Why We Need Confidence Intervals

Abstract. The estimation approach to statistical analysis aims to quantify the effect of interest as an ‘‘estimate’ ’ of a clinically relevant quantity and to quantify the uncertainty in this estimate by means of a confidence interval (CI). As such, results expressed in this form are much more informative than results presented just as p values. This article focuses on the principles rather than the mathematics of CIs and discusses inter-pretation of CIs and some common misuses. CIs can be constructed for almost all analyses. They are especially useful for avoiding misinterpre-tation of nonsignificant results of small studies. CIs should be provided routinely for the main results of trials and observational studies. A dramatic change in published reports of medical research over the last 20 years has been the wide adoption of confidence intervals (CIs) as a standard part of the presentation of the quantitative results of studies. The steady increase in the use of CIs, which has been encouraged by the fact that some leadin