Multivariable analysis is a combination of various statistical methods that are designed to test hypotheses, the relationship between the factors studied in a research, and certain features that do not have a quantitative description. This technique allows one to determine the degree of interaction of factors and their influence on certain processes. The method of multivariable analysis is most often used to determine the relationship between a continuous quantitative variable and nominal qualitative features. In essence, this method is a test of various hypotheses about the equality of various arithmetic samples. Thus, it can also be considered as a criterion for comparing several samples. However, the results will be identical if only two elements are used for comparison. Using the t-test, this analysis method allows researchers to study the problem of hypotheses in more detail than any other known method.
Multivariable analysis is especially useful for epidemiological research, as the latter requires the study of many different factors that influence the spread of a disease and its mortality rates. For a detailed epidemiological study, the researcher should assess those factors that control the circumstances of the experiment and affect the final result. Factors can also mean methods and levels of processing values ​​that characterize a specific manifestation of a certain condition. Thus, sometimes situations in epidemiology may arise when it is necessary to compare two or more different samples with each other. In this case, it would be most logical to apply a multivariate correlation-regression analysis based on the study of the hypothesis and the relationship of various factors in the degree of regression.
In epidemiological studies, researchers must account for biases and factors that could possibly go unnoticed to avoid false results due to the multivariable and probability-based nature of the correlations between a disease and its causes (Hill, 1965). To avoid such mistakes, Bradford Hill (1965) created criteria for causality and association that provide guidelines for the correct study of epidemiological problems. Applying these criteria to multivariable analysis, one can see that it pertains to Hill’s standards of temporality, plausibility, and consistency. Thus, it becomes possible to conduct epidemiological studies that review the influence of different factors on the process of spread of a disease, its mortality rates, or any other related feature (Hill, 1965). Multivariable analysis minimizes bias and possibility of false results through combining quantitative and qualitative variables.
Seeing as the epidemiological research studies features of certain populations and their relevance to the spread of disease, how it manifests, and mortality rates, it is incredibly important to account for potential mistakes. The proper research design with correctly chosen study method allows to minimize the risks of mistakes and false results, thus it is a crucial review point in criticizing an epidemiological study. Understanding how the reviewed study type might affect the processes of data gathering, processing, and analyzing can help the reviewer determine whether the research is vulnerable to bias and confounding.
Biases can occur during the information collection and sampling, as well as due to the interfering factors. However, biases at any stage lead to the result that all or a significant part of the compared groups differ not only in the analyzed factor, but also in some unaccounted factors. This may seriously affect the final results; therefore, a scientific critique of a research must assess how the bias is evaluated and tackled to determine whether the study was conducted correctly. Meanwhile, evaluating confounder in a study is important to determine and correctly assess the risks a specific factor has on the population’s health. Thus, it is important to control confounding in order to isolate the effect of a particular threat, such as a food additive, pesticide, or a new drug. For prospective studies, it is difficult to recruit and sample volunteers with similar background, and in cross-sectional and repeated studies, dependent variables can behave in similar ways for dissimilar reasons. Due to the inability to control the quality of volunteers, confounding is a particular problem for epidemiological studies, and it requires thorough evaluation.
Reference
Hill, A. B. (1965). The environment and disease: Association or causation?. Proceedings of the Royal Society of Medicine, 58(5), 295–300.