Support for a point-of-sale cigarette display ban among smokers: findings from the international tobacco control (ITC) Netherlands survey

Longitudinal data were obtained from the ITC Netherlands Survey. The surveys were administrated via the internet by the research firm Kantar Public (previously TNS NIPO) which used a quota sample of respondents from a probability-based web database to retrieve a representative sample of Dutch smokers aged 15 years and older [30]. Respondents were categorized as smokers if they were currently smoking at least monthly and if they had smoked at least 100 cigarettes in their lifetime [31]. Sampling weights and tailored replenishment samples ensured representativeness by compensating for attrition effects [32]. Respondents received incentives for participation in the form of points for each answered question, which could be exchanged for gift certificates.

For our analyses, we used data from survey wave 4 (May to June 2010; N = 2060), wave 5 (May to June 2011; N = 2101), wave 6 (May to June 2012; N = 2022), wave 7 (May to June 2013; N = 1970), wave 8 (May to June 2014; N = 2008), and wave 9 (November to December 2015; N = 1720). Attrition ranged from 17.1 to 23.9% between survey waves. We excluded quitters from our analyses. Exclusion due to smoking cessation ranged from 12.6 to 25.6% between survey waves. Figure 1 shows the number of smokers of the initial cohort and of the replenishment samples that remained in the study for each wave, leading to a total number of smokers per survey wave. To answer research question 1, all smokers from survey waves 4-9 were included in the analyses.

In the analyses for research question 2 we used the two most recent survey waves (waves 8 and 9). Respondents were included in the analyses if they were categorized as a smoker in both waves. In wave 9, 1017 of 1565 smokers from wave 8 participated. Respondents were excluded if they had more than five missing values on the independent variables (out of 22 variables) or if they had not filled in the outcome variable on wave 9. This left 844 smokers eligible for the analyses of research question 2.

Support for a PoS cigarette display ban was measured by asking ‘Do you support complete bans on displays of cigarettes inside shops and stores?’ [25]. This measure used a three-point scale with the response options (0) not at all, (1) somewhat, and (2) a lot. This variable was dichotomized for the analyses by combining the last two response options.

Respondents were asked whether they were (1) male or (2) female and were classified in one of the following age groups: (1) 15-24, (2) 25-39, (3) 40-45, or (4) 55 years and older. Education was divided into three categories: (1) low (Primary education and lower pre-vocational secondary education), (2) moderate (Middle pre-vocational secondary education and secondary vocational education), and (3) high (Senior general secondary education, (pre-) university education and higher professional education).

Knowledge about the health risks of smoking was measured by asking eight questions. The following format was used: ‘The following are a few health effects and diseases. Based on what you know or believe, does smoking cause ...’ [33]. This question was asked for heart disease, impotence in male smokers, lung cancer, blindness, mouth and throat cancer, stroke, lung-cancer in non-smokers from secondhand smoke, and heart disease in non-smokers from secondhand smoke (α in wave 8 = 0.84). Respondents could answer with (1) yes, (0) no, and (0) don’t know.

Noticing anti-tobacco information was assessed by asking if respondents had noticed advertising or information about the dangers of smoking or advertising that encouraged quitting in the last 6 months [34]. There were five answering categories ranging between (1) never and (5) very often.

Risk perception was assessed by measuring perceived susceptibility and perceived severity. Perceived susceptibility was obtained by the question ‘If you continue to smoke the amount you do now, how likely do you think it is that you will develop lung cancer in the future?’ The scale ranged between (1) very low and (5) very high. Additionally, all smokers were asked about perceived severity via the question ‘If you develop lung problems due to smoking, how serious would you find this?’ The scale ranged between (1) not at all serious and (5) extremely serious.

Attitude towards quitting smoking was asked via the questions ‘If you quit smoking within the next 6 months, this would be…’: (1) very foolish to (5) very wise and sensible, (1) very disagreeable to (5) very agreeable, and (1) very negative to (5) very positive. The three answers were summed into one attitude score (α in wave 8 = 0.83).

