It is crucial for policy makers to monitor the evolution of tobacco smoking prevalence. In France, this monitoring is based on a series of cross-sectional general population surveys, the Health Barometers, conducted every five years and based on random samples. A methodological study has been carried out to assess the reliability of a monitoring system based on regular quota sampling surveys for smoking prevalence.
Design / Outcome Measures
In 2010, current and daily tobacco smoking prevalences obtained in a quota survey on 8,018 people were compared with those of the 2010 Health Barometer carried out on 27,653 people. Prevalences were assessed separately according to the telephone equipment of the interviewee (landline phone owner vs “mobile-only”), and logistic regressions were conducted in the pooled database to assess the impact of the telephone equipment and of the survey mode on the prevalences found. Finally, logistic regressions adjusted for sociodemographic characteristics were conducted in the random sample in order to determine the impact of the needed number of calls to interwiew “hard-to-reach” people on the prevalence found.
Current and daily prevalences were higher in the random sample (respectively 33.9% and 27.5% in 15-75 years-old) than in the quota sample (respectively 30.2% and 25.3%). In both surveys, current and daily prevalences were lower among landline phone owners (respectively 31.8% and 25.5% in the random sample and 28.9% and 24.0% in the quota survey). The required number of calls was slightly related to the smoking status after adjustment for sociodemographic characteristics.
Citation: Guignard R, Wilquin J-L, Richard J-B, Beck F (2013) Tobacco Smoking Surveillance: Is Quota Sampling an Efficient Tool for Monitoring National Trends? A Comparison with a Random Cross-Sectional Survey. PLoS ONE 8(10): e78372. https://doi.org/10.1371/journal.pone.0078372
Editor: Nandita Mitra, University of Pennsylvania, United States of America
Received: June 18, 2013; Accepted: September 19, 2013; Published: October 23, 2013
Copyright: © 2013 Guignard et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: The study was financially supported by the French Institute for Health Promotion and Health Education (INPES). The funder had no role in study design, data collection and analysis, or preparation of the manuscript, and allowed the publication of the study.
Competing interests: The authors have declared that no competing interests exist.
Tobacco smoking constantly decreased in France between 1970 and 2005. Nevertheless, the most recent survey carried out in 2010 at a national level revealed an increase compared with 2005, non-significant for men, but significant for women, especially concerning those aged 45–64 . On the whole, tobacco smoking prevalence has increased from 31.5% in 2005 to 33.7% in 2010 in 15–75 years-old, and the prevalence of daily smoking from 27.1% to 29.1% in the same time. Taking into account the huge efforts initiated in France in the late 1970s and reinforced in 1991 (Evin Law) in tobacco control policies, the last being large price increases between 2002 and 2004 and smoke-free laws in public places in 2007-2008, it reveals crucial for policy makers to monitor the evolution of these indicators.
The acknowledgement of this evolution is based on a series of multi-thematic repeated cross-sectional general population surveys, the Health Barometers of the French Institute for Health Promotion and Health Education (Institut National de Prévention et d’Education pour la Santé - INPES). These surveys, conducted every five years, are based on random samples of about 30,000 people. A random survey carried out properly is usually considered more accurate than a quota survey since it covers more socially diverse and representative populations. Furthermore, it is known that tobacco smoking depends on social background. However, since the methodology of random surveys is complex, and due to economic and organizational reasons, it is difficult to conduct this kind of survey more frequently, and quotas surveys have sometimes been considered as an acceptable alternative to probability sample surveys . Consequently, a methodological study has been carried out by INPES to assess the reliability of a monitoring system based on less costly regular quota sampling surveys for tobacco smoking prevalence, despite their limits.
All the data were analyzed anonymously. Participants provided their verbal informed consent to participate in the study at the beginning of the questionnaire. Interviewers keyboarded the participant’s consent as part of the computer-assisted telephone interviews (CATI). For participants under 18, verbal consent was obtained from the head of the family in the same way. The population-based surveys process and the consent procedure were approved by the French individual data protection authority (Commission nationale de l’informatique et des libertés [CNIL]), that considered that written consent was not requested for such surveys.
The 2010 Health Barometer fieldwork was conducted by the GfK-ISL Survey Institute. In total, 27,653 people aged 15 to 85 living in continental France were interviewed through CATI between October 2009 and July 2010.
