Lifetime smoking patterns: A transition probability analysis

An analytical framework for studying the population dynamics and lifetime patterns of cigarette smoking is developed. The key idea is that changes in the distribution of the population among smoking and nonsmoking categories can be viewed as an age-sex-dependent Markov process and that transition probability matrices for successive ages can be linked to generate the complete smoking history of a cohort as it moves along its age path. The framework is used to analyse data from 1981 and 1986 Canadian surveys of smoking habits: transition probabilities are estimated from the data and some interesting results are obtained that could not be obtained by more conventional forms of analysis. The movements of the population among smoking/nonsmoking categories are examined and the lifetime smoking patterns implicit in the 1981–1986 data are derived and discussed. It is emphasized that a lifetime perspective on smoking behaviour is important for assessing health risks because smoking effects tend to be cumulative, especially with regard to the respiratory system.