Discovering subgroups using descriptive models of adverse outcomes in medical care.

OBJECTIVES: Hospital discharge databases store hundreds of thousands of patients. These datasets are usually used by health insurance companies to process claims from hospitals, but they also represent a rich source of information about the patterns of medical care. The proposed subgroup discovery method aims to improve the efficiency of detecting interpretable subgroups in data. METHODS: Supervised descriptive rule discovery techniques can prove inefficient in cases when target class samples represent only an extremely small amount of all available samples. Our approach aims to balance the number of samples in target and control groups prior to subgroup discovery process. Additionally, we introduce some improvements to an existing subgroup discovery algorithm enhancing the user experience and making the descriptive data mining process and visualization of rules more user friendly. RESULTS: Instance-based subspace subgroup discovery introduced in this paper is demonstrated on hospital discharge data with focus on medical errors. In general, the number of patients with a recorded diagnosis related to a medical error is relatively small in comparison to patients where medical errors did not occur. The ability to produce comprehensible and simple models with high degree of confidence, support, and predictive power using the proposed method is demonstrated. CONCLUSIONS: This paper introduces a subspace subgroup discovery process that can be applied in all settings where a large number of samples with relatively small number of target class samples are present. The proposed method is implemented in Weka machine learning environment and is available at http://ri.fzv.uni-mb.si/ssd.

Diagnosing mitral valve prolapse by improving the predictive power of classifiers.

Mitral valve prolapse (MVP) has been described as one of the most common cardiac valvular abnormalities in industrialized countries, and can result in sudden death. This study focused on various feature selection mechanisms that might improve the predictive power of a classifier to diagnose MVP. The experiment included selection mechanisms using classical greedy feature selection approaches (forward selection and backward elimination), a genetic algorithm (GA) approach and a cellular automaton (CA) approach. The main aim of this latest approach is to use CA with GA for the data transformation phase of the knowledge discovery process. The CA-GA approach produced better results than the classical greedy approaches. The subsets of features produced by the GA and CA approaches were most appropriate for the decision tree classifier, for diagnosing MVP with the highest overall class accuracy. More importantly, the CA and GA approaches were also capable of generalizing some important knowledge concerning MVP diagnosis.

Accuracy of intelligent medical systems.

Intelligent medical systems are a special kind of medical software in general, and just as any medical software system they should make accurate presumptions. However, accuracy of intelligent medical systems is highly dependent on various factors such as: choosing an appropriate basic method (i.e. decision trees, neural networks), induction method (i.e. purity measures) and appropriate support methods (i.e. discretization, pruning, boosting). In this paper we present the results of extensive research of the above alternatives on 54 UCI databases and their influence on the accuracy of decision trees, which constitute one of the most desirable forms of intelligent medical systems. We also introduce new hybrid purity measures that on some databases outperform other purity measures. The results presented here show that the selection of the right purity measure with the proper discretization method and application of the boosting method can really make a difference in terms of higher accuracy of induced decision trees. Thereafter choosing the appropriate factors that can increase the accuracy of the induced decision tree is a very demanding and time-consuming task.

Supporting medical decisions with vector decision trees.

The article presents the extension of a common decision tree concept to a multidimensional - vector - decision tree constructed with the help of evolutionary techniques. In contrary to the common decision tree the vector decision tree can make more than just one suggestion per input sample. It has the functionality of many separate decision trees acting on a same set of training data and answering different questions. Vector decision tree is therefore simple in its form, is easy to use and analyse and can express some relationships between decisions not visible before. To explore and test the possibilities of this concept we developed a software tool--DecRain--for building vector decision trees using the ideas of evolutionary computing. Generated vector decision trees showed good results in comparison to classical decision trees. The concept of vector decision trees can be safely and effectively used in any decision making process.

Intelligent systems for nursing education.

Health care is one of the fastest growing areas in terms of care, treatment and the exploitation of new technology in Slovenia. There is a great need for new approaches ensuring that education and work of health care professionals will be built upon the state of the art in nursing. As a consequence the educational, governmental and "industrial" institutions from Slovenia, UK, Italy and Greece have determined to work on above problem. EU agreed to support the project under the Phare Tempus Framework and the aim of this paper is to present an educational approach based on intelligent systems and its application in nursing education.

