Abstract
The incidence rate difference (IRD) is a parameter of interest in many medical studies. For example, in vaccine studies, it is interpreted as the vaccine-attributable reduction in disease incidence. This is an important parameter, because it shows the public health impact of an intervention. The IRD is difficult to estimate for various reasons, especially when there are quantitative covariates or the duration of follow-up is variable. In this paper, the authors propose an approach based on weighted least-squares regression for estimating the IRD. It is very easy to implement because it boils down to performing ordinary least-squares regression analysis of transformed variables. Furthermore, if the outcome events are repeatable, the authors propose that data on all events be analyzed instead of first events only. Four versions of the Huber-White robust standard error are considered for statistical inference. Simulation studies are used to examine the performance of the proposed method. In a variety of scenarios simulated, the method provides an unbiased estimate for the IRD, and the empirical coverage proportion of the 95% confidence interval is very close to the nominal level. The method is illustrated with data from a vaccine trial carried out in the Gambia in 2001-2004.