ABSTRACT
After having collected a sufficient number of standardized data checked for reliability from patients with chronic non specific lung disease, the following subgroups of pulmonary arterial mean pressure were defined: A = PAP less than 2.7 kPa [= 20 mmHg]; B = 2.7 less than PAP less than 4 kPa; C = PAP greater than 4 kPa [= 30 mmHg]. Within these subgroups a diagnostic classification was also carried out: chronic obstructive lung disease, thrombo-embolic lung disease, diffuse interstitial lung disease. The aims of data processing were: clustering correctly individual cases in one of the I-II-III subgroups based on non-invasive variables only; To define equations predictive of quantitative values of pulmonary arterial mean pressure from non-invasive data; To find regularity of evolution of elevated pulmonary arterial mean pressure in chronic non specific lung disease and detect risk factors and/or beneficial interventions if such exist at all. Two different statistical methods were applied: conditional probability (Bayes theorem) and discriminant analysis. The predictive power of each depends mainly on the seriousness of signs and symptoms denoted by the medical user. Three probability-functions have been defined between severity of single variables and their specificity, sensitivity and probability of pulmonary hypertension. By application of these functions the user can select different quantitative values of variables which determine the proportion of false/true, positive/negative cases. After having chosen the desired values of these proportions, multivariate equations are defined by stepwise regression analysis optimized for predictive power.(ABSTRACT TRUNCATED AT 250 WORDS)
