An Appraisal of Logistic Regression Models Reported in Published Articles in Two Thai Health Journals

Abstract

           This study was to determine the using and reporting of multiple logistic regression analysis in Journal of the Medical Association of Thailand and Journal of Health Science from 2002 to 2007. All original papers were searched by manual and electronic data based for the use of multiple logistic regression, matching keywords as follows; “logistic” “regression” and “multivariable”. Multiple logistic regression was used 6.1 percent (145/2,370), in these papers, reporting of logistic regression analysis as follows; used for the primary research question 48.3 percent (70/145) and sample size calculation was reported 12.9 percent (9/70), assessment of multicollinearity 3.5 percent (5/145), testing for interactions 2.8 percent (4/145), test of model and independent variables significantly 2.1 (3/145) and 95.2 percent (138/145) respectively, and goodness-of-fit measures 4.8 percent (7/145). In addition, no studies reported test of conformity of variables to a linear gradient. These problems would make the reported results potentially inaccurate, misleading, or difficult to interpret. Multiple logistic regression has become a standard statistical method in health and medical researches. It is recommended that researchers and statisticians pay greater attention to guidelines concerning the use and reporting of multiple logistic regression.

Key words: reporting, multiple logistic regression analysis