Comparison of three databases with a decision tree approach in the medical field of acute appendicitis.

Decision trees have been successfully used for years in many medical decision making applications. Transparent representation of acquired knowledge and fast algorithms made decision trees one of the most often used symbolic machine learning approaches. This paper concentrates on the problem of separating acute appendicitis, which is a special problem of acute abdominal pain from other diseases that cause acute abdominal pain by use of an decision tree approach. Early and accurate diagnosing of acute appendicitis is still a difficult and challenging problem in everyday clinical routine. An important factor in the error rate is poor discrimination between acute appendicitis and other diseases that cause acute abdominal pain. This error rate is still high, despite considerable improvements in history-taking and clinical examination, computer-aided decision-support and special investigation, such as ultrasound. We investigated three different large databases with cases of acute abdominal pain to complete this task as successful as possible. The results show that the size of the database does not necessary directly influence the success of the decision tree built on it. Surprisingly we got the best results from the decision trees built on the smallest and the biggest database, where the database with medium size (relative to the other two) was not so successful. Despite that we were able to produce decision tree classifiers that were capable of producing correct decisions on test data sets with accuracy up to 84%, sensitivity to acute appendicitis up to 90%, and specificity up to 80% on the same test set.

Finding the right decision tree's induction strategy for a hard real world problem.

Decision trees have been already successfully used in medicine, but as in traditional statistics, some hard real world problems can not be solved successfully using the traditional way of induction. In our experiments we tested various methods for building univariate decision trees in order to find the best induction strategy. On a hard real world problem of the Orthopaedic fracture data with 2637 cases, described by 23 attributes and a decision with three possible values, we built decision trees with four classical approaches, one hybrid approach where we combined neural networks and decision trees, and with an evolutionary approach. The results show that all approaches had problems with either accuracy, sensitivity, or decision tree size. The comparison shows that the best compromise in hard real world problem decision trees building is the evolutionary approach.

Towards more optimal medical diagnosing with evolutionary algorithms.

Efficiency in hospital performance is becoming more and more important. Studies showed that diagnosis can considerably reduce the inefficiency, so one of the most important tasks in achieving greater hospital efficiency is to optimize the diagnostic process. For the best of the patient the diagnostic process has to be optimized regarding the number of the examinations and individualized in order to maximize accuracy, sensitivity and specificity. In addition the duration of the diagnostic process has to be minimized and the process has to be performed on the most reliable equipment. The main contribution of our paper is the introduction of the integrated computerized environment DIAPRO enabling the diagnostic process optimization. The DIAPRO is based on a single approach--evolutionary algorithms.

Power of heterogeneous computing as a vehicle for implementing E(3) medical decision support systems.

A computer system of PCs, workstations, minicomputers etc. connected together via a local area network or wide area network represents a large pool of computational power. Our aim is to use this power for the implementation of an E(3) (efficiency, effectiveness, efficacy) medical decision support system, which can be based on different models, the best providing an explanation together with an accurate and reliable response. One of the most viable among models is decision trees, already used for many medical decision-making purposes. In this paper, we present a parallel implementation of a genetic algorithm on a heterogeneous computing system for the induction of decision trees with the application on solving the mitral valve prolapse syndrome. Our approach can be considered as a good choice for different real-world decision making, with respect to the advantages of our model, especially the great computational power.

The art of building decision trees.

Decision support systems that help physicians are becoming a very important part of medical decision making. They are based on different models and the best of them are providing an explanation together with an accurate, reliable, and quick response. One of the most viable among models are decision trees, already successfully used for many medical decision-making purposes. Although effective and reliable, the traditional decision tree construction approach still contains several deficiencies. Therefore we decided to develop and compare several decision support models using four different approaches. We took statistical analysis, a MtDeciT, in our laboratory developed tool for building decision trees with a classical method, the well-known C5.0 tool and a self-adapting evolutionary decision support model that uses evolutionary principles for the induction of decision trees. Several solutions were evolved for the classification of metabolic and respiratory acidosis (MRA). A comparison between developed models and obtained results has shown that our approach can be considered as a good choice for different kinds of real-world medical decision making.

Using intelligent search for finding medical sites.

