Search

First results and methods of assessment used in the application of the AEDMI project structured medical interview in a sample of 824 patients are reported. The sensitivity, specificity, and predictivity of the most frequent signs and symptoms with regard to the diagnosis of high blood pressure have been computed. Positive predictivities between 2% and 40% were found. The diagnostic value of these variables can be applied in the context of the hypothesis of the project which is based on the stepwise procedure used by physicians to establish a clinical diagnosis. The objective probabilities represented by the aforementioned parameters will be used to build a system to help clinical decision making.

A project to develop a clinical decision aid system is proposed, based on a multicentric design and on the application of epidemiological methods to the analysis of clinical data bases and clinical judgment. The knowledge of the prevalences of the most common diseases in the treated virtual population allows the assessment of diagnostic predictivities, and the provision of objective probabilities as a reference for the expert system. The characteristics of the system proposed are be described.

The development of an expert system for the detection of adverse drug reactions based on the epidemiological approach and careful monitorization of a random sample of inpatients is proposed. Clinical parameters at the time of admission to hospital are taken as a starting point to assess new manifestations appearing during hospitalization. The newly developed signs and symptoms are sequentially introduced in a data base and analyzed using causality algorithms and statistical tests. Any statistically significant observation made or causality association established are entered into a medical knowledge data base. The expert system proposed may be used as a consultation tool or therapeutical decision aid.

A model of structure for a complementary expert system to the AEDMI project (An Epidemiological Approach to Computerized Medical Diagnosis) is proposed. Starting from a diagnosis of abnormality in a system or apparatus, the model identifies the organ affected, and determines which of the possible illnesses is the one we are faced with in a particular case. The resoning is based on a tree-like structure divided into a set of three levels, assuming that greater sensitivity and less specificity of the questions, clinical variables, and the signs and symptoms to be taken into consideration is found at level I than at level III. Each time that the system is consulted for a real patient, the system offers the possibility to register all the parameters in a data base. The modules, rules, and diagnoses are registered and compared with physicians1 judgment. The study of conflicting patterns helps to analyze clinical judgment and rule-based reasoning techniques in greater depth. The system is being implemented for more than one year and the data from 825 patients introduced are being currently assessed.