Clancy, Thomas R. MA, MBA, RN; Delaney, Connie White PhD, RN, FAAN; Morrison, Bernice MSN, RN; Gunn, Jody K. MSN, RN
In its 2004 report, "Keeping Patients Safe: Transforming the Work Environment of Nurses," the Institute of Medicine cited that documentation and paperwork impose a heavy demand on nurses' time spent in hospitals.1 Studies indicate that nurses spend an estimated 13% to 28% of total shift time documenting both administrative and patient care activities.2-5 Documentation performed by nurses includes both external regulatory requirements, such as advanced directives and consent forms, and clinical forms including nurses' notes, care plans, assessment forms, education, and so forth. In addition to being time consuming, nurses claim that documentation is often redundant and irrelevant. There seems to be consensus among hospital nurses that paperwork is a major source of frustration and takes valuable time away from patient care.6,7
Not only is excessive documentation frustrating to nurses, it is also expensive. If the average annual salary and benefits of a registered nurse is $49,171,8 then a conservative estimate of annual documentation costs per nurse ranges from $6,393 to $13,768 (13-28% of $49,171). Multiply these labor costs by the total full-time nursing equivalents in a midsize community hospital, and overall documentation expenses will likely be in millions. Clearly, streamlining documentation workflow to eliminate redundancy and waste should take center stage in transforming the work environment of nurses and produce cost savings.
Hospitals and Complex Adaptive Systems
The work environment surrounding hospital-based nurses is embedded in a network of numerous constituents connected through a rich social network collectively known as complex adaptive systems (CASs).9 Complex adaptive system is characterized by "nonlinear interactive components, emergent phenomena, continuous and discontinuous change, and unpredictable outcomes."10 Nonlinear means that cause and effect are disproportionate. A small stimulus may end in amplified system behavior, whereas a large stimulus may have little or no impact. This phenomenon results from information becoming either trapped or amplified in feedback loops that run through multiple hospital departments.11 A familiar example is the apparently minor medication prescription error that propagates throughout the organization and results in serious patient harm.
Complex adaptive systems are ubiquitous and occur in such diverse areas as weather patterns,12-15 cell formation,16,17 animal schooling or flocking behavior,18 economic markets,19 and human social networks.20,21 The constituents in a complex hospital system are represented by endogenous "agents" (patients, nurses, physicians, technicians) and exogenous agents (competitor hospitals, insurance payers, regulatory agencies). Agents within a CAS can be represented by entities such as hospital departments, equipment, or even medical records.
Complexity and CASs are closely related but distinguishable. Complexity is described as "not simple," complicated, involved, and with many component parts. The assembly of equipment, intravenous tubing, and supplies surrounding an intensive care unit patient is visually complex. However, it is the dense web of human to human, machine to human, and machine to machine interactions that creates a CAS. Complexity, in part, drives the formation of CASs.
Complex adaptive systems yearn for order. When unstable, complex systems adapt and transition to a new state through self-organization and emergence.16 An example is the "work-around" in which staff, dissatisfied with a change, subtlety yet collectively detour around a new process. Care planning is one such instance where the complex nature of a process has historically resulted in self-organization and the emergence of work-arounds. If the forms used are cumbersome and the information collected considered irrelevant, care plans are often completed in a manner that, at most, meets the minimum requirements of regulatory agencies. Care planning approached from this perspective is nonvalue added and amplifies nurses' frustration with growing paperwork demands.
Viewed from a CAS perspective, much of the frustration in care planning is the result of combinatorial explosion. The concept of combinatorial explosion arises when a system is presented with a multitude of potential "states."9,11 A state is a condition, situation, or circumstance at a given time. Patients with heart failure admitted through the emergency room may arrive in an acute state. However, once treatment has been provided, the acute state will likely transition to a stable state. In healthcare the combination of potential states a patient may experience is the driving force behind combinatorial explosion. For example, there are 167 North American Nursing Diagnosis Association (NANDA)22 approved diagnoses. The "state space" for all combinations of NANDA diagnosis is calculated by increasing 2 (either the patient has the diagnosis or does not) to the 167th power. Furthermore, each diagnosis is linked to a variety of signs, symptoms, and etiologies. The result is a near infinite number of potential diagnostic combinations to be considered when formulating a care plan. This is combinatorial explosion.