The design was a two-stage random sampling stratified by size of urban area and geographical area: the first stage of the sampling involved selecting households by using random digit dialling (RDD). The second stage involved selecting a member in the household through the Kish method , without replacement in case of refusal (neither at the household, nor at the individual level). In this way, 24,709 people contacted through their landline phones were interviewed. RDD can “catch” households that do not appear in the directory. These households represent about 30% of French households, and these individuals are more likely to smoke than people registered in the directory, the difference being partly explained by their different socioeconomic structures [4,5]. In addition, previous studies have shown a higher prevalence in individuals from households with a cell phone and without a landline than in the others [6,7]. In France, these “mobile-only” households account for 12% of all households. An additional sample of 2,944 mobile-only individuals were interviewed using the same process of generation and selection as for the landline phone sample.
Specific efforts were made to successfully reach households: up to 40 attempts were made for each generated phone number, since it has already been shown that “hard-to-reach” people, like non-respondents, have specific health behaviour [8-12]. As well, people who firstly refused to participate were contacted a second time by specially trained interviewers. This protocol allowed to recover 20% of the final sample. Details of the survey methodology have been previously published elsewhere [13-15].
To assess comparability between random sampling and quota sampling, another telephone survey was conducted between September and December 2010 by the BVA Survey Institute on behalf of the French Institute for Health Promotion and Health Education. In total, 8,018 people aged 15 to 85 years-old were interviewed. The design was a quota sampling based on the French population structure in terms of gender crossed by age, and occupation of the head of the household, after stratification by size of urban area and geographical area. Reference data came from the 2008 Employment Survey, which is based on the French population census.
Since the quota method is designed with the only aim of having final margins equal to those of a defined population, French quota surveys are generally conducted on samples of landline telephone numbers randomly selected from the main French landline directory. However, mainly because young adults are more and more difficult to reach in the context of telephone surveys, several survey agencies, including BVA, have added a cell random digit-dialling sample (whether individuals also have a landline phone or not). In this study, “cell-only” households ended up representing 6% of the whole sample. Nevertheless, unlike the Health Barometer Survey, households with unlisted landline phone numbers were not included in the quota sampling.
Moreover, in the quota survey, phone numbers with no answer at the first call or a refusal by the person answering were not dialled again and were replaced. Lastly, in the quota sampling, there was no random selection within the household. The interviewee was either the person who hung up if he/she met the quota variables or any person meeting these selection criteria in the household.
Sociodemographic variables were collected in both random and quota sample surveys: age, gender, geographical region, size of urban area (rural village, fewer than 20,000 inhabitants, 20,000 to 100,000 inhabitants, 100,000 to 200,000 inhabitants, more than 200,000 inhabitants, urban area of Paris), level of education (less than high school, high school graduate, college graduate) and occupational status (employed, student, unemployed, retired, other non-working).
Smoking status (current smoker or daily smoker) was assessed using two questions: “do you smoke tobacco, at least occasionally?”, and (if yes) “do you smoke every day?”
Given the survey period was different in each survey, the year and month when smokers had started to smoke and the year and month when former smokers had quit were asked in the quota survey. Thus, smoking status in January 2010 could be assessed at a time when half of people had been interviewed as part of the Health Barometer.
The analysis was carried out on 25,990 individuals from the Health Barometer and 7,621 from the quota survey all aged 15 to 75, the age group on which evolution of smoking prevalence has been assessed in France until now. Data from both surveys have been weighted according to gender crossed by age, size of urban area, geographical region and level of education. Distinct weights were calculated in the Health Barometer for the whole sample, landline phone owners, listed numbers. The 2008 French Employment Survey was used as a reference for margins.
First, both surveys were compared in terms of socioeconomic characteristics: gender, age, size of urban area, level of diploma and occupational status. Socioeconomic data were also crossed with the phone equipment and the needed number of calls in the Health Barometer. Crude and weighted percentages are shown. Next, prevalence of current and daily tobacco smoking was assessed in both surveys: in the whole sample (whether they have a landline and/or a cell phone) and among landline phone owners (with or without a cell phone) in the quota survey; in the whole sample, among landline phone owners and among people listed in the telephone directory in the Health Barometer. Design-based Pearson chi-square tests with Rao-Scott second order correction were used to test differences in proportions. Differences in prevalence according to the phone equipment and the survey mode were controlled using logistic regressions adjusted for sociodemographic variables. Lastly, in order to explain potential differences, the prevalence obtained according to the needed number of calls was estimated in the Health Barometer among landline phone owners, and was modelled using logistic regression adjusted for gender, age, level of education, occupational status and kind of phone equipment.
