A method is described which used the Rasch model for combining information across administrations of the same items in different tests to different student populations. The method was used in conjunction with the composite point biserial and mean square fit to provide more valid information about item quality. Over 700 items were administered to groups of students in grades 5 through 9. Each group was comprised of 150 to 250 students, and was located in one of three school districts. Using the level of the achievement group relative to the equal interval Rasch item calibration made it possible to superimpose all the administrations onto a single item characteristic curve (ICC). Findings included: (1) neither the point biserial nor the mean square fit was a reliable indicator of item quality, and neither appeared to be a substitute for the ICC; (2) the method of combining information across administrations significantly increased the amount and utility of information available about the actual performance of an item. (RL)