Review of the graphs suggested that for some of the nine latter ANOVAs, some of the N grades were not justified because the model patterns along the full range of intensities were visually very dissimilar. For example, model A might give consistently lower values than model B at low intensities and consistently higher values than model B at high intensities. Because of such visually apparent “crossover” patterns, we calculated four more ANOVAs, each including all six or seven models of that manufacturer, but using only the upper half of the range of values (including 15 rather than 30 lots).
Next, the 20 most commonly used models from the three 1996 FCE surveys were selected to compare the actual deviations for low, medium, and high intensities of Po2, Pco2, and pH to demonstrate the actual differences among models observed. First we tabulated the range of differences for 7 models; then we counted the number of model pairs for all 20 models that exceeded the following arbitrary limits: 4 mm Hg, 6 mm Hg, and 6% for Po2; 3 mm Hg and 3% for Pco2; and 0.030 U for pH at these same intensities. antibiotics online
Finally, the FCE data were compared with tonometered blood data of Scuderi et al, who reported the differences among four blood gas analyzers (AVL995, COR178, IL1312, and RAD330) by measuring tonometered blood at 16 levels of Po2 (their Table 3). They did not report Pco2 or pH values. From their data, we selected the 11 blood Po2 levels that we could reasonably match with 11 survey FCE lot levels (47 vs 43.4, 56 vs 57.4, 66 vs 65.1, 75 vs 78.1, 85 vs 86.4, 94 vs 94.4, 104 vs 108, 113 vs 111.5, 122 vs 123.1, 142 vs 152.3, and 189 vs 175.5 mm Hg). The deviations in mm Hg of each of these four analyzer models at each level of tonometered blood (their data) were paired with FCE deviations (our data) of the same models (44 pairs) and regression analysis was performed to compare the model biases of blood and FCE.