Sociodemographic characteristics of the samples
Before weighting, males are under-represented in the Health Barometer compared with the quota survey (44.9% vs 48.5%, p-value<0,001) (Table 1). Undergraduate respondents are less represented in the quota survey (41.3%) than in the random survey (50.1%, p-value<0,001). Besides, after weighting, unemployed people are about twice as frequent in the Health Barometer (8.2%) as in the quota survey (4.5%, p-value<0,001).
|Quota (N=7,621)||HB (N=25,990)||HB landline phone (N=21,952)||HB listed (N=11,433)|
|crude %||weighted %||crude %||weighted %||crude %||weighted %||crude %||weighted %|
|15-24 years old||12,4%||16,3%||13,2%||16,7%||11,0%||16,8%||9,8%||17,1%|
|25-34 years old||15,7%||16,6%||16,5%||16,7%||13,6%||16,6%||11,1%||16,4%|
|35-44 years old||22,6%||18,7%||20,7%||19,0%||20,8%||19,0%||18,6%||19,1%|
|45-54 years old||20,1%||18,3%||17,8%||18,6%||18,7%||18,6%||19,0%||18,6%|
|55-64 years old||16,8%||16,6%||19,4%||16,8%||21,6%||16,8%||23,9%||16,8%|
|65-75 years old||12,4%||13,4%||12,4%||12,2%||14,3%||12,1%||17,5%||12,0%|
|Size of urban area|
|Less than 20,000 inhab.||16,3%||17,5%||18,8%||17,3%||19,3%||17,3%||20,7%||17,5%|
|20,000 to 100,000 inhab.||11,8%||12,7%||12,5%||12,7%||12,1%||12,6%||12,0%||12,7%|
|100,000 to 200,000 inhab.||5,7%||5,6%||5,1%||5,6%||4,8%||5,6%||4,3%||5,5%|
|More than 200,000 inhab.||22,4%||22,0%||21,9%||22,4%||20,4%||22,3%||16,5%||22,1%|
|Urban area of Paris||15,5%||16,2%||14,0%||16,3%||13,2%||16,2%||8,7%||15,9%|
|Less than high school||41,3%||60,0%||50,1%||59,8%||49,9%||59,6%||52,1%||59,5%|
|High school graduate||18,3%||17,7%||18,2%||17,6%||17,9%||17,7%||17,2%||17,7%|
With regard to the type of phone equipment in the 2010 Health Barometer, there are more individuals aged 55 years old or more among the ones who are listed in the directory (41.4% vs 24.3% among unlisted respondents or cell-only ones, p-value<0,001). Respondents who are listed in the directory less often live in the urban area of Paris (8.7% vs 18.1% of the ones who live elsewhere, p-value<0,001) and more often live in rural areas (37.8% vs 19.8%, p-value<0,001). After weighting, unemployed people are less represented among landline phone owners (5.7% vs 16.3% of the ones who only have a cell phone, p-value<0,001).
In the 2010 Health Barometer, “hard-to-reach” respondents are more frequently males: 53.3% of the respondents who were interviewed after more than 20 calls are males vs 44.2% of the ones for whom one or two calls were sufficient. “Hard-to-reach” respondents were also younger, more often employed and had a higher level of education (Table 2).
|1-2 calls||3-5 calls||6-10 calls||11-20 calls||more than 20 calls|
|15-24 years old||11,2%||17,3%||20,6%||20,5%||18,7%|
|25-34 years old||14,6%||15,1%||15,8%||19,5%||23,9%|
|35-44 years old||17,6%||18,8%||19,4%||19,4%||22,4%|
|45-54 years old||17,9%||19,4%||18,7%||18,5%||18,3%|
|55-64 years old||20,1%||17,2%||16,2%||13,7%||12,2%|
|65-75 years old||18,7%||12,2%||9,4%||8,3%||4,5%|
|Less than high school||62,5%||60,2%||59,0%||57,0%||55,4%|
|High school graduate||17,2%||17,4%||17,9%||17,8%||19,2%|
Prevalence of smoking according to the survey and the phone equipment
On the whole, the current smoking prevalence among 15–75 years old is 33.9% in the Health Barometer and 30.2% in the quota survey (test of difference of proportions: p-value<0.001) (Table 2). The daily smoking prevalence is 27.5% in the Health Barometer and 25.3% in the quota survey (p-value<0.01).
