PubMed1993
The purpose of this study was to investigate the amount of diagnostic information contained in a set of routine lung function studies and to attempt to determine which tests could be omitted without significant loss of discrimination. Cluster analysis was performed on a set of physiological and questionnaire data, collected prospectively in 1,542 male patients, referred consecutively for measurement of forced expired volumes, static lung volumes and measurements of the transfer factor for carbon monoxide. A respiratory questionnaire was completed for each patient. A physician assigned the patients to a rigorously defined diagnostic category, based on supporting clinical information, as well as pulmonary function, apart from 241 patients with unusual diagnoses and those in whom the criteria did not apply satisfactorily. This diagnosis was never included as a classification variable. Basing the classification on three independent measurements, total lung capacity, the ratio of forced expiratory volume in one second (FEV1) to vital capacity, and the transfer factor for carbon monoxide, the computer generated six groups: one normal, one showing an isolated gas exchange defect, and four with varying degrees of restriction and obstruction. This classification performed well in separating the patients with the clinical diagnoses of chronic airflow obstruction, bronchial asthma and interstitial lung disease from those with ischaemic and valvular heart disease and other miscellaneous disorders. Omitting total lung capacity resulted in some loss of specificity, but valid information was still obtained. The inclusion of all the static and dynamic lung volumes and of carbon monoxide transfer coefficient made little difference.(ABSTRACT TRUNCATED AT 250 WORDS)