See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/261410467

A Survey of Agent-Based Modeling of Hospital Environments

Article in IEEE Access · January 2014

DOI: 10.1109/ACCESS.2014.2313957

CITATIONS READS

15 1,441

2 authors:

Marcia Friesen R. D. Mcleod

University of Manitoba University of Manitoba
90 PUBLICATIONS 787 CITATIONS 157 PUBLICATIONS 1,746 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Marcia Friesen on 28 April 2015.

The user has requested enhancement of the downloaded file.

Received March 6, 2014, accepted March 22, 2014, date of publication March 26, 2014, date of current version April 4, 2014.
Digital Object Identifier 10.1109/ACCESS.2014.2313957

A Survey of Agent-Based Modeling of Hospital

Environments
MARCIA R. FRIESEN AND ROBERT D. MCLEOD, (Member, IEEE)
Department of Electrical and Computer Engineering, University of Manitoba, Winnipeg, MB, Canada
Corresponding author: R. D. McLeod ([email protected])

ABSTRACT Agent-based modeling has become a viable alternative and complement-to-traditional analysis
methods for studying complex social environments. In this paper, we survey the role of agent-based modeling
within hospital settings, where agent-based models investigate patient flow and other operational issues as
well as the dynamics of infection spread within hospitals or hospital units. While there is a rich history of
simulation and modeling of hospitals and hospital units, relatively little work exists, which applies agent-
based models to this context.

INDEX TERMS Agent based modeling, emergency departments, institutional environments.

I. INTRODUCTION systems, most often people) imbued with properties: charac-

This paper surveys the application of agent based modeling teristics, behaviours (actions), and interactions that attempt
(ABM) and simulation of complex social dynamics within to capture actual properties of individuals with a high degree
the institutional scale of a hospital. Hospitals are a promis- of diversity and fidelity. In the most general context, agents
ing area of continued ABM research with the concomitant are both adaptive and autonomous entities who are able to
potential for substantive outcomes. Healthcare around the assess their situation, make decisions, compete or cooperate
world deals with a perennial pressure to find cost efficiencies, with one another on the basis of a set of rules, and adapt future
and target areas include the optimization of healthcare pro- behaviours on the basis of past interactions. Agent properties
cesses and flow, reducing emergency department (ED) wait are determined by the modeler and are ideally derived from
times and length of stay, and reduce admission times. Within actual data that reasonably describe agents’ behaviours – i.e.
these areas, hospitals rely on the experience of practitioners their movements and their interactions with other agents.
for improvements in triage procedures, diverting low-acuity The emergence of a data culture, also called ‘big data’ and
patients, reconfiguring the healthcare staffing model, and associated ‘big data analytics’, offers new opportunities to use
reorganizing operational units both physically and procedu- real world data, even in near real time, as inputs into ABMs.
rally. The modeler’s task is to determine which data sources best
Simulation offers a potential to identify improvements and govern agent profiles in a given ABM institutional simulation.
new understandings in how a facility operates. Simulation The foundational premise and the conceptual depth of
has the potential to models real-world variability, lessens ABM is that simple rules of individual behaviour will aggre-
testing and implementation costs of planned changes, and gate to illuminate complex and/or emergent group-level phe-
helps to minimizes the risk of errors in implementing changes. nomena that are not specifically encoded by the modeler and
Patient tracking through their stay in the hospital by using that cannot be predicted or explained by the agent-level rules.
technologies such as radio frequency identification (RFID) In essence, ABM has the potential to reveal a whole that is
and improved electronic reporting and dashboards are one greater than the sum of its parts [1], [2].
example of initiatives that can be integrated with simulation ABMs provide a natural description of a system that can be
studies to generate valuable information on social dynamics calibrated and validated by representative expert agents, and
within the institution. is flexible enough to be tuned to high degrees of sensitivity
In general agent based modeling is ‘bottom-up’ systems in agent behaviours and interactions. As such, they play a
modeling from the perspective of constituent parts. Systems vital role as an information translation vehicle. The lexicon
studied are modelled as a collection of agents (in social used to develop an ABM is the lexicon of area experts and of
2169-3536 2014 IEEE. Translations and content mining are permitted for academic research only.
Personal use is also permitted, but republication/redistribution requires IEEE permission.
VOLUME 2, 2014 See http://www.ieee.org/publications_standards/publications/rights/index.html for more information. 227
 M. R. Friesen, R. D. Mcleod: Survey of Agent-Based Modeling of Hospital Environments

