Factors Associated with the Development of Recurrent Acute Otitis Media - the Significance of Choosing the Right and Accurate Dependent and Independent Variables in the Multivariate Analysis

Abstract

The results of univariate and multivariate statistical analysis were compared in identifying predictive factors of the development of recurrent acuta otitis media (RAOM) after an initial episode of acute otitis media (AOM) in 121 children. Univariate correlations between the development of RAOM and potential risk factors were analysed, and variables at p <0.10 were incorporated into the stepwise multiple logistic regression analysis. The comparisons between the univariate and multivariate analysis in identifying the predictive factors were made and the importance of changing the dependent variables in the multivariate analysis was analysed. It seems that univariate analysis is over-sensitive, but multivariate analysis is over-conservative in finding possible predictors of RAOM. Choosing the right and accurate dependent and independent variables in the multivariate analysis is extremely important, when this method is used.