Why questionnaire scores are not measures : a question-raising article

Any person is provided by characteristics that can be neither located in body parts nor directly observed (so-called "latent" variables): these may be behaviors, attitudes, perceptions, motor and cognitive skills, knowledge, emotions, and the like. PRM frequently faces variables of this kind, the target of many interventions. Latent variables can only be observed through representative behaviors (e.g., walking for independence, moaning for pain, social isolation for depression, etc.). To measure them, behaviors are often listed and summated as items in cumulative questionnaires ("scales"). Questionnaires ultimately provide observations ("raw scores") with the aspect of numbers. Unfortunately, they are only a rough and often misleading approximation to true measures for various reasons. Measures should satisfy the same measurement axioms of physical sciences. In the present article, the flaws hidden in questionnaires' scores are summarised, and their consequences in outcome assessment are highlighted. The report should inspire a critical attitude in the readers and foster the interest in modern item response theory, with reference to Rasch analysis.