CIN: Computers, Informatics, Nursing • Vol. 22, No. 6, 001–006 • © 2004 Lippincott Williams & Wilkins, Inc.
F E A T U R E A R T I C L E
The N-CODES Project The First Year EILEEN S. O’NEILL, PhD, RN NANCY M. DLUHY, PhD, RN PAUL J. FORTIER, DSc HOWARD MICHEL, PhD
INTRODUCTION
[QA1]
[QA2]
Rapid proliferation of new knowledge, expanding expectations, and rapidly changing and uncertain practice environments demand new approaches to support nurses’ clinical decision making. The Nursing Computer Decision Support (N-CODES) project, a collaborative effort by the Colleges of Nursing and Engineering at the University of Massachusetts, Dartmouth, addresses this need. The purpose of the project is to develop a prototype for a point-of-care system that will make relevant client information available to acute care nurses as they make decisions. The system is being designed to be particularly helpful to novice nurses, assisting them to make more focused assessments, anticipate deleterious client reactions, and initiate appropriate early actions. Nurses have frontline responsibility for managing symptoms, monitoring complications, and anticipating preventative measures. Now, as patient acuity rises and cost-effective treatment becomes mandatory, nurse clinicians must work “faster” and in a “smarter” manner, making complex decisions on an almost continuous basis. Although evidence-based knowledge and standardized guidelines are available in some areas to support clinical decisions, real-time access is a major problem inhibiting their use. The N-CODES project addresses this problem by developing a prototype to deliver clinical knowledge via a handheld computer using wireless access to a central server. The clinical knowledge base will provide guidance in several decision areas such as identifying probable exCIN: Computers, Informatics, Nursing
Clinical decision making is a complex task, and particularly challenging for the novice nurse. Little assistance is available, and decision supports such as standardized guidelines are difficult to access in the hectic flow of practice. The Nursing Computer Decision Support (NCODES) project, directed by investigators for nursing and computer engineering, addresses this problem by developing a prototype of a point-of-care system to deliver clinical knowledge via a handheld computer. This article reports on the progress made during the first year of the project. The Nursing investigators have developed a novice-nurse decision-making model, a comprehensive knowledge development process, and a series of computerized practice maps. The focus of engineering has been on designing the database architecture and the knowledge representation, extraction, and discovery algorithms used to mimic nursing knowledge and clinical decision processes in software. But the success of an interdisciplinary collaborative project depends on establishing tasks and boundaries, clarifying persepectives and language, and developing a productive process. Therefore, along with the progress of each discipline, strategies used to promote collaboration are discussed. KEY WORDS Acute care • Computerized decision support system • Interdisciplinary collaboration • Nurse clinical decision making • Novice nurse
planations given current data, recommending nursing interventions, highlighting possible complications, and outlining pertinent patient teaching. The software will use rule-based reasoning on the basis of individual client data. The technology will provide real-time decision support
From the University of Massachusetts, Dartmouth, MA. This project is funded by the National Science Foundation grant EIA 0218909. Corresponding author: Eileen S. O’Neill, PhD, RN, University of Massachusetts, 285 Old Westport Rd, Dartmouth, MA 02747 (e-mail:
[email protected]). • November/December 2004
1
for the multiple, concurrent cases and sequential decisions characterizing nursing practice. Nurses will be able to consider a full range of alternative explanations, determine additional data needs, or verify the appropriateness of a selected strategy. A fully developed system will have the capacity to maintain a history of decisions and outcomes. Innovations in high-density handheld computers, the sophisticated design of software inference systems, advances in data-based management, and progress in synthesizing and mapping nursing knowledge makes this project timely and feasible. Some aspects of the project are discussed in depth in other papers.1,2 This article presents an overview of the first year of the project. This phase included developing theoretical direction, selecting the clinical problems to be addressed, grappling with evidence issues, constructing practice maps of selected clinical conditions, building the initial architecture, and overriding it all, shaping a cohesive working team.
The team consists of four investigators; two from nursing and two from computer engineering. In addition, each investigator works with a number of research assistants (RAs) who are master’s- and doctorate-level students. During the first year, seven RAs—three engineering students and four nursing students—worked on the project. Several authors3,4 have outlined the factors that prevent different disciplines from working together. These factors include poor communication among different disciplines and lack of understanding of the conceptual basis for practice of each other’s discipline. In the initial N-CODES meetings it became clear to the investigators that we not only lacked understanding of each other’s discipline but also had little shared professional language. This prompted the four faculty investigators to take several steps. First this group decided to meet weekly. Initially these meetings were used to address diverse styles of thinking and different approaches to the proposed project tasks and to resolve language differences. For example, the usage of the terms case and rule differs significantly in nursing and engineering language, and it took us a few meetings to realize this. These meetings have also been essential to configure the fluid nature of nursing practice into practice maps appropriate for coding. During these early meetings the knowledge engineering process of acute sinusitis was developed. Within the initial category of respiratory conditions, acute sinusitis was chosen because it was fairly simple to model compared with other clinical problems. The group worked together on wording simple data rules so they could be coded as logical expressions, developing an engineering design for the rulebased inference engine, designing an architecture to
hold the meta rules, and designing a data structure to represent the client-specific information. The process was implemented in a manner similar to the “spiral” process of software engineering in which nursing rules were proposed and worded jointly, a model was designed, and problems were analyzed by the engineers. This resulted in the development of refined rules and Phase I architecture. The process was repeated many times as nursing knowledge was added to the system. Once the initial architecture, data rules, and meta rules were finalized, they were coded and a prototype decision tree was developed (Figure 1). To promote dialogue and disciplinary understanding between the RAs, five of the RAs—three from engineering and two from nursing—participated in a graduatelevel clinical reasoning course, which is described in another publication.5 The students have studied decision analysis as well as reflective practice. The nursing students provided rich examples of decision making to further their overall understanding of this complex process. The engineering RAs developed slides depicting decision trees, discussed their work in class, and wrote a reflective paper on their learning in a nursing course. It is clear that all RAs in the course gained a greater understanding of each other’s conceptual domain. This was captured in statements from the engineering students, like, “How do nurses keep all that information in their head?” The nursing students initially felt that the engineering students had little to learn from them, as engineers were “so smart.” The nursing students developed a new appreciation of the complexities of their practice when seen through the eyes of another discipline. At the end of the course all students expressed excitement at their personal discoveries and collaborative explorations. All students came away with an understanding of clinical reasoning, decision analysis, and reflective thinking. The engineering students remarked that nurse thinking “came alive” for them in the classroom. All RAs on the project have also begun to develop a common professional vocabulary, which clearly was missing in the beginning. This course will be recommended for all RAs participating in the project in future. To further promote collaboration, the University has supported our efforts by providing a space for the team to work. Here members have a communal workspace without issues of territoriality. Leadership is shared among the coinvestigators as dictated by the focus of the work. Another factor that has led to professional cohesion is the stability of the team.6 This has allowed us the time to understand the problems and work on potential solutions. We have also been able to coauthor descriptions of the work being developed. In addition to the total group’s endeavors, nursing and engineering have made significant individual progress, which has contributed to the overall project.
2
• November/December 2004
BUILDING THE TEAM
CIN: Computers, Informatics, Nursing
[QA7]
FIGURE 1. Architecture.
NURSING ACTIVITIES The nursing team has accomplished three major activities during the first year, the development of a theoretical framework of nurse decision making, the construction of an evidence framework, and the creation of seven practice maps.
Theoretical Framework To construct a clinical decision support system (CDSS) useful in practice, it is necessary to understand how nurses make decisions and how novices develop clinical reasoning skills. While decision frameworks typically address one finite decision point, nursing practice involves an ongoing series of cascading decisions. To systematically examine serial decision events, a theoretical framework of nurse decision making was developed. The framework guides understanding of how nurses represent knowledge implicitly in memory to construct explicit knowledge representations for the CIN: Computers, Informatics, Nursing
computer that fit nurses’ working models.7 The framework consists of two models, a clinical decision-making model that is grounded in information-processing theory and one that resembles previous models,8–10 and a new model of novice-nurse clinical reasoning development. The Novice Clinical Reasoning Model (NCRM) synthesizes theoretical and research thinking in this area (Figure 2). The NCRM proposes a process by which the novice nurse can develop a working knowledge. [QA3] Kennedy11(p193) describes working knowledge as the “organized body of knowledge that is used spontaneously and routinely in the context of one’s work.” Owing to anxiety and knowledge limitations, the novice initially has a limited perception of the situation. Cognitive processing at this stage is deliberate and ruledriven. But over time, with repeated practice experiences, the novice begins to develop a complex system of organized clinical patterns. These patterns form the foundation of working knowledge. The model also includes factors in the practice situation that promote the development of working knowledge, such as the availability of experienced nurses and supportive leadership. • November/December 2004
3
FIGURE 2. Novice Clinical Reasoning Module (NCRM).
Evidence Issues Early in the project the nursing team had to confront the issue of what is the “best evidence.” Most existing support systems use either textbooks or available clinical guidelines as the support base for diagnoses and in[QA4] terventions. But is this approach adequate? When nurses are relying on a system to help them make critical decisions, what evidence should be used to develop and maintain the system? Several other concerns arose such as locating the best evidence to answer clinical questions, deciding what should be done in the gray zones of nursing practice where answers are not apparent, and determining what merit experiential information should have as evidence. In response to these concerns, the nursing team developed a comprehensive framework for nursing knowledge acquisition, validation, and synthesis as well as a knowledge-mapping framework. The framework outlines types of evidence, levels of confidence in different types of evidence, and a tracking grid to determine the value of collective evidence. It also includes the use of clinical experts and a network of acute care nurses to incorporate recent and regional practice patterns. Each data rule—the IF … THEN statements in the program—has an evidence ranking of “strong,” “sufficient,” or “marginal.” For example, in the pulmonary emboli map, the data rule “If high probability PE, then consider adequate hydration” is rated as sufficient because two nursing texts were the only evidence source found to support this rule. The process of accessing knowledge to support the evidence base, and of codifying and evaluating the quality of this knowledge, has 4
CIN: Computers, Informatics, Nursing
led to new insights into evidence-based practice. In addition, the process has exposed gaps in our knowledge, which required creative strategies to bridge. As datarule construction proceeds, the nursing team is continually refining the method.
Practice Maps Decision-support developers often do not select clinical problems that clinicians find most difficult.6 To avoid this mistake, the nursing team examined previous research and discussed with clinicians their most frequent and difficult decisions. On the basis of this data, respiratory problems were chosen as the content focus for the prototype. The nursing team then developed a typology to identify the acute and chronic respiratory problems that needed to be mapped. Over the course of the first year, practice maps for acute care respiratory problems have been developed. To date, the nursing team has created seven practice maps, which contain the data and meta rules for a given clinical condition. Each map represents a collection of IF … THEN data rules that represent the knowledge regarding an acute clinical condition. For example, “If pneumonia, then consider providing adequate caloric intake” is a data rule. Meta rules are strategy or process rules that suggest the best approach for selecting sets of data rules. These rules mimic how the nurse moves from one line of reasoning to another. Meta rules have the same format as data rules but indicate which data rules should be explored first, which second, and which not at all, for example, “if pneumonia, then monitor for • November/December 2004
complications.” This meta rule leads to a set of data rules that outline the complications of pneumonia. During the first year the nursing team approached practice map development using the theoretical framework by listing rules under a number of categories such as risk assessment and interventions. Then each member of the team developed rules under a certain category. Next the rules were laid out on a large board and the team discussed each rule, each category, and the meta rules to move between categories. Currently the team is experimenting with an alternative approach using nursing problems such as nutritional needs as a guide for rule development. Figure 3 depicts a section of the practice map for pneumonia. The nursing team has discovered that knowledge mapping is a laborious process. The bottleneck that develops between knowledge engineers and domain experts in the transfer of knowledge has been reported previously.12,13 In this project, the nursing team is mapping its own clinical knowledge. However, it is clear to us that examining how we actually think about clinical problems is a challenging task. Each practice map has gone through many cycles of building, testing, and rebuilding.
ENGINEERING ACCOMPLISHMENTS Many of the major accomplishments during the first year were based on the ability of the groups as a whole to configure the nursing knowledge needed for the CDSS as previously described. The primary focus of engineering was to design the database architecture and the knowledge representation, extraction, and discovery algorithms used to capture and mimic nursing knowledge and clinical decision processes in software. In engineering terms, the architecture has three major components: data rules, meta rules, and state information. These components equate to knowledge rules, process rules, and the patient chart in nurse terminology. The architecture also includes structure that allows rules to be grouped together to manage the temporal nature of the nursing process. The logical arrangement of these blocks forms a decision tree, through which there are many correct paths. The tree is traversed as follows: The data rules produce or act on patient information. That information is stored in the state. If the state changes, meta rules operate to enable or disable the blocks of data rules. Thus the architecture as well as the
FIGURE 3. A section of the practice map for pneumonia.
