Methods of the International Tobacco Control (ITC) EUREST-PLUS ITC Europe Surveys

Abstract

Background

The EUREST-PLUS ITC Europe surveys aim to evaluate the impact of the European Union’s Tobacco Products Directive (EU TPD) implementation within the context of the WHO FCTC. This article describes the methodology of the 2016 (Wave 1) and 2018 (Wave 2) International Tobacco Control 6 European (6E) Country Survey in Germany, Greece, Hungary, Poland, Romania and Spain; the England arm of the 2016 (Wave 1) and 2018 (Wave 2) ITC 4 Country Smoking and Vaping (4CV) Survey; and the 2016 (Wave 10) and 2017 (Wave 11) ITC Netherlands (NL) Survey. All three ITC surveys covering a total of eight countries are prospective cohort studies with nationally representative samples of smokers.

Methods

In the three surveys across the eight countries, the recruited respondents were cigarette smokers who smoked at least monthly, and were aged 18 and older. At each survey wave, eligible cohort members from the previous waves were retained, regardless of smoking status, and dropouts were replaced by a replenishment sample.

Results

Retention rates between the two waves of the ITC 6E Survey by country were 70.5% for Germany, 41.3% for Greece, 35.7% for Hungary, 45.6% for Poland, 54.4% for Romania and 71.3% for Spain. The retention rate for England between ITC 4CV1 and ITC 4CV2 was 39.1%; the retention rates for the ITC Netherlands Survey were 76.6% at Wave 10 (2016) and 80.9% at Wave 11 (2017).

Conclusion

The ITC sampling design and data collection methods in these three ITC surveys allow analyses to examine prospectively the impact of policy environment changes on the use of cigarettes and other tobacco products in each country, to make comparisons across the eight countries.

Introduction

Created in 2002, the International Tobacco Control Policy Evaluation (ITC) Project has conducted prospective cohort studies in multiple countries from varying regions of the world—the first-ever international cohort study of tobacco use. One of the ITC Project’s primary objectives is to measure the impact of the World Health Organization (WHO) Framework Convention on Tobacco Control (FCTC).¹ With a current portfolio of 29 countries, the ITC Project, centred at the University of Waterloo, Canada, evaluates the psychosocial and behavioural effects of tobacco control policies and regulations of the WHO FCTC on tobacco users, former tobacco users and non-users. Through the use of statistical models, the ITC Project is capable of measuring and quantifying the impact of such policies—at the national level and, where sample size permits, at the sub-national level.

The ITC 6 European (ITC 6E) Country Survey has been conducted as part of the larger Horizon2020 project, European Regulatory Science on Tobacco: Policy Implementation to Reduce Lung Disease (EUREST-PLUS), which aims to monitor and evaluate the impact of the European Union’s Tobacco Products Directive (EU TPD) implementation within the context of the WHO FCTC at a continental level.² The EU TPD entered into force in 2014, with a compliance date of 20 May 2016 for Member States (MS) to transpose it into National legislation.³ Under the latest version of the EU TPD of 2014,³ EU MS were required to implement new regulations that included among others requirements on:

• Reporting of ingredients and emissions, as well as the prohibition of tobacco products with characterizing flavours;

• Labelling and package design restrictions, including minimum dimensions for packages, as well as minimum standards for graphic warning labels;

• Security measures targeting a reduction of both illegally smuggled or counterfeit products;

• New cross-border regulations;

• E-cigarette product regulations such as restrictions on nicotine amounts, product design elements;

• Regulation of other tobacco products and herbal products for smoking.

The ITC Project has eight EU MS cohorts across three survey projects, i.e. ITC 6E Countries Survey, the England arm of the ITC 4 Country Smoking and Vaping (4CV) Survey and the ITC Netherlands (NL) Survey. The ITC methodology allows for comparable and robust cross-country analyses across the eight EU MS to evaluate European tobacco control policies. Specifically:

• Two waves of the ITC 6E Survey, within the context of EUREST-PLUS, conducted from June to September 2016 and February to May 2018, respectively, were completed in Germany, Greece, Hungary, Poland, Romania and Spain.

