Reliability, precision or reproducibility of the measurements. Methods of assessment, utility and applications in clinical practice

Abstract

Reliability (accuracy, consistency and reproducibility) is a psychometric property, which is related to the absence of measurement error, or, to the degree of consistency and stability of the scores obtained through successive measurement processes with the same instrument. Thus a greater variability of results will lower the accuracy or reliability of instrument used, fact that is transverse from the laboratory to the clinical practice. It is determined by applying the reliability coefficient, which is the correlation between the scores obtained by the subjects in two parallel forms of a test. Assuming that the two forms of the test are parallel (measure the same), the scores of the subjects under study should be the same in both applications. In this way, when the correlation is 1, the reliability or precision is maximum. On the other hand, reliability could be influenced by the observer (the one that measures), the measuring instrument (by that with which it is measured), and by the observed (by what is measured). Therefore, the variability of each of these components must be taken into account when planning the measurement of the variable under study, in such a way to reduce measurement biases as much as possible. The most common ways to determine reliability are the models of parallel forms, test-retest and two halves. This manuscript focuses on the concepts of measurement and the various statistical techniques used for this, as a step prior to application in the clinic. Therefore, the aim of this manuscript is to generate a consultation document related to the reliability or reproducibility of the measurement process.