Watching symptoms and illnesses through the eyes of Multiple Correspondence Analysis: a case study in toxicology
Résumé
The goal of the reported research was the design of a computerized tool aimed at aiding an emergency specialist, in a toxicological emergency unit, to quickly identify, at the admission time, the various poisons ingested by a comatose patient. This medical decision making problem has been proved to be computationally intractable since a lot of patient cases sharing the same clinical table have different diagnoses in terms of psychotropes combinations. The paper explores the idea that the outcome of Multiple Correspondence Analysis (MCA), a mathematical data analysis method, can be thought of as visual and analytical aids for the physician facing this decision problem. We argue that the expert's clinical reasoning can be enhanced by the factorial maps and some computerized results provided by MCA. Using a learning database of 505 diagnosed cases, we realized a whole decision aiding system called TOXSYMEDIA. A test-base of 97 patients was used to partially assess the system. The method and resulting tool revealed to be appreciable to early inform the physician about the possible combination of ingested psychotropes.