USE OF AN ARTIFICIAL NEURAL-NETWORK IN ESTIMATING PREVALENCE AND ASSESSING UNDERDIAGNOSIS OF ASTHMA

An artificial neural network was trained in the recognition of asthmat ics in a general practice population, employing cross-validation on a subset of 350 patients of known asthmatic status. The trained network was then run on the data from 3139 patients whose asthmatic status was unknown. Using the values from the test set as estimates of sensitivi ty and specificity, the number predicted positive was adjusted to allo w for false positives and false negatives to give an estimate of asthm a prevalence and the minimum under-diagnosis rate that this suggested for the population. Using different data sets and network structures, prevalence rates of approximately 16-21% were measured providing evide nce, even after allowing for maximum variability in the estimates, con sistent with under-diagnosis of at least a small percentage (0.7-4.0%) . To provide a more precise estimate of the rate of this under-diagnos is and associated prevalence, a larger training and testing set of mor e accurately labelled cases is planned.