Decision Trees - A CMI Tool in Nursing Education.
The use of artificial intelligence and machine learning in the form of decision trees in supporting nursing education is examined and discussed.
As a response to the need for the nurses educated in nursing informatics we started an EU Project financed by Phare Tempus grant called NICE (Nursing Informatics and Computer Aided Education). The aim of the NICE project is:
The development and introduction of a new short cycle degree courses in Nursing Informatics at the university colleges in the partner countries. The project will also produce teaching materials, computer aided tools for nursing education, and establishment of a multimedia teaching laboratory enabling teleconferencing, telemedical applications and distant learning.
The NICE consortium consists of following institutions and countries:
The University College of Nursing, Maribor, Slovenia,
The NICE project is currently in the final year and most of the main outcomes have been already achieved. The curriculum is agreed and will be given into the experimental use shortly. In addition, the drafts of the teaching materials have been produced, as have been plans for the classroom. Educational tools have been developed, and one of them the DecTree will be presented.
Computer managed tools for nursing education Computer Managed Instruction (CMI) has been used in nursing education (Habjani, Kokol, Zorman & Japelj 1998) since the late 1960s (Kohl 1995, Hebda 1988). It is due to the accessibility and self paced format that CMI is very well suited for both students and practicing nurses (Kohl 1995, Hebda 1988), while learning can occur at the learner's own pace and time. In addition, CMI supports also continuing education and distant learning.
The early applications of CMI employed room sized mainframes at large institutions, but the proliferation of microprocessors in the 1990 expanded the depth and breadth of instructional computing (Williamson 1994, Cambre & Castner 1993). Recent studies (Hebda 1988, Jelovsej 1993, Athappilly 1994, Haus 1996) show that students using CMI have better average examination scores, improved ability for critical thinking (Williamson 1994) and enhanced computer literacy, facilitated decision making skills and positively affected achievements (Belfry 1988). In spite of these advantages still many students and faculty (Khoiny 1995) are still reluctant to utilise CMI, but Haus (Haus 1996) reports that the perception and attitude toward computer managed instruction positively changes after actual use of CMI software packages.
Athappilly, Durban and Woods (1994) list three benefits for using CMI and multimedia tools in the educational process:
1) quality multimedia presentation reduces the cost, in spite that initial investments are large the reduction of participants and instructors time are significant;
2) the effectiveness of the teaching and learning is improved because of greater motivation, retention, and mastery of learning;
3) production is improved because of increased satisfaction and enjoyment of learning.
Decision Trees in the Education of Nurses
The majority of the current teaching tools for nursing education are based on the so called concept of Drill and Practice (Conrick 1998), motivated by the research of Skinner (Skinner 1953) The major advantage of such type of learning is the immediate feedback to a student. There is no waiting period for correction and therefore students do not practice their mistakes. But some researchers suggest that after the novelty effect of drill and practice wears off and the motivational power is lost. The wear effect can be overcome if the educational package is adaptive and can be individualised. This can be achieved with the use of artificial intelligence and automated learning (Stuart & Norvig1995), which in addition offers the possibility to analyze the mistakes and explain the problem to a student. Thereafter we decided to employ the concept of decision trees to improve the learning process in nursing education.
Decision Trees: a brief overview
The use of techniques from artificial intelligence especially machine learning is a common practice in various medical applications. Machine learning deals with the discovery of hidden knowledge, unexpected patterns and new rules. Therefore decision trees have been and are still being used to extract knowledge from data, in order to make decisions in cases where explicit human knowledge cannot be used efficiently (Kokol, Zavranick, Milan, Mali & Kancler 1996).
The algorithm for learning a decision tree is trivial and the representation of accumulated knowledge can be easily understood. Namely, the decision trees do not give us just the decision in a previously unseen case - they also give us the explanation of the decision, and that is essential in educational settings.
A decision tree is induced on a training set, which consists of training objects (instances). Each training object is completely described by a set of attributes and a class label (category, outcome). Classes are mutually exclusive, what means that the training object can belong to only one class. Attributes can be continuous (numeric) or discrete. Continuous attributes are not suitable for learning a tree, so they must be mapped into a discrete space. A decision tree contains nodes and edges (links). There are two types of nodes. Each internal node (non-terminal node) has a split, which tests the value of the chosen attribute for the training objects, that have come into this node and according to that splits the training set. Each internal node has at least two child nodes. External nodes, also called leaves or terminal nodes, are labelled with outcomes. Nodes (internal and external) are connected with edges. Edges are labelled with different outcomes of test, performed in the source node. Number of edges that come out of the node depends on the number of possible outcomes of the test.
