Data source

GATS is a nationally-representative, cross-sectional survey of non-institutionalized adults aged ≥15 years, conducted using standardized protocols and data collection instruments⁸. Selected survey respondents live in households and are interviewed in-person by a trained interviewer with a questionnaire of tobacco-use and other tobacco control indicators. The Turkey GATS in 2008, 2012, and 2016 used a multi-stage, geographically clustered sample design; sample weights were calculated for each survey that allowed for the computation of weighted estimates that were representative of the overall population, as well as by gender⁸. The 2008 and 2012 sample clusters were also stratified by urban/rural residence; in the 2016 survey, selected clusters were determined without distinction between urban and rural residence. One adult from each participating household was randomly selected to complete the survey.

Overall, response rates were 90.9% in 2008, 90.1% in 2012, and 82.2% in 2016. Household response rates were 93.7%, 86.5% and 83.2% in 2008, 2012, and 2016, respectively; individual response rates were >97% each survey year. In total, the final sample size for each round of GATS Turkey was 9030 persons (weighted n=51.2 million) in 2008, 9851 persons (weighted n=54.5 million) in 2012, and 8760 persons (weighted n=60.9 million) in 2016.

Measures

The dependent variables were current tobacco smoking and interest in quitting tobacco smoking. Current tobacco smoking was defined as daily or less than daily smoking, based on the response to the question: ‘Do you currently smoke tobacco on a daily basis, less than daily, or not at all?’. Interest in quitting tobacco smoking was assessed among current tobacco smokers, and was defined as planning to quit within the next month, thinking about quitting within the next 12 months, or intending to quit someday, but not within the next 12 months; a response of either ‘not interested in quitting’ or ‘don't know’ were considered not to be interested in quitting smoking.

Independent variables included year of the survey, sex, age (15–24, 25–44, 45–64, and ≥65 years), residence (urban/rural), education level (no formal/less than primary; primary/less than secondary; secondary or high school; college/university degree or higher), wealth index (quintiles), awareness of anti-tobacco warnings and advertisements on media (health warning labels on cigarette packages; billboards; newspapers or magazines; radio; and local television), awareness of any advertisements promoting tobacco products, belief in smoking as a cause of severe health consequences, and living with other adults who smoke tobacco. The wealth index, categorized into quintile rankings from one (lowest) to five (highest), was constructed using principal component analysis based on household assets such as electricity, cell telephone, car, and refrigerator^9,10. Awareness of advertisements promoting tobacco products was defined as having seen or been exposed to advertisements promoting tobacco products, within the previous 30 days, in at least one location: stores selling cigarettes, television, radio, billboards, newspapers/magazines, the internet, public transportation, or on building walls. Severe health consequences included participants’ belief that smoking causes at least one of the following conditions: heart attack, cancer, stroke, or other serious illness. Living with other adults who smoke tobacco was defined based on the household questionnaire question: ‘Does this person currently smoke tobacco, including cigarettes, hand-rolled cigarettes, pipes, cigars and water pipes?’, which was asked for each adult living in the household. Year of the survey was included as a proxy for secular changes over time, including tobacco control measures and other environmental changes.

Statistical analysis

Each round of GATS data was adjusted for non-response and weighted to represent the national adult population aged ≥15 years. Descriptive statistics, including point estimates and 95% confidence intervals (CI) for the unadjusted prevalence of current tobacco smoking and interest in quitting smoking were calculated, both overall and stratified by each characteristic of interest. Relative change, defined as percent change in the unadjusted prevalence between each pair of survey years was calculated [e.g. relative change between 2012 and 2016 = (GATS 2016 estimate – GATS 2012 estimate)/GATS 2012 estimate]. Chi-squared tests were used to assess significant differences (p<0.05) in the relative change in unadjusted prevalence between survey years.

Logistic regression analyses were conducted using pooled data of all three rounds of GATS to evaluate predictors of tobacco smoking and interest in quitting smoking. We estimated unadjusted (PR) and adjusted (APR) prevalence ratios and 95% confidence intervals for predictors of current tobacco smoking and interest in quitting smoking¹¹. Adjusted models for each dependent variable included all a priori characteristics of interest.

