Email updates

Keep up to date with the latest news and content from BMC Medical Research Methodology and BioMed Central.

Open Access Highly Accessed Research article

Calculating confidence intervals for impact numbers

Mandy Hildebrandt12*, Ralf Bender13, Ulrich Gehrmann1 and Maria Blettner2

Author Affiliations

1 Department of Medical Biometry, Institute for Quality and Efficiency in Health Care (IQWiG), Dillenburger Str. 27, 51105 Cologne, Germany

2 Institute for Medical Biometry, Epidemiology and Informatics (IMBEI), Johannes-Gutenberg-University Mainz, 55101 Mainz, Germany

3 Medical Faculty of the University of Cologne, Joseph-Stelzmann-Str. 9, 50931 Cologne, Germany

For all author emails, please log on.

BMC Medical Research Methodology 2006, 6:32  doi:10.1186/1471-2288-6-32

Published: 12 July 2006

Abstract

Background

Standard effect measures such as risk difference and attributable risk are frequently used in epidemiological studies and public health research to describe the effect of exposures. Recently, so-called impact numbers have been proposed, which express the population impact of exposures in form of specific person or case numbers. To describe estimation uncertainty, it is necessary to calculate confidence intervals for these new effect measures. In this paper, we present methods to calculate confidence intervals for the new impact numbers in the situation of cohort studies.

Methods

Beside the exposure impact number (EIN), which is equivalent to the well-known number needed to treat (NNT), two other impact numbers are considered: the case impact number (CIN) and the exposed cases impact number (ECIN), which describe the number of cases (CIN) and the number of exposed cases (ECIN) with an outcome among whom one case is attributable to the exposure. The CIN and ECIN represent reciprocals of the population attributable risk (PAR) and the attributable fraction among the exposed (AFe), respectively. Thus, confidence intervals for these impact numbers can be calculated by inverting and exchanging the confidence limits of the PAR and AFe.

Examples

We considered a British and a Japanese cohort study that investigated the association between smoking and death from coronary heart disease (CHD) and between smoking and stroke, respectively. We used the reported death and disease rates and calculated impact numbers with corresponding 95% confidence intervals. In the British study, the CIN was 6.46, i.e. on average, of any 6 to 7 persons who died of CHD, one case was attributable to smoking with corresponding 95% confidence interval of [3.84, 20.36]. For the exposed cases, the results of ECIN = 2.64 with 95% confidence interval [1.76, 5.29] were obtained. In the Japanese study, the CIN was 6.67, i.e. on average, of the 6 to 7 persons who had a stroke, one case was attributable to smoking with corresponding 95% confidence interval of [3.80, 27.27]. For the exposed cases, the results of ECIN = 4.89 with 95% confidence interval of [2.86, 16.67] were obtained.

Conclusion

The consideration of impact numbers in epidemiological analyses provides additional information and helps the interpretation of study results, e.g. in public health research. In practical applications, it is necessary to describe estimation uncertainty. We have shown that the calculation of confidence intervals for the new impact numbers is possible by means of known methods for attributable risk measures. Therefore, estimated impact numbers should always be complemented by appropriate confidence intervals.