Alternative data treatment principles for categorical ADL data.

Abstract

Scaling methodology represents a problem in assessments of activities of daily living (ADL) and little is known about how the results of these assessments are affected by data treatment principles and statistical methods. The aims of this paper are to: (i) describe alternative ways of transforming a response pattern on ADL into a single number; and (ii) to present and compare different ways of analysing both changes in ADL capacity from one occasion to another and also differences in ADL between one group and another. Three datasets based on assessments with the ADL Staircase were used. Four different data treatment principles were described and the development of a novel principle to transform response patterns into ranks was put forward. Thereafter, different paired-data cases and two-sample cases were analysed, using different statistical standard methods to explore possible variations in results. The results demonstrated a few marked differences among P values, no matter which data treatment principle or statistical method was used. That is, different principles and methods yield similar results in terms of P values, although there are important differences as regards selection bias. Principles and methods respecting the ordinal character of ADL data encourage the use of non-parametric methods and the novel rank principle presented here is a useful alternative.