Performance of a patient-dedicated, on-demand blood gas monitor in medical ICU patients.

We examined the performance characteristics of a new bedside blood gas monitor. This monitor's fluorescent pH, PCO2, and PO2 sensors are embedded in a cassette, which is calibrated in vitro and then inserted into the patient's radial artery tubing set. In 50 medical ICU patients, 683 paired monitor and conventional blood gas analyzer values were obtained. Performance was assessed via calculations of bias (mean monitor and analyzer difference) and its standard deviation (SD), plots of monitor and analyzer differences against the means (of monitor and analyzer), and linear regression analysis of the sequential changes in monitor values versus the corresponding sequential changes in analyzer values. The ex vivo calibration, assessed using the initial paired blood samples, showed a bias +/- SD of 0.02 +/- 0.02 for pH, -0.1 +/- 1.9 mm Hg for PCO2, and 4.3 +/- 6.0 mm Hg for PO2. For all paired samples (n = 683), the biases +/- SD were 0.004 +/- 0.023 for pH, 0.6 +/- 2.4 mm Hg for PCO2, and 2.7 +2- 6.4 mm HG for PO2. The PO2 bias increased as PO2 increased. The standard deviations (imprecision) of both PCO2 and PO2 also increased as the magnitudes of these variables increased. Sequential changes in monitor values versus the corresponding sequential changes in analyzer values revealed regression lines close to the line of identity. Serum sodium had no effect on pH bias. Daily drift of the sensors was inconsequential, with values of -0.01/d for pH, 1.7 mm Hg/d for PCO2, and 1.1 mm Hg/d for PO2. We conclude that the performance of this monitor is comparable to that of conventional blood gas analyzers.