the institution under consideration (e.g. a hospital), reflect- nosocomial infection spread [10]). ABMs in healthcare have
ing the world in a real and specific a manner as possible. also examined economic models of healthcare, removed from
In essence, one builds a laboratory where the behaviours of scale of the patient itself; these models are not surveyed here.
individuals are similar to those in the real-world emergency
department and the one observes what happens when the rules A. SYSTEM ATTRIBUTES
of behaviour and interactions are changed. The underlying When designing an ABM for hospital applications, there are
ABM engine may be quite complex and utilize the most choices in system attributes that become design decisions
advanced processing and hardware techniques available, but unique to the context and objectives of the model. An ABM
this level of detail is not required in developing the model or is inherently agent-centric, and the model arises from the
in the analysis of its output. consideration and definition of the agent’s environment, the
Although simulation and modeling in healthcare facilities agent’s characteristics, and the agent’s interactions with other
is not new, agent based modeling within these settings is a agents.
relative newcomer. This survey paper focusses on hospital
ABMs, which is an agent centric approach as opposed to more • Commercial / Homegrown: At present a large number
established areas of simulation which tend to the process of ABMs are developed as one-offs or custom models,
oriented. The key differences between modeling techniques dedicated to the objective at hand. These offer advan-
such as discrete event, system dynamics, network analysis, tages associated with data fusion, accelerators through
and ABM are well-documented and to date, the majority of multicore, cluster, high performance computing (HPC)
research in healthcare simulation has utilized Monte Carlo, optimization as well as general purpose computation
discrete event simulation (DES), and system dynamics rather on graphics processing units (GP-GPU). Disadvantages
than ABMs [3]–[5]. are the considerable overhead in developing one’s own
Yet, ABMs are considered to be a very promising and code, inclusive of code verification. The benefits of
complementary technique by which to simulate hospital a commercial platform are a proven code base and
dynamics, with arguments for their more widespread use user community. Just as with many other areas where
within healthcare will depend on more widely adopted and simulation plays a crucial role in product development,
more effective conceptualization and implementation tools eventually the benefits of a commercial product usu-
[6]. Some researchers claim that the ‘‘signature’’ success ally outweighs the advantages of a homegrown solution.
of ABMs in public health is in the study of epidemics and There are however intermediary code bases that are typ-
infectious disease dynamics [5], [7], where the successes of ically open and community supported. These are usually
ABMs have demonstrated the importance of the role of social verified to some degree but usually not to the degree of
networks, human movement patterns, transportation systems, a commercial offering. All forms of ABM development
and the disease dynamic itself. This overwhelming amount of have associated learning curves. The largest and most
research in applying ABMs to the study of large scale infec- popular commercial ABM offering is that within Any-
tious disease spread (e.g. influenza, STIs) is not addressed Logic (anylogic.com). Opensource ABM frameworks
here. ABMs applied to institution-scale environments (rather include Repast (http://repast.sourceforge.net/), NetL-
than regional scales) are nonetheless emerging as an excellent ogo (http://ccl.northwestern.edu/netlogo/), and Swarm
vehicle for modeling hospitals due to their inherent ability to (swarm.org).
leverage social network analysis in a similar manner to social • Environment:
interactions of a large scale infectious disease. ◦ The topography or layout upon which agents oper-
The remainder of this paper is organized as follows. ate is an initial decision in ABM development.
Section II surveys the application of ABMs to hospital Environments can be real-world, synthesized, or
and similar institutional settings. Section III discusses data abstracted. Real-world environment can be captured
sources that may be useful in extending the models more fully. from hospital floor plans, while synthesized envi-
Section IV provides reference examples that encompasses ronments can be generated by the modeler with
many of the phenotypes of a typical hospital centric ABM. simplifications or assumptions compared to real
Section V provides a summary. floor plans. The environment can also be abstracted
entirely as a data point in the overall model and
II. ABMS WITHIN HOSPITALS assigning the agent to discrete non-physical loca-
Agent Based Modeling has seen a tremendous growth in tions within the computer code. However, the strong
many areas over the past 15 years and more recently one benefit of ABM is to allow for real-world envi-
of these areas being hospital and healthcare settings. The ronments to enhance the validity and credibility of
primary application of ABMs to hospital environments exam- the model, to ease the interpretation of simulation
ine patient flow (e.g. emergency departments) [8] and other results, and to assist in knowledge transfer.
hospital operational issues, and using ABM to examine the ◦ Most ABM simulation suites include some means
dynamics of infection spread within a hospital (e.g. the hos- of visualization of the agent within the envi-
pital’s role in an influenza epidemic [9] and the dynamics of ronment, and this benefit of ABM over other