Social norms for quitting smoking were assessed by asking ‘Thinking about the people who are important to you, how do you think most of them would feel about your quitting smoking within the next 6 months?’ [35]. Based on a five-point scale respondents could (1) strongly disapprove to (5) strongly approve.

Self-efficacy to quit smoking was measured by asking how sure they were to succeed if they decided to give up smoking completely in the next 6 months [29]. This was measured on a five-point scale ranging from (1) not at all sure to (5) extremely sure. Smokers were also asked how easy or hard it would be to quit smoking if they wanted to. Answering categories ranged between (1) extremely difficult and (5) not at all difficult. The two answers were summed into one self-efficacy score (α in wave 8 = 0.72).

Intention to quit smoking was measured by asking if the respondents were planning to quit smoking within the next 6 months [35, 36]. A five-point scale ranging between (1) very unlikely and (5) very likely was used.

Smoking-related covariates were level of addiction to tobacco, and ever having tried to quit smoking. The Heaviness of Smoking Index (HSI) was used to measure the level of addiction to tobacco. The HSI is based on the time before smoking the first cigarette of the day (61 + min, 31-60 min, 6-30 min, 5 min or less) and on the number of cigarettes smoked per day (0-10, 11-20, 21-30, 31+). This measure ranges between 0 and 6, with a higher score indicating a higher level of addiction to tobacco [37]. Also, the respondents were asked whether they had ever tried to quit smoking, (1) yes or (2) no.

Data were analyzed using the statistical software program SPSS 23. All statistical estimates and tests presented were weighted for gender and age [38].

Trends in outcome measure (research question 1) were tested with Generalized Estimating Equations (GEE) analyses [39], while controlling for gender, age, ever having made a quit attempt, education, and level of addiction to tobacco. Missing data values on these variables were imputed multiple times using the full conditional specification method (with linear regression for scalar covariates) [40]. The number of imputations was set according to the percentage of cases that were incomplete [41]. Moreover, starting from this number of imputations, we systematically increased the number of imputations by 10 until results hardly differed. This resulted in 80 imputations. Also, we adjusted for the time a respondent participated in the cohort (time-in-sample) since this may influence responses [42]. The dependent variable was dichotomous and therefore the binomial distribution and the logit link was used [43]. Survey wave was the repeated measure variable. The interaction between educational level and survey wave was assessed in a separate analysis (research question 3).

T-tests and Chi-square analyses were run to test differences in independent variables (measured in 2014) between supportive and non-supportive smokers of a cigarette display ban (measured in 2015). The associations between independent variables on wave 8 with support for a PoS cigarette display ban on wave 9 (research question 2) were examined by performing multiple logistic regression analyses, while controlling for gender, education and age. To examine total effects of predictor variables, we should not correct for possible mediators or descending proxies thereof [44]. For this reason, the variables were added in successive steps: first the awareness variables were entered (knowledge about the health risks of smoking, perceived severity, perceived susceptibility, and noticing anti-tobacco information), second the motivational variables (attitude, social norms and self-efficacy), and in the last step intention to quit smoking. The analysis results for the variables as reported, are those for the step in which they were first entered into the analysis model. Only 477 out of the eligible 844 respondents completed all independent variables. For the remaining 367 respondents missing values were filled in by multiple imputation. The 884 subjects available for the analysis therefore contained no persons with missing values on the outcome, but with possibly missing values on predictor variables that were filled in by multiple imputation. This method improves the statistical power by increasing the sample size for the multiple logistic regression [40]. Following the same procedure as for the trend analysis, the number of imputations was set at 100. The imputed values for the outcome variable were not employed in the analyses, since excluding cases with missing values on the outcome variable yields more stable estimates [45]. Because the variables education and age each were represented in the analyses through multiple dummy variables and SPSS provided only p-values for these separate dummy variables when analyzing multiple imputed datasets, a correction for multiple testing was applied. More specifically, for all tests involving the variables education and age, the Holm correction was applied to the significance level α = 0.05 [46]. Interactions between educational level and independent variables were added in a separate regression analysis to determine whether there were differences between educational levels in the predictors for support for a PoS cigarette display ban (research question 3). This analysis was also done and reported with the same steps as delineated above, but then also adding each time the interaction terms with education.

