Agreement between two tympanometers. A methodological study of instrument comparison

When old instruments are replaced by new ones, agreement between them becomes important when comparing measurements. Their precision, too, is important. A common, however inappropriate, statistical approach in these cases is correlation analysis. As an alternative, a simple statistical method is suggested, in which differences between measurements of either instrument are plotted against their means. The distribution of the differences is tested against 0 (one-sample t-test), and for good agreement P > 0.05. This statistical method was used in a group of 32 adults in which measurements taken with an older automatic tympanometer and a new digital one were compared. Static compliance and and gradient showed good agreement, while middle ear pressure showed a significant bias between instruments explained by phase delay. This disagreement was insignificant for clinical purposes, but may be important in scientific studies. The same statistical principle applies determining the precision of each instrument. The new digital instrument showed higher precision than the older one.