Using Bayesian evidence synthesis to quantify uncertainty in population trends in smoking behaviour.

Area of Science:

• Public Health
• Biostatistics
• Epidemiology

Background:

• Simulation models are crucial for tobacco control policy and disease burden estimation.
• Parameter uncertainty assessment is often incomplete in existing tobacco control models.
• Accurate uncertainty quantification is vital for reliable model-based forecasts.

Purpose of the Study:

• To demonstrate a Bayesian approach for model calibration that quantifies parameter uncertainty.
• To improve the accuracy of simulation models for population behaviors.
• To provide decision-makers with a clearer understanding of model uncertainty.

Main Methods:

• Developed and applied a Bayesian approach to calibrate a smoking behavior simulation model.
• Utilized Australian data to inform the model calibration process.
• Quantified parameter uncertainty using Bayesian inference.

Main Results:

• Observed an increasing smoking cessation rate in Australia since the late 20th century.
• In 2016, the smoking cessation rate was 4.7 quit-events per 100 person-years.
• Individuals quitting before age 30 transitioned to never-smoker status at approximately 2% annually.

Conclusions:

• The Bayesian approach effectively quantifies parameter uncertainty in tobacco control models.
• The method provides a clearer picture of uncertainty for policy-related decisions.
• This approach can serve as a blueprint for modeling other complex population behaviors.