Deviation of a single observation from the hypothetical true value. Guidelines for computing the magnitude of misclassification of risk individuals due to this deviation

Abstract

In clinical work as well as in intervention studies, decisions concerning further follow-up or treatment of individuals are often based on a single measurement of each variable considered. If the measured variable for a certain individual varies considerably with time due to biological intra-individual variation and methodological variation, we would intuitively be sceptical to such decisions. The intra-individual variations, i.e. a composition of both biological and methodological variations, could be regarded as noise in the measurements. A more precise definition of noise would be: 'Noise is defined as random variations from a person's true value (hypothetical) in a time interval for which the true value is assumed constant.' Methodological variation, in particular, will sometimes lead to misclassifications (with respect to a given cut-off point) of a single measurement versus a person's true value. In this paper we will present formulas and tables which will help the reader to quantify such misclassifications in his/her own material.

Key Words:

• intra-individual variation
• misclassification
• noise
• true value