A simplified 4-parameter model of volumetric capnograms improves calculations of airway dead space and slope of Phase III

To evaluate a compact and easily interpretable 4-parameter model describing the shape of the volumetric capnogram, and the resulting estimates of anatomical dead space (VD_AW) and Phase III (alveolar plateau) slope (S_III). Data from of 8 mildly-endotoxemic pre-acute respiratory distress syndrome sheep were fitted to the proposed 4-parameter model (4p) and a previously established 7-parameter model (7p). Root mean square error (RMSE) and Akaike information criterion (AIC), as well as VD_AW and S_III derived from each model were compared. Confidence intervals for model’s parameters, VD_AW and S_III were estimated with a jackknife approach. RMSE values were similar (4p: 1.13 ± 0.01 mmHg vs 7p: 1.14 ± 0.01 mmHg) in the 791 breath cycles tested. However, the 7p overfitted the curve and had worse AIC in more than 50% of the cycles (p < 0.001). The large number of degrees of freedom also resulted in larger between-animal range of confidence intervals for 7p (VD_AW: from 6.1 10^–12 to 34 ml, S_III: from 9.53 10^–7 to 1.80 mmHg/ml) as compared to 4p (VD_AW: from 0.019 to 0.15 ml, S_III: from 3.9 10^–4 to 0.011 mmHg/ml). Mean differences between VD_AW (2.1 ± 0.04 ml) and S_III (0.047 ± 0.004 mmHg/ml) from 7 and 4p were significant (p < 0.001), but within the observed cycle-by-cycle variability. The proposed 4-parameter model of the volumetric capnogram improves data fitting and estimation of VD_AW and S_III as compared to the 7-parameter model of reference. These advantages support the use of the 4-parameter model in future research and clinical applications.