Considering only landline phone owners, smoking prevalences are also higher in the random survey than in the quota survey (31.8% vs 28.9% for current smoking, p-value<0.001; 25.5% vs 24.0% for daily smoking, p-value<0.05). When keeping only the listed numbers, current smoking prevalence remains higher in the Health Barometer than in the quota survey (31.2% vs 28.9%, p-value<0.01) but not daily smoking prevalence (Table 3). The difference in current smoking prevalence is still significant when respondents from the Health Barometer converted to interview after initial refusal are excluded from the analysis (31.0% vs 28.9%, p-value<0.05).
|Health Barometer (HB)||Quota Survey|
|Whole sample1||Landline phone owners2||Listed numbers3||Whole sample||Landline phone owners|
|Prevalence of current smoking prevalence||33.9%*** [33.2-34.5]||31.8%*** [31.0-32.5]||31.2%** [30.0-32.4]||30.2% [29.1-31.4]||28.9% [27.7-30.0]|
|Prevalence of daily smoking||27.5%** [26.8-28.1]||25.5%* [24.8-26.2]||24.9%ns [23.7-26.0]||25.3% [24.2-26.4]||24.0% [22.9-25.1]|
Factors related with current and daily tobacco smoking
As shown in other studies , a low level of diploma and unemployment are related with current and daily tobacco smoking in the pooled database (Table 4). Cell-only owners are also more likely to smoke after adjusment for sociodemographics (ORa=1.7 [1.6-1.8] for current and daily smoking compared with landline phone owners). After adjustement for sociodemographics and phone equipment, an association with the survey mode (random vs quotas) remains for current smoking (ORa=1.1 [1.1-1.2] for the random survey compared with the quotas survey, p-value<0.001), but not for daily smoking. The odds-ratios do not change when the respondents who accepted to answer after initial refusal are removed from the logistic model
|Current smoking||Daily smoking|
|%||aOR||95% CI||%||aOR||95% CI|
|Male (ref.) (n=15372)||36.8||- 1 -||30.0||- 1 -|
|15-24 years-old (ref.) (n=4377)||38.7||- 1 -||27.9||- 1 -|
|25-34 years-old (n=5492)||47.0||1.0||[0.9-1.2]||39.1||1.1*||[1.0-1.3]|
|35-44 years-old (n=7087)||40.5||0.8***||[0.7-0.9]||34.3||0.9||[0.8-1.0]|
|45-54 years-old (n=6172)||34.5||0.6***||[0.6-0.7]||29.6||0.8***||[0.7-0.8]|
|55-64 years-old (n=6322)||20.7||0.4***||[0.3-0.4]||17.3||0.4***||[0.4-0.5]|
|65-75 years-old (n=4161)||9.7||0.2***||[0.2-0.2]||7.4||0.2***||[0.2-0.3]|
|Less than high school (ref.) (n=16128)||34.0||- 1 -||28.9||- 1 -|
|High school graduate (n=6123)||34.2||0.9***||[0.8-0.9]||27.2||0.8***||[0.8-0.9]|
|College graduate (n=11294)||29.7||0.7***||[0.7-0.7]||21.8||0.6***||[0.6-0.6]|
|Employed (ref.) (n=19427)||37.7||- 1 -||31.2||- 1 -|
|Other non-working (n=1971)||35.6||1.1**||[1.0-1.3]||31.5||1.2***||[1.1-1.4]|
|No (ref.) (n=30189)||30.2||- 1 -||24.4||- 1 -|
|Quotas (ref.) (n=7621)||30.2||- 1 -||25.3||- 1 -|
As expected, in the model restricted to landline phone owners interviewed in the Health Barometer, being unlisted in the directory is related to a higher risk of being a smoker or a daily smoker (ORa=1.2 [1.1-1.3], p-value<0.001) after adjustment for gender, age, level of diploma and occupational status (Table 5).
