ORIGINAL ARTICLES   

Minerva Pneumologica 2005 March;44(1):17-28

Copyright © 2005 EDIZIONI MINERVA MEDICA

language: English

Validation of a knowledge-based expert system for the diagnostic assessment of dyspnoea

Torchio M., Pollet C., Battista S., Ardizzone M., Seidenari C., Cappello N., Molino G.

Aim. The paper describes the validation of an expert system (DYSPNEA) supporting the diagnostic assessment of dyspnoic patients, realized using EMPTY, a domain independent shell assisting the development of clinical decision support systems.
Methods. EMPTY considers major structural constraints (taxonomies and description of diseases and data) and functional requirements of clinical reasoning (built on inference mechanisms based on if-then rules, frames and evidence). Clinical validation was performed by a group of physicians on a sample of 300 inpatients with dyspnoea and provided, on the basis of the available data, a reference diagnosis used for the evaluation of system accuracy.
Results. The system provided reliable (that is correct or approximate) diagnostic conclusions in approximately 90% of eligible cases, a figure reasonable for a decision support system based on clinical data, assuming that the final decision can be only achieved under user responsibility, in most cases after differential diagnosis based on invasive and costly investigations.
Conclusion. The expert system DYSPNEA is accurate and as such reliable for clinical use. The evaluation of the educational impact of the system provides some evidence in favour of the use of the support system also in the field of medical education.