Table 1 shows characteristics of smokers between 2010 and 2015. The smokers’ educational level increased over the years.

Table 2 shows that support for a PoS display ban of cigarettes increased from 28.9% in 2010 to 42.5% in 2015. A GEE analysis confirmed the overall linear trend (Odds Ratio (OR) = 1.40, 95% Confidence interval (CI): 1.25, 1.58).

A separate GEE analysis including interaction terms of wave by educational level revealed that the support for a PoS cigarette display ban increased less among moderately educated smokers than among highly educated smokers (OR of interaction = 0.87, 95% CI: 0.78, 0.98, p = 0.011 < α_adjusted = 0.017). This result can also be seen in Table 2 that shows that the support among moderately educated smokers increased from 2010 to 2015 by 7.6% whereas the support for highly educated smokers increased by 20.1%. No significant interactions were found between wave and low versus moderate education (OR = 1.07, 95% CI: 0.97, 1.18, p = 0.155 > α_adjusted = 0.025), nor between wave and low versus high education (OR = 0.94, 95% CI: 0.84, 1,05, p = 0.238 > α_adjusted = 0.05).

The bivariate analyses from Table 3 show that smokers who support a PoS cigarette display had more knowledge about the health risks of smoking, had a higher perceived severity of smoking-related lung problems, had a more positive attitude towards quitting smoking, perceived stronger social norms for quitting, had a higher self-efficacy for quitting smoking, and a higher intention to quit smoking.

The results from the multiple logistic regression (Table 4) revealed that support for a PoS cigarette display ban was significantly associated with more knowledge about the health risks of smoking, a higher perceived severity, a more positive attitude towards quitting smoking, stronger social norms for quitting smoking, a higher self-efficacy for quitting smoking, and a higher intention to quit smoking.

A separate analysis including interaction terms with educational level and all independent variables did only show almost significant interactions. There was a nearly significant interaction between attitude towards quitting smoking and moderate versus high education (OR of interaction = 0.49, 95% CI: 0.25, 0.96, p = 0.038 > α_adjusted = 0.025), as well as a nearly significant interaction between attitude towards quitting smoking and moderate versus low education (OR of interaction = 0.46, 95% CI: 0.24, 0.90, p = 0.024 > α_adjusted = 0.017). Attitude towards quitting smoking was a stronger predictor for high educated smokers (OR = 2.07, p = 0.008, 95% CI: 1.21, 3.55), and for low educated smokers (OR = 2.10, p = 0.006, 95% CI: 1.23, 3.57) as compared to moderate educated smokers (OR = 0.97, p = 0.886, 95% CI: 0.64, 1.46).

Several limitations should be taken into account when interpreting the results. First, the sample differed over the years on age, educational level, level of addiction to tobacco, and ever having made a quit attempt which may indicate that the sample is not entirely representative of the Dutch population of smokers. Therefore, we adjusted for these sample characteristics in the analyses and applied weights. Second, to cope with the high number of missing values we applied multiple imputation procedures to increase the statistical power. There is never complete certainty about the correctness of the imputation model but a rather substantial set of variables was used in this model and care was taken to take a large number of imputations to maximize the efficiency of the pooled estimates. Third, the measurements of wave 9 took place from November to December 2015, whereas in prior years measurements took place from May to June. This difference can give a somewhat distorted image of the trends in support. Fourth, this paper addressed support for a ban of cigarette displays. Since the ITC Netherlands survey did not include a question on support for a complete ban of tobacco displays, we could not study a display ban of other tobacco or nicotine products.

To increase support further, educational campaigns may focus on explaining the upcoming PoS tobacco display ban as well as on improving knowledge about the health risks of smoking, increasing perceived severity, stimulating a more positive attitude towards quitting smoking, changing social norms about quitting smoking, increasing self-efficacy for quitting smoking, and stimulating quitting. Such educational campaigns may be especially important in the Netherlands where the knowledge and general concern about the health risks of smoking and secondhand smoke in smokers is low compared to other countries [48].