Multiple logistic regression models—what are they?

Abstract

Multiple logistic regression provides a way of fitting statistical models to predict a categorical outcome or dependent variable from a combination of explanatory variables. This allows researchers to investigate the association between the outcome variable and a number of explanatory variables using odds ratios, which we covered in a previous article (Campbell, 2004). The simplest form is what is usually known as a binary logistic regression model, where the outcome variable has only two values. There are more complicated versions for outcome variables with more than two values—see, e.g., Kirkwood and Sterne (2003, pp. 212–213). Binary logistic regression is widely used in healthcare research, as many outcome variables have or can be interpreted as having two categories, such as ‘yes’ or ‘no’, ‘present’ or ‘absent’, or ‘case’ or ‘not a case’. For example, the outcome variable in Fabian et al. is non-attendance at childbirth and parenthood education classes, where ‘yes’ corresponds to ‘nonattendance’ and ‘no’ to attendance. The authors