Care planning is further complicated when the combination of nursing interventions and outcomes is considered. For instance, the fourth edition of Nursing Interventions Classification (NIC)23 lists more than 500 nursing interventions that may be linked to NANDA diagnosis. Furthermore, each intervention has multiple nursing care activities associated with it. Add nursing outcomes to the mix and the potential combination of nursing diagnoses and interventions and outcomes becomes infinite. The human capacity to process information, though, is quite limited. Studies have shown that, on average, human short-term memory can hold only 5 to 9 pieces of information simultaneously.23 Thus, the sheer amount of information that must be processed can potentially lead to overload and ineffective decision making.
Care Planning in a CAS
Historically, the nursing profession attempted to deal with care planning and combinatorial explosion through the creation of critical paths and exhaustive lists of discrete nursing care activities or tasks.23 These nonstandardized lists were published in textbooks and used as references during the creation of care plans. Many nursing departments developed their own unit specific, standardized care plans that focused on the most frequently occurring diagnosis, interventions, and outcomes. Subsequently, these large data sets were embedded in computerized clinical documentation systems in an attempt to automate the care-planning process.24
Although comprehensive, care plans that attempt to describe every nuance of care are impractical. Lengthy paper care plans are time consuming and take nurses away from direct patient care. Rather than creating long lists, nurses tend to document a small number or "cluster" of similar interventions for a specific diagnosis.23,25 For example, when validating the NIC, McCloskey-Dochterman and Bulechek23 found that 6 interventions out of 336 were used 75% of the time in pilot organizations. From a complex systems viewpoint, such behavior is not surprising. Because of the enormous state space created by combinatorial explosion, nurses use heuristics created through their own mental models to efficiently capture the most vital interventions needed to achieve a desired outcome.1
Selecting the right intervention, however, is highly dependent on the accuracy of information presented (nurse to nurse report, documented nursing notes, discussions with the patient and family members) and the nurse's unique "mental model" derived from patient assessment cues. Mental models act as constructs to manage information processing in highly complex situations.11,26 They are built on feedback from previous life experiences, cultural norms, education, and other important factors. Mental models allow nurses to efficiently process information by categorizing events and experiences through heuristics or "rules of thumb." These heuristics have been studied extensively by Persut and Fowler in the Outcome-Present-State-Test (OPT) Spellout model of clinical reasoning.26 The OPT Spellout model "unpacks" the essential decision-making strategies used by experienced nurses working in a complex system. These key processes include reflection, framing, cue logic, testing, and judgment to focus on the most essential outcomes.
Two fundamental problems occur with the inflow of information and the use of mental models in complex systems. First, the human capacity to interpret accurately the structure and dynamics of complex systems from the information at hand is grossly overestimated. Because of limitations on attention, memory, and time, most individuals are unable to infer correctly the dynamics of all, but the simplest, causal maps.25 Second, studies have shown that the filtering of information through individual mental models leads to significant misperceptions of system feedback.11 System complexity results in limited information, confounding variables, and ambiguity. The filtering of this information through personal mental models leads to erroneous inferences on system dynamics, judgment errors, bias, and impediments to learning.11
Standardized Nursing Languages
Care planning is an important component in the provision of quality patient care. However, balancing the need to provide a comprehensive plan with the realities of today's fast paced, complex hospital systems is problematic. As a means to address this problem, many nursing departments are turning to standardized nursing languages (SNLs).22,23,27,28 Standardized nursing languages benefit nurses by linking evidenced-based nursing diagnosis with interventions and outcomes and minimizing the bias imposed by individual mental models. An SNL provides a consistent terminology for nurses to describe and document assessments, interventions, and the outcomes of their actions.23 Currently, there are 13 American Nurses Association recognized terminologies that are supported through the Nursing Information & Data Set Evaluation Center.29
The taxonomic structure of an SNL fits naturally into the hierarchical structure of CASs. For example, the Nursing Intervention and Classification System23 is built on a foundation that begins with defining groups of discrete nursing activities that fall into specific interventions (microlevel). These clusters are based on an exhaustive survey of nursing literature, focus groups, and expert opinion.23 In a similar manner, interventions are grouped into broad "classes" at the mesolevel. Classes are then grouped into 7 separate and distinct domains at the macrolevel.
