Bayesian analysis on smoking prevalence and age of smoking initiation in Nepalese population: Findings from the STEPS national surveys

Introduction Smoking prevalence and age of smoking initiation (AOI) are two important variables for tobacco control programs. The study aimed to compare the prevalence of smoking between three WHO STEPS (STEPwise approach to surveillance) surveys and the AOI between males and females, using the Bayesian approach. Methods We made three null hypotheses (H0) at a 5% level of significance: the smoking prevalence in the 2019 WHO STEPS survey is similar to the previous two surveys (2008 vs 2019, and 2013 vs 2019); mean AOI between males and females is similar within 2019 survey. Both classical and Bayesian hypotheses were tested. In the Bayesian hypothesis, the Bayes factor (BF) and robust analyses were performed through the Markov chain simulation-based estimation method. Results We found no difference in smoking prevalence between the 2013 and 2019 surveys (BF0- =56.59). In contrast, there is strong evidence of the difference (BF0- =2.38×10-43) in smoking prevalence between the 2008 and 2019 surveys. Next, there is no evidence of a difference in the mean log AOI between males and females (BF01=12.54). The sequential analysis showed strong to very strong evidence for the H0 for AOI (BF10 1), respectively. Conclusions Our findings go beyond classical hypothesis testing on smoking behaviors and highlight the importance of the BF for the decision-making process in the tobacco control program. Further, the findings suggest that immediate efforts should be made to understand the underlying cause behind the stationary prevalence rate of the smoking population in the last five years.