37

Downloaded by [ Faculty of Nursing, Chiangmai University 5.62.158.117] at [07/18/16]. Copyright © McGraw-Hill Global Education Holdings, LLC. Not to be redistributed or modified in any way without permission.
Planning, Design, and Implementation
of Information Technology in
Complex Healthcare Systems
Thomas R. Clancy

• OBJECTIVES
. Define complex systems in the context of General Systems Theory.
1
2. Describe complex adaptive systems, a special case of complex systems.
3. Illustrate challenges and solutions for planning and designing information tech-
nology in complex healthcare systems.
i. Wicked problems
ii. High-reliability organizations
iii. Structured and agile design methods
4. Provide examples of tools and methods used to plan and design information sys-
tems in complex healthcare processes.
i. Computer-Aided Software Engineering
ii. Discrete Event Simulation
iii. Network Analysis Tools

• KEYWORDS
Complex adaptive systems
Wicked problems
Agile design
Discrete event simulation
Network science
Nurse informaticist
Electronic Medical Record
Systems analyst

today are among some of the most complex systems in

INTRODUCTION the world. They are composed of diverse specialties and
The introduction of information technology into the multiple providers linked through a complex information
workflow of clinicians requires thoughtful planning, network. Understanding and predicting the impact of new
design, and implementation. Healthcare organizations of technology on healthcare system behavior is an ongoing

525

c37.indd 525 20-02-2015 18:52:17

 526    P art 6 • N ursing I nformatics —C omplex A pplications

Downloaded by [ Faculty of Nursing, Chiangmai University 5.62.158.117] at [07/18/16]. Copyright © McGraw-Hill Global Education Holdings, LLC. Not to be redistributed or modified in any way without permission.
challenge for nurse informaticists. That is because com- networks. To achieve goals, systems must transform
plex systems are unpredictable and what works in one inputs into outputs. An input might include the entry of
organization does not guarantee success in another. information into a computer where it is transformed into
Nurse informaticists play a critical role in the success- a series of binary digits (outputs). These digits can then
ful deployment of healthcare information technology. Their be routed across networks to end users more efficiently
unique blend of clinical knowledge and information systems than other forms of communication. As simple systems
expertise make them ideal systems analysts. As complex interact, they are synthesized into a hierarchy of increas-
systems research discovers new findings, nurse informati- ingly complex systems, from sub-atomic particles to entire
cists must remain informed on how new knowledge can civilizations (Skyttner, 2001). Levels within the hierarchy
be translated into practice settings. The objective of this have novel characteristics that apply universally upward to
chapter is to provide a review of complex systems from the more complex levels but not downward to simpler levels.
context of a general systems theory and then provide an In other words, at lower or less complex levels, systems
overview of new strategies to plan, design, and implement from different fields share some characteristics in com-
information systems in healthcare organizations. mon. For example, both computer and biological systems
communicate information by encoding it, either through
computer programs or via the genome respectively. It is
the inter-relationships and inter-dependencies between
GENERAL SYSTEMS THEORY and across levels in hierarchies that create the concept of
A system can be defined as a set of interacting units or ele- a holistic system separate and distinct from its individual
ments that form an integrated whole intended to perform components. As commonly noted, a system is greater than
some function (Skyttner, 2001). An example of a system simply the sum of its individual parts.
in healthcare might be the organized network of provid-
ers (nurses, physicians, and support staff), equipment, and Complex Systems
materials necessary to accomplish a specific purpose such as
the prevention, diagnosis, and treatment of illness. General System behavior can be related to the concept of com-
systems theory (GST) provides a general language which plexity. As systems transition from simple to complex it is
ties together various areas in inter-disciplinary communi- important for nurse informaticists to understand that sys-
cation. It endeavors toward a universal science that joins tem behavior also changes. This is especially important in
together multiple fields with common concepts, principles, the management of data and information for clinical and
and properties. Although healthcare is a discipline in and administrative decision-making. Table 37.1 presents some
of itself, it integrates with many other systems from fields examples of differences between simple and complex sys-
as diverse as biology, economics, and the physical sciences. tems in healthcare.
Systems can be closed, isolated, or open. A closed As healthcare systems become increasingly com-
system can only exchange energy across its borders. For plex it becomes progressively more difficult to predict
example, a greenhouse can exchange heat (energy) but not how changes in provider workflow are impacted by new
physical matter with its environment. Isolated systems information technology. This is especially important for
cannot exchange any form of heat, energy, or matter across those individuals responsible for planning, designing, and
their borders. Examples of closed and isolated systems are implementing information systems in the healthcare envi-
generally restricted to physical sciences such as physics ronment of today.
and chemistry. An open system is always dependent on The evolution of information systems in healthcare
its environment for the exchange of matter, energy, and is replete with failed implementations, cost overruns,
information. Healthcare systems are considered open and and dissatisfied providers. It is important to understand,
are continuously exchanging information and resources however, that healthcare systems of today are some of the
throughout their many integrating sub-systems. Some most complex in the world. The unpredictable nature of
fundamental properties collectively comprise a general complex healthcare systems results from both structural
systems theory of open systems. and temporal factors. The structure of complex systems
is composed of numerous elements connected through
a rich social network. In hospitals, those interactive ele-
Open Systems
ments include both internal entities such as patients,
All open systems are goal seeking. Goals may be as fun- nurses, physicians, technicians, and external entities, such
damental as survival and reproduction (living systems) as other hospitals, insurance payers, and regulatory agen-
to optimizing the flow of information across computer cies. The way information flows through this complex

c37.indd 526 20-02-2015 18:52:18

 Chapter 37 • Planning, Design, and Implementation of Information Technology in Complex Healthcare Systems    527

Downloaded by [ Faculty of Nursing, Chiangmai University 5.62.158.117] at [07/18/16]. Copyright © McGraw-Hill Global Education Holdings, LLC. Not to be redistributed or modified in any way without permission.
  TABLE 37.1    Simple Versus Complex Healthcare Systems
Simple System Example Complex System Example

There are a small number of A one physician, one There are a large number of A large academic health center
elements in the system. nurse practice. elements in the system. with multiple providers and
Elements are parts of the specialties.
system (people, places,
and things).
There are few interactions Interactions occur primar- Many interactions between Thousands of interactions occur
between the elements. ily between the nurse elements. daily between patients, provid-
and physician, and ers, and machines (computers).
patient.
Interaction between ele- There is direct commu- Interactions between ele- Interactions occur between mul-
ments is highly orga- nication between the ments are loosely orga- tiple providers. The medical
nized. The number of nurse, physician, and nized. There are multiple center may have a sophisticated
sub-systems is small and patient. These interac- levels in the hierarchy EHR with many applications
fixed. Information flows tions may be enabled of sub-systems where integrated within it.
primarily in two or three through a single, prac- information flows in many
directions. tice-based EHR. directions.
System behavior varies little. Because patient acuity System behavior is probabi- Because patient acuity can vary
Day-to-day activities are varies little, provider listic and it is impossible to considerably, provider workload
highly predictable. workload remains the predict with certainty. can change significantly from
same from day to day. day to day.
The system does not evolve The practice is mature and The system evolves over time. The medical center has multiple
over time, is largely is not accepting new The system is largely open specialties that accept diverse
closed, and is slow to patients. and constantly adapting to disease conditions that require
adapt to environmental the external environment. ongoing research and new
changes. knowledge.
The system is not robust to Changes in reimburse- Because of its diversity of Because there are many providers
environmental shocks. ment or competitive services, the system can with multiple specialties, the
pressures can quickly withstand environmental system can withstand significant
eliminate the practice shocks. environmental changes that
in the marketplace. may impact one area.

Web of channels depends on the technology used, organi- At the micro-system level, the relationship between
zational design, and the nature of tasks and relationships cause and effect is somewhat predictable. However, as the
(Clancy & Delaney, 2005). organizational structure expands to include meso- and
As systems evolve from simple to complex systems a macro-system levels, causal relationships between stimu-
hierarchy of levels emerges within the organization. From lus and response become blurred. With increasing levels
a theoretical standpoint, complex systems form a “fuzzy,” of complexity, the rich network of interactions between
tiered structure of macro-, intermediary, and micro-­ nurses, physicians, pharmacists, and other providers at
systems (Clancy & Delaney, 2005). In hospital environ- multiple levels within the organization quickly expands
ments, for example, nursing units can be understood as the variability and range of potential states a system might
micro-systems within the health system. When nursing exhibit. This is the driving force behind the unpredictabil-
units are combined into divisions (e.g., critical care, medi- ity in complex system behavior.
cal-surgical, or maternal-child), the nursing units’ internal
processes expand beyond their boundaries (for example,
Complex Adaptive Systems
the medication usage process) to create intermediary or
meso-systems. Major clinical and administrative systems Complex adaptive systems (CAS) are special cases of
emerge at the macro-level through the aggregation of complex systems; the key difference being that a CAS can
interactions occurring at lower levels. learn and adapt over time (Clancy, Effken, & Persut, 2008).

c37.indd 527 20-02-2015 18:52:18

 528    P art 6 • N ursing I nformatics —C omplex A pplications

Downloaded by [ Faculty of Nursing, Chiangmai University 5.62.158.117] at [07/18/16]. Copyright © McGraw-Hill Global Education Holdings, LLC. Not to be redistributed or modified in any way without permission.
From biologic to man-made systems, CAS’s are ubiq- relationships can improve communication both vertically
uitous. The human body, comprised of its many sub- and horizontally.
systems (cardiovascular, respiratory, neurologic), is a Complex adaptive systems exhibit specific charac-
CAS and continuously adapts to short- and long-term teristics that are different than simple systems and the
changes in the environment. The same principles can terminology used to define them may be unfamiliar to
be applied to social organizations such as clinics or hos- healthcare providers. Table 37.2 provides a descrip-
pitals. Organizational learning is a form of adaptation tion of terminology used to describe complex system
and it has the capacity to change culture. New reporting behavior. Each term in the table builds on the previous
relationships can alter the structure of a social network term to demonstrate how system behavior becomes
and act as the catalyst for adaptation to environmental unpredictable.
change. For example, the movement from hierarchical As healthcare systems become increasing complex,
organizational structures to quasi-networked reporting those individuals challenged with planning, designing,

