A stepwise regression approach applied to evaluation of nonlinear patterns in breathing mechanics

A multiple linear stepwise regression approach was applied to respiratory data from intensive care patients. The influence of different nonlinear viscoelastic properties on the prediction of pressure at the airway opening taken as dependent variable was studied by entering and removing potential predictive variables. These variables were derived from measured ventilatory flow data so as to quantitatively describe elastic and resistive nonlinearities attributed to typical pathophysiological mechanisms. It was found that for a better characterization of breathing mechanics, volume-dependent lung elastance must be introduced and that turbulence effects can sometimes arise. The resulting regression model still corresponds to the classical one-compartment model of breathing mechanics, but two parameters are: airflow-dependent resistance and lung volume-dependent elastance.