Smoking is a leading cause of death and many fatal diseases in Australia, killing an estimated 15,531 people annually. Smoking and cessation are seen as dynamic processes where people quit, relapse, and quit again, often with repeated cycles across many years. This study demonstrates how to apply hidden Markov modeling (HMM) methodology to examine the dynamic nature of smoking and cessation in Australian adults. The HMM is widely used analytic technique for examining the probability of movement/transition through (discrete) stages over a specific time interval. This approach enabled us to assess smoking (and cessation) transitions simultaneously over 6 year time period. The data were from the Household, Income, and Labour Dynamics Australia (HILDA) Survey. Three waves of the HILDA survey were utilised in the study (wave 7: 2007, wave 9: 2009; and wave 11: 2011). Study samples were restricted to those who responded to tobacco use questions at baseline (N = 11,235).

We constructed a non-homogeneous first-order hidden Markov model to estimate the probability of moving from a current status of smoking or non-smoking to a later smoking status (e.g., transition from current smoker status into non-smoker status). In addition, we identified potential factors predicting the transition probabilities, including time-varying (psychological distress) and time-invariant covariates (gender, age, and country of birth). The impacts of these covariates on the transition probabilities were examined using a logit link function.

We found that non-smokers were more likely to retain across time than smokers (W7 to W9 and W9 to W11). More specifically, those who self-reported their non-smoking status had approximately 97% chance of being in the same status, whereas those who reported they are current smokers had about 80% chance of being in the same status. We also found that considerable transitions from the Current Smoker status to the Non-Smoker status (19.7% to 20.9%), indicating that there was about a 20% chance of transitioning from the Current Smoker status to the Non-Smoker status. Finally, logistic regression analyses revealed that age was a significant predictor of the transition from the Current Smoker (W9) to the Non-Smoker status (W11) (OR = .990, p = .016). However psychological distress was not a significant predictor of the transition probabilities, whilst it was significantly associated with smoking status at each time-point (W7/W9/W11). These findings provide a novel insight into the nature of smoking transitions in Australia. Further research to understand predictors of these transitions will inform anti-smoking strategies.

References
Begg, S, Vos, T, Stevenson, C, Stanley, L, Lopez, A 2007, ‘The burden of disease and injury in Australia 2003’, cat no. PHE 82, Australian Institute of Health and Welfare, Canberra.

Yeh, H, Ellerbeck, EF, & Mahnken, JD, 2012, ‘Simultaneous evaluation of abstinence and relapse using a Markov chain model in smokers enrolled in a two-year randomized trial’, BMC Medical Research Methodology, vol. 12, pp. 95-104.