Also referred to as observational bias
. A Dictionary of Epidemiology
, sponsored by the International Epidemiological Association
The International Epidemiological Association is an organization that aims at promoting epidemiologic research worldwide. It is run by an executive council and regional councilors from all WHO regions. The organization hosts World Conferences in Epidemiology every three years and regional...
, defines this as the following:
“1. A flaw in measuring exposure, covariate, or outcome variables that results in
different quality (accuracy) of information between comparison groups. The occurrence of information biases may not
be independent of the occurrence of selection bias
Selection bias is a statistical bias in which there is an error in choosing the individuals or groups to take part in a scientific study. It is sometimes referred to as the selection effect. The term "selection bias" most often refers to the distortion of a statistical analysis, resulting from the...
2. Bias in an estimate arising from measurement errors.”
Information bias, essentially, refers to bias arising from measurement error.
Misclassification thus refers to measurement error. There are two types of misclassification in epidemiological research: non differential
misclassification and differential
Non differential misclassification
Non differential misclassification is when all classes, groups, or categories of a variable (whether exposure, outcome, or covariate) have the same error rate or probability of being misclassified for all study subjects. The traditional assumption has been that, in the case of binary or dichotomous variables, this would result in an underestimate
of the hypothesized relationship between exposure and outcome. This has more recently been challenged however in that results of individual studies represent a single estimate and not the average of repeated measurements and thus can be farther (or nearer) from the null value (i.e. zero) than the true value.
Differential misclassification occurs when the error rate or probability of being misclassified differs across groups of study subjects. For example, the accuracy of blood pressure measurement may be lower for heavier than for lighter study subjects, or a study of elderly persons may find that reports from elderly persons with dementia are less reliable than those without dementia. The effect(s) of such misclassification can vary from an overestimation to an underestimation of the true value. Statisticians have developed methods to adjust for this type of bias, which may assist somewhat in compensating for this problem when known and when it is quantifiable.