Complex adaptive systems also form a "fuzzy," tiered structure of microlevel, mesolevel, and macrolevel.9 The repeated replication and recombining of individual constituents at lower levels to form new, higher levels is a characteristic of CASs referred to as "self-similarity." Systems characterized by self-similarity have been studied extensively in both the natural and social sciences.30-32 Self-similar systems start with an elegant design at the microlevel and then continually replicate and recombine. A good example is the evolution of social organizations such as hospitals. At its lowest level, the design of a human cell is replicated in the development of proteins and other compounds. This same design is then replicated to build human organs and then recombined to form systems (cardiovascular, respiratory, neurologic) and, eventually, the human body. Individuals are combined to form families, families are combined to form communities, and community members combine to form organizations.
Self-similar systems exist because they work. Through the process of natural selection, the elegant design of constituents at the microlevel survives by replicating themselves at higher, more abstract levels. However, the repeated replication and recombining of one level upon another creates a cascading effect of complexity and combinatorial explosion that surpasses human information processing capabilities.9,11,32 Formulating a care plan requires nurses to consider activities and interventions at the microlevel (patient response to specific interventions), the mesolevel (response to the interaction of broad classes of interventions), and the macrolevel (the overall patient effect considering all health domains). To efficiently sift through this mountain of data, experienced nurses search for patterns or linkages that connect to specific nursing diagnoses.26 These clusters of activities capture the essence of the patient's condition and formulate the basis of the care plan.
The design of SNLs fits naturally into the information processing structure of nurses working within a CAS. The grouping of similar activities into interventions that link to specific nursing diagnosis collapses multiple layers of self-similarity into a manageable number of information processing units. For example, rather than searching for the multiple discrete activities that surround the care of patients in respiratory acidosis, the NIC systems provide 1 intervention (Acid Base Management: Respiratory Acidosis, no. 1913) with a cluster of 20 activities. Other SNLs provide a similar methodology.23
Process Cycle Efficiency and SNLs
The degree of complexity within an organization has been cited as the greatest determinant of nonvalue added cost.33 Complex organizations tend to have many service offerings and processes with multiple steps. In hospitals, product offerings include the number of clinical service lines (orthopedic, cardiovascular, maternal-child, and so forth) or the diagnostic and treatment modalities available (cardiac catheterization, magnetic resonance imaging, computed tomography, etc). The higher the product offerings, the higher the organizational complexity.
Product offerings contribute to increased complexity when each offering requires its own unique design or "platform." The word platform is taken from manufacturing and indicates the layout of equipment on the factory floor used to produce a specific product. Frequent platform changes to accommodate different products increase complexity and cost by increasing nonvalue added steps in a process.33 An analogy to frequent platform changes in manufacturing can be made in hospitals by comparing the difference in care-planning processes from nursing unit to nursing unit. For example, if clinical service lines are viewed as product offerings and each service line has its own unique platform for care planning (different forms, policies, procedures, and so forth), then system complexity is high. On the other hand, if a common platform is used for all service line care plans, then complexity is greatly simplified. This is how SNLs reduce complexity and lower nursing labor costs.
By designing one common platform that accommodates multiple, different nursing diagnoses, interventions, and outcomes, SNLs provide an efficient and meaningful tool for care planning. The percent improvement in efficiency by moving to an SNL can be measured by calculating "process cycle efficiency." Again, taking a lesson from manufacturing;
Equation (Uncited)Image Tools
In manufacturing, lead-time is defined as how long it takes for any item of work to be completed. By substituting "steps" for "time," the process efficiency equation can be modified in the following manner to represent the decision-making efficiency in care planning;
Equation (Uncited)Image Tools
In the care-planning process, "total process" steps indicate the number of discrete units of information that must be processed to formulate a care plan. Total "value added" represents those steps that most directly impact the decision on which diagnosis, interventions, and outcomes are to be used. A process cycle efficiency of less than 10% indicates significant waste. Most service organizations such as hospitals have an overall process cycle efficiency of 5% to 20%.33 If used appropriately, hospitals that migrate toward an SNL should see significant improvements in the process efficiency of care planning (see case history).
