Enhancing the estimation of PaCO2 from etCO2 during ventilation through non-invasive parameters in the ovine model

Background In mechanically ventilated neonates, the arterial partial pressure of CO2 (PaCO2) is an important indicator for the adequacy of ventilation settings. Determining the PaCO2 is commonly done using invasive blood gas analyses, which constitute risks for neonates and are typically only available infrequently. An accurate, reliable, and continuous estimation of PaCO2 is of high interest for medical staff, giving the possibility of a closer monitoring and faster reactions to changes. We aim to present a non-invasive estimation method for PaCO2 in neonates on the basis of end-tidal CO2 (etCO(2)) with inclusion of different physiological and ventilation parameters. The estimation method should be more accurate than an estimation by unaltered etCO(2) measurements with regard to the mean absolute error and the standard deviation. Methods Secondary data from 51 preterm lambs are used, due to its high comparability to preterm human data. We utilize robust linear regression on 863 PaCO2 measurements below or equal to 75 mmHg from the first day of life. etCO(2) along with a set of ventilation settings and measurements as well as vital parameters are included in the regression. Included independent variables are chosen iteratively by highest Pearson correlation to the remaining estimation deviation. Results The evaluation is carried out on 12 additional neonatal lambs with 246 PaCO2 measurements below or equal to 75 mmHg from the first two days of life. The estimation method shows a mean absolute error of 3.80 mmHg with a 4.92 mmHg standard deviation of differences and a standard error of 0.31 mmHg in comparison to measured PaCO2 by blood gas analysis. Conclusions The estimation of PaCO2 by the proposed equation is less biased than unaltered etCO(2). The usage of this method in clinical practice or in applications like the automation of ventilation needs further investigation.