Trends in the accuracy of self-reported cigarette smoking in Canada

INTRODUCTION: Self-reported data are commonly utilized to track trends in tobacco use and assess the impact of policy changes. Estimates derived from self-reported measures, especially for behaviours considered socially undesirable, can be influenced by reporting biases. Enhanced tobacco control measures may have exacerbated social desirability bias resulting in increased underreporting of tobacco use. Our aim was to assess the accuracy of self-reported cigarette smoking. METHODS: Using repeated cross-sectional data from six cycles of the Canadian Health Measures Survey (CHMS) conducted between 2007 and 2019, we first compared cigarette use prevalence (daily and occasional) based on self-reports and urinary cotinine concentrations and calculated sensitivity and specificity using urinary cotinine concentrations with a cut-point of 50 ng/ml as a measure of true cigarette use. Second, we used logistic regressions to examine if sensitivity and specificity varied by individual/household characteristics and time. RESULTS: Over the 12-year study period, the average sensitivity for self-reported smoking status was 89.8% (95%CI 85,5, 92.9) and ranged from 86.0% (95% CI 79.0, 91.0) to 92.7% (95%CI 89.1, 95.2). The average specificity was 97.5% (95% CI 96.0, 98.4) and ranged from 97.1 (95%CI 93.3, 98.8) to 98.6 (95%CI 97.6, 99.2). Sensitivity and specificity estimates did not vary appreciably by sex or household income, with one exception. Sensitivity estimates among younger respondents (12-19 years) were generally lower, not precisely estimated, and varied substantially between cycles. CONCLUSIONS: Self-reported responses from survey data can be used to provide accurate point estimates and trends in cigarette smoking prevalence in Canada.