Abstract
BACKGROUND: The purpose of this computer model study was to develop a mathematical model of steady-state gas exchange and incorporate these equations in a computer simulation that could accurately predict the venous admixture, or physiologic shunt fraction (Q̇s/Q̇t), and PaO2. METHODS: A computer model was constructed to include interactions of O2 and CO2 dissociation curves and end-capillary, mixed-venous, and arterial blood gas values. Predictions of the computer model are used to assess the variability of the 5 most common indices of gas exchange-P(A-a)O2, PaO2/PAO2, PaO2/FIO2, estimated venous admixture (EVA), and Respiratory Index ([P(A-a)O2]/PaO2). The ability of the model to predict Q̇s/Q̇t and PaO2 was validated using bias and limits of agreement between model-generated data and traditional indices of gas exchange in 50 sets of data from 37 patients. RESULTS: The magnitude of the change in PaO2/PAO2 is less than that of P(A-a)O2. However, for any given Q̇s/Q̇t value, PaO2/PAO2 varies widely. If Q̇s/Q̇t is constant, PaO2/PAO2 must vary as FIO2 changes. If PaO2/PAO2 is constant Q̇s/Q̇t must vary with changes in FIO2. In hemodynamically stable patients, EVA is a good estimate of true Q̇s/Q̇t. CONCLUSIONS: The O2-tension-based indices are poor indicators of true Q̇s/Q̇t. Using bias and limits of agreement, the computer model accurately predicted the observed Q̇s/Q̇t in critically ill patients. [Respir Care 1996;41(7):586-594.]
Get full access to this article
View all access options for this article.