• Waves 1 and 2 of the ITC 4CV Survey were conducted from July to November 2016 and February to July 2018, respectively, in Australia, Canada, the USA and England. The ITC 4CV Survey is an expansion of the original ITC Four Country (4C) Survey, which had focussed on evaluating the impact of tobacco control policies on smoking‐related beliefs, attitudes and behaviours.^4–7 The ITC 4CV Survey addresses important issues regarding the use of this broad array of products, with special attention to the use of nicotine vaping products (NVPs). An important specific objective was to examine how policies on smoked tobacco, NVPs and other alternative nicotine products may influence the use of these classes of products and transitions among them by present and former smokers.^8,9

• The ITC NL Survey’s main objectives are to measure the effectiveness of tobacco control policy measures on Dutch smokers, to investigate the psychological determinants of smoking, and to compare smoking behaviour and the impact of policies in the Netherlands with other ITC countries.¹⁰ Waves 10 and 11 of the ITC NL Survey were respectively conducted from November to December 2016 and from November to December 2017.

Methods

In all three ITC survey projects, the primary sample of respondents when recruited were aged 18 or older, and were smokers defined as smoking at least monthly and having smoked over 100 cigarettes in their lifetime. For cross-country analyses involving the data from EN and/or NL, the samples in each of those countries were limited to current smokers within the baseline wave (analogous to the Wave 1 of the ITC 6E Survey) so that comparisons across all countries could be based on the same target population of baseline smokers.

The ITC 6E Survey respondents were sampled using an area probability sampling design and interviewed face-to-face via computer-aided personal interviews (CAPIs); respondents of the other two surveys were sampled from the panels of commercial firms, and invited to fill out questionnaires online. Survey weights were calibrated to benchmarks from nationally representative surveys.

ITC 6E Survey sampling and data collection

For sampling purposes at Wave 1 of the ITC 6E Survey, each country was divided into its major geographic regions based on the Nomenclature of Territorial Units for Statistics or Nomenclature des unités territoriales statistiques (NUTS regions), the standard geocoding of the European region overseen by the EU. There is a hierarchical subdivision of regions within each country that creates up to three administrative levels (NUTS 1, NUTS 2 and NUTS 3). The ITC 6E sample utilized the NUTS 2 regions for Greece, Hungary, Poland, Romania and Spain, while using the NUTS 1 regions for Germany.^11–13

Geographic strata were formed by taking the NUTS regions outlined above and crossing them with degree of urbanization (urban, intermediate and rural). The strata were conceptually made up of clusters, each the size of an enumeration area. Clusters were then allocated to strata in numbers proportional to the population size of adults (aged 18 and older). One hundred clusters within each country were then selected by stratified random sampling with at least two clusters in each stratum. In each selected cluster, interviewers used a random walk method to select dwellings to approach and sampled up to one male smoker and one female smoker within each selected dwelling. Within each cluster, ten adult smokers were chosen for interview. Trained interviewers conducted face-to-face interviews using a CAPI mode.^11–13

At Wave 2 of the ITC 6E Survey, efforts were made to re-contact and interview all of the Wave 1 respondents who had agreed to be re-contacted, regardless of smoking status. Dropouts were replaced by adult smokers recruited wherever possible by the same method as in Wave 1 and in the same cluster, from dwellings not approached at the time of Wave 1.

ITC 4CV Survey sampling and data collection

The 2016 ITC 4CV1 sample comprised the following cohorts: re-contact smokers and former smokers who participated in the previous wave of the ITC 4C Project, regardless of vaping status, newly recruited current smokers and former smokers (quit in the past 24 months) from country-specific panels, regardless of vaping status, and newly recruited current vapers (using a vaping device at least weekly) from country-specific panels. By design, among the newly recruited current and former smokers the age group 18–24 was over-sampled in Canada, England and the USA.^8,9

