A community-acquired pneumonia is a frequent infection, especially for people with chronic diseases and for old people. It is one of the most common causes of mortality related to infectious diseases (the first in EEUU).
There are a lot of microorganisms causing pneumonia. The diagnosis depends on the identification of the microorganism causing pneumonia. Nowadays, with the available diagnostic methodology, it is still very difficult to determine which of the microorganisms is the infecting agent in a particular pneumonia case. The research focused to determine which are the microorganisms causing pneumonia only succeeds in the 50% of the cases [1, 5]. Errors in diagnosis can be fatal, for instance, to diagnose pneumococcal pneumonia in a patient with legionella pneumonia may produce the death of the patient by the delay of an adequate treatment.
Despite the uncertainty of the diagnosis, a treatment has to be speedily administrated to avoid a negative evolution of the severity of the illness or in some cases the death.
In our approach we make two main assumptions:
1) Existence of a previous diagnosis: Normally a pneumonia is caused by only one microorganism, but symptoms and signs are not specific enough to determine which one. Diagnosis usually gives evidence for two or three microorganisms possibly causing pneumonia. We will assume in this work that such diagnosis already exists. It can be obtained from another knowledge-based system such as Pneumon-IA [6].
2) Independence of treatments: We can independently find the best treatment for every microorganism appearing in a diagnosis. Moreover, these treatments can be combined to give a treatment covering all the possible causes. By ``covering'' we mean that a treatment is specific for a particular microorganism, in other words, it ``attacks'' the microorganism.
Besides the uncertainty of the diagnosis, data about the patient is in many cases also uncertain and incomplete.
In the subsequent sections we will describe in detail how a solution to this problem has been implemented. We begin with a conceptual structure of the problem. After that we deal with the main ideas used in the real implementation of Terap-IA, a knowledge-based system for the automatic generation of treatments. Finally the conclusions and results obtained up to date are presented.