228 VOLUME 2, 2014

M. R. Friesen, R. D. Mcleod: Survey of Agent-Based Modeling of Hospital Environments

modeling techniques has been accentuated with the gaps within the data and how those limitations
affordability and accessibility of high performance impact the veracity of the dataset for the ABM’s
desktop computing and graphical processing. Visu- objective. Pre-processing is generally required for a
alization of specific instances of the process allows single dataset as well as the consolidation of varied
verification of the model setup, simulation in datasets. While data is often technically available,
progress, and simulation results. Where a simu- political barriers may exist to access the data. The
lation requires a very large number of iterations area of real data is likely the area where ABMs
to generate meaningful findings, the visualization within healthcare facilities will more fully evolve
methods are halted while data accumulated. as they install in-house systems to capture the data
(e.g. patient flows) themselves, which will support
• Agents:
the ability to fine-tune ABMs. Such systems may
◦ The selection of agents is a foundational task of the include electronic records, dashboards, as well as
ABM developer. In most hospital ABMs, the logical technologies such as RFID. In the case of RFID,
selection of agents includes patients and hospital both inanimate and animate agents can be tracked.
staff members. Basic ABMs for hospital EDs may ◦ The assignment of rules that govern the interactions
only include patients, nurses, and physicians [6], between agents is the other foundational task of the
while more detailed ABMs include allied healthcare ABM developer, in order to capture the processes
providers who also consult within a hospital, and within the ABM, i.e. the process within the hospital.
potentially reaching as far as including visitors and Here, the ABM’s impact is evident in the natural
facility personnel not directly involved in health- inclusion of expert guidance to establish valid and
care delivery (e.g. maintenance staff). Furthermore, reliable agent interaction rules, formulated directly
an explicit decision should be made to include or in the lexicon of the hospital environment and in
exclude inanimate objects as agents within a hospi- the real-world topography of the practitioners (e.g.
tal ABM. Where the ABM is developed to model nurses and physicians in the hospital). The role of
infection spread (vs. patient flow), researchers have real data in the assignment of agent behavioral rules
considered the role of equipment and hospital fix- is just as significant as in the assignment of agent
tures as vectors for infection [11], including medi- characteristics or profiles.
cal instruments, bed capacity, allied areas relevant
• Interventions: Whether the hospital ABM was
to the main ABM focus (e.g. diagnostic services
developed to examine patient flow, infection spread
within an ED ABM). Inanimate agents are mod-
dynamics, or another purpose, the key objective in
eled without explicit agency or any decision making
developing an ABM is to introduce policy changes or
capability. Besides their role as vectors in infec-
interventions (agent profile changes, agent behavioral
tion spread, the availability and utilization strategies
changes, topography changes, or others) in order to
of inanimate agents (e.g. bed capacity, equipment
investigate ‘‘what if’’ scenarios. In patient flow ABMs,
availability) can also be illuminated via ABM.
interventions may include topography re-configurations
◦ The assignment of characteristics or profiles to the
of the ED or procedural reorganization such as low-
agents is another foundational task of the devel-
priority patient diversions within and between hospitals.
oper. The relevant factors for agent profiles are
In an infection spread ABM, interventions may include
determined by the objective of the ABM and may
agent hygiene behaviours and rules of contact.
include distributions of sex, age and other demo-
• Validation & Verification: There are emerging guidelines
graphic factors, physical origin and destination
addressing the importance and techniques to validate
within the topography and beyond the topography,
ABMs [12], including micro-face validation, macro-face
and risk factors associated with, for example, infec-
validation, output validation, backcasting to known data,
tion spread. The power of ABM is accentuated
and comparison of output to other modeling methods.
within today’s emerging big data culture, where
the sources of real data for agent characteristics
are numerous and varied. Data sets may or may B. EARLY HOSPITAL ABMs
not have been generated for the purpose at hand. Some early simulation models within healthcare settings were
Data sources may include hospital information sys- not specifically denoted as ABM but carry all the character-
tems, census data, government databases in the case istics of ABM. In 2006, researchers discussed a simulation
of publicly-funded health systems (e.g. Canadian model of an ED, recognizing the strengths of ABM as a
Institute for Health Information), cellular service means of communication across disciplines, indicating that
records that can be used to approximate physical part of their validation process was consultation with area
trajectories of Smartphone users upon a topography, experts (doctors and nurses). The model provided a means
and even Smartphone apps that are GPS-enabled. evaluating ‘‘what if scenarios’’, specifically, alternative triage
The developer must be aware of limitations and methods. Their work was also one of the first to recognize

VOLUME 2, 2014 229

 M. R. Friesen, R. D. Mcleod: Survey of Agent-Based Modeling of Hospital Environments

the importance of visualization as a strong element of ABM

and the requirement of using as real data as possible of
available [13].
Earlier in 2001, others argued for the use and require-
ments of ABM for hospital management, although they do
not appear to have built a working model [13]. Another
early paper investigated the role of an ABM within an ED,
deriving it from a more traditional intelligent software agent
perspective or multi agent system (MAS) perspective [14].
That work introduces processes, treatments, individual agents
and protocols for their interaction.

