Classification of Bias

Bias is an important factor to be considered while performing research. It can be defined as a structured variation in the actual results of a study or an analysis. The role of bias is to signify error in research. A study or research is highly prone to error. The errors are of two types they are random error and systematic error. The systematic error gives rise to bias. 

The direction of bias:

Bias can be referred to as any activity at any phase of analysis that produces results or inferences that are substantially distinct from the truth. The different types of biases based on their direction are:

1. Positive bias: In this case, the value, effect or data observed are greater than the actual or casual data. It can be typically termed as away from the null. Here, the perceived value is nearer to 1.0 than the actual value.
2. Negative bias: In this case, the value, effect or data observed are lesser than the actual or casual data. It can be typically termed null.  Here, the perceived value is lower to 1.0 than the actual value.

Types of Bias:

The types of biases in epidemiology can be categorised into three major types. They are

• Selection bias
• Information bias
• Confounding

Selection Bias:

This is due to the error that occurred while selecting a group of the population for analysis and study. The major false expression that occurs here is the sample population chosen for the study will not resemble or match the target population. This type of bias can be applied at any stage of research. They are divided into four groups they are:

1. Disproportionate statement of the eligible population: It can be referred to as ascertainment bias. It arises whenever the category of patients examined does not correspond to the incidents in the population. The different types of biases under this are:
2. Spectrum bias
3. Healthcare access bias
4. Length-biased sampling
5. Neyman bias
6. Competing risks
7. Survivor treatment selection bias

Some of the specific biases under selection bias are Berkson’s bias, Inclusion bias, Exclusion bias, Relative control bias, Matching and Friend control bias.

• Inadequate sampling frame precision:

Non-random sampling bias possesses a large portion of this type of bias. The parameters used in the sampling the targe are misdone or miscalculated which results in the deviation from the target population. The common bias which falls under this are:

• Publication bias
• Citation bias
• Dissemination bias
• Post hoc analysis.

• Irregular diagnostic practices in the target population

This occurs as a result of deviations in the diagnosis or detection of diseases. The bias that falls under this type of bias is:

• Detection bias
• Diagnostic suspicion bias
• Unmasking-detection signal-bias
• Mimicry bias

• Study process execution:

This is a bias that occurs during the process of analysis or study. During this process, the 3 most prevalent biases are:

• Non-response bias
• Losses to follow up
• Missing information in multivariable analysis

Information bias:

Information bias occurs during the collection and processing of data. The three different types of biases that come under information bias are:

1. Misclassification bias: This lacks specificity in processing the data. It is further classified into differential misclassification bias and non-differential misclassification bias. The most common types of biases that exhibit misclassification are reporting bias, recall bias, observer bias and detection bias.
2. Ecological fallacy: It occurs when a result obtained for a group is calibrated for a single individual of that group.
3. Regression to the mean: The bias called regression dilution bias is highly correlated with this. 
4. Other information biases are the Hawthorne effect, Lead time bias, Workup bias, Protopathic bias, Temporal ambiguity and the Will Rogers phenomenon.

Confounding:

Confounding can also be called integration or mixing effect. Here, the variables, effects, and risk factors are mixed. It is of two types. They are:

1. Confounding by group: It is generated or observed in an ecological study when the incidence of sensitivity in each group is associated with disease risk in non-exposed individuals of the same population.
2. Confounding by indication: If an intervention is prompted by a presumed high risk, a poor prognosis, or just certain symptoms, this bias is produced.

Specific biases in the trial:

The specific biases in the trial are as follows:

• Allocation of intervention bias
• Compliance bias
• Contamination bias
• Differential maturing
• Lack of intention to treat analysis

Conclusion

Bias is a deviation or error in data collection, evaluation, assessment, recording, presentation or review which directly affects the result of research. This can reduce the quality of a study or research. Various strategies and methods are developed to reduce and control the bias. New studies are also made to optimise the errors and bias in research that provides clear and efficient results. Works are also carried out in reducing the misconceptions about the strategies. Controlling, mitigating and reducing the bias can provide quality study results.