You are here

Article Text

Article menu
Research paper
Proximity to a tobacco store and smoking cessation: a cohort study
  1. Jaana I Halonen1,
  2. Mika Kivimäki1,2,
  3. Anne Kouvonen3,4,
  4. Jaana Pentti1,
  5. Ichiro Kawachi5,
  6. S V Subramanian5,
  7. Jussi Vahtera1,6
  1. 1Finnish Institute of Occupational Health, Kuopio, Turku and Helsinki, Finland
  2. 2Department of Epidemiology and Public Health, University College London Medical School, London, UK
  3. 3School of Sociology, Social Policy and Social Work, Queen's University Belfast, Belfast, UK
  4. 4UKCRC Centre of Excellence for Public Health (NI), Queen's University Belfast, Belfast, UK
  5. 5Department of Society, Human Development and Health, Harvard School of Public Health, Boston, Massachusetts, USA
  6. 6Department of Public Health, University of Turku, Turku University Hospital