C. AGENT BASED MODELS FOR PATIENT FLOWS

FIGURE 1. An agent based nosocomial model within healthcare models.
An evolving literature exists with respect to applying ABMs,
alone or in complement to other techniques, to the operations relative to topography and agents. HAI ABMs may be useful
of EDs. In general, this literature addresses system-level per- in assessing the effectiveness of different infection control
formance dynamics, quantified in terms of patient safety [15], protocols or policies, intervention costs, as well as shedding
economic indicators [15], [16], staff workload and scheduling light on potential confinement failures which would accom-
[8], [17], [19], and patient flows. While much of the literature pany widespread infection dynamics [10], [27].
addresses system-level operational concerns during periods Several of the models oriented to nosocomial infections
of typical operation or stasis, there is also a literature on are known as ‘individual based models’, in which agents
modeling of healthcare operations during critical incidents are limited by definition to individuals (persons). One such
like disease outbreaks and terrorist attacks [20]. example is a mathematical individual based model for study-
More recently, others have modeled improvements to ing infection spread in a nursing home [28]. By contrast,
patient flow using an ABM running on a High Performance the notion of an agent based model expands the definition
Compute resource [21]. The ABM was built with NetLogo of agent beyond an individual person, to include inanimate
and representative of the role for which an ABM is well objects that can act as vectors of transmission for nosocomial
suited. The objective of the study was to provide impact infections. This concept is supported by a significant body
values of alternative policies for patient diversion. Not unex- of evidence that non-person agents play a significant role as
pectantly, the results indicated that patients that do not require infection transmission vectors [29], [30], including the CDC’s
urgent attention and are fast tracked or diverted improves the overview of SARS related information [31], which states ‘‘the
capacity of the ED and reduces the Length of Stay of patients virus also can spread when a person touches a surface or
that remain in the ED. More extensive consideration of ABMs object contaminated with infectious droplets and then touches
for patient flow in EDs are developed by the same researchers his or her mouth, nose, or eye(s). In addition, it is possible that
[22], [23], including the utilization of an ABM within a deci- the SARS virus might spread more broadly through the air
sion support system for EDs [24]. A contrasting technique (airborne spread) or by other ways that are not now known’’
to model patient flows would be DES [17]. In similar work, (pp. 1).
the role of re-triage in improving ED patient flow is exam- Nosocomial agent based modeling initiatives focus upon
ined [25]. The results are not unexpected and lend additional explicit modeling of structure and behaviour extending the
credibility to the use of ABMs in healthcare modeling by agent based model to include individuals, inanimate objects,
facilitating and modeling ‘‘what if’’ scenarios. In other work, and locations, in order to investigate an organization’s poli-
a pseudo-agent based approach is introduced into a DES in an cies and practices in the event of a serious nosocomial infec-
attempt to capture the representative strengths of each mod- tion outbreak. Much of the current efforts in nosocomial
eling approach for simulating an emergency department [26]. ABMs set the framework for potential future efforts in mod-
In that work, the importance of interaction at the agent level eling and evaluation of organization’s documented infection
is illustrated, not typically captured with DES. control plans (policies and practices). For example, best prac-
tices [32] are available for healthcare practitioners and pol-
D. AGENT BASED MODELS FOR icy makers dealing with health care-associated infections in
NOSOCOMIAL INFECTIONS patient and resident populations. This may be a useful refer-
Fig. 1 illustrates where modeling nosocomial infections, or ence to model, as a means of identifying and evaluating their
hospital-acquired infections (HAIs) would be characterized effectiveness. At this time, best practice documents typically
within the range of healthcare models. reflect ‘‘consensus positions on what the committee deems
In general, the modeling of HAI or nosocomial infections prudent practice and are made available as a resource to the
is perhaps the best suited area for ABMs within healthcare public health and healthcare provider’’ (p. ii). Clearly this
institutions. This is largely a consequence of being able to is also an opportunity for ABM models to contribute to a
address all of the model components included in Fig. 1 collaborative, multi-stakeholder effort.