  TABLE 37.2    Complex Adaptive Systems Terminology

Term Description Example

Combinatorial A system with a large number of alternative The medication management process involves many
Complexity states. different providers, medications, times, routes, and
patients. The many potential combinations of these
at any one time can be enormous.
Dynamic The degree to which system behavior becomes Computer provider entry systems can create a network
Complexity more complex over time. of interactions between the prescription, transcrip-
tion, dispensing, and administration process that is so
complex that errors may become lost in the system.
Over time, if not discovered, the effects of the error
can amplify and result in serious harm to a patient.
Feedback The iterative process where system outputs Growth in the complexity of CPOE systems is driven
loop back and impact system inputs. by feedback. Since CPOE systems are composed of
combinations of many different applications, new
combinations of them create new features adding to
overall complexity. Technology actually creates itself
out of itself
Exponential Growth The doubling of an output in a fixed period of The computational processing capacity of computers
time. Exponential growth can be positive or doubles about every 18 months. This phenomenon,
negative. known as Moore’s Law, has held steady for many
years.
Non-linearity When stimulus and response are unequal. A The introduction of information technology into the
large stimulus may have little effect on out- traditional workflow of providers can elicit a
comes while a small stimulus may generate significant backlash if not carefully managed.
a large response. The long-term negative effects of provider resistance
can far exceed the initial costs of purchasing the
system.
Self-organization The “coming together” of system entities (pro- Work-arounds are a form of self-organization. For
viders) to achieve a goal without the guid- example, if a new bar code medication manage-
ance or influence of a central authority. ment system interferes with the workflow of nurses,
they may self-organize and apply shortcuts without
knowledge by the manager.
Emergence New patterns of behavior arising from A spike in medication errors from work-arounds may
self-organization emerge as a new pattern on managerial reports.

(continued)

c37.indd 528 20-02-2015 18:52:18

 Chapter 37 • Planning, Design, and Implementation of Information Technology in Complex Healthcare Systems    529

Downloaded by [ Faculty of Nursing, Chiangmai University 5.62.158.117] at [07/18/16]. Copyright © McGraw-Hill Global Education Holdings, LLC. Not to be redistributed or modified in any way without permission.
  TABLE 37.2    Complex Adaptive Systems Terminology (continued)
Term Description Example

Stochastic The distribution of events is a mixture of deter- The sequence of times between medication administra-
ministic and random processes. tions is a mixture of scheduled (deterministic) and
PRN (random) processes. The distribution is neither
fully deterministic nor random.
Chaos Chaos describes a class of systems in which The behavior of a system, from simple to complex lies
small changes to the initial conditions of the on a continuum from fully deterministic to fully ran-
system create deterministic (non-random), dom. Chaos appears random but in fact is determin-
but very complex behavior. istic. It is found primarily in natural systems (weather
patterns, biologic systems, chemistry) but also in
man-made systems. Chaotic behavior often shows
itself in the sequence of events in a process, i.e., inter-
arrival rate of admissions to a nursing unit.
Power Law A power law distribution is characterized Pareto’s Law where 80% of the output is caused by 20%
Distributions by a few events of enormous magnitude of the inputs is an example of a power law distribu-
disbursed among many events with much tion. For example, 80% of the complaints surround-
smaller impact. ing a new information system may be verbalized by
only 20% of the users.

and implementing information technology must be distances through large cable networks connecting insti-
knowledgeable of complex adaptive system principles as tutions. As rivers fan out to create deltas, smaller chan-
well as the tools and methods to analyze their behavior. nels form and flow slows and travels a shorter distance.
And as computer networks branch out to users within
organizations, the flow of digital information also slows
IT IS ALL ABOUT FLOW: FAST AND and travels shorter distances as it supplies multiple users.
To optimize and maintain flow, fast and long flow must
LONG, SHORT AND SLOW equal short and slow flow. In natural systems such as riv-
A key concept to understand is that all systems, whether ers, the ratio of channels that are slow and short make up
they are natural (rivers, trees, weather) or man-made about 80% of total channels, while the remaining 20% are
(stock markets, the Intranet, health systems), have cur- fast and long. And this ratio shows up not only in rivers
rents that flow through them. Oxygen and blood flow but in tree branches, the human circulatory system, and
through animate systems such as trees or humans while computer networks! Although flow is faster in large river
water and electricity flow in currents through inanimate channels than in smaller tributaries, the total flow is main-
systems such as rivers or lightening. Patients flow through tained by the creation of numerous tributaries that as a
hospitals and clinics while information flows through whole are equal to the flow of the larger channels (Bejan
computer networks and electronic health records. & Zane, 2012).
Without flow, a system cannot exist and to survive, it must This fundamental concept can also be applied to the
evolve in a direction that improves its access to designs flow of electronic data and the design of information sys-
that improve the flow of currents through them. In nature, tems. When information is delayed, either as a result of
the branching pattern in trees is not accidental. Rather the poorly designed systems and cumbersome processes and
configuration of branches is the result of natural selec- policies, it is usually a mismatch between “short and slow”
tion and the search for optimal designs to transport water and “fast and long” information flow. If it is overly time
and oxygen from the ground back to the atmosphere. This consuming to enter provider orders into an EHR (short
treelike branching pattern is so efficient that it is also seen and slow) then it delays those orders from being com-
in other structures such as river basins, the human respi- municated in a timely manner through the health sys-
ratory system, computer networks, and road systems. tems computer network to the pharmacy (long and fast).
Just as water flows fast for long distances through wide As a result, orders stack up (develop queues) and delay
river channels, digitized information flows fast for long the flow of information. Whenever, delays occur, nurse

