Health care is information sensitive and information rich industry. Future doctors and managers need to feel comfortable using information technology (IT) tools in all aspects of their work. IT helps access the increasing amount of electronic data available, but searching and appraisal skills are needed to filter from this huge amount of data, the relevant and valid information. However, often the size and complexity of the data exceed even human capabilities and skills. Then again computers can assist to search for the most relevant information for the problem at hand. For this purpose computers have to perform very complex and often tedious computations based on the results of research from other scientific disciplines, particularly mathematics, theory of information, mathematical statistics and artificial intelligence. Unfortunately, partly due to inability and unwillingness of mathematicians to communicate with the medical community, often very interesting and potentially powerful results achieved in the above mentioned disciplines are made accessible and understandable neither to managers nor to physicians. Though „high mathematics“ can look like a distant world for non-mathematicians, we shall try to show (almost without any mathematics) that it is worthwhile to find a common language and to combine the expert knowledge of managers, physicians and mathematicians for the benefit or improving the quality of decision-making. Finally, some results achieved in the research project concerning the support of decision-making which is being solved in our research laboratory will be presented. Their utilization is intended not only for real managerial problems but also for educational purposes. The purpose of the lecture will be manifold: • to point out formal analogies of decision-making problems in health care management (or management generally) and in clinical decision-making • to stress the importance of bringing together managers, physicians and mathematicians in order to solve problems of evaluating information and selecting those its components which are most valuable for decision-making • to make both the researchers and practitioners in health care management and in clinical medicine acquainted with some results in artificial intelligence (AI), particularly in pattern recognition and feature selection which are directly applicable to the solution of the discussed problem . A typical problem which both managers and physicians often encounter is the problem of too many potential inputs into their respective decision-making problems. This phenomenon has been extensively studied in mathematics and in artificial intelligence. It can be stated that in order to make reliable decisions (or more exactly to learn to make them based on the past experience and the available data) the need for the amount of data dramatically grows with the number of inputs. Mathematically it means that the sample size required grows exponentially with the data dimensionality. This problem is very relevant particularly to the field of medicine as the process of medical or economic data acquisition is usually both time consuming and costly. Consequently, the data sets acquired in medicine are usually too small with respect to their dimensionality. Each of the considered fields (managerial and clinical decision-making) has its own specificity and accordingly different ways of treating the problem. On the other hand, we shall try to show that that the methods developed recently in the field of statistical pattern recognition to solve the problem of feature selection can enrich the methodology of selecting the most useful information in medical decision-making.

Fakulta managementu VŠE Praha