Mortality rate and confidence interval estimation in humanitarian emergencies

Surveys are conducted frequently in humanitarian emergencies to assess the health status of the population. Most often, they employ complex sample designs, such as cluster sampling. Mortality is an indicator commonly estimated in such surveys. Confidence limits provide information on the precision of the estimate and it is important to ensure that confidence limits for a mortality rate account for the survey design and utilise an acceptable methodology. This paper describes the calculation of confidence limits for mortality rates from surveys using complex sampling designs and a variety of software programmes and methods. It contains an example that makes use of the SAS, SPSS, and Epi Info software programmes. Of the three confidence interval methods examined--the ratio command approach, the modified rate approach, and the modified proportion approach--the paper recommends the ratio command approach to estimate mortality rates with confidence limits.