c37.indd 529 20-02-2015 18:52:18

 530    P art 6 • N ursing I nformatics —C omplex A pplications

Downloaded by [ Faculty of Nursing, Chiangmai University 5.62.158.117] at [07/18/16]. Copyright © McGraw-Hill Global Education Holdings, LLC. Not to be redistributed or modified in any way without permission.
informaticists should look for imbalances between “short Wicked problems cannot be solved by the common
and slow” and “fast and long” information flow. rubric of defining the problem, analyzing solutions, and
making a recommendation in sequential steps; the rea-
son being, there is no clear definition of the problem. By
INFORMATION SYSTEM PLANNING engaging all stakeholders in problem solving, those people
most affected participate in the planning process and a
The inter-relatedness of sub-systems characteristic of com- common, agreed approach can be formulated. Of utmost
plex healthcare processes requires participation by many importance is understanding that different problem solu-
different stakeholders in the planning cycle. In complex tions require different approaches. For example, bar-coded
systems, outputs from one process often become inputs to medication management is one strategy for reducing med-
many others. A new admission on a nursing unit (input) ication errors. However, to be effective, this strategy must
can generate orders (output) to pharmacy, radiology, respi- be implemented as part of an overall strategy for reducing
ratory care, as well as many other departments. As vari- medical errors. This strategy might be adopting concepts
ous levels of care (clinic, hospital, long-term care) become and principles from high-reliability organizations (HRO).
increasingly connected, interoperability of computer sys- A high-reliability organization assumes that accidents (or
tems becomes crucial. In addition, as clinical and admin- errors) will occur if multiple failures interact within tightly
istrative areas become more specialized, it is essential that coupled, complex systems. Much of the research regard-
domain experts participate in the planning process. ing HRO has been fueled by well-known catastrophes
such as the Three Mile Island nuclear incident, the space
Wicked Problems shuttle Challenger explosion, and the Cuban Missile Crisis
(Weick & Sutcliffe, 2001).
Problems encountered in complex systems are commonly
described as “wicked.” A wicked problem is difficult or
impossible to solve because of incomplete, contradictory, High-Reliability Organizations
and changing requirements that are often difficult to rec- Key attributes of HROs include a flexible organizational
ognize (Rittel, Horst, & Melvil, 1973). Wicked problem are: structure, an emphasis on reliability rather than efficiency,
• Very difficult to define or formulate aligning rewards with appropriate behavior, a perception
that risk is always present, sensemaking (an understand-
• Not described as true or false, but as better or worse
ing of what is happening around you), heedfulness (an
• Have an enumerable or exhaustive set of alternative mutual understanding of roles), redundancy (ensuring
solutions there is sufficient flex in the system), mitigating decisions
• Inaccessible to trial and error; solutions are a “one (decision-making that migrates to experts), and formal
shot” deal rules and procedures that are explicit. Employing infor-
mation technology, such as bar-coded medication admin-
• Unique and often a symptom of another problem
istration, must be implemented in the context of an overall
Provider order entry (POE) systems often exhibit strategy to become an HRO. For example, BCMA empha-
wicked problem behavior because stakeholders have sizes reliability over efficiency and integrates formal rules
incomplete, contradictory, and changing requirements. and procedures into the medication management process.
For example, although POE systems have been shown to Collectively, these features reduce the probability of mul-
reduce cycle time for the entire medication management tiple failures converging simultaneously.
process on a global level, individual providers may spend On the other hand, planning for clinical decision sup-
more time locally, having to enter orders via a keyboard. port (CDS) applications may require a different approach
Here global and local benefits contradict each other. In than implementation of BCMA. CDS systems link health
addition, commercial vendors cannot allow “test driving” observations with health knowledge to influence health
their application before purchase (trial and error). Because choices by clinicians for improved healthcare (Garg et al.,
of the enormous expense associated in implementation 2005). These applications cover a broad range of systems
of POE applications, once a system has been installed it from simple allergy alerts to sophisticated algorithms for
is very difficult and costly to remove it (a one shot deal). diagnosing disease conditions. The cognitive sciences
Finally, because practice patterns vary, each provider inform and shape the design, development, and assess-
wants the system customized to their workflow (each ment of information systems and CDS technology. The
problem is unique). But because there are many providers sub-field of medical cognition focuses on understand-
there are an enumerable number of alternative solutions. ing the knowledge structures and mental processes of

c37.indd 530 20-02-2015 18:52:18

 Chapter 37 • Planning, Design, and Implementation of Information Technology in Complex Healthcare Systems    531