|Current smoking||Daily smoking|
|%||aOR||95% CI||%||aOR||95% CI|
|Male (ref.) (n=9630)||35,6||- 1 -||28,6||- 1 -|
|15-24 years-old (ref.) (n=2414)||37.8||- 1 -||26.2||- 1 -|
|25-34 years-old (n=2980)||45.3||1.0||[0.8-1.1]||37.4||1.1||[0.9-1.3]|
|35-44 years-old (n=4558)||39.6||0.7***||[0.6-0.8]||32.9||0.8*||[0.7-1.0]|
|45-54 years-old (n=4110)||32.3||0.5***||[0.5-0.6]||27.5||0.7***||[0.6-0.8]|
|55-64 years-old (n=4752)||19.2||0.3***||[0.3-0.4]||16.0||0.4***||[0.3-0.5]|
|65-75 years-old (n=3138)||9.2||0.2***||[0.1-0.2]||6.7||0.2***||[0.2-0.3]|
|Less than high school (ref.) (n=10927)||32.2||- 1 -||27.2||- 1 -|
|High school graduate (n=3925)||34.3||0.9||[0.9-1.0]||26.4||0.8***||[0.8-0.9]|
|College graduate (n=7044)||28.8||0.7***||[0.7-0.8]||20.2||0.6***||[0.5-0.6]|
|Employed (ref.) (n=12378)||36.8||- 1 -||30.2||- 1 -|
|Other non-working (n=1204)||28.2||1.0||[0.9-1.2]||24.7||1.1||[0.9-1.2]|
|Type of phone line||***||***|
|Listed number (réf.) (n=11433)||29.2||- 1 -||23.2||- 1 -|
|Listed number except for direct marketing (n=2745)||31.2||0.9||[0.8-1.0]||23.4||0.9**||[0.8-1.0]|
|Unlisted number (n=7774)||35.4||1.2***||[1.1-1.3]||29.2||1.2***||[1.1-1.3]|
|Number of calls needed||***||***|
|1-2 (ref.) (n=6733)||28.2||- 1 -||22.5||- 1 -|
|Interview after initial refusal|
|No (ref.) (n=17462)||31.8||- 1 -||25.3||- 1 -|
The prevalence of current and daily tobacco smoking also dramatically increases with the number of calls needed before the interview is completed. Prevalence is about ten points higher for both indicators between people interviewed after just one or two calls and those for whom more than twenty calls were needed. Even though this association is strongly mediated by socioeconomic characteristics, a small effect of the needed number of calls remains after adjustment for the hardest-to-reach people, in particular above twenty calls (ORa=1.1 [1.0-1.3] for current smoking compared with people interviewed after just one or two calls, p<0.05; ORa=1.2 [1.0-1.3] for daily smoking, p-value<0.05). On the other hand, no relation was shown for people who accepted to answer after initial refusal at the household level.
Discussion / Conclusion
Facing the tobacco epidemic, monitoring trends in tobacco use is a crucial issue worldwide [17,18]. This issue raised major methodological concerns. On the one hand, former studies have shown how far estimates from different methodologies could differ within several countries . On the other hand, a previous analysis carried out in 1995 showed no significant difference between random and quota sampling for tobacco smoking prevalence in France . In our study, tobacco smoking prevalence as assessed in the Health Barometer is markedly higher than the one obtained in the quota sampling survey. The difference is almost four points for current smoking (33.9% vs. 30.2%) and more than two points for daily smoking (27.5% vs. 25.3%). When restricting the analysis to the listed landline phone owners, a one-point difference remains for current smoking, but not for daily smoking.
Even if a 2-point difference in daily smoking appears to be slight, it corresponds to the significant increase in smoking prevalence that was observed between 2005 and 2010 in France with the Health Barometer , and it must be taken into consideration. The insistence coefficient, i.e., the number of calls needed to contact hard-to-reach populations, could partly explain the observed differences.
Most authors agree that increased reporting of substance use is a sign of improved validity in the methodology, because these behaviors tend to be underreported [21-27]. Thus, random sampling appears to be now more effective than quota sampling, by making it possible to interview hard-to-reach populations, whose proportion is currently increasing in France  and who are more often smokers than the ones who easily answer to surveys. On the whole, some specific populations are under-represented in standard quotas surveys, for example: people with a low level of education, unemployed people... Furthermore, due to the recent evolution in telephony, this analysis confirms the need to include from now on, in prevalence survey sampling, individuals who are impossible to contact in classic phone quota surveys, i.e. mobile-only owners, unlisted landline owners, etc. Lastly, random surveys covering the whole population are needed to enable accurate international comparisons .