In England, the new recruit sample for Wave 1 of the ITC 4CV Survey consisted primarily of cigarette smokers and former smokers (≤2years), and ∼550 additional e-cigarette users aged 18 years and older. The new recruits were invited from the panel of Ipsos MORI UK. Ipsos MORI’s online panels are subject to a rigorous recruitment procedure aimed at ensuring accuracy, consistency and non-duplication. To join, panel applicants are validated by means of vetting procedures, using a variety of opt-in recruitment channels. Shortly after joining, panel members’ survey-taking behaviour is tested, with those most likely to make intentional or unintentional errors on future surveys deactivated. Subsequently, panelists’ behaviour is monitored and tracked across all surveys. Ipsos employs purging procedures based on these data to remove suspect respondents from eligible sampling pools. In view of the scope and scale of the 4CV research, approved panel partners that met Ipsos quality control procedures were also used in England to support the project (Statement on panel description provided by Ipsos MORI on 6 September 2018).^8,9

ITC NL Survey sampling and data collection

The ITC NL sample was designed to be nationally representative of Dutch smokers aged 15 and older, with the age group 15–30 being oversampled. TNS NIPObase panel members are actively recruited by TNS NIPO. Web panel members were recruited by phone or mail, but not through the internet. We selected respondents that met the inclusion criteria for the ITC NL Survey through a screening procedure.¹⁴ All chosen respondents were invited to complete the survey using computer-assisted web interviews. The sampling design calls for the population to be stratified into 12 geographic strata corresponding to the 12 Dutch provinces. The numbers of respondents to be sampled in each of the geographic strata were assigned using allocation proportional to the estimated size of the adult population. The sampling design also calls for the population to be stratified into 2 age strata: 15≤ age ≤ 30 and age > 30. Quotas were then assigned to each age stratum so that young smokers were over-sampled and formed about 44% of the web sample, whereas they only represented about 25% of Dutch smokers. Within each geographic and age stratum, respondents were randomly selected from the screened TNS NIPObase until the corresponding quota was met. More details of the sampling in the first few waves can be found elsewhere.¹⁵

Survey weights

Cross-sectional survey weights were constructed for each of the relevant survey waves, and longitudinal survey weights were constructed for the two waves of data being examined in each country. The survey weights were calibrated to geographic and demographic benchmarks constructed from census data and nationally representative surveys of prevalence. These nationally representative surveys in the 6E countries were the 2014 Special Eurobarometer¹⁶ for the ITC 6E Wave 1 cross-sectional weights and the Waves 1 and 2 longitudinal weights, and the 2017 Special Eurobarometer¹⁷ for the ITC 6E Wave 2 cross-sectional weights.^11,12 The Eurobarameter sample sizes were ∼1000 in each country and year. Calibration variables were sex, age group (18–39, 40–54 and 55+), NUTS region and urbanization (rural, intermediate and cities); to obtain estimated benchmarks, a weighted logistic regression of smoking status (binary) was carried out on the calibration variables within each country.

The nationally representative surveys used for benchmarks for the Waves 1 and 2 of the ITC 4CV England data were the Office of National Statistics Opinion and Lifestyle Survey (2015 and 2017) (sample size ∼1100 per month) and the Smoking Toolkit Surveys (sample size ∼1700 per month) over the same periods. Calibration variables for smokers were sex, age group (18–25, 25–29, 30–34, 35–39, 40–44, 45–49, 50–54, 55–59 and 60+), education (three levels), region (nine values), each crossed with use vs. non-use of vaping.

For The Netherlands, calibration benchmarks for numbers of smokers by province and demographic groups (sex crossed with age group (18–24, 25–39, 40–54 and 55+) were calculated by combining population figures from the Centraal Bureau voor de Statistiek (January 2016 and 2017) and prevalence estimates from the Continuous Survey of Smoking Habits (CSSH 2014–17; sample size ∼18 000 per year).

