A confidence interval (CI), calculated using statistical methods, gives a range of likely values for the parameter of interest (e.g. average, percentage or mortality rate). There is usually random variation present, such as variation in the number of deaths each year, and it is useful to have a range for the parameter of interest as well as a single value to get an idea of the range of the likely values and the degree of variability. The usual CI calculated is the 95% CI, in which we are 95% confident that the interval obtained (from the sample) will contain the true underlying measure of interest (of your population of interest). The interval also takes into consideration the number of people on which the estimate is based, so that if there are many people surveyed the interval tends to be narrower (and therefore more useful). If the CI is wide then there is a high degree of uncertainty around the parameter of interest, and caution should be used when interpreting the findings. This often occurs when the estimate is based on a small number of events.
Also see: Bias, Causality, Confounding, Effect Modification, Interaction and Small Numbers.