Reported Prevalences of Former Smokers in Survey Data: The Importance of Differential Mortality and Misclassification

Abstract

The purpose of this study was to determine whether the reported increase in the percentage of never smokers in the group of men born between 1924 and 1938, as observed in survey data in the Netherlands, can be attributed to differential mortality or to misclassification. In a spreadsheet program, the mortality experiences of the men born in 1924 and 1938 were simulated with the use of lifetable techniques. Due to differential mortality, the percentage of true never smokers would be expected to increase from 9% to 9.3% for men born in 1938 and from 9% to 10.3% for men born in 1924. In the survey, the percentage of people who reported that they had never smoked increased from 9% to 36% for the men born between 1924 and 1938. Differential mortality can only explain a very small part of the reported increase in the percentage of “never” smokers. Misclassification of former smokers as never smokers is the most likely cause for the reported increase. This is reason for concern, because even a small percentage of mis classified former smokers may bias the estimated relative risk of mortality associated with current smoking to a considerable extent. Because former smoking status cannot be measured correctly, the percentage of former smokers may be calculated by modeling the percentage of ever smokers. Am J Epidemiol 1994;140:52–7.