Additionally, using multivariable logistic regression models that were stratified by each level of each characteristic of interest and adjusted for all other covariates, adjusted prevalence estimates and 95% confidence intervals were calculated; the marginal effects were calculated to represent the adjusted prevalence differences (APD) in the outcome measures between years. The unadjusted and adjusted PRs and APDs were considered as statistically significant if the two-side p value was <0.05.

All data preparation and analyses were conducted using SAS v9.4 and SAS-callable SUDAAN v11.0.1.

Current tobacco use

Overall, the estimated prevalence and weighted number of adults who were classified as current tobacco smokers was 31.2% (95% CI: 30.0–32.6; n=16.0 million), 27.1% (95% CI: 25.8–28.3; n=14.8 million), and 31.6% (95% CI: 30.2–33.0; n=19.2 million) in 2008, 2012, and 2016, respectively (Table 1).

In unadjusted analyses, a significant relative decrease in the prevalence of current tobacco smoking of >10% was observed for most demographic subgroups between 2008 and 2012, followed by varying significant increases in the prevalence between 2012 and 2016 (Table 1). Among men, the prevalence of current tobacco smoking significantly decreased by 13.5% between 2008 and 2012, from 47.9% (95% CI: 45.9–50.0) to 41.5% (95% CI: 39.4–43.5), with an overall relative decrease of 8% between 2008 and 2016 (44.1%, 95% CI: 42.2–46.0). Among women, a 13.7% decrease in the prevalence of current tobacco smoking between 2008 and 2012 was followed by a 46.4% relative increase between 2012 and 2016, resulting in an overall increase of approximately 26% between 2008 and 2016, from 15.2% (95% CI :14.0–16.5) to 19.2% (95% CI: 17.5–21.0). Statistically significant relative increases in the prevalence of current tobacco smoking were also noted among adults aged 15–24 years (26.1% increase) between 2008 and 2016. Significant relative decreases between 2008 and 2016 in the prevalence of current tobacco smoking were also noted among adults 25–44 years (7.3%), among those who completed primary or secondary/high school (13.1% and 11.3%, respectively), and among adults who were aware of anti-tobacco advertisements on billboards (23.6%), in newspapers or magazines (11.6%), or on the radio (14.8%), and among adults who lived with other adults who smoke tobacco (13.2%).

Interest in quitting

The percentage and number of adults who currently smoke tobacco and were interested in quitting smoking decreased significantly from 53.0% (95% CI: 50.1–55.8; n=8.5 million) in 2008 to 32.8% (95% CI: 30.2–35.5; n=6.2 million) in 2016 (Table 2). Between 2008 and 2012, relative changes in the interest in quitting smoking varied by demographic subgroup, with significant decreases noted among those aged 15–24 years (14.4%), rural residents (2%), among those with college or higher levels of education (11.7%), those in the lowest and highest wealth quintiles (9.1% and 2.7%, respectively), and among those with awareness of anti-tobacco advertisements on billboards or in newspapers/magazines (2.7% and 1.1%, respectively). Between 2012 and 2016, and, overall, between 2008 and 2016, interest in quitting smoking among current smokers significantly declined by more than 20% in every demographic subgroup.

Multivariate analyses

Adjusted for all other variables, the prevalence of tobacco smoking was significantly lower in 2012 (APR=0.77; 95% CI: 0.72–0.81) and in 2016 (APR=0.90; 95% CI: 0.85–0.96) than in 2008, among women compared to men (APR=0.38; 95% CI: 0.36–0.40), among residents of rural areas compared to their urban counterparts (APR=0.86; 95% CI: 0.81–0.91), among those in the highest two wealth quintiles compared to those in the lowest wealth quintile (Quintile 4: APR=0.90; 95% CI: 0.83–0.97; Quintile 5: APR=0.88; 95% CI: 0.80–0.96), among those that were aware of anti-tobacco advertisements on local television (APR=0.93; 95% CI 0.87–1.00), and among those that believed smoking causes severe health consequences compared to those who did not (APR=0.69; 95% CI: 0.60–0.79) (Table 3). The adjusted prevalence of tobacco smoking was higher among survey respondents with higher levels of education than no formal schooling/less than primary (primary/less than secondary: APR=1.15; 95% CI: 1.04–1.28; secondary or high school: APR=1.36; 95% CI: 1.22–1.52; college/university: APR=1.16; 95% CI: 1.02–1.31), among those who were aware of health warning labels on cigarette packages compared to those who did not (APR=1.58; 95% CI: 1.45–1.73), and among those that lived with other adults who smoke tobacco (APR=1.55; 95% CI: 1.49–1.62). Compared to young adults aged 15–24 years, adults aged 25–44 years (APR=1.63; 95% CI: 1.51–1.74) or 45–64 years (APR=1.36; 95% CI: 1.25–1.47) were more likely to smoke tobacco, while those ≥65 years were less likely to smoke tobacco (APR=0.63; 95% CI: 0.55–0.72).