In summary, the information processing capabilities of nurses working in hospitals is limited because of the cascading effect of complexity and combinatorial explosion. Tools that collapse layers of complexity into clusters of similar units (activities, interventions, classes, domains, and so forth) support nurses' own heuristics for information processing in CASs. Standardized nursing languages fit naturally into the architecture of CASs because their taxonomic structure closely mirrors the levels of complexity in hospitals. By providing a common platform, SNLs allow nurses to formulate evidence-based care plans for multiple service offerings in a highly efficient manner. Further improvements in care-planning efficiency can be anticipated with the inclusion of standardized languages into the Systemized Nomenclature of Medicine Clinical Terms, which are now being integrated into many electronic health record systems.29 The search and storage capabilities of electronic health record systems should allow electronic linkages between assessment cues, nursing diagnosis, interventions, and outcomes and expedite the care-planning process.
The following actual case history in a midsized community hospital illustrates the benefits of using an SNL.
Mercy Hospital is a 234-bed community hospital located in Iowa City, Iowa. As part of its ongoing performance improvement plan, a hospital-wide audit of how well nurses complied with documentation standards was conducted in 2001. The audit set an initial threshold of 90% compliance with documentation requirements such as completion of nursing assessments, patient health histories, advance directives, and care plans. Results of the audit indicated that a number of nursing units fell below the threshold on the completion of patient care plans. Interviews with nursing staff indicated that the current care-planning forms were cumbersome, time consuming, and redundant. There was consensus that care planning was an external requirement that added little value in the care of the patient. Units that failed to meet the minimum threshold for the completion of care plans were thus challenged to find solutions to the problem.
One such unit, Fifth Floor Telemetry, convened a performance improvement team to study the problem and search for alternative solutions. Fifth Floor Telemetry provides nursing care to patients who require cardiac monitoring after an acute event. The patient population consists of postoperative coronary artery bypass, postangioplasty, recent myocardial infarction, and a number of other cardiac-related diagnoses. This 34-bed nursing unit has a high-patient acuity level with frequent admissions, transfers, and discharges. The work is demanding, and like other areas of the hospital, nurses found themselves under considerable time pressure.
Clarifying and Understanding the Problem
The 2 most frequently occurring medical diagnoses on Fifth Floor Telemetry had standard critical paths with integrated care plans. Compliance with the hospital-wide audit seemed within an acceptable range for these plans. However, the alternative care plan consisted of an 11-page document that contained 3 separate columns listing diagnoses (NANDA), outcomes, and interventions, respectively. As compliance thresholds on the critical path were within acceptable limits, the team decided to concentrate on improving the 11-page form first.
Although interventions were listed on the document, closer inspection revealed that these statements met NIC's criteria for "activities." Nursing Interventions Classification defines activities as discrete actions that assist patient status or behavior toward a desired goal.23 The sets of activities make up interventions. The team realized that by using NIC interventions, these activities could be collapsed into broad evidence-based interventions to reduce complexity. A hospital-wide common care-plan platform existed. However, the diagnoses, interventions, and outcomes varied by nursing unit. The Fifth Floor Telemetry care plan contained 9 NANDA diagnoses, 62 activities, and 51 outcomes (desired goals). Furthermore, most diagnoses, activities, and outcomes were decomposed into numerous check boxes containing discrete information. A similar situation existed on other nursing units.
The nurse manager on Fifth Floor Telemetry and the hospital's nurse informaticist analyzed a sample of care plans selected from each of the various nursing units. They discovered that although there were numerous diagnoses, interventions, and outcomes listed on unit specific care plans, most nurses documented only a small cluster of these items. For example, the average number of diagnoses, interventions, and outcomes selected were 2, 9, and 6, respectively. Using the formula previously described, the decision-making efficiency of the current care-planning processes (Table 1) can be calculated in the following manner.
Value added steps include the actual diagnoses, activities, and outcomes that nurses ultimately documented on the care-plan form. Total steps in the process include all diagnoses, interventions, and outcomes listed on the care plan. Individual units of information listed in check boxes under separate diagnoses, activities, and outcomes were also counted and considered as separate steps.
The total information processing steps (of the entire 11-page form) required by nurses to select diagnoses, interventions, and outcomes, respectively were 170 + 67 + 140 = 377 steps. The average number of steps/diagnosis was 170/11 = 15; steps/outcome, 67/51 = 1.31; and steps/intervention, 140/62 = 2.3. Because nurses chose an average of only 2 diagnosis, 6 interventions, and 9 outcomes, the value added steps were calculated as 2(15) + 6(1.31) + 9(2.3) = 59 steps. Thus decision-making efficiency was 59/377 = 16%.