Downloaded by [ Faculty of Nursing, Chiangmai University 5.62.158.117] at [07/18/16]. Copyright © McGraw-Hill Global Education Holdings, LLC. Not to be redistributed or modified in any way without permission.
clinicians during such activities as decision-making and   TABLE 37.3    System Design and Life Cycle
problem solving (Patel, Arocha, & Kaufman, 2001).
Optimizing the capabilities of CDS systems to allow Step Description
better decision-making requires an understanding of the
1 Identify problems, opportunities, and objectives
structural and processing patterns in human information
processing. For example, knowledge can be described as 2 Determine human information requirements
conceptual or procedural. Conceptual knowledge refers to 3 Analyze system needs
a clinicians’ understanding of specific concepts within a 4 Design the recommended system
domain while procedural knowledge is the “how to” of an 5 Develop and document software
activity. Conceptual knowledge is learned through mind- 6 Test and maintain the system
ful engagement while procedural knowledge is developed
7 Implement and evaluate the system.
through deliberate practice. If CDS planners are not care-
ful, they may inadvertently design a system that transforms
a routine task such as checking a lab value (procedural
knowledge) into a cumbersome series of computer entries. method used is the System Design Life Cycle (SDLC). The
If the clinician is simultaneously processing conceptual SDLC acts as a framework for both software development,
knowledge (problem solving and decision-making) and and implementation and testing of the system. Table 37.3
complex procedural tasks, it will place an unnecessary presents the broad steps in the SDLC (Kendall & Kendall,
burden on working memory and create frustration. 2006)
In summary, the frequent occurrence of wicked prob- An SDLC approach prescribes the entire design, test-
lems in complex healthcare systems highlights the chal- ing, and implementation of new software applications as
lenges faced by information system planners today. one project with multiple sub-projects (see Table 37.3).
Successful planning for the introduction of information Project deadlines can extend over many months and in
technology requires participation by a diverse group of some cases years. The method encourages the use of stan-
stakeholders and experts. There is no “cookie cutter” solu- dardization (for example, programming tools, software
tion in system planning. Each application must be aligned languages, data dictionaries, data flow diagrams, and so
with an overall organizational strategy which drives the forth). Extensive data gathering (interviews, question-
implementation approach. naires, observations, flowcharting) occurs before the start
of the project in an effort to predict overall system behav-
ior, after full implementation of the application.
INFORMATION SYSTEM DESIGN
The characteristic behavior of complex systems and the Agile Design
emergence of wicked problems have prompted system
planners to search for new methods of system design and As healthcare systems have become increasingly com-
implementation. The introduction of new information plex, the SDLC approach has come under fire as being
technology in healthcare organizations involves the inte- overly rigid. Behavior, characteristic of complex sys-
gration of both new applications and incumbent legacy tems (sensitivity to initial conditions, non-linearity, and
systems. For example, a new POE application will need to wicked problems), is often unpredictable, especially after
interface with existing pharmacy, radiology, and lab sys- the introduction of information technology in clinical
tems. Although, an initial “starter set of orders” usually workflow. Equally challenging is the difficulty of trialing
accompanies the POE application, the task of interfacing the impact of new information systems before having to
it with appropriate departmental systems, creating new actually purchase them. Site visits to observe a successful
order sets, and developing CDS generally falls to an inter- application in one facility are no guarantee of success in
nal implementation team with vendor support. Over time, another. Many healthcare organizations have spent count-
this development team will design and build a POE system less millions in failed system implementations.
that is much different than the original application. To overcome these problems, healthcare organizations
are turning to “agile” methods for design and implemen-
tation of information systems (Kendall & Kendall, 2006).
Structured Design
Agile design is less prescriptive than structured meth-
Historically design and implementation of healthcare ods and allows for frequent trial and error. Rather than
information systems relied heavily on structured methods. mapping the entire project plan up front, agile methods
The most common structured design and implementation clearly define future milestones but focus on short-term

c37.indd 531 20-02-2015 18:52:18

 532    P art 6 • N ursing I nformatics —C omplex A pplications

Downloaded by [ Faculty of Nursing, Chiangmai University 5.62.158.117] at [07/18/16]. Copyright © McGraw-Hill Global Education Holdings, LLC. Not to be redistributed or modified in any way without permission.
successes. To do so the implementation team and soft- solutions evolve into a highly integrated, complex informa-
ware developers collaboratively evaluate and prioritize the tion system. Ironically, this is the same process that many
sequence of task to implement first. Priority is assigned to natural systems have used to evolve into their current state.
tasks that can be accomplished within a short time frame,
with a minimum of cost while maintaining high quality
standards. This “timeboxing” of projects forces the team
THE SYSTEMS ANALYST TOOLBOX
to search for simple, but elegant solutions. Developers Whether structured or agile design methods are used dur-
often work in pairs to cross-check each other’s work and ing project implementation, systems analysts rely heavily
distribute the workload. Communication is free flowing on various forms of modeling software. Modeling applica-
between developers and the implementation team. Once tions visually display the interaction and flow of entities
a solution is developed it is rapidly tested in the field and (patients, providers, information) before and after imple-
then continuously modified and improved. There is a phi- mentation of information technology. Models link data
losophy that the interval between testing and feedback be dictionaries with clinical and administrative workflow
as close as possible to sustain momentum in the project. through logical and physical data flow diagrams. Models
Agile design is one example of how to manage informa- can be connected through a hierarchy of parent and child
tion technology projects in complex systems. Rather than diagrams to analyze system behavior locally (at the user
trying to progress along a rigid project schedule, simple level) and globally (management reports). There are many
elegant solutions are rapidly designed, coded, and tested on commercial modeling tools available. Table 37.4 presents
a continuous basis. Over time, layer upon layer of elegant various types of modeling software and how they are used.