Ethics review

For the ITC 6E Survey, study procedures and material including the survey questionnaire were approved by the ethics research committee at the University of Waterloo (Ontario, Canada), and ethics committees in Germany (Ethikkommission der Medizinischen Fakultät Heidelberg), in Greece (Medical School, University of Athens—Research and Ethics Committee), in Hungary (Medical Research Council – Scientific and Research Committee), in Poland (State College of Higher Vocational Education—Committee and Dean of the Department of Health Care and Life Sciences), in Romania (Iuliu Hatieganu University of Medicine and Pharmacy), and in Spain (Clinical Research Ethics Committee of Bellvitge, Hospital Universitari de Bellvitge, Catalonia).

Results

ITC 6E Survey outcomes

The ITC 6E Survey was designed to produce a nationally representative sample in all six participating countries. To achieve this, the project sought to sample ∼1000 smokers from each country. The total sample sizes (combining respondents recruited at Wave 2, and those recruited at Wave 1 who also completed the Wave 2 survey) for the six countries are 1010 for Germany, 1010 for Greece, 1000 for Hungary, 996 for Poland, 1003 for Romania and 1008 for Spain; for a grand total of 6027 (see table 1).

Table 1.

Retention rates and sample size at each wave, by country

Country	Respondents at Wave 1 (N)	Retention rate at Wave 2	Respondents continuing at Wave 2 (N)	Replenishment respondents at Wave 2 (N)	Respondents at Wave 2 (N)
Germany	1003	70.5%	707	303	1010
Greece	1000	41.3%	413	597	1010
Hungary	1000	35.7%	357	643	1000
Poland	1006	45.6%	459	537	996
Romania	1001	54.4%	545	458	1003
Spain	1001	71.3%	714	294	1008
Overall	6011	53.2%	3195	2832	6027
England	3536	40.7%	1438	2822	4260
The Netherlands	1213	80.7%	979	277	1256

Wave 1 to Wave 2 retention rates varied greatly across the six countries, from over 70% in Germany and Spain to only 36% in Hungary. Accordingly, the countries varied in reciprocal fashion in the degree of sample replenishment accomplished at Wave 2. Table 2 shows the breakdown of retention into contact rate and cooperation rate. Difficulties with retention can arise for numerous reasons, including that interviewers may be unable to locate a Wave 1 respondent via their initial contact attempt (usually over the phone). Although interviewers would still attempt to reach these past respondents at their homes, household visits can introduce other obstacles, such as: (i) the household members refusing to provide follow-up information; or (ii) contacts having moved. Certain countries have higher rates of transient migration than others; transient migration can create difficulties with follow-up interviews.

Table 2.

Re-contact outcome rates by 6E country

Country	Respondents at Wave 1 (N)	Contact made at Wave 2	Contact Rate at Wave 2	Respondents continuing at Wave 2 (N)	Cooperation rate at Wave 2 (N)
Germany	1003	815	81.3%	707	86.7%
Greece	1000	479	47.9%	413	86.2%
Hungary	1000	Not available		357
Poland	1006	487	48.4%	459	94.3%
Romania	1001	644	64.3%	545	84.6%
Spain	1001	754	75.3%	714	94.7%
Overall	6011	–	–	3195	–

Table 3 shows response rate calculations for replenishment at Wave 2 of the ITC 6E Survey. We take the replenishment unconditioned individual response rates to be the product of Rows 14 and 15, and these are given in row 16 as 26.1% for Germany (n = 303), 37.2% for Greece (n = 597), 61.9% for Hungary (n = 643), 70.4% for Poland (n = 537), 44.5% for Romania (n = 458) and 66.0% for Spain (n = 294). The overall individual response rate for Wave 2 (2018) replenishment for all 6 countries was 47.0% (n = 2832). These rates are broadly consistent with but a little higher than those observed in Wave 1, which were 20.7% for Germany, 32.5% for Greece, 59.6% for Hungary, 49.3% for Poland, 36.0% for Romania and 48.0% for Spain.

Table 3.

Replenishment outcome rates by country

	Germany	Greece	Hungary	Poland	Romania	Spain
1. Number of addresses approached/attempted	1645	1825	1441	743	1487	933
2. Number of addresses where contact has been made	1438	1470	1260	731	1264	848
3. Number of contacted addresses with eligibility determined	633	733	1009	607	807	744
4. Number of contacted addresses with no eligible respondents	287	233	458	119	366	193
5. Number of contacted addresses with eligible respondents	346	481	551	488	441	284
6. Number of addresses with eligible respondents, members selected	342	481	539	486	433	283
7. Eligibility rate for households, given determination (5./3.)	0.547	0.656	0.546	0.804	0.546	0.382
8. Estimated eligible households among attempted (7.*1.)	899	1198	787	597	813	356
9. Number of individuals selected for interview	441	645	711	621	549	354
10. Number of individual refusals or break offs	73	43	59	76	82	56
11. Number of completed interviews	303	597	643	537	458	294
12. Household contact rate (2./1.)	0.874	0.805	0.874	0.984	0.850	0.909
13. Household cooperation rate, given eligible (6./5.)	0.988	1.00	0.978	0.996	0.982	0.996
14. Household response rate (6./8.)	0.380	0.402	0.685	0.814	0.533	0.795
Individual cooperation rate [11./(11.+10.)]	0.806	0.933	0.916	0.876	0.848	0.840
15. Individual response rate, given selection (11./9.)	0.687	0.926	0.904	0.865	0.834	0.831
16. Individual response rate (6.*11.)/(8.*9.)	0.261	0.372	0.619	0.704	0.445	0.660

ITC 4CV Survey (England) outcomes

At Wave 1, the sample in England consisted of 3773 smokers or former smokers (aged 25+, n = 2886; aged 18–24, n = 887) and an additional 551 at least weekly vapers.⁹ At Wave 2, the sample consisted of 4939 smokers or former smokers (aged > 25, n = 3240; aged 18–24, n = 1105) and an additional 594 at least weekly vapers. The retention rate from Waves 1 to 2 was 39.7%.

The response rate for 2016 England ITC 4CV England Wave 1 (2016) was 15.2%, for a total sample size of 3536 respondents. At Wave 2 (2018), the replenishment response rate was 44.1%, which yielded a sample of 2882 new recruits and a total sample size of 4260.

ITC NL Survey outcomes

The retention rates for the ITC NL Survey were 76.6% at Wave 10 (2016) and 80.9% at Wave 11 (2017). Response rates for NL are 67% for 2016 (Wave 10; total sample size n = 1213) and 69% for 2017 (Wave 11; replenishment sample size n = 277 and total sample size n = 1256; numbers provided by NL team).

Conclusion

This article describes the methods of approximately contemporaneous cohort surveys in eight countries of the EU, designed to evaluate the policies of the TPD. The strengths of this study include:

• (i) a survey tool that allows detailed evaluation of the TPD policies with respect to smoking behaviour and cessation;

• (ii) the prospective cohort study design;

• (iii) simultaneous data collection in six European countries; and the possibility of near-simultaneous comparisons among eight European countries;

• (iv) the representativeness in each country of the smoking population.

Some limitations arise from differences in data collection and outcomes across countries and waves:

• (i) the fact that different recruitment and data collection methods are used in England and The Netherlands compared with the ITC 6E Countries; in England and the Netherlands, the samples were taken from commercial web panels, subject to geographic and demographic quotas to guarantee representativeness on these dimensions, and respondents filled in the questionnaire online; in the ITC 6E countries, data collection was carried out face-to-face by interviewers;

• (ii) the different seasons for the fieldwork of the waves in the ITC 6E Countries (summer vs. winter/spring);

• (iii) the different retention rates observed in various countries, varying between 36 and 81%.

These differences must be kept in mind when interpreting differences across countries and across waves.

Although the coverage of the smoker populations is a strength, and the study is ideal for examining the impact of policy changes on smoking cessation, the fact that non-smokers are not recruited means that we cannot in this study observe whether policy changes influence initiation of cigarette smoking or use of other nicotine delivery products.

Nevertheless, the ITC sampling design and data collection methods in these three survey projects across eight countries allow analyses to examine prospectively the impact of policy environment changes on the users of cigarettes and NVPs in each country, and to make comparisons across countries, the results of which are outlined in the subsequent manuscripts of this special supplement in the European Journal of Public Health.