For the three manufacturers, statistical comparisons were made comparing each model with other models of that manufacturer. Nine ANOVAs were performed, each revealing statistically significant differences. The subsequent Scheffe tests yielded 53 N grades, 15 D grades, and 85 V grades. Thus, the mean values for nearly two thirds of the paired comparisons revealed a statistical model dissimilarity. After adding the results from the four additional ANOVAs (used because of visually apparent near mid-range graphic crossing), there were 38 N grades, 18 D grades, and 97 V grades, indicating an even greater incidence of dissimilarity than recognized by the earlier ANOVAs. Examples of the necessity for adding graphic review to the initial ANOVAs are given in Figs 4-6 and their legends. buy allegra
Table 2 shows statistical comparisons among models of the same manufacturer. For the seven COR models, only the 278 model is not dissimilar from the 288 model for all three analytes. The 170 and 178 models are not dissimilar for Pco2 and pH, nor are the 238 and 840 models. Otherwise all models differ from each other for at least two analytes. Comparing all COR models for Po2, average differences between models are <1 mm Hg for seven comparisons, 1 to 3 mm Hg for six comparisons, 3 to 5 mm Hg for seven comparisons, and 5 to 6 mm Hg for one comparison. For the full range of Pco2, average differences between models are <1 mm Hg for 19 comparisons and 1 to 2 mm Hg for 2 comparisons; while at Pco2 values above 40 mm Hg, average differences are <1 mm Hg for 11 comparisons, 1 to 2 mm Hg for 8 comparisons, and 2 to 3 mm Hg for 2 comparisons. For pH, average differences between models are <0.001 U for 4 comparisons, 0.001 to 0.003 U for 10 comparisons, 0.003 to 0.005 for 5 comparisons, and 0.005 to 0.006 for 2 comparisons.
Table 2—Dissimilarity Between Models of Blood Gas Analyzers by Manufacturer
|Chiron Corning Models|
|Instrumentation Laboratory Models|