Juan A. Fernández del Pozo

A Graphical Decision-Theoretic Model for Neonatal Jaundice:

Background. Neonatal jaundice is treated daily at all hospitals. However, the routine, urgency, and case load of most doctors stop them from carefully analyzing all the factors that they would like to (and should) take into account. This article develops a complex decision support system for neonatal jaundice management. Methods. The problem is represented by means of an influence diagram, including admission and treatment decisions. The corresponding uncertainty model is built with the aid of both historical data and subjective judgments. Parents and doctors were interviewed to elicit a multiattribute utility function. The decision analysis cycle is completed with sensitivity analyses and explanations of the results. Results. The construction and use of this decision support system for jaundice management have induced a profound change in daily medical practice, avoiding aggressive treatments—there have been no exchange transfusions in the past 3 years—and reducing the lengths of stay at the hospital. More information is now taken into account to decide on treatments. Interestingly, after embarking on this modeling effort, physicians came to view jaundice as a much more difficult problem than they had initially thought. Comparisons between real cases and system proposals revealed that treatments by nonexpert doctors tend to be longer than what expert doctors would administer. Conclusion. The system is especially designed to help neonatologists in situations in which their lack of experience may lead to unnecessary treatments. Different points of view from several expert doctors and, more interestingly, from parents are taken into account. This knowledge gives a broader picture of the medical problem— incorporating new action criteria, new agents to intervene, more uncertainty variables—to get an insight into the suitability of each therapeutic decision for each patient situation. The benefits gained and the usefulness perceived by neonatologists are worth the increased and time-consuming effort of developing this complex system. Although specially designed for a specific hospital and for neonatal jaundice management, it can be easily adapted to other hospitals and problems.

Parameter Control of Genetic Algorithms by Learning and Simulation of Bayesian Networks — A Case Study for the Optimal Ordering of Tables

Parameter setting for evolutionary algorithms is still an important issue in evolutionary computation. There are two main approaches to parameter setting: parameter tuning and parameter control. In this paper, we introduce self-adaptive parameter control of a genetic algorithm based on Bayesian network learning and simulation. The nodes of this Bayesian network are genetic algorithm parameters to be controlled. Its structure captures probabilistic conditional (in)dependence relationships between the parameters. They are learned from the best individuals, i.e., the best configurations of the genetic algorithm. Individuals are evaluated by running the genetic algorithm for the respective parameter configuration. Since all these runs are time-consuming tasks, each genetic algorithm uses a small-sized population and is stopped before convergence. In this way promising individuals should not be lost. Experiments with an optimal search problem for simultaneous row and column orderings yield the same optima as state-of-the-art methods but with a sharp reduction in computational time. Moreover, our approach can cope with as yet unsolved high-dimensional problems.

Optimal Decision Explanation by Extracting Regularity Patterns

When solving decision-making problems with modern graphical models like influence diagrams, we obtain the decision tables with optimal decision alternatives. For real-life clinical problems, these tables are often extremely large, hindering the understanding of the reasons behind their content. KBM2L lists are new structures that simultaneously minimise memory storage space of these tables, and search for a better knowledge organisation. In this paper, we study the application of KBM2L lists in finding and thoroughly studying the optimal treatments for gastric nonHodgkin lymphoma. This is a difficult clinical problem, mainly because of the uncertainties involved. The resultant lists provide high-level explanations of optimal treatments for the disease, and are also able to find relationships between groups of variables and treatments.

Sensitivity Analysis in IctNeo

IctNeo is a complex decision support system to manage neonatal jaundice, the situation in which bilirubin accumulates when the liver does not excrete it at a normal rate. This system finds a maximum expected utility treatment strategy based on an influence diagram (ID). Due to the computational intractability of such a large ID, IctNeo incorporates some procedures to the standard evaluation algorithm.

Finding and Explaining Optimal Treatments

Influence diagrams are modern decision-theoretic representations that can be used to model medical decision-making problems. The output of evaluating an influence diagram are decision tables with optimal decision alternatives. For real-life clinical problems the resulting tables can be really big, so that understanding what they say is nearly impossible. KBM2L lists are new list-based structures suitable for minimising memory storage space of these tables, and at the same time searching for a better knowledge organisation. In this paper, we study the application of KBM2L lists for finding the optimal treatments for gastric non-Hodgkin lymphoma.

Knowledge Organisation in a Neonatal Jaundice Decision Support System

The tables containing the optimal decisions obtained when solving real decision-making problems under uncertainty are often extremely large. Tables can be considered as multidimensional matrices (MMs) and computers manage them as lists, where each position is a function of the order chosen (or base) for the matrix dimensions. In this paper, we propose turning the decision tables into minimum storage lists. Evolutionary computation is required to minimise the number of list entries (items). The optimal list includes the same knowledge as the original list, but it is compacted, which is very valuable for explaining expert reasoning. We illustrate the ideas using our decision support system IctNeo (Bielza et al., 2000) for neonatal management, outputting excellent results. The methodology is so general that it also applies to any table considered as a knowledge base (KB).

An Interactive Framework for Open Queries in Decision Support Systems

We have recently introduced a method for minimising the storage space of huge decision tables faced after solving real-scale decision-making problems under uncertainty [4]. In this paper, the method is combined with a proposal of a query system to answer expert questions about the preferred action, for a given instantiation of decision table attributes. The main difficulty is to accurately answer queries associated with incomplete instantiations. Moreover, the decision tables often only include a subset of the whole problem solution due to computational problems, leading to uncertain responses. Our proposal establishes an automatic and interactive dialogue between the decision support system and the expert to extract information from the expert to reduce uncertainty. Typically, the process involves learning a Bayesian network structure from a relevant part of the decision table and the computation of some interesting conditional probabilities that are revised accordingly.

Dominance measuring methods for the selection of cleaning services in a European underground transportation company

Dominance measuring methods are a recent approach for dealing with complex decision-making problems with imprecise, incomplete or partial information within multi-attribute value/utility theory. These methods compute pairwise dominance values and exploit the information included in the dominance matrix in different ways to derive measures of dominance intensity to rank the alternatives under consideration. We review dominance measuring methods proposed in the literature, describing how their possible drawbacks have been progressively overcome, and comparing their performance with other existing approaches, like surrogate weighting methods, the adaptation of classical decision rules to encompass an imprecise decision context, SMAA or Sarabando and Dias’ method. An example of the selection of cleaning services in a European underground transportation company is used to illustrate dominance measuring methods in a real complex decision-making problem.