Outcome measurement in behavioural sciences: a view on how to shift attention from means to individuals and why.

Outcome studies in biomedical research usually focus on testing mean changes across samples of subjects and, in so doing, often obscure changes in individuals. These changes, however, may be very informative in studies in which large or homogeneous samples are unavailable and mechanisms of action are still under scrutiny, as is often the case for trials on new exercise treatments. Single measurements of change may be valid, provided that the reliability of the measurement process (i.e. instrument and procedure combined) was ascertained in a dedicated study in which appropriate statistical models and algorithms were applied. This article summarizes the conceptual steps that allow the calculation of the minimal real difference (MRD, also known as the minimal detectable change), on the basis of the assessment of reliability using an intraclass correlation coefficient. Individuals who show changes beyond the MRD can be reliably considered as changed. A numerical example of an MRD calculation is shown, and the pros and cons of the mean change and the MRD approach are discussed.