  TABLE 37.4    The Systems Analyst Toolbox

Simulation
Method Description Healthcare Example

Flowcharts Flowcharts represent one of the most basic tools used by sys- Prior to the installation of a new clinical
tem analysts to describe the flow of entities (information, documentation system, the current
patients, and providers) in a process. Flowcharting soft- paper documentation process was flow
ware can be found as a stand-alone commercial product or charted and then compared to the new
as a feature in word processing, spreadsheet, and project automated process using standard
management software. ­flowchart symbols.
CASE Tools Computer-Aided Software Engineering is a suite of software CASE tools could be used to develop and
applications used for systems design and analysis. CASE implement a package of software appli-
tools improve communication, integrate life cycle activi- cation used in the development of an
ties, and are used extensively in the design and implemen- electronic medical record.
tation of new applications.
Discrete Event Discrete event simulation (DES) utilizes mathematical formu- Analyzing cycle time, patient flow, bottle-
Simulation las to show how model inputs change as a process evolves necks, and non-value-added activities
over time. Typically the model is built visually using stan- before and after installation of a com-
dard flowchart symbols. For example, rectangles may rep- puterized clinical documentation sys-
resent processes and diamonds, decision points. tem in an emergency department.
Simulation can be easily incorporated
into popular methods of performance
improvement such as Six Sigma and Lean.
Network Analysis The unit of measure in network analysis is the pattern of Network analysis tools could be used to
relationships that exist between entities and the informa- quantify access to computer help desks
tion that flows between them. Network analysis combines for providers by showing how central-
theories from sociology and information science (network ized or decentralized the distribution of
theory). The field characterizes entities as “nodes” and the Centers are for organizations within a
relationship or link between them as “ties.” The pattern and health system.
strength of ties between nodes is then plotted on a graph
where the flow of information can be analyzed.

c37.indd 532 20-02-2015 18:52:18

 Chapter 37 • Planning, Design, and Implementation of Information Technology in Complex Healthcare Systems    533

Downloaded by [ Faculty of Nursing, Chiangmai University 5.62.158.117] at [07/18/16]. Copyright © McGraw-Hill Global Education Holdings, LLC. Not to be redistributed or modified in any way without permission.
Discrete Event Simulation across multiple inter-dependent hospital departments,
has a much greater impact on a large hospital than on a
Two software applications that are well suited for the analy- smaller one. It takes significantly more resources to inves-
sis of complex systems are discrete event simulation (DSA) tigate and make changes in larger hospitals than smaller
and network analysis (NA). DSA is a software application ones because of the number of people, departments, types
that allows analysts to flow chart processes on a computer of equipment, and so forth that are involved. Thus increas-
and then simulate entities (people, patients, information) ing size (or complexity) promotes “fragile” behavior in
as they move through individual steps. Simulation allows health systems: they break easily.
analyst to quickly communicate the flow of entities within One can get a sense of the impact of size and com-
a process and compare differences in workflow after the plexity on the fragility of organizations by scanning The
introduction of new technology. DSA applications con- Centers for Medicare and Medicaid’s Hospitals Consumer
tain statistical packages that allow analysts to fit empiri- Assessment of Healthcare Providers (HCAHPS) sur-
cal data (process times, inter-arrival rates) to theoretical vey scores (HCAHPS Online). The HCAHPS provide
probability distributions to create life-like models of real a standardized survey instrument and data collection
systems. Pre- and post-implementation models can be methodology for measuring patients' perspectives on
compared for differences in cycle time, queuing, resource hospital care. The HCAHPS is administered to a random
consumption, cost, and complexity. DSA is ideal for agile sample of patients continuously throughout the year in
design projects because processes can be quickly modeled hospitals. From “Communication with Nurses” to “Pain
and analyzed prior to testing. Unanticipated bottlenecks, Management,” of the 10 hospital characteristics publically
design problems, and bugs can be solved ahead of time to reported, scores decrease with an increase in hospital bed
expedite the agile design process. size in nearly every category. Just as an elephant falling five
feet suffers significantly more damage than a mouse, large,
Network Analysis complex organizations suffer exponentially greater harm
than smaller ones for similar events.
Network analysis applications plot the pattern of relation- In contrast, certain systems are “anti-fragile” to nega-
ships that exist between entities and the information that tive events and actually become stronger as they are
flows between them. Entities are represented as network stressed. A good example, among many, is the human
nodes and can identify people or things (computers on body’s immune system, which creates new antibodies
nursing units, handheld devices, servers). The informa- when it is exposed to antigens. The concepts illustrated
tion that flows between nodes is represented as a tie and from these biologic “learning systems” can be translated
the resulting network pattern can provide analysts with into man-made systems through continuous learning
insights into how data, information, and knowledge move organizations. By learning from stressors (unanticipated
throughout the organization. Network analysis tools mea- events) and then implementing and revising policies and
sure information centrality, density, speed, and connected- procedures to prevent them from occurring again the
ness and can provide an overall method for measuring the health system actually becomes stronger. To illustrate one
accessibility of information to providers. NA graphs can example of the concept of fragility the following actual use
uncover power laws (see Table 37.2) in the distribution of case is described.
hubs in the organizations network of computers and serv-
ers. This can be beneficial in investigating the robustness
of the computer network in the event a key hub crashes. DESIGNING SYSTEMS WITHIN
AN INCREASING COMPLEX
ORGANIZATIONAL FRAGILITY WORLD—A CASE SCENARIO
The potential impact of growing healthcare complexity In 2005, a large health system located in the Southwest
is enormous. Beyond a certain level, organizational com- initiated a plan to implement a single vendor, enterprise-
plexity can decrease both the quality and financial perfor- wide EHR across its 10 hospitals. The system contained
mance of a health system. In his book, Antifragile, Taleb one large academic medical center of 800 beds and 9 hos-
describes the concept of fragility in complex systems such pitals ranging in size from 25 to 475 beds. These hospitals
as healthcare (Taleb, 2012). As certain systems become were distributed throughout the state where 50% were in
increasingly complex, unexpected events can create an large urban areas and the remainder in rural locations.
exponentially negative impact. For example, the death of Although the network of hospitals acted as an integrated
a patient, as the result of a medication error propagating delivery system, each implemented their EHR according