Unlike some other approaches the representation of a decision tree can be easily understood by a human. All tests in internal nodes of a tree can be determined so the importance of attributes can be obtained from the decision tree. This that is the way to take advantage of the decision trees even without using them for their primary task ñ decision making.
According to above, decision trees can support the nursing education process in four ways:
1) to represent the knowledge and decision making as a simple two-dimensional hierarchical model;
2) to outline important factors needed for successful decision making
3)to enable a nurse to use the decision tree (in the paper form or as a computer program) to learn, support and test their own decision making in new situations and
4) with new cases to construct the decision tree (using automatic learning) for their own cases,
A Sample Decision Tree
As an example of a theory presented above we generated a decision trees dealing with incontinence. The tree shown bellow enables one to decide about the type of incontinence according to related factors and defining characteristics.
Figure 1. The decision tree dealing with incontinence
(0) Neuropathy Preventing Transmission Of Reflex
The decision tree in Figure 1 is built hierarchically from left to right. Leaves are marked with outcomes. Internal nodes are marked with attributes and have successors. The most important attribute is Neuropathy Preventing Transmission Of Reflex. In our case each attribute has only two values 0 (absent) or 1 (present).
The example of decision making derived from the decision tree presented on Figure 1 follows:
Absence of Neuropathy Preventing Transmission Of Reflex and
indicates the definition Without incontinence.
On the other hand
Absence of Neuropathy Preventing Transmission Of Reflex and
indicates the definition Reflex incontinence.
In addition to decision making and hierarchical importance structure of attributes some other interesting facts can be revealed from the above decision tree. To decide about the various forms of incontinence the tree does not use all 44 different related factors, but just a subset of them. This fact was not evident to students (not even to lecturers) by using traditional teaching methods, but has become very clear after using our approach ñ as a consequence the decision making has been much simplified.
Using the above decision tree we classified 12 test objects 11 were classified correctly (92%), and just one classification failed: the total incontinence was classified as functional.
To improve and better support nursing education with the use of Computer Managed Instruction concept we employed artificial intelligence in more specifically machine learning and decision trees. It is our belief that in such way the use of computers in nursing education can become still more successful.
List of References
Athappilly K K, C. Durben, S. Woods, 1994, Multimedia Computing., Reisman, Harrisburg.
Ball M. J. 1995, Nursing Informatics. Springer, New York.
Belfry J M, 1988 A Review of the Effect of Computer Assisted Instruction in Nursing Education, Computers in Nursing, March.
Cambre M, L. J. Castner, J 1993 The status of interactive video in nursing education environments, FITNE, Atlanta,.
Checkland P &Scholes J. 1990 Soft Systems Methodology in Action, John Wiley &Sons, Chichester.
Conrick M. 1998 Computer Based Education: more than just a package. Australian Electronic Journal of Nursing Education 4:1 October.
Habjani, A , Kokol P, Zorman M, Japelj, M, 1998 CArE: A software package for computer aided nurse education, unpublished paper.
Haus, K, 1996 Changes in Perception of Computer Aided Instruction, Research report, (HYPERLINK http://parsons.ab.umd.cdu/~khaus/paper.html http://parsons.ab.umd.cdu/~khaus/paper.html)
Hebda T, 1998 A Profile of the Use of CAI within baccalaureate nursing education, Computers in Nursing, November.
Khoiny, F. 1995 Factors that contribute to computer assisted instruction effectiveness, Computers in Nursing, July.
Kohl J E, 1995. Computer Assisted Instruction: Implications for Achievement and Critical Thinking, Research report, (http://www.nursing.ab.umd.edu/students/~jkohl/cai.htm).
Kokol, P., Zavranick, J., Zorman, M,. Mali, I & Kancler, K 1996 Participative design, decision trees, automatic learning and medical decision making, in Brender, J (ed) Medical informatics Europe 96, (Studies in health technology and informatics vol 34), IOS Press Amsterdam.
Skinner, B. 1953 Science of Human Behaviour. New York, MacMillan.
Stuart J., Norvig, P 1995 Artificial intelligence: a modern approach, Englewood cliffs, Prentice-Hall.
Williamson M, 1994 High tech training, Byte, December
Last modified on: