Understanding perceived availability and importance of tobacco control interventions to inform European adoption of a UK economic model: a cross-sectional study

Data for this study were drawn from the following two sources: (i) the EQUIPT stakeholder interview survey capturing views on availability and importance of tobacco-control interventions [21]; and (ii) published data on cost-effectiveness of tobacco control interventions [22].

First, a stakeholder interview survey (face-to-face, by phone or Skype) was conducted between April and July 2014 as part of the EQUIPT project (http://​equipt.​eu) [21, 23]. Further information about the interview survey is documented elsewhere [16, 21, 23]. A list of stakeholders categorized into five groups was drawn by the EQUIPT team from Germany, Hungary, the Netherlands, Spain, and the UK based on their previous knowledge; and then those stakeholders were asked to provide additional names (snowballing). This list served as the sampling frame for the interview survey. The five categories of stakeholders included: decision makers (e.g. senior officials at the department of health or directors of public health services), purchasers of services or pharmaceutical products (e.g. service commissioners or top officials at health insurance funds), professionals or service providers (e.g. leading physicians in smoking cessation, physicians/psychologists, or coordinators of local health programs), evidence generators (e.g. academic and researchers with experience in health-technology-assessments, reimbursement procedures, health care costing, general public health and health promotion research, and specific research to tobacco control and smoking cessation), and advocates of health promotion (e.g. leaders of charities, NGO’s or patient organizations).

A researcher from each country interviewed the stakeholders in the local language and filled out the questionnaire, which was developed originally in English and then translated to each of the local languages. To ensure the consistency of the survey across countries, a training workshop was organised in Maastricht in January 2014, followed by a few online training sessions prior to the interviews. A pilot study was conducted to improve the actual interview survey. The interviews were based on pre-defined questionnaire administered by trained interviewers. Thus, interviewer validity was achieved to a reasonable degree. Interview guide and survey questionnaire are provided in Additional file 1.

The questionnaire included several questions: stakeholders’ profiles, motivational factors, and stakeholders’ opinion. Perception of availability and importance of tobacco-control interventions were captured by the following two questions, respectively: (a) “Would you tell me whether these [show the list] tobacco control measures and smoking cessation interventions are available in …[name country]…?” with possible answers being ‘yes’, ‘no’, and ‘don’t know’; and (b) “Could you also indicate on a scale from 1 to 3 —1 meaning ‘not important’ and 3 meaning ‘important’ – to what extent you think the following interventions are considered important in addressing smoking behaviour?” Respondents were allowed to state ‘don’t know’ as a response in addition to one of the three numbers (1, 2, and 3). A total of 30 interventions grouped into five categories – 3 pharmacological, 11 behavioural, 3 combined pharmacological and behavioural, 7 nonconventional, and 6 population-level - were shown to stakeholders. The intervention list was informed by a previous review conducted as part of the NICE Tobacco ROI project [15].

Secondly, cost-effectiveness data on the selected interventions was sourced from a published study in the UK by Owen and colleagues who comprehensively reviewed the economic evidence of key public health interventions including tobacco-control [22]. This single review study was selected because it provided summarised cost-effectiveness data on most of the tobacco-control interventions listed in the stakeholder survey questionnaire in a comparable way, and thus allowed a head-to-head comparison between the two.

Brunel University Research Ethics Committee (UK) reviewed this research and gave full ethical clearance. Respective authorities in other countries - Ethik-Kommission, Bayerische Landesärztekammer (Germany), Egészségügyi Tudományos Tanács, Tudományosés Kutatásetikai Bizottság (Hungary), Medisch-ethische toetsingscommissie (METC) azM/UM (Netherlands), and Parc de SalutMAR - Clinical Research Ethics Committee (Spain) - also provided clearance. Written consent was obtained from all respondents.

Three separate analyses were carried out. Firstly, descriptive analyses with all possible responses were conducted to explore stakeholders’ perceptions of availability and importance of tobacco control interventions in each country and between the countries. Kappa statistics were calculated to evaluate the agreement on the perceived availability with all three possible responses and importance with all four possible responses between stakeholders in each country, from two selected countries, and from all five countries. A kappa statistic or a kappa coefficient is commonly used to measure the agreement between multiple observers; it ranges from − 1 to 1, the latter indicating perfect agreement [24]. If it is less than or equal to 0, there is less or no agreement between raters other than what would occur by chance. Kappa values of 0.01–0.20, 0.21–0.40, 0.41–0.60, 0.61–0.80, and 0.81–0.99 refer to slight agreement, fair agreement, moderate agreement, substantial agreement, and almost perfect agreement, respectively [24]. A significant kappa (p-value < 0.05) means that the estimated value of the kappa coefficient significantly differs from zero. [24]. Confidence intervals around kappa coefficient were calculated by a bootstrap method as the computation involved more than three raters [25]. Combined kappa coefficients from all possible responses are reported. Moreover, a comparison of perceived importance when combining ‘somewhat important’ (response 2 on the questionnaire) and ‘important’ (response 3) values between the 30 interventions was presented in a visual display. Such comparison was conducted as part of the decision to choose the interventions for the EQUIPT model.

Secondly, to assess whether stakeholders’ views about availability of interventions were correlated with their views about importance of interventions, a descriptive analysis showing distributions of availability and importance was conducted. This relationship was further analysed using an ordered logistic regression [26]. The ‘don’t know’ responses were excluded from the regression analysis. The model included importance (not important, somewhat important, or important) as the dependent variable and availability (yes or no), countries (the UK as the reference country), stakeholder roles (evidence generators as the reference role), and gender (male as the reference group) as covariates.

Thirdly, a perceived importance mean score for each intervention was computed after excluding the ‘don’t-know’ responses. The 3-point Likert scale response on the question related to perceived importance of an intervention was recoded from [1, 2 3] to [0, 1, 2] respectively. Then the mean score was calculated as the arithmetic average of the responses across all stakeholders. This score was finally used to rank the interventions. The other set of ranking involved scrutiny of the cost-effectiveness data from Owen et al. [22]. To do this, first the median costs per quality adjusted life years (QALYs), the only reported measure in that study, were extracted for the interventions that were in common with those listed in the EQUIPT stakeholder survey. Then, interventions were rank ordered using the lowest median costs per QALY and moving upwards. This ranking was then compared with the ranking based on the stakeholders’ views to determine where similarities or differences occurred. Kendall rank (tau) correlation coefficient was estimated to evaluate the difference between the concordance and the discordance probabilities of the mean-importance-score ranking and the cost-effectiveness-measure ranking. The tau coefficient is normally used when a distribution of a variable is not assumed and a sample size is small [27]. The coefficient ranges from − 1 to 1 meaning complete disagreement and complete agreement between the mean importance score and the rankings based on cost-effectiveness data, respectively. The zero value refers to no relationship between the two variables. The null hypothesis states that the two variables are independent to each other. All analyses were accomplished in Stata version 13.1 by StataCorp LP, Texas, USA.

Of the 167 stakeholders who were invited to participate in the interview survey, 93 respondents (55.7%) completed the survey and were included in the analysis—17 from Germany (18.3%), 16 from Hungary (17.2%), 28 from the Netherlands (30.1%), 18 from Spain (19.4%), and 14 from the UK (15.1%). Fifty-eight (62.4%) were males.

Twenty-nine were decision makers (31.2%), 7 were purchasers of services or pharmaceutical products (7.5%), 18 were professionals or service providers (19.4%), 25 were evidence generators (26.9%), and 14 were advocates of health promotion (15.1%). The distributions of stakeholder types across the five countries were not largely different except the UK that had a large proportion of decision makers (approximately 64% compared with the remaining countries that had around 17–33%). The complete compositions of stakeholder types are exhibited in a Additional file 2: Table S1.

Perceptions on availability of 11 selected interventions are shown in Table 1 (data on all interventions shown in Additional file 2: Table S2). Between the countries, the perceptions on availability varied considerably for almost all of the interventions. The interventions that > 70% stakeholders viewed as ‘available’ in each country were: advertising restrictions/bans, product labelling and information/health warnings on tobacco products, restrictions on sales to minors, restrictions on smoking in workplaces and public places, tax increase, self-help manuals, nicotine replacement therapy, brief physician advice, and group and counselling with or without pharmacotherapies. Stakeholders did not know about the availability of some widely-evaluated interventions such as varenicline (33% of 93 stakeholders), bupropion (29%), computer tailored programs (20.4%), and community pharmacy-based services (17.2%).

In each country, a number of interventions were deemed important by > 58% of stakeholders (Table S2). These included nicotine replacement therapy, brief physician advice, the 5-step protocol advice (i.e. ask, advice, assess, assist, and arrange), group and individual counselling by specially trained professionals with and without medications (e.g. NRT or bupropion), advertising restrictions/bans, restrictions on sales to minor, restrictions on smoking in workplaces and public places, and tax increase. Non-conventional therapies were rated important by stakeholders in Germany and Hungary compared to the remaining countries. Non-conventional interventions were viewed to be much less important than the rest of the interventions as only 1.1–6.5% of the whole sample viewed that they were important. Figure 1 shows perceptions on importance of the 11 interventions (somewhat important and important responses combined). The pattern of perceptions was similar when only ‘important’ response was analysed.

Generally, stakeholders tended to agree to a lesser extent on the availability compared with the importance (Table 2). All combined kappa coefficients were statistically significant with p-values < 0.001. Stakeholders from Germany, Hungary, and the Netherlands had lower agreement on the availability of interventions than those from Spain and the UK. Slight agreement on availability was observed for every pair of countries except for Spain-UK pair (fair agreement). In general, stakeholders from all countries had slight agreement on the intervention availability (combined kappa = 0.0980; 95% confidence interval = [0.085, 0.111]).

Within country, variations on importance of interventions included slight to fair agreement (Germany, Hungary, the Netherlands, and Spain) and moderate agreement (UK) (combined kappa in the UK = 0.4246). Every pair of the countries showed fair agreement on importance except Hungary-Spain pair (slight agreement). Stakeholders from all countries had fair agreement on the importance of interventions (combined kappa = 0.2386; 95% CI = [0.208, 0.253]).

Perceived availability of interventions was positively associated with their perceived importance for the following interventions once stakeholder roles, gender and countries were controlled for (p-values < 0.05): community pharmacy-based services, computer tailored programs, internet-based interventions, stage-based interventions, and brief advice by a general practitioner and medication (Table 3). The highest correlation was observed for stage-based interventions.

Table 4 exhibits the scores of all 30 interventions with their categories. The top three interventions were 1) restrictions on smoking in workplaces and public places (score: 1.89; N = 92), 2) individual counselling by specially trained professionals with medication (e.g. NRT) (score: 1.86; N = 90), and 3) advertising restrictions/bans (score: 1.83; N = 92).

Table 5 puts the stakeholders’ perceived importance scores into perspective by comparing their rank order with that by published cost-effectiveness evidence. A total of 10 interventions were included for this comparison. The top three interventions by the perceived importance score were (1) individual counselling by professionals with medication (score: 1.86); (2) group counselling by professionals with medication (score: 1.81); and (3) nicotine replacement therapy (score: 1.74). These first three interventions would also be most cost-effective based on the published review. The large differences in importance of interventions between stakeholder views and cost-effectiveness evidence were apparent for self-help manuals, telephone counselling, community pharmacy-based services and mass media campaigns. These interventions ranked lower by importance score than by cost-effectiveness criteria. Brief advice by a general practitioner ranked marginally higher by importance score than by cost-effectiveness criteria. Kendall rank correlation coefficient was 0.40 with the p-value of 0.11, indicating that ranking based on the importance score did not statistically agree with the ranking based on cost-effectiveness evidence (95% CI = [− 0.09, 0.89]).