Some applications of latent trait analysis to the measurement of ADL.

Abstract

The use of latent trait methods for detection of biased items used in scales is discussed and two examples given. In the first, items measuring functional impairment in elderly community residents are tested for possible sex bias, and items predicted on the basis of clinical judgment to be clearly sex-biased are correctly identified. In the second example, taken from a cross-national study of elderly residents in long-stay institutions in New York and London, scale items suspected of bias due to interviewer variability and to cross-national differences in institutional environments are identified. It is shown that estimated rates of impairment are affected by presence of variant (biased) items. We argue that latent trait methods are useful for identifying biased items and may have wide application in gerontological research.