Predictive validity and measurement issues in documenting quit intentions in population surveillance studies.

Introduction: Discrete classification of smokers by intention to quit is desirable in many public health and clinical settings.

Methods: Two methodological studies examine measurement properties of measures of discrete-time intention to quit smoking used in population-based tobacco surveillance surveys: an ecological comparison of rates of positive intention in relation to the form of measure used and a prospective analysis examining predictive validity of self-reported quit intentions using multiple possible points of dichotomization of an ordinal measure of intention to quit. The prospective analysis used a repeated measures design and follow-up to 1 year for 2,047 smokers in the Ontario Tobacco Survey cohort.

Results: The estimated percent of smokers intending to quit was significantly higher using the Stages of Change intention measure, relative to another single question measure. Significant dose-response effects were found. The sooner one intended to quit the more likely one was to make an attempt or achieve at least 30 days abstinence in the next 6 months. Intending to quit in a month or later was not associated with cessation during follow-up among respondents without prior attempts. Examination of cutpoints revealed no value, which maximized both positive and negative prediction. Regardless of quit attempt history, greatest predictive validity was found where respondents stated that they had no intention at all.

Discussion: Measures of intentions quit smoking in specific time periods and expressed as dichotomies have limited psychometric properties but utility in applied research. Our findings suggest a possible measurement effect warranting caution in comparisons across studies.