Smoke or fog? The usefulness of retrospectively reported information about smoking.

Abstract

To investigate the reliability and validity of retrospectively reported information on smoking. Nationally representative retrospective data from longitudinal surveys and contemporaneous data from repeated cross-sectional surveys were used. Adult respondents to three of the four samples of the National Longitudinal Surveys Original Cohort 1966-68; the National Longitudinal Survey of Youth 1979; and various waves of the US National Health Interview Survey. Reliability was investigated by calculating kappa statistics for repeated measures of ever-smoking and annual-smoking status. Validity was investigated by comparing smoking prevalence rates generated by retrospective data with contemporaneously measured rates. Kappa statistics indicated the repeated measures of ever-smoking status show substantial agreement; repeated measures of annual-smoking status show moderate agreement. Retrospective reports on smoking behavior produced prevalence rates that match reasonably well with those from contemporaneous reports of smoking behavior. Retrospective data on smoking can be an important resource for tobacco addiction research.