Assuming that the average numbers of diagnoses, outcomes, and interventions were 2, 6, and 9 respectively, the decision-making efficiency of the care-plan form on Fifth Floor Telemetry was only 16%. Stated differently, the 84% of the time nurses used to formulate care plans was wasted in processing 11 pages of information, of which only 16% was documented on. The nurse manager and nurse informaticist both agreed that a more efficient and effective care-planning process was needed to improve compliance on the unit.
In reviewing the form, it was noted that diagnoses, interventions, and outcomes had originally been decomposed into 377 steps. The historic rationale behind this was to provide an exhaustive list of the most frequently occurring nursing diagnoses and then link them with appropriate interventions and outcomes. In this way, nurses would not have to write lengthy, nonstandardized narrative notes in the chart. Although this strategy eliminated narrative care plans, capturing this enormous "state space" created an 11-page form. If the many information processing steps could be combined into evidence-based clusters, the complexity of the care-planning process would be greatly simplified.
Selecting a Solution
The challenges of balancing the need for an effective, yet efficient, care-planning process in complex systems such as hospitals are not new to nursing. Both the manager for Fifth Floor Telemetry and the nurse informaticist recognized that the solution lay, in part, in the extensive research in the development of SNL conducted over the last 25 years.22,23,27,28 After reviewing the available SNLs,29 a decision was made to develop a pilot proposal to link NANDA diagnoses with the Nursing Interventions (NIC) and Nursing Outcomes (NOC) Classification system (NNN) currently under development at the Center for Nursing Classification and Clinical Effectiveness located at the University of Iowa College of Nursing.23 A proposal was drafted and approved by the Vice President of Nursing to pilot integrating the NNN SNL into the care-planning process on Fifth Floor Telemetry.
Implementing the Solution
The Fifth Floor Telemetry team reconvened and, through a series of meetings, developed a timeline for the implementation of the NNN. Early steps in the timeline focused on providing both unit and hospital-wide nursing staff with education on SNLs. This was important because leading professional organizations such as the American Nurses Association and the Association of Critical Care Nurses have brought the adoption of SNLs to the forefront of nursing practice.29 The staff development coordinator of Fifth Floor Telemetry provided most of this education through staff meetings and committee presentations. A key decision point hinged on whether the hospital would be purchasing an automated clinical documentation system within the next year. If so, the team would need to focus on building the SNL into the computerized workflow of the application. If not, the team would focus on creating a reasonable paper system to bridge the gap between the present and future date of system installation.
Although the information system plan for the hospital supported an automated clinical documentation system, the estimated date of implementation was not scheduled for 2 years. Thus, for the short term, the team was challenged with developing a paper system that would meet the objectives of the project. At first glance, the enormity of the task at hand was overwhelming. There are 167 NANDA diagnoses, 330 NOC outcomes, 514 interventions, and more than 12,000 activities. The potential state space for all combinations of entities would quickly result in combinatorial explosion and intractable complexity. Fortunately, a solution to the problem lay in the hidden dynamics of CASs.
A common behavior exhibited by CASs follows a "power law" distribution.34,35 A power law is a mathematical distribution graphically characterized as a declining relationship between the occurrence and frequency of an event. Often referred to as "Pareto's Law" or the "80/20" rule, power law behavior results from the many cause and effect relationships and balancing feedback loops within a complex system. For example, the distribution of medical diagnoses on Fifth Floor Telemetry shows a declining curve of high frequency (myocardial infarction, postsurgical coronary artery bypass surgery, angina pectoris, and so forth) to low-frequency diagnoses. However, roughly 80% of the cases fall into only 20% of the diagnoses. This is power law behavior.
The remarkable ability of complex systems to adapt and exhibit power law behavior allows organizations to prioritize and focus resources into a manageable number of entities. For instance, using the 80/20 rule, the Fifth Floor Telemetry team was able to limit diagnoses, interventions, and outcomes to those that occurred most frequently (80% of the time) on the unit. Rather than creating care plans with extensive lists, the team developed one form per diagnosis that contained links to commonly used interventions (NIC) and outcomes (NOC). The number of steps required to formulate a care plan was reduced by collapsing the numerous activities and indicators listed on the previous form into NIC interventions and NOC outcomes. In addition, nurses selected only those diagnoses they needed. A total of 27 care plans were developed, which reflected just the high-frequency diagnoses cared for on the unit. For those diagnoses that did not fit within the unit's top 80%, a generic care plan form was developed.
