Transferability of a Computer System for Medical History Taking and Decision Support in Dyspepsia. A Comparison of Indicants for Peptic Ulcer Disease

Abstract

The transferability of a British data base for differential diagnosis of dyspepsia using data obtained by computer interrogation was tested in 467 Swedish patients. The diagnostic value for peptic ulcer disease of symptoms such as frequent night pain relieved by food or antacids, smoking, family history of ulcer, food relief pain, male sex, and episodic pain was shown to be reproducible. However, for a number of symptoms their value for the diagnosis of peptic ulcer disease could not be reproduced in Swedish patients. The combined value of indicants was tested using a computer based algorithm for calculating diagnostic probabilities. The performance of this algorithm was poor when British data were applied to Swedish patients bus reclassification of the Swedish patients on their own data base showed promising results.

Crean and colleagues in Glasgow have developed a computer system for automated interrogation of patients with dyspepsia (1,2,3). The system utilises a large number of questions to obtain information regarding a maximum of 160 diagnostic indicants. The symptoms elicited from a patient can be compared with those of a large number of previosly examined patients and the probabilities of ten different diagnoses can be calculated. The calculation of diagnostic probabilities is based on scores reflecting the diagnostic value of different symptoms in different diseases.

After careful translation of questions the system has been transferred for use in Sweden. The present report is based on data from patients seen during the first two years with the system at a Swedish hospital. The aim of the study is to investigate the transferability to another geographical area of the system and its data base for diagnostic inference, in particular with respect to peptic ulcer.