Sample size estimation in occupational mortality studies with use of confidence interval theory.

In occupational epidemiology, the standardized mortality ratio is firmly entrenched as an analytic tool. Sample size determinations have usually been made in a hypothesis testing framework. However, since interval estimation is the usual goal of analysis, it may be more appropriate to use confidence interval theory to estimate the required sample size. This paper gives the method and some results for determining the required "expected number" of deaths, based on specifications concerning the confidence interval for the true standardized mortality ratio.