In models assessing predictors of interest in quitting smoking among adults who currently smoke tobacco, there was no significant difference between 2008 and 2012, but the prevalence of interest in quitting smoking was significantly lower in 2016 than in 2008 (APR=0.66; 95% CI: 0.60–0.73) (Table 3). Other significant predictors of any interest in quitting smoking included age, with an increase in interest noted among persons 45–64 years compared to persons aged 15–24 years (APR=1.14; 95% CI: 1.03–1.25), living in a rural area (APR=1.13; 95% CI: 1.05– 1.22), completion of primary school or higher levels of education (primary: APR=1.21; 95% CI: 1.02–1.43; secondary: APR=1.25 (95% CI: 1.05–1.48; college or higher: APR=1.36; 95% CI: 1.13–1.63), awareness of anti-tobacco warnings on cigarette packages (APR=1.30; 95% CI: 1.12–1.51) and on television (APR=1.16; 95% CI: 1.04–1.30), and a belief that smoking causes severe health consequences (APR=1.57; 95% CI: 1.17–2.11).

Model-adjusted prevalence estimates and differences in the model-adjusted prevalence of current tobacco smoking between survey years are presented in Table 4. Overall, the model-prevalence of tobacco smoking was 33.9% (95% CI: 32.6–35.4) in 2008, 26.1% (95% CI: 24.9–27.3) in 2012, and 30.6% (95% CI: 29.2–30.2) in 2016. We observed a significant 7.9% (95% CI: -9.6 – -6.1) decrease in the prevalence of current tobacco smoking between 2008 and 2012 and a 4.5% (95% CI: 2.8–6.3) significant increase between 2012 and 2016, with a significant decrease between 2008 and 2016 (adjusted prevalence difference, APD= -3.3%; 95% CI: -5.3 – -1.4). A similar pattern of change was observed among most subgroups in the analysis.

Among adults who currently smoke tobacco, there was little change in interest in quitting smoking between 2008 and 2012, but a significant decline in interest between 2012 and 2016 in all subgroups (Table 5). Overall, the model-adjusted prevalence of interest in quitting smoking was 52.0% (95% CI: 49.0–55.0) in 2008, 53.9% (95% CI: 51.0–56.7) in 2012, and 34.5% (95% CI: 31.7–37.3) in 2016. We observed significantly less overall interest in quitting smoking in 2016 than in 2008 (APD= -17.5%; 95% CI: -21.8 – -13.4).

Strengths and limitations

The analysis was strengthened by use of GATS data, which are nationally representative and utilize standardized protocols and procedures. Using standard protocols and procedures increases the reliability of comparing estimates over time and with other countries. However, despite these strengths, the study is still subject to at least four limitations. First, this study did not assess the number and distribution of television channels with anti-tobacco media and/or content of the anti-tobacco media, which may vary depending on geographical area or among channels preferred by different population subgroups. Second, GATS did not collect information on internet or social media usage throughout the study period, so we could not assess the impact of these sources of tobacco promotion or anti-tobacco advertising on the prevalence of current tobacco smoking or interest in quitting smoking or respondent ability to purchase products online such as heated tobacco products and electronic cigarettes. The third limitation of this study was the use of self-reported survey data, which may have introduced social desirability reporting bias resulting in artificially lower prevalence of tobacco smoking or higher interest in quitting smoking. However, as tobacco smoking is relatively common in Turkey, social acceptance of smoking is also relatively high, and therefore unlikely to have substantially impacted the findings. Lastly, the data for this study are from repeated cross-sectional surveys; while these data can be used to analyze population-level changes over time, they cannot be used to assess individual-level changes.