c37.indd 533 20-02-2015 18:52:18

 534    P art 6 • N ursing I nformatics —C omplex A pplications

Downloaded by [ Faculty of Nursing, Chiangmai University 5.62.158.117] at [07/18/16]. Copyright © McGraw-Hill Global Education Holdings, LLC. Not to be redistributed or modified in any way without permission.
  TABLE 37.5    EHR Performance Measures then work with commercial vendors to create flowcharts
that describe their existing workflows and revise them to
• User satisfaction (survey tool) align with the EHR’s new workflow.
• Staffing utilization and overtime (staffing reports) There are a number of disadvantages of the SDLC that
• Percent utilization of EHR by providers (management can lead to an increase in organizational fragility in com-
reports)
plex systems such as healthcare. These include:
• Provider productivity (patient visits per day)
• Pharmacy to provider callbacks (pharmacy system reports)
• Number and severity of medication errors (risk manage-
• It is virtually impossible to anticipate all of the dif-
ferent scenarios that may arise in advance.
ment reports)
• Evaluation of documentation quality—legibility, com- • User needs and technology may change during the
pleteness, appropriateness implementation phase especially if the timeline is
• Compliance with clinical decision support reminders and delayed.
alerts
• Volume of lab or imaging procedures ordered (lab and
• Workflows are often designed in silos (nursing,
pharmacy, radiology) and do not communicate well
radiology system reports)
• Chart pulls (medical record departmental reports)
when integrated in an overall workflow.
• Release of information (ROI) response time and backlog
Of the nine hospitals that used the SDLC many of
• Billing turnaround times (revenue cycle reports)
• Claim rejections
these issues emerged during implementation of the EHR.
• CPT E&M code levels For example, because of its size and complexity, the 800-
• Documentation of services (CPT codes) bed medical center had a protracted EHR implementation
• Patient satisfaction that spanned three years. Because the entire system was
fully designed in advance, it was difficult to accommodate
changes in technology and new service requirements that
occurred during the three-year implementation. The sys-
to their own process and timeline. Thus the success of tem could not readily adapt to the emergence of wireless
various EHR implementation strategies used by each hos- mobile devices (tablets, phones, and so forth), consumer
pital could be assessed by measuring specific performance informatics (patient portals), and Web-based services.
outcomes post-implementation (see Table 37.5) Although clinical documentation was now electronically
Complete implementation of the EHR occurred over stored it was not standardized in a way that accommodated
a three-year period with one hospital at a time going reporting requirements for the Accountable Care Act and
“live” according to a master schedule. One year post- “meaningful use.” And because workflow design had been
implementation each hospital then compared pre- and developed through individual departments (nursing, radi-
post-performance metrics to assess how well strategic ology, pharmacy, and so forth) communication was slow
objectives were met. and cumbersome. Provider order entry systems, clinical
Result of the one-year post-study varied considerably documentation, and medication administration all took
between hospitals and it is beyond the scope of this chap- longer than previous paper systems.
ter to review all of the results. However, it is important The aforementioned use case is an example of a frag-
to note that one hospital significantly outperformed the ile organizational design. When the system was stressed,
remaining nine hospitals on nearly all performance met- such as in a sudden increase in admissions, it quickly
rics. This high-performing, community hospital had a bed became overwhelmed and “broke.” Medical errors, system
size of 475 and utilized an “agile” design strategy as com- downtime, resources, and cost all exponentially increased
pared to the other hospitals that followed a more tradi- rather than decreased in response to unanticipated events.
tional systems development lifecycle (SDLC). In contrast to other hospitals, the high-performing
As previously described, traditional EHR implemen- hospital utilized an agile design process when imple-
tation strategies have followed the SDLC (see Table 37.3) menting their EHR. One team of clinicians, vendor
using a structured design process. The SDLC typically consultants, administrators, and hospital programmers
maps out the complete planning, design, and implemen- developed a project timeline that “loosely” guided the
tation of projects in advance. This prescriptive process system over a one-year period. This allowed flexibility for
attempts to anticipate all possible future scenarios and unanticipated events that occurred along the way and did
then designs and build the system to accommodate them. not tie the team down to a ridged plan. Rather than imple-
Nursing, pharmacy, radiology, and other clinical areas menting the EHR on all units simultaneously, the same

c37.indd 534 20-02-2015 18:52:18

 Chapter 37 • Planning, Design, and Implementation of Information Technology in Complex Healthcare Systems    535