CIN: Computers, Informatics, Nursing
• November/December 2004
5
individual data rules and the meta rules captures nursing knowledge. As the architecture was being developed, the engineering group also began to look at ways to access the database wirelessly. The portability from a PC to a wireless device involves altering the implementation of the software and dealing with the limited functionality and constrained memory of the wireless device. The work is ongoing.
Acknowledgments The authors thank the research assistants, Elizabeth Chin, Veronica Coutu, Elizabeth Kelley, Rekha Madiraju, Sun Dip Pranhan, Jessica Ryan, and Beena Sarangarajan, for their contributions to the first year of the project.
REFERENCES
Interdisciplinary collaborative research presents challenges but also provides many benefits. Disciplines can no longer find solutions to the complex problems society faces without collaborative efforts. The pooling of intellectual, financial, and energy resources promotes creativity. Disciplines also learn from one another, making more substantial contributions to healthcare. However, interdisciplinary research requires attention to perspectives, language, and methods within each discipline. As of this writing the N-CODES group is looking forward to testing the architecture of the system with a network of practicing nurses. These acute care nurses with differing levels of experience will help challenge, expand, refine, and validate the current practice maps. They also will assist the nursing team in refining methods to build maps in the future. The logistics of integration will also be a demanding task. How do we ensure that the CDSS fits smoothly into the flow of practice? We anticipate that the nurse network will also help with issues related to usefulness and applicability.
1. O’Neill ES, Dluhy NM, Chin E. A framework of clinical decision making to guide the development of a decision support system for novice nurses. J Adv Nurs. In press. 2. O’Neill ES, Dluhy NM, Fortier P, Michel M. Knowledge acquisition, synthesis, and validation: a model for decision support systems. J Adv Nurs. 2004;47(2):34–142. 3. Schein E. Professional Education. New York, NY: McGraw Hill; 1972. 4. Mariano C. The case for interdisciplinary collaboration. Nurs Outlook. 1989;37(6):285–288. 5. O’Neill ES. Strengthening graduate students’ clinical reasoning. Nurse Educ. 1999;24(2):11–15. 6. Heathfield HA, Wyatt J. Philosophies for the design and development of clinical decision support systems. Methods Inf Med. 1993;32(1):1–8. 7. Hasman A, Safran C, Takeda H. Quality of healthcare: informatics foundations. In: Yearbook of Medical Informatics. 2001:141–152. 8. Charlin B, Tardif J, Boshuizen H. Scripts and medical diagnostic knowledge. Acad Med. 2000;75(2):182–190. 9. Custers E, Regehr G, Norman GR. Mental representations of medical diagnostic knowledge: a review. Acad Med. 1996;71(10, suppl):S55–S61. 10. Elstein AS, Schwarz A. Clinical problem solving and diagnostic decision making: selective review of the cognitive literature. Br Med J. 2002;324(7339):729–732. 11. Kennedy M. Working knowledge. Knowledge Creation Diffusion Util. 1983;5(1):193–211. 12. Musin M, Vander Lei. Knowledge engineering for clinical consultation programs: modeling the application area. Methods Inf Med. 1989;28(1):28–35. 13. Clayton P. Hripcsak G. Decision support in healthcare. Int J Biomed Comput. 1995;39(1):59–66.
6
• November/December 2004
CONCLUSION
CIN: Computers, Informatics, Nursing
[QA5]
[QA6]
AUTHOR QUERIES TITLE: The N-CODES Project: The First Year AUTHOR: Eileen S. O’Neill, PhD, RN, Nancy M. Dluhy, PhD, RN, Paul J. Fortier, DSc ,Howard Michel, PhD QA1. This sentence has been deleted as the information appears in the grant information line under the affiliation. Please check. QA2. Some edits have been made to avoid the “work smarter” construction. Please check if they retain the intended sense. QA3. Please check if the edits made retain the sense intended. QA4. See QA3 QA5. Please update this reference, if possible. QA6. The publisher name and location are missing. Please provide the same. QA7. Is this legend explanatory enough? If not please provide a more detailed one. Also check the seventh line in the second box in the left of the figure.