A monitoring system for smoking prevalence based on quota surveys would assume that their biases are constant from one period to another, which would mean that hard-to-reach populations have the same trend in prevalence as the rest of the population. Another possibility is to set up a quota survey including an insistence coefficient in order to include a number of hard-to-reach people.
To conclude, the present analysis by itself does not completely allow to decide on the possibility to use quotas surveys in order to monitor tobacco smoking prevalence. A new comparison between random and quotas survey is needed to see if the evolutions observed in both systems are consistent.
Conceived and designed the experiments: RG JLW JBR FB. Performed the experiments: RG JLW JBR FB. Analyzed the data: RG JBR. Contributed reagents/materials/analysis tools: RG JLW JBR FB. Wrote the manuscript: RG JLW JBR FB.
- 1. Beck F, Guignard R, Richard JB, Wilquin JL, Peretti-Watel P (2011) Increasing trends in smoking in France: main results of the French Health Barometer. Weekly. Epidemiol Bull 21-22: 101-103.
- 2. Cumming RG (1994) Is probability sampling always better? A comparison of results from a quota and a probability sample survey. Community Health Stud 14(2): 132-137.
- 3. Kish L (1949) A procedure for objective respondent selection within the household. J Am Stat Assoc 44: 380-387. doi:https://doi.org/10.1080/01621459.1949.10483314.
- 4. Orden SR, Dyer AR, Liu K, Perkins L, Ruth KJ et al. (1992) Random digit dialing in Chicago CARDIA : comparison of individuals with unlisted and listed telephone numbers. Am J Epidemiol 135: 697-709. PubMed: 1580246.
- 5. Gautier A, Beck F, Marder S, Legleye S, Riandey B et al. (2006) Mobile only: results of a method of partially random generated numbers. In: P. LavalléeL. Rivest. Survey Methods. European and North American Practices. Paris: Dunod, coll. Sciences Sup. pp. 60-63.
- 6. Hu SS, Balluz L, Battaglia MP, Frankel MR (2011) Improving public health surveillance using a dual-frame survey of landline and cell phone numbers. Am J Epidemiol 173(6): 703-711. doi:https://doi.org/10.1093/aje/kwq442. PubMed: 21343246.
- 7. Beck F, Legleye S, Peretti-Watel P (2004) Using the telephone in general population surveys on drugs. In: T. DecorteDJ Korf. European studies on drugs and drug policy. Brussels: VUB Press. pp.113-140.
- 8. Firdion JM (1995) Effet du rang d’appel et de la présence du conjoint dans une enquête par téléphone. Population 5: 1281-1314.
- 9. Turner CF, Lessler JT, Gfroerer JC (1992) Survey measurement of drug use, methodological issues. US Departement of Health and Human Services Pub, n°92-1929. Washington, DC: Government. Printing Office. 413 pp.
- 10. Faugier J, Sargeant M (1997) Sampling hard to reach populations. J Adv Nurs 26: 790-797. doi:https://doi.org/10.1046/j.1365-2648.1997.00371.x. PubMed: 9354993.
- 11. Van Loon AJ, Tijhuis M, Picavet HS, Surtees PG, Ormel J (2003) Survey non-response in the Netherlands: Effects on prevalence estimates and associations. Ann Epidemiol 13(2): 105-110. doi:https://doi.org/10.1016/S1047-2797(02)00257-0. PubMed: 12559669.
- 12. Caetano R, Ramisetty-Mikler S, McGrath C (2003) Characteristics of non-respondents in a US national longitudinal survey on drinking and intimate partner violence. Addiction 98(6): 791-797. doi:https://doi.org/10.1046/j.1360-0443.2003.00407.x. PubMed: 12780367.
- 13. Beck F, Gautier A, Guignard R, Richard JB (2013) Health Barometer 2010 Methods. French Institute for Health Promotion and Health Education (INPES). Available: . http://www.inpes.sante.fr/Barometres/barometre-sante-2010/pdf/Health-Barometer-2010-Methods.pdf. Accessed 24 September 2013.