In this paper we present an intelligent search tool, which we developed in order to automate search and evaluation of websites. We used TFIDF heuristics to determine term frequency and decision trees to evaluate the quality of sites. Training set for the decision tree contained manually evaluated websites. Each website was described by the combination of various attributes, complexity metrics and the evaluation. The intelligent search tool is equipped with a user-friendly interface, which enables people to exploit the tool to its limits with minimum effort, in their quest for information. For testing purposes, we looked for sites with telemedical content. The set of sites, which was the result of using the intelligent search tool, has been evaluated by a group of students.

The influence of class discretization to attribute hierarchy of decision trees.

Decision support systems that help physicians are becoming very important part of medical decision making. They are based on different models and the best of them are providing an explanation together with an accurate, reliable and quick response. One of the most viable among decision-making models is the concept of decision trees, already successfully used for many medical decision making purposes. Although effective and reliable, the traditional decision tree construction approach still contains several deficiencies. Therefore we decided to develop and compare several decision supporting models, each of them built with different discretization of attributes and decision classes. For the construction of decision trees we used MtDeciT, in our laboratory developed tool for building decision trees using the classical induction method. All solutions were evolved for determining the influence of basic properties of child and his/her parents to length of successful breastfeeding. A comparison between developed models and obtained results has shown that the way of discretization obviously plays a great role in the reliable and accurate real-world medical decision making.

Power of heterogeneous computing as a vehicle for implementing E3 medical decision support systems.

A computer system of PCs, workstations, minicomputers etc., connected together via a local area network or wide area network represents a large pool of computational power. Our aim is to use this power for the implementation of E3 (efficiency, effectiveness, efficacy) medical decision support system, which can be based on different models; the best of them are providing an explanation together with an accurate and reliable response. One of the most viable among models are decision trees, already used for many medical decision making purposes. In this paper we would like to present a heterogeneous implementation of genetic algorithm for the induction of decision trees with emphasis on solving the mitral valve prolapse syndrome.

A tutor for nursing process education.

Computer managed instruction (CMI) has been used in nursing education since the late 1960's. It is due to the accessibility and self paced format that CMI is very well suited for both students and practicing nurses, while learning can occur at the learner's own pace and time. In addition CMI supports also continuing education and distant learning. The aim of this paper is to present CArE--a software package for Computer Aided Nurse Education in particular for teaching the basics of the nursing care process, developed as a result of the Phare TEMPUS project called NICE (Nursing Informatics and Computer Aided Education).

The limitations of decision trees and automatic learning in real world medical decision making.

The decision tree approach is one of the most common approaches in automatic learning and decision making. It is popular for its simplicity in constructing, efficient use in decision making and for simple representation, which is easily understood by humans. The automatic learning of decision trees and their use usually show very good results in various "theoretical" environments. The training sets are usually large enough for learning algorithm to construct a hypothesis consistent with the underlying concept. But in real life it is often impossible to find the desired number of training objects for various reasons. The lack of possibilities to measure attribute values, high cost and complexity of such measurements, unavailability of all attributes at the same time are the typical representatives. There are different ways to deal with some of these problems, but in a delicate field of medical decision making, we cannot allow ourselves to make any inaccurate decisions. We have measured the values of 24 attributes before and after the 82 operations of children in age between 2 and 10 years. The aim was to find the dependencies between attribute values and a child's predisposition to acidemia--the decrease of blood's pH. Our main interest was in discovering predisposition to two forms of acidosis, the metabolic acidosis and the respiratory acidosis, which can both have serious effects on child's health. We decided to construct different decision trees from a set of training objects, which was complete (there were no missing attribute values), but on the other hand not large enough to avoid the effect of overfitting. A common approach to evaluation of a decision tree is the use of a test set. In our case we decided that instead of using a test set, we ask medical experts to take a closer look at the generated trees. They examined and evaluated the decision trees branch by branch. Their comments on the generated trees can be found in this paper. The comments show, that trees generated from available training set mainly have surprisingly good branches, but on the other hand some are very "stupid" and no medical explanation could be found. Thereafter we can conclude, that the decision tree concept and automatic learning can be successfully used in real world situations, constrained with the real world limitations, but they should be used only with the guidelines of appropriate medical experts.