Technical note: The risk ratio, an alternative to the odds ratio for estimating the association between multiple risk factors and a dichotomous outcome

Publisher Website
Google Scholar
Abstract
The objectives were (1) to explain why the risk ratio (RR) is an appropriate measure of association when the outcome of interest is dichotomous (e.g., displaced abomasum or no displaced abomasum) in both cohort studies and randomized trials; and (2) to outline an applied method for estimating the RR using currently available software. Interest in the association between multiple risk factors and a yes or no outcome is very common in the dairy industry; historically, logistic regression, which reports odds ratios (OR), was the method available in common statistical packages to evaluate this kind of association. However, the OR can overestimate the magnitude of the response in cohort studies and randomized trials when the outcome frequency is large. In addition, the interpretation of odds is not intuitive; fortunately, recent advances in statistical software have allowed the estimation of the RR. Because SAS software (SAS Institute Inc., Cary, NC) is commonly used to analyze data, this technical note outlines the basic programming code that may be used to estimate the RR from raw data. Example data from a prospective cohort study was used to compare the OR and RR of developing a displaced abomasum or ketosis or metritis based on multiple predictors, their interaction, and a random effect (e.g., herd).