- 14. Beck F, Gautier A, Guignard R, Richard JB (2011) Une méthode de prise en compte du dégroupage total dans le plan de sondage des enquêtes téléphoniques auprès des ménages. In: ME TremblayP. LavalléeM. El Hadj Tirari. Pratiques et Méthodes de sondage. Paris: Dunod, coll. Sciences Sup. pp. 310-314.
- 15. Husky MM, Guignard R, Beck F, Michel G (2013) Risk behaviors, suicidal ideation and suicide attempts in a nationally representative French sample. J Affect Disord (In press).
- 16. Peretti-Watel P, Constance J, Seror V, Beck F (2009) Cigarettes and social inequalities in France: is tobacco use increasingly concentrated among the Poor? Addiction 104(10): 1718-1728. doi:https://doi.org/10.1111/j.1360-0443.2009.02682.x. PubMed: 19681803.
- 17. World Health Organization (2003) WHO Framework Convention on Tobacco Control. Geneva: WHO. 36 pp.
- 18. Delnevo CD, Bauer UE (2009) Monitoring the tobacco use epidemic III. The host: data sources and methodological challenges. Prev Med 48: S16-S23. doi:https://doi.org/10.1016/j.ypmed.2008.09.008. PubMed: 18851990.
- 19. Bogdanovica I, Godfrey F, McNeill A, Britton J; Bogdanovica I, Godfrey F, McNeill A, Britton J (2011) Smoking prevalence in the European Union: a comparison of national and transnational prevalence survey methods and results. Tob Contr 20(1): e4. doi:https://doi.org/10.1136/tc.2010.036103. PubMed: 20966129. Epub 2010 Oct 21. doi:10.1136/tc.2010.041996. PubMed: 20966129.
- 20. Guilbert P, Baudier F, Arwidson P (1999) Comparison of two types of behavior and attitude surveys on alcohol, tobacco and illegal drug use. Rev Epidemiol Sante Publique 47(2): 129-138. PubMed: 10367300.
- 21. Turner CF, Ku L, Rogers SM, Lindberg LD, Pleck JH et al. (1998) Adolescent sexual behavior, drug use, and violence: increased reporting with computer survey technology. Science 280: 867-873. doi:https://doi.org/10.1126/science.280.5365.867. PubMed: 9572724.
- 22. Gmel G (2000) The effect of mode of data collection and of non-response on reported alcohol consumption: a split-sample study in Switzerland. Addiction 95: 123-134. doi:https://doi.org/10.1046/j.1360-0443.2000.95112313.x. PubMed: 10723837.
- 23. Midanik LT, Greenfield TK (2003) Telephone versus in-person interviews for alcohol use: results of the 2000 National Alcohol Survey. Drug Alcohol Depend 72: 209-214. doi:https://doi.org/10.1016/S0376-8716(03)00204-7. PubMed: 14643937.
- 24. Greenfield TK, Midanik LT, Rogers JD (2000) Effects of telephone versus face-to-face interview modes on reports of alcohol consumption. Addiction 95: 277-284. doi:https://doi.org/10.1046/j.1360-0443.2000.95227714.x. PubMed: 10723856.
- 25. Miller JW, Gfroerer JC, Brewer RD, Naimi TS, Mokdad A et al. (2004) Prevalence of adult binge drinking: a comparison of two national surveys. Am J Prev Med 27: 197-204. doi:https://doi.org/10.1016/j.amepre.2004.05.004. PubMed: 15450631.
- 26. Aquilino WS, Lo Sciouto LA (1990) Effect of interview mode on self-reported drug use. Public Opin Q 54: 362-393. doi:https://doi.org/10.1086/269212.
- 27. Grucza RA, Abbacchi AM, Przybeck TR, Gfroerer JC (2007) Discrepancies in estimates of prevalence and correlates of substance use and disorders between two national surveys. Addiction 102: 623-629. doi:https://doi.org/10.1111/j.1360-0443.2007.01745.x. PubMed: 17309538.
- 28. Beck F, Guilbert P, Gautier A, Arwidson P (2008) L’acharnement téléphonique dans les enquêtes est-il justifié ? In: P. GuilbertD. HazizaA. Ruiz-GazenY. Tillé. Méthodes de sondages - Cours et cas pratiques - Master, écoles d'ingénieurs. Paris: Dunod, coll. Sciences Sup. pp. 254-259.