Introduction
In epidemiology, statistical precision and information exactness are extremely vital. As such, accuracy on the associations between outcomes of interest and exposure should be free from any form of distortion. However, different factors may hinder information validity resulting in distorted exposure/outcome associations. This essay discusses bias and confounding as sources of lack of statistical precision.
Discussion
Bias
Bias in medical research can be termed as any systematic error, which when not restricted may have predictable impacts on results and, therefore, hinder statistical precision by either inflating or deflating outcomes.
There are various types of bias, including selection, detection, observer, reporting/recall, response, and publication biases (Lakshminarayan, 2016; Page, et al., 2016).
The observer may favor certain variables depending on prior affiliations or knowledge of the variables. The observation bias can be avoided by adopting scientific methodologies, especially blinding techniques.
Reporting bias results from inaccuracy related to the human incapacity to recall past events. The inaccuracy in recalling oftentimes results in untrue conclusions.
Selection bias results from deliberate or involuntary omission/deletion of key information/observation leading to incorrectness in association and untrue interpretation.
Outstanding comparable characteristics among the different types of bias is that they are all caused by human errors and have significant impacts on exposure-results associations. On the other hand, bias types differ in that some biases such as observer bias happen during data collection stages while others like the reporting bias and selection bias happen at the end of the study.
Confounding
On a hypothetical ground that both variable X and variable Y can independently cause Z, confounding will occur, for instance, when variable X causes Z in the presence of Y but inaccuracy results in the distortion of the association between X and Z. The variable Y in such a case is the confounder and it distorts the association between variable X and the result Z.
For an empirical example, taking food with a high acid content, and eating sugary foods are some of the causes of dental carriers. In carrying out epidemiological studies, the effects of the two variables on dental health should be separated. Otherwise, confounding may occur when the association between dental carriers and eating food with high acid levels are distorted by a high intake of sugary foods.
Certain conditions (which constitute confounding characteristics) must hold for confounding to occur (Skelly, Dettori, & Brodt, 2012). First, the confounder variable, such as variable Y in the example above, must have an autonomous association (risk factor) with the outcome, which is the result Z in the case above (Tilaki, 2012).
Second, confounding variables should never play any role in enhancing or inhibiting the activities of the exposure variable on the disease. As such, Y in the case above should not be an intermediate between X and the result Z.
For instance, the pH levels in a mouth inhibit or enhance the activities of acids in the teeth’ corrosion processes. A person with high pH levels reduces the acidity in food and, therefore, reduces the effects of acids. In such a case, the pH levels in a mouth cannot be termed as a confounding factor between intake of food with high acidity levels and dental carriers.
Third, the possible effects of confounding must be evident through the extent/degree of misrepresentation of association. As such, the extent/rate at which Y distorts the association between X and Z depends on the association between Y and Z.
Conclusion
It is evident that epidemiological studies are prone to incorrectness. Bias and confounding are some of the causes of errors that can hinder statistical precisions. There are different types of bias, including selection, detection, observer, reporting/recall, response, and publication. Moreover, it is apparent that certain conditions must hold for confounding to occur.
References
Lakshminarayan, N. (2016). In How Many Ways May a Health Research Go Wrong? Journal of ICDRO, 8(1), 8-13. Web.
Page, M. J., Higgins, J. P., Clayton, G., Sterne, J. A., Hróbjartsson, A., & Savović, J. (2016). Empirical Evidence of Study Design Biases in Randomized Trials: Systematic Review of Meta-Epidemiological Studies. PLoSE ONE, 1-26. Web.
Skelly, A. C., Dettori, J. R., & Brodt, E. D. (2012). Assessing Bias: the importance of considering confounding. Evidence-Based Spine Care-Journal, 3(1), 9-12. Web.
Tilaki, K. H. (2012). Methodological Issues of Confounding in Analytical Epidemiologic Studies. Caspian Journal Internal Medicine, 3(3), 488–495.