Bias

[reading time: 3 minutes]

In scientific research numerous causes can produce the inaccuracy of the results. Random factors can alter the results of a research but also other kinds of factors linked, for example, to the design, the conduct and the reporting of the study.

Many of these errors are called bias, defined as a systematic error or deviation (distorsion) from reality, in the results or in the inference.

The researchers have classified different types of bias, so much so that currently more than two hundred types have been cataloged, classified in different categories. A parameter of clinical relevance has been assigned to each type of bias in order to be able to easily interpret each bias during the conduct of a study1.

To mention only a few examples, general bias but also very specific ones have been identified, of both methodological and interpretative type, or linked to the modality of the reporting, that is the way the results of a study are described.

There are also the implicit biases, particularly difficult to recognise as they are unconscious and non intentional. Despite the best intentions of the researchers, these biases can have serious repercussions, for example, resulting in an unequal treatment of marginalized groups.

Particularly dangerous is the confirmation bias, that is the tendency of the human being to search and valorize the information that validates the theory they already believe in, while ignoring, belittling or avoiding the evidence to the contrary.

The biases can be, for example, associated to the different operations carried out by the researchers, like the review of the literature and the writing of the manuscript for publication, the design of the study protocol, the choice of the subject, the execution of the intervention, the measuring of the outcomes and the data analysis. For further information please refer to the volume edited by Francesco Cerritelli and Diego Lanaro.¹
The following examples are for mere illustrative purposes:

• selection bias: the methods used to select the population sample favor one of the compared groups, or they neglect a significant part of the population;
• performance or procedure bias: systematic differences in the intervention administered or introduction of factors other than the planned intervention;
• dropout bias: if a relevant number of subjects withdraw from the study while the experimentation is still ongoing the remaining sample size could no longer be representative of the entire population. In the case in which the dropouts happen in a different measure within the compared groups, the measurements of the results may result unbalanced;
• publication bias: only the results of studies which have reached acceptable conclusions are communicated selectively while the others are omitted.

In particular in the field of the qualitative studies some specific bias have been identified. Here are some examples:

• compliance bias, that is the tendency of a respondent to agree with anything stated by the interviewer or the moderator;
• social desirability bias, for which the subjects provide answers they consider more socially acceptable than others, with the intent to appear more compliant and to not be rejected.

Various strategies and techniques have been created to minimize the risk of bias in the research, however, due to their characteristics and to the context in which they may manifest, it is difficult to draw up a universal handbook to avoid all of them. The most important point is to recognize their presence in order to be able to manage them.

Only as a way of example, we mention the ROBIS², a guide elaborated to identify the biases in the systematic reviews. It is destined mainly to the developers of the guidelines, to the authors of the summaries, systematic reviews or critical evaluations. To reach an estimate of the risk of bias, four categories are evaluated (intervention, diagnosis, prognosis, etiology) implementing a three-step procedure.

Bibliography

1. Cerritelli F, Lanaro D. Elementi di ricerca in osteopatia e terapie manuali. Napoli: Edises, 2018.
2. Whiting P, Savović J, Higgins JP, Caldwell DM, Reeves BC, Shea B, Davies P, Kleijnen J, Churchill R; ROBIS group. ROBIS: A new tool to assess risk of bias in systematic reviews was developed. J Clin Epidemiol. 2016 Jan;69:225-34.