Downloaded by [ Faculty of Nursing, Chiangmai University 5.62.158.117] at [07/18/16]. Copyright © McGraw-Hill Global Education Holdings, LLC. Not to be redistributed or modified in any way without permission.
team went from unit to unit and designed the system the potential combinations of them grow exponentially.
along the way. The more potential combinations of technologies there
The team designed the system in a modular format are (combinatorial complexity), the higher the probability
rather than building the entire system in advance. The of discovering new uses for them. Thus, technology actu-
modular format allowed system designers to break work- ally creates itself out of itself (Arthur, 2009).
flow up into natural use cases. For example, one nurs- The relentless march of technology creates both
ing module might be “New Patient Assessment” while benefits and challenges for nurse informaticists. As the
another might be “Patient Discharge.” By using the modu- sheer number of technologies grows, the combinations
lar format, designers could make changes to one part of of them become ever more complex. Simply look at the
the EHR without having to go back and make multiple complexity of EHR systems in hospitals today. These
changes to other areas. systems are a mixture of new and legacy systems con-
As the design team went from one department to the nected through a complex network of customized inter-
next it used a rapid design and implementation strategy. faces. Equally complex is the acceleration in how quickly
The team would meet with key departmental content the underlying processes these technologies execute
experts, review the existing workflow, and then quickly changes. New knowledge supporting evidence-based
design a new workflow. The new workflow would be rap- practices is growing so fast that new CDS algorithms
idly piloted on the unit over a one- to two-week period programmed into today’s EHRs can become outdated in
while continuously making changes based on feedback. a matter of weeks.
This process allowed the team to build upon the success of Nurse informaticists play a vital role in the success-
previous units and make recommendations on best prac- ful planning, design, and implementation of information
tices. In other words, the team was continuously learning technology. However, to achieve that success, nurse infor-
from their mistakes and correcting them as they advanced matics must have an in-depth knowledge of complex sys-
through the organization. tems and strategies to manage its behavior.
The agile design process used by this hospital is an
example of an “anti-fragile” organization. Stressors or
unanticipated events actually strengthened the system as
the design evolved. That is because an organization that REFERENCES
uses feedback to continuously learn, just as in biologic sys- Arthur, B. (2009). The nature of technology: What it is and
tems, adapts more readily to change and becomes stron- how it evolves. New York, NY: Simon and Schuster.
ger. In high-performing organizations, the design phase Bejan, A., & Zane, J. P. (2012). Design in nature: How the
never actually ends. It is continuous and always adapting constructal law governs evolution in biology, physics,
and learning from unanticipated events. technology, and social organization. New York, NY:
Doubleday.
Clancy, T. R. (2010). Technology and complexity: Trouble
brewing? Journal of Nursing Administration, 40(6),
SUMMARY AND CONCLUSIONS 247–249.
Clancy, T. R., & Delaney, C. (2005). Complex nursing
Healthcare complexity is growing at an exponential rate
­systems. Journal of Nursing Management, 13(3),
(Clancy, 2010). This, in part, is because new technology 192–201.
itself is composed of combinations of existing technolo- Clancy, T. R., Effken, J., & Persut, D. (2008). Applications of
gies. For example, POE systems interface with existing complex systems theory in nursing education, research
applications (technologies) in pharmacy, lab, and radiol- and practice. Nursing Outlook, 56(5), 248–256, e3.
ogy. Each individual departmental application is composed Garg, A. X., Adhikari, N. K., McDonald, H., Rosas-Arellano,
of further sub-components that are themselves technolo- M.P., Devereaux, P. J., Beyene, J., ... Haynes, R. B. (2005).
gies. This recursive process continues until it reaches the Effects of computerized clinical decision support sys-
most basic parts of the system. If each component of this tems on practitioner performance and patient outcomes:
hierarchical tree is considered a technology, then new, A systematic review. Journal of the American Medical
Association, 293(10), 1223–1238.
novel technologies form from new combinations. For
Kendall, K. E., & Kendall, J. E. (2006). Systems analysis and
example, voice-activated clinical documentation recently design (6th ed.). Upper Saddle River, NJ: Pearson Prentice
emerged by combining Voice over Intranet (VoIP) tech- Hall.
nology from other industries with existing EHR systems. Patel, V. L., Arocha, J. F., & Kaufman, D. R. (2001). A primer
Thus as we see new technologies grow at a linearly rate, on aspects of cognition for medical informatics. Journal

c37.indd 535 20-02-2015 18:52:18