage (triage is the
involve numerous factors and interactions that impact the assessment of a patient’s urgency for medical treatment
system. In this project, a simulation model is developed to and is divided into 5 categories 1-5 in increasing order of
investigate potential impacts of changing the following seriousness.); record retrieval; physician assessment;
aspects of ED (number of beds; number and rate of patient imaging and laboratory studies; x-rays or MRI; treatment
arrivals; acuity of illness or injury of patients; hospital planning; nursing activity; procedures (e.g. suturing and
staffing arrangements; and access to inpatient beds). The casting); decision to discharge or admit; access to
project combines implementation of basic data structures inpatient beds and physicians. These stages generally
(priority queues, lists, arrays) with the design and occur in a sequential manner.
implementation of several interesting interacting classes.

Figure 1. Patient Flow through ED

 Patients generally arrive according to some The treatment time and admission status is determined
discrete probability distribution. The probability based on the category. The treatment times for Category 4
distribution is condition-dependent and recognizes: patients are based on a Pearson VI distribution with scale
Multitrauma; Blood/Immune; Cardiac/Vascular; and two shape parameters 355, 1.64, and 5.72 with the
Diabetes/Endocrine; DNW Prior to Triage; following density function:
Drug/Alcohol/Poisoning; ENT/Oral;
Environmental/Temperature/MISC; Gastrointestinal; GP
referred /Hosp transfer; Injury; Neurological; Eye;
Nonemergent Review; Obstetrics/Gynaecology;
Paediatric; Pain; Psychiatric/Behavioural; Regional
Problems; Renal; Respiratory; and Urinary/Reproductive.
As an example, patients with a gastrointestinal condition
arrive with interarrival times according to a Weibull
distribution with scale and shape parameters 180 and
0.914 respectively. where B(p, q) is the beta function.

The patients, once being admitted to a treatment room, are

then placed in other sub-queues as determined by their
presenting condition. These sub-queues include staffing
Once a patient arrives, he/she is assigned a triage category resources and diagnostic testing and imaging.
and then assigned a treatment time and inpatient A patient may also be admitted as inpatient if required.
admission status dependent on the triage category. The The PDDT follows an exponential distribution with mean
final assignment is the post discharge decision time PDDT 156 and the following density function:
(only if the patient has a positive admission status), which
follows a single probability distribution.
The percentages of each triage category for each condition
are also studied; an example is shown in Figure 2.
Staffing resources queues are priority based allowing pre-
emption, that is, patient treatment may be interrupted for
treatment of a higher priority with the intent of returning
to complete treatment.
Patients follow paths through the ED as shown in Figure
3. As can be seen paths are highly dependent on triage
category.

Due to the possibility of the patient presenting with any

condition or disease and being in any stage of that disease,
patients have a variable time in the system. There are
guidelines that dictate as to the expected length of time
required for treatment of any presenting conditions by
breaking up the patients into categories based on
Figure 2. Category Distribution for Gastrointestinal
characteristics including condition, time of day, triage
Triage Categories
category, admission requirements, and productivity rates
of the resources.
The assignment of the next arrival to enter the system is Category 3-5 patients under supervision. Generally,
determined by a priority rule. Patients are triaged and seen interns can see 1-2 patients simultaneously. Junior and
according to priority. The College for Emergency senior residents can treat Category 3-5 patients with
Medicine gives the Triage Scale and the recommended supervision as required and form part of the team treating
maximum time between arrival in the ED queue and the Category 1 and 2 patients. Junior and senior residents can
commencement of treatment. The assumption is that if see 2-3 and 2-4 patients simultaneously respectively.
these patients are not seen within this time their condition The typical ED has 24 standard treatment areas
will degenerate requiring further time in the system when ranging in equipment available for use. There are an
they are finally treated. The categories that patients are additional 13 corridor positions for stretchers and 3
sorted into are given in Table 1, along with the recliner chairs to be used as overflow treatment areas in
recommended and desired times for the patient to be cases of critical overcrowding. The ED consists of
assessed by a physician. resuscitation beds (Category 1 only); acute beds (Category
2-5); subacute beds (Category 3-5); minor procedure
Table 1. Maximum Waiting Time for Various rooms (Category 2-5).
Categories
3. Solution Design (Tentative, Incomplete)

3.1 Patient and Bed Classes

The member data of patient class represents the time
of arrival, the assigned triage category, the time when
the patient has been assigned a bed, and the time of
discharge.
The bed object only needs to know whether a
bed is available, or which patient is using the bed. A
reference to a patient object can encode both pieces of
The resources in ED are both physicians and beds.
information - a NULL reference indicates that the
Patients could only enter the ED system from the waiting
bed is available.
room if both resources are available. Physicians possess
The bed that the patient can enter is dependent
multitasking patient treatment capabilities. Physicians in
on their category and patients wait in the resource queue
the ED include interns, junior residents, senior residents,
until a doctor is available to take them to the bed and
registrars and consultants. The registrars and consultants
perform an initial consultation. If all beds are full,
supervise junior staff, consult patients as required and
Category 2 patients are routed to the corridor positions,
initiate and oversee the treatment of Category 1 and 2
otherwise they enter beds as normal. Once treatment is
patients. Interns are first year doctors who are able to treat
completed, the patients are discharged and doctors are
released into the resource pool, available to see the next You should be able to demonstrate what happens if
waiting patient. more beds are available, measure the utilization of beds,
model the busier times in the emergency room (e.g.
3.2 Queue Classes Saturday night) when the number of patients and the
The simulation contains multiple queue types. All arrivals severity of their problems increase, etc. You should also
join a single queue in the waiting room. This queue is a look at data and compare the behavior for the different
priority-FIFO queue with patients being seen in order of sets of input parameters.
priority (Category 1 through to Category 5) with the
patients within a category group being seen in order of The user should be able to vary the discrete probability
arrival. Lower category patients are bumped down the distributions of three events: patient arrivals, the triage of
queue every time a higher priority patient enters the ED. the patient’s problem, and the estimated length of
When both a bed and a doctor are available the patient at treatment in minutes. The user should also be allowed to
the top of the queue enters an ED bed. The patient choose the number of available beds and the length of the
remains in the bed for the treatment time and, if the simulation run.
patient is to be admitted, the PDDT is generated. All
patients are then discharged home; or to an inpatient bed.

3.3 Assumptions
• registrars and consultants are considered always
available to initiate treatment on Category 1 and 2 patients
and are sufficient to consult patients as required and
supervise the junior staff;
• all patients are considered equal for inpatient bed
placement-triage categories no longer dominate queue
position but rather transfer to an inpatient bed depends on
availability in the target ward and length of time waiting
for bed;

3.4 Outputs
• utilization of the trainee doctors
• percentage of patients seen within recommended time
• utilization of the different bed types
• number of patients waiting

3.5 Collecting Statistics

There are several options for collecting the results
of the simulation. The easiest is to save the results
to a file. To make this file usable for further analysis,
each line should represent the data for one
discharged patient, with individual data items
separated by tabs. This format is readable by most
spreadsheet applications.
Another option is to create a dynamically growing array
of patient discharge data objects. The STL vector class is
well suited for this use. In any case, eventually, you
should be able to see and plot the results for at least
several different runs of the simulation.

3.5 Designing Experiments and Analyzing Data

You have to continue with designing a suite of
simulation runs, where all decisions - the length of the
run, the number of beds, and the discrete probability
distributions are read from a file and several runs
are done during one execution of the program.
At this point, you also work on designing the
proper user interface, use input files in a meaningful
setting, and work on the design of experiments.