Role of oximeter error in the diagnosis of shunts

We have analyzed the role that instrument error plays in the oximetric diagnosis of shunts. First, we review the accuracy specifications of instruments currently available for measuring oxyhemoglobin saturation during cardiac catheterization. Then we derive a mathematical model of a left-to-right shunt and illustrate the relationships between the magnitude of shunts and the saturation step-ups they generate. Using the shunt model and statistical methods, we show how oximeter error affects the likelihood of success in detecting shunts of a given magnitude. By computing the probabilities of false negative and false positive diagnoses, the model can be used to determine how the likelihood of successful shunt detection is influenced by factors such as shunt flow, the minimum saturation step-up considered indicative of a shunt, oximeter error, hemoglobin concentration, and the number of blood samples. A spreadsheet is included for those who wish to analyze the diagnostic capabilities of their own instrumentation and clinical methods. Conclusions. (1) Modern instruments for use during cardiac catheterization analyze blood samples with a measurement error ranging from 2.5 to 1% saturation or better. (2) Until more information is available regarding the accuracy of such devices in measuring oxygen content, the detection of shunts should be made on the basis of step-ups in saturation rather than oxygen content (even though step-ups in oxygen content have the potential advantage of being independent of the oxygen-carrying capacity of blood). (3) Averaging the analyses of two or more blood samples from each site and using an oximeter with 1% error greatly improve the probability of reaching the correct diagnosis. (4) Oximeter error of 2.5% or greater is unacceptable by today's standards. (5) With the instrumentation currently available, oximetry can potentially detect saturation step-ups as small as 3.6% and still keep both false negative and false positive diagnoses below 10% if venous blood is adequately mixed.