Using lung cancer mortality to indirectly approximate smoking patterns in space

Smoking is the leading cause of lung cancer. Non-smoking factors have been associated with the disease. Existing Swiss survey data only capture the country partially and temporal coverage does not allow for a time lag between exposure to tobacco and lung cancer outbreak. Knowledge about the distribution of tobacco-use is essential to estimate its contribution to disease burden. Bayesian regression models were applied to estimate spatial smoking patterns. Data were provided from the Swiss Health Survey (14521 participants). Regression models with spatial random effects (SREs) were employed to obtain smoking proxies based on mortality rates and SREs adjusted for environmental exposures. Population attributable fractions were estimated to assess the burden of tobacco-use on lung cancer mortality. Correlation between observed smoking prevalence with smoking proxies was moderate and stronger in females. In the absence of sufficient survey data, smooth unadjusted mortality rates can be used to assess smoking patterns in Switzerland.