Statistical justification for the use of composite scores in quality of life assessment

In clinical trials, the patient's quality of life (QOL) is usually assessed by means of a number of questions (or items). These items are often grouped to form subscales, composite scores (obtained by grouping a number of subscales), or a total score to measure different components of QOL, different dimensions of QOL, or overall QOL. In practice, it is not only easy to perform statistical analysis on these subscales or composite scores but also easy to interpret the results. The use of subscales and/or composite scores, however, lacks statistical justification. This paper proposes an approach using principal components analysis and factor analysis to select: 1. An appropriate number of composite scores, 2. The subscales to be grouped in each composite score, and 3. Optimal weights for obtaining composite scores. The proposed method is illustrated through an example concerning QOL assessment of an antihypertensive therapy recently published by Testa et al. (1).

Duke Scholars

Author Shein-Chung Chow Biostatistics & Bioinformatics, Division of Biostatistics