A Report of the Management of Continuous Independent Variables in Logistic Regression Analysis in Medical and Public Health Journals in Thailand

Abstract

Background and Objective: The management of continuous independent variables before being inputted in logistic regression analysis can affect the independent variables in the final model.Improper grouping of continuous independent variables can cause losing the informational meaning, decreasing the predictive ability and increasing type 1 error. From these reasons, we were interested in studying the handling of these continuous independent variables in the logistic regression analysis in Thai medical and public health journals.

Methods: This descriptive study reviewed the management of the continuous independent variables in logistic regression analysis which reported in medical and public health journals in Thailand. The journal article selection was initiated by classifying the medical and public health journals that published in the database of Thai-Journal Citation Index Center into 4 groups. We chose the journal articles from those which used binary logistic regression analysis with at least 1 continuous independent variable into this study. There were 119 from a total of 1997 research articles, published between January 1, 2013 and December 31, 2015.

Results: Of 119 articles, 4 articles were reported an investigation of the linear relationship between the continuous independent variables and the logit but only 3 articles were reported the results of the test. There were 115 articles which did not mention investigating on linear relationship between continuous independent variables and the logit. Of these, there were 341 continuous independent variables which 71.8 percent were grouped, before an analysis, to the categorical variables and 59.6 percent did not have any referenced criteria.

Conclusion: There were rarely presented investigations of the linear relationship between the continuous independent variables and the logit and details on management of the continuous independent variables. Many continuous independent variables were grouped by unknown criteria.