Decision support by computer analysis of selected case history variables in the emergency room among patients with acute chest pain

Publisher Website
Google Scholar
Abstract
A computer system to be used in the emergency room has been developed for estimating the risk of acute coronary heart disease (ACHD). The system uses data on 38 case history and clinical variables collected consecutively over a year from 918 patients with acute chest pain. A statistical procedure based on Bayes' formula is used to estimate disease probabilities. A quadratic scoring rule was used for variable selection. The score increased markedly until 15–20 variables had been added, reached a maximum after inclusion of about 30 variables and then deteriorated slightly. Thus, the number of variables carrying additional information on the presence/absence of ACHD seems to be much larger than the number normally utilized by doctors and by other decision support systems. Reclassification into two groups, those with and without ACHD, gives a diagnostic accuracy of 89%. We conclude that analysing detailed case histories by computer is a promising decision support system for use in the emergency room as a supplement to ECG analysis.

This publication has 0 references indexed in Scilit: