Computer technology of accounting of confounders in the risk assessment in comparative studies on the base of the method of standardization

Abstract

Here was performed an analysis of the accounting of the impact of concomitant variables (confounders), introducing a systematic error in the assessment of the impact of risk factors on the resulting variable. The analysis showed that standardization is an effective method for the reduction of the shift of risk assessment. In the work there is suggested an algorithm implementing the method of standardization based on stratification, providing for the minimization of the difference of distributions of confounders in groups on risk factors. To automate the standardization procedures there was developed a software available on the website of the Institute of Industrial Ecology, UB RAS. With the help of the developed software by numerically modeling there were determined conditions of the applicability of the method of standardization on the basis of stratification for the case of the normal distribution on the response and confounder and linear relationship between them. Comparison of results obtained with the help of the standardization with statistical methods (logistic regression and analysis of covariance) in solving the problem of human ecology, has shown that obtaining close results is possible if there will be met exactly conditions for the applicability of statistical methods. Standardization is less sensitive to violations of conditions of applicability.