The process for care planning under the new system starts as before, with nurses compiling assessment data to formulate nursing diagnoses. Nurses then select the specific diagnosis-driven care plans (the average number is 2) and indicate on the form which interventions and outcomes are appropriate. As the patient progresses, interventions and outcomes are documented as indicated. If outcomes are not achieved as planned, nurses document the reasons why in the care-plan evaluation notes. Complete copies of the NANDA definitions, NIC interventions, and NOC outcomes are available to nurses on the unit for reference. These references contained the sets of activities that comprised interventions on the paper care plans as well as the complete set of diagnoses, interventions, and outcomes in the NNN data set. References books were used extensively at the beginning of the implementation as nurses became familiar with the NNN language.
Comparing decision-making efficiency between the previous system and the new system shows a 64% improvement assuming an average of 2 diagnoses, 6 outcomes, and 9 interventions per patient (Table 2). The primary reason for the improvement is the collapse of nursing activities and indicators into broad nursing interventions and outcomes, respectively.
In October 2004, a 5-question follow-up survey of 15 (25% of total) nurses on Fifth Floor Telemetry indicated a high level of satisfaction with the new system. Three questions resulted in nearly 100% agreement among those surveyed. Of the 15 nurses, all either agreed or strongly agreed that the new system was easier to use, had an adequate number of choices for diagnoses, and saved time. There was a mixed response on whether the new format improved care-plan updates and the level of understanding on how to choose outcomes and interventions. A survey to determine whether the new system achieves thresholds for JCAHO care planning and documentation standards will be conducted later in 2005.
If the NNN system was implemented house-wide, paper savings are estimated at approximately 9 sheets (11 sheets − 2 sheets) per inpatient admission for a total of 67,500 sheets of paper (7,500 admissions × 9 sheets/admission).
The subtle yet pervasive growth in documentation requirements for nurses can, in part, be attributed to an exponential rise in health system complexity. Multiple service lines, improved access to information, integrated delivery systems, and advances in technology all contribute to an increasingly complicated world. Although relatively new to social organizations such as hospitals, CASs have been around since the beginning of time. Regardless of the field of study, CASs share many common characteristics. Combinatorial explosion, self-similarity, and power law distributions all appear in biological, physical, and social systems. Understanding the nature of CASs is rapidly becoming crucial in the design of effective and efficient hospital processes. Process complexity has become the primary driver of nonvalue added cost in organizations such as hospitals.
Hospital nurses work in a CAS. Care planning is a vital component in the practice of nursing. If improvements in the care-planning process are to be successful, tactics must seek solutions that reduce complexity. An SNL is one such tactic because of its ability to collapse the multiple layers of activities and indicators into evidence-based interventions and outcomes. The integration of standardized languages through Systemized Nomenclature of Medicine Clinical Terms into electronic health records will further improve the efficiency of care planning by automatically searching large data sets for the appropriate combination of nursing diagnoses, activities, interventions, and outcomes. In effect, SNLs are rapidly becoming the key to developing expert nursing systems that can link a series of nursing diagnoses, interventions, and outcomes based on patient assessment cues. Rather than nurses having to search an enormous state space for such information, embedded algorithms within the electronic health record will prune complex data sets to formulate a plan of care. Such systems will never replace nurses' decision-making skills; however, they will greatly expedite the process.
The historic development of SNLs has focused on finding a common tool that represents what nurses do.22,23,27,28 There is also a need today for further research in how SNLs improve efficiency and reduce expenses by simplifying the decision-making process in CASs such as healthcare. Standardized nursing languages drive the formation of data repositories by providing a structure for the storage of comparable information within and across healthcare organizations. The discovery and sharing of new knowledge extracted from these large data sets is vital to managing the rising complexity of today's healthcare organizations.
The authors thank Jan Johnston, MSN, RN; Jill Auderer, MSN, RN; and Sue Grindle, BS, RNC.
© 2006 Lippincott Williams & Wilkins, Inc.