Hierarchical structure in ADL and IADL: analytical assumptions and applications for clinicians and researchers.

The results of a Canadian study have shown that a set of 12 (I)ADL items did not meet the criteria of Guttman's scalogram program, questioning the assumption of hierarchical ordering. In this article, the hierarchical structure of (I)ADL items from the Canadian elderly sample is retested with another scalogram program, and both the epidemiological research and clinical applications of hierarchical scales are discussed. For retesting the hierarchical structure, Mokken's Stochastic Scale Program was used. When item distributions were not too skewed, the existence of one general dimension with two (correlated) subdimensions were clearly found. The analyses for scalability showed that, except in the case of extremely skewed item distributions, the data set met the criteria of Mokken's Stochastic Scale Program, indicating a hierarchical structure of the items. The scalability of a set of items is affected by the sample studied, the choice of response options for each item, the selection and number of items in the scale, and the type of scaling procedure. The practical assumptions of hierarchy should be restricted to program planning and epidemiological research purposes only. For clinical assessment, a finer grading is desirable and all questions on an ADL/IADL inventory should be asked for each individual.