Methods: One hundred and eighteen participants were recruited into a RCT designed to compare the delivery of smoking cessation advice and nicotine replacement therapy, delivered by trained dental hygienists, with brief advice alone given at the recruitment visit by the consultant. Participants completed questionnaires at baseline eliciting smoking habits and readiness to quit. Potential predictors extracted from the baseline data included demographics, smoking habits, level of physiological dependency on nicotine, as defined by the Fagerstrom Test for Nicotine Dependence (FTND) and the Heaviness of Smoking Index (HSI), motivation and intention to quit (Stage of Change), together with the allocated group. After 6 months, the primary outcome of smoking status and the secondary outcomes of any quit attempts in the previous 3 months, quit attempt of °Ý1 week and reduction in number of cigarettes smoked were examined (data available for 71 participants) and classification trees used to try to identify significant factors in predicting primary and secondary outcomes.
Results: The primary outcome of quitting smoking was predicted for young females who had not previously had a prolonged quit attempt (°Ý1 week). This model only correctly identified 4/9 quitters. Models for predicting secondary outcomes of (a) any quit attempts and (b) reduced number of cigarettes smoked/day included the predictors: pack years, FTND, age, Stage of Change and previous quitting history. Both of the models had sensitivity, specificity and positive predictive values of between 74% and 90%. The derived model for predicting a prolonged quit attempt included more predictors (including allocated group) but had relatively poor sensitivity (53%).
Conclusion: Using classification trees identified important combinations of variables for attempting to predict quitting smoking and other positive changes in smoking behaviour.