230 VOLUME 2, 2014

M. R. Friesen, R. D. Mcleod: Survey of Agent-Based Modeling of Hospital Environments

Despite nosocomial modeling’s natural fit with the ABM III. VISUALIZATIONS
approach, it is a fairly recent area of exploration for health- Agent based models are well suited to visualization and
care ABMs [33]–[35]. One of the earlier simulation efforts animation as means of informally verifying and validating
modeled antibiotic resistance in hospitals, contrasting and an the model, as well as communicating with practitioners and
individual based model with that of a differential equation policy-makers, in that visualization renders potentially com-
based model, including consideration of where they can be plex dynamics more intelligible to the recipient of the infor-
used in conjunction with one another [36]. Another study mation. YouTube is an excellent platform for presenting a
investigated the spread of a nosocomial pathogen in a dial- model’s or project’s progress. Fig. 2 illustrates a prototype
ysis unit using a Monte Carlo individual based model [10]. emergency department. A large waiting area is depicted,
The dialysis unit is a very good example of where agent where one person is identified at a higher acuity level than
based models may be particularly useful as ‘‘the frequency of the others. Detailed modeling allows for modeling of social
patient visits and intimate, prolonged physical contact with distancing within a waiting area, as well as obtaining esti-
the inanimate environment during dialysis treatments make mates of ED length of stay, influenced by policy or staffing
these facilities potentially efficient venues for nosocomial interventions modeled via ABM. Two videos illustrating the
pathogen transmission’’ (pp. 1176). In related paper [27], the potential use of visualization with ABMs within healthcare
same authors developed a fairly abstracted nosocomial ABM facilities are appended to this paper as on-line videos. One
within an intensive care unit, advocating for a ‘‘conceptually video illustrates patient flow through an ED, and the other the
simple discrete element (agent-based or cellular automata) use of an ABM in evaluating the efficacy of an RFID/RTLS.
models [that] can explicitly address ‘geographic’ considera-
tions and probabilistic transmission dynamics germane to the
spatially intricate environments and small population sizes
characteristic of ICUs’’ (pp. 174). In another nosocomial
ABM of an intensive care unit, operational and epidemi-
ological features are considered in an attempt to estimate
the effect of understaffing and overcrowding on infection
spread [37]. The ABM simulated contact-mediated pathogen
transmission, which should allow one to establish quantita-
tive relations between patient flow, staffing conditions and
pathogen colonization in patients. Another individual based
approach investigated the role of cohorting, with the aim
of minimizing the possible interactions between individuals
within a ward [38]. In a relatively recent nosocomial ABM, a
combination of differential equation models and probabilistic
models are used for each agent in order to simulate changes,
FIGURE 2. Agent based model of an emergency department.
over time, in the bacteria sub-populations within the agent’s
body [39]. As with many ABM efforts, work is ongoing in AnyLogic, a commercial simulation suite of tools has also
terms of validation and verification. In order to construct been fairly widely deployed for emergency department simu-
biologically plausible transmission risk models that can guide lation, with an example at http://www.youtube.com/watch?v=
cross infection control, researchers have developed an RFID LaHdn3GBIWM (Fig. 3).
tracking system in an ED by which to extract agent contact It is very likely that hospital simulations and their visu-
data on the understanding of the critical role that contact alizations will continue to improve, driven largely by ini-
patterns play in cross-infection control [40]. This type of tiatives in ‘serious games’ and the gamification of models
high-fidelity individual data, topography, as well as contact and simulations. An example of the level of detail one can
patterns is ideally suited for an ABM as well. envision can be found at http://www.youtube.com/watch?v=
7CwoMsVyo2Y. The animation utilized Flexsim, which is a
E. MISCELLANY discrete event simulator, but similar high quality animations
Other scenarios that have been investigated using ABMs can be extracted from ABMs as well (Fig. 4).
within healthcare settings include optimization of computer
terminals [41], serving as another illustration of incorporating IV. ENHANCEMENTS TO HOSPITAL ABMs
inanimate objects as agents within the ABM. The use of ABMs tend to be labour intensive and are often deployed
electronic devices including stationary workstations, mobile for specific experiments or studies. Although time consum-
workstations, tablets, and Smartphones in healthcare delivery ing, they generate vast quantities of data for each run. Typ-
is evolving very rapidly, and will likely develop momentum ically, the many runs are used to extract statistics that can
that will outpace the insights of ABMs in this area. Other be used to demonstrate the impact of the policy or inter-
ABMs are oriented to interactions between hospitals where vention being simulated. This massive data generator also
patient diversion on response to load was modeled [42]. offers the potential to be mined and used in machine learning

VOLUME 2, 2014 231

 M. R. Friesen, R. D. Mcleod: Survey of Agent-Based Modeling of Hospital Environments

ABM and computational fluid dynamics model. In another

instance, a hybrid ABM-DES model for emergency medical
services in a city is conjectured, although an actual model
or simulation has not been reported [45]. In a more pedes-
trian optimization, resource planning for placement of RFID
readers in an RFID system may be integrated into the ABM
as a means of estimating errors associated with the tracking
system [46].

V. SUMMARY
This paper reviews the current status of and advocates for the
increased use of ABMs within healthcare settings, particu-
larly within hospitals. In this context, ABMs of nosocomial
infection spread are among the most advanced and numerous
at this time, with an emerging body of work associated with
ABMs investigating patient flow and other operational pro-
FIGURE 3. Agent based model of an emergency department using cesses in hospitals. However, ABMs are not without their dis-
AnyLogic.
advantages as well. Some of these disadvantages are related
to developing robust validation and verification techniques
which the ABM research community agrees upon; this is a
difficulty faced by many simulation modalities. Other diffi-
culties arise from the challenge of generating accurate models
of agent behaviours and interactions, as well as data extracted
from the systems being modeled. The emergence of ABMs
will likely be within a more integrated simulation and analysis
suite, often combined with other established techniques as
demonstrated within the more recent literature. The role of
ABMs as useful simulation vehicles within healthcare facili-
ties is still in its infancy, but offers tremendous potential for
the better understanding and optimization of these complex
systems.

