Influence of nonlinearities on estimates of respiratory mechanics using multilinear regression analysis

Abstract

To investigate the influence of nonlinearities on estimates of respiratory mechanics, differing patterns of mechanical ventilation patterns were analyzed from 8 puppies and 14 children. Respiratory mechanics were calculated using multiple linear regression to fit a linear single-compartment model, a volume-dependent single-compartment model (VDSCM), and a flow-dependent single-compartment model. The ratio of the compliance of the last 20% of the dynamic volume-pressure (V-P) curve to the total compliance (C20/C) and the contribution of a volume-dependent elastance to total elastance [%E2 = E2 (VT)/[(E1 + E2)VT], where E1 + E2 is total elastance, E2 is the volume-dependent component, and VT is tidal volume] were used as the indexes of over-distension. By positioning the dynamic loops on the static V-P curves, ventilation patterns were classified as overdistended or nonoverdistended. In the overdistended group, the C20/C was significantly lower (0.71 +/- 0.10 vs. 0.92 +/- 0.16; P < 0.0001) and %E2 was significantly higher (43.4 +/- 15.0 vs. 0.51 +/- 18.02%, P < 0.0001) than in the nonoverdistended group. The mode of ventilation (pressure controlled vs. volume controlled) and the resistive pressures that resulted in widening of the dynamic V-P loop were found to alter C20/C but not %E2. When the respiratory system was overdistended, i.e., ventilated up to the flattened portion of the V-P curve, the VDSCM gave more accurate estimates of respiratory mechanisms. Furthermore, %E2 calculated from VDSCM is a useful parameter for estimating respiratory system overdistension that is not affected by resistive pressures.