FIGURE 4. ‘‘Gamified’’ simulation visualization using Flexsim.

REFERENCES
[1] E. Bonabeau, ‘‘Agent-based modeling: Methods and techniques for simu-
or pattern classification algorithms. For example, instead of lating human systems,’’ PNAS, vol. 99, no. 3, pp. 7280–7287, 2002.
having emergency physicians travel through the ED to see [2] N. Hupert, W. Xiong, and A. Mushlin, ‘‘The virtue of virtuality: The
patients in individual treatment rooms, the patients would promise of agent-based epidemic modeling,’’ Trans. Res., vol. 151, no. 6,
pp. 273–274, 2008.
travel through the ED to visit the (stationary) physician, with [3] A. Borshchev and A. Filippov, ‘‘From system dynamics and discrete event
this policy generated via a genetic program combined with an to practical agent based modeling: Reasons, techniques, tools,’’ in Proc.
ED ABM [43]. 22nd Int. Conf. Syst. Dyn. Soc., 2004, pp. 1–23.
[4] N. Mustafee K. Katsaliaki, and S. J. E. Taylor, ‘‘Profiling literature in
Another enhancement to ABMs and to simulation in gen- healthcare simulation,’’ Simulation, vol. 86, nos. 8–9, pp. 543–558, 2010.
eral arises when data analysis augments the simulation. For [5] D. A. Luke and K. A. Stamatakis, ‘‘Systems science methods in pub-
example, researchers have analyzed data to identify best sce- lic health: Dynamics, networks, and agents,’’ Annu. Rev. Public Health,
vol. 33, pp. 357–376, Jan. 2012.
narios extracted from discrete event simulation of an ED [44].
[6] P. Escudero-Marin and M. Pidd, ‘‘Using ABMS to simulate emergency
Although scenarios were extracted from a DES as compared departments,’’ in Proc. Winter Simul. Conf., 2011, pp. 1239–1250.
to an ABM, the same type of enhanced data analysis are [7] S. Barnes, B. Golden, and S. Price, ‘‘Applications of agent-based model-
beginning to emerge in ABM, borrowing heavily from non- ing and simulation to healthcare operations management,’’ in Handbook
of Healthcare Operations Management. New York, NY, USA: Springer-
parametric methods in operations research. Although ABMs Verlag, 2013, pp. 45–74.
are a useful paradigm for aiding to the understanding of [8] S. S. Jones and R. S. Evans, ‘‘An agent based simulation tool for scheduling
a complex system on their own, this significant existing emergency department physicians,’’ in Proc. AMIA Annu. Symp., 2008,
pp. 338–342.
contribution will be augmented with the integration of data
[9] M. Laskowski, B. C. P. Demianyk, J. Witt, S. N. Mukhi, M. R. Friesen,
analytics. and R. D. McLeod, ‘‘Agent-based modeling of the spread of influenza-like
ABMs may also be useful in hospital facility design where illness in an emergency department: A simulation study,’’ IEEE Trans. Inf.
Technol. Biomed., vol. 15, no. 6, pp. 877–889, Nov. 2011.
additional importance may concern the role of the HVAC
[10] J. R. Hotchkiss, P. Holley, and P. S. Crooke, ‘‘Analyzing pathogen trans-
system within various departments. This would imply a mission in the dialysis unit: Time for a (schedule) change,’’ Clin. J. Amer.
hybrid of simulation techniques, likely encompassing an Soc. Nephrol., vol. 2, no. 6, pp. 1176–1185, 2007.

232 VOLUME 2, 2014

M. R. Friesen, R. D. Mcleod: Survey of Agent-Based Modeling of Hospital Environments

[11] J. R. Hotchkiss, D. G. Strike, and P. S. Crooke, ‘‘Pathogen transmission [33] R. D. McLeod, M. Laskowski, M. R. Friesen, and B. Podaima, ‘‘Agent
and clinic scheduling,’’ Emerging Infectious Diseases, vol. 12, no. 1, based models for nosocomial infections: Modeling of hospital-acquired
pp. 159–162, 2006. infections within a hospital,’’ in Proc. Rep. Public Health Agency Canada,
[12] W. Rand and R. T. Rust, ‘‘Agent-based modeling in marketing: Guidelines 2009, pp. 1–3.
for rigor,’’ Int. J. Res. Marketing, vol. 28, no. 3, pp. 181–193, 2011. [34] B. S. Ong, M. Chen, V. Lee, and J. C. Tay, ‘‘An individual-based
[13] R. Sibbel and C. Urban, ‘‘Agent-based modeling and simulation for hos- model of influenza in nosocomial environments,’’ in Proc. ICCS, 2008,
pital management,’’ in Cooperative Agents: Applications in the Social pp. 590–599.
Sciences, N. J. Saam and B. Schmidt, Eds. Dordrecht, The Netherlands: [35] Y. Meng, R. Davies, K. Hardy, and P. Hawkey, ‘‘An application of
Kluwer Academic, 2001, pp. 183–199. agent-based simulation to the management of hospital-acquired infection,’’
[14] A. Daknou, S. Hammadi, H. Zgaya, and H. Hubert, ‘‘Agent based opti- J. Simul., vol. 4, no. 1, pp. 60–67, 2010.
mization and management of healthcare processes at the emergency [36] E. D’Agata, P. Magal, D. Olivier, S. Ruan, and G. F. Webb, ‘‘Modeling
department,’’ Int. J. Math. Comput. Simul., vol. 2, no. 3, pp. 285–294, antibiotic resistance in hospitals: The impact of minimizing treatment
2008. duration,’’ J. Theoretical Biol., vol. 249, no. 3, pp. 487–499, 2007.
[15] A. K. Kanagarajah, P. A. Lindsay, A. M. Miller, and D. W. Parker, ‘‘An [37] J. Ferrer, M. Salmon, and L. Temime, ‘‘Nosolink: An agent-based approach
exploration into the uses of agent-based modeling to improve quality to link patient flows and staff organization with the circulation of nosoco-
of health care,’’ in Proc. Int. Conf. Complex Syst., Boston, MA, USA, mial pathogens in an intensive care unit,’’ Proc. Comput. Sci., vol. 18, no. 1,
Jun. 2006, pp. 1–10. pp. 1485–1494, 2013.
[16] D. Blachowicz, J. H. Christiansen, A. Ranginani, and K. L. Simunich, [38] L. Milazzo, J. L. Bown, A. Eberst, G. Phillips, and J. W. Crawford,
‘‘Future EHR ROI: Agent-based modeling and simulation offers a new ‘‘Modelling of healthcare associated infections: A study on the dynamics
alternative to traditional techniques,’’ J. Healthc Inf. Manag., vol. 22, no. 1, of pathogen transmission by using an individual-based approach,’’ Comput.
pp. 39–45, 2008. Methods Programs Biomed., vol. 104, no. 2, pp. 260–265, 2011.
[17] T. J. Coats and S. Michalis, ‘‘Mathematical modelling of patient flow [39] R. Pasupathy, S. H. Kim, A. Tolk, R. Hill, and M. E. Kuhl, ‘‘A hybrid
through an accident and emergency department,’’ Emergency Med. J., agent-based and differential equations model for simulating antibiotic
vol. 18, no. 3, pp. 190–192, 2001. resistance in a hispital ward,’’ in Proc. Winter Simul. Conf., Dec. 2013,
pp. 1419–1430.
[18] T. Ruohonen, P. Neittaanmaki, and J. Teittinen, ‘‘Simulation model for
[40] D. W. Lowery-North et al., ‘‘Measuring social contacts in the emergency
improving the operation of the emergency department of special health
department,’’ PloS, vol. 8, no. 8 p. e70854, 2013.
care,’’ in Proc. IEEE Winter Simul. Conf., Mar. 2006, pp. 453–458.
[41] M. R. Poynton et al., ‘‘Computer terminal placement and workflow in an
[19] C. W. Spry and M. A. Lawley, ‘‘Evaluating hospital pharmacy staffing
emergency department: An agent-based model,’’ in Proc. CSSS, Aug. 2007,
and work scheduling using simulation,’’ in Proc. Winter Simul. Conf., Dec.
pp. 1–3.
2005, pp. 4–7.
[42] M. Laskowski and S. Mukhi, ‘‘Agent-based simulation of emergency
[20] V. Mysore et al., ‘‘Agent modeling of a Sarin attack in Manhattan,’’ in Proc.
departments with patient diversion,’’ Electronic Healthcare. Berlin,
1st Int. Workshop ATDM, 2006, pp. 108–115.
Germany: Springer-Verlag, 2009, pp. 25–37.
[21] M. Taboada, E. Cabrera, F. Epelde, M. L. Iglesias, and E. Luque, ‘‘Using [43] M. Laskowski, ‘‘A prototype agent based model and machine learning
an agent-based simulation for predicting the effects of patients derivation hybrid system for healthcare decision support,’’ Int. J. E-Health Med.
policies in emergency departments,’’ Proc. Comput. Sci., vol. 18, no. 1, Commun., vol. 2, no. 4, pp. 67–90, 2011.
pp. 641–650, 2013. [44] A. Al-Refaie, R. H. Fouad, M.-H. Li, and M. Shurrab, ‘‘Applying simu-
[22] E. C. Flores, ‘‘Agent based simulation to optimise emergency depart- lation and DEA to improve performance of emergency department in a
ments,’’ M.S. thesis, Dept. d’Arquitectura de Comput. i Sistemes Oper- Jordanian hospital,’’ Simul. Modelling Practice Theory, vol. 41, pp. 59–72,
atius., Univ. Autonomade Barcelona, Barcelona, Spain, 2010. Feb. 2014.
[23] E. Cabrera, E. Luque, M. Taboada, F. Epelde, and M. L. Iglesias, ‘‘ABMS [45] A. Anagnostou, J. Eatock, and S. J. E. Taylor, ‘‘Nosopolis: Towards a
optimization for emergency departments,’’ in Proc. Winter Simul. Conf., hybrid agent-based discrete event simulation tool for emergency medical
2012, p. 89. services improvement,’’ in Proc. Winter Simul. Conf., 2012, p. 365.
[24] M. Taboada, E. Cabrera, M. L. Iglesias, F. Epelde, and E. Luque, ‘‘An [46] M. Laskowski, B. Demianyk, M. R. Friesen, and R. D. McLeod, ‘‘Uncer-
agent-based decision support system for hospitals emergency depart- tainties inherent in RFID tracking systems in an emergency department,’’
ments,’’ in Proc. Comput. Sci., vol. 4. 2011, pp. 1870–1879. in Proc. WHCM, Jun. 2010 pp. 1–6.
[25] M. H. Rahmat, M. Annamalai, S. A. Halim, and R. Ahmad, ‘‘Agent-based
modelling and simulation of emergency department re-triage,’’ in Proc.
BEIAC, Apr. 2013, pp. 219–224.
[26] M. E. Lim, A. Worster, R. Goeree, and J.-É. Tarride, ‘‘Simulating an emer- MARCIA R. FRIESEN received the Ph.D. degree
gency department: The importance of modeling the interactions between
in biosystems engineering from the University
physicians and delegates in a discrete event simulation,’’ BMC Med. Inf.
of Manitoba, Winnipeg, MB, Canada, in 2009.
Decision Making, vol. 13, no. 1, pp. 59–60, 2013.
She is a multidisciplinary Design Engineering
[27] J. R. Hotchkiss, D. G. Strike, D. A. Simonson, A. F. Broccard, and
P. S. Crooke, ‘‘An agent-based and spatially explicit model of pathogen
Researcher with active research in engineering
dissemination in the intensive care unit,’’ Critical Care Med., vol. 33, no. 1, education, and computer modeling and simulation.
pp. 168–176, 2005. She is currently an Associate Professor of Design
[28] C. van den Dool, M. J. M. Bonten, E. Hak, J. C. M. Heijne, and J. Wallinga, Engineering with the University of Manitoba. She
‘‘The effects of influenza vaccination of health care workers in nursing is also serving as the President of the Associa-
homes: Insights from a mathematical model,’’ PLoSMedicine, vol. 5, tion of Professional Engineers and Geoscientists of
no. 10, p. 200, 2008. Manitoba.
[29] B. Bean, B. M. Moore, B. Sterner, L. R. Peterson, D. N. Gerding, and
H. H. Balfour, ‘‘Survival of influenza viruses on environmental surfaces,’’
J. Infectious Diseases, vol. 146, no. 1, pp. 47–51, 1982.
[30] Y. Thomas et al., ‘‘Survival of influenza virus on banknotes,’’ Appl. Envi- ROBERT D. MCLEOD received the B.Sc. and
ron. Microbiol., vol. 74, no. 10, pp. 3002–3007, 2008. M.Sc. degrees in electrical engineering and the
[31] (2007). CDC SARS Fact Sheet [Online] Available: Ph.D. degree from the University of Manitoba
http://www.cdc.gov/sars/about/fs-SARS.pdf in 1981, 1983, and 1985, respectively, where he
[32] (2008, Jun.). Best Practices for Surveillance of Health Care- is a Professor of Electrical and Computer Engi-
Associated Infections in Patient and Resident Populations, Ontario neering. His research interests include agent-based
Ministry of Health and Long-Term Care, Provincial Infectious modeling in diverse application areas with empha-
Diseases Advisory Committee (PIDAC) [Online]. Available: sis on leveraging auxiliary data sources originally
http://www.publichealthontario.ca/en/eRepository/Surveillance_3- intended for alternate purposes.
3_ENGLISH_2011-10-28%20FINAL.pdf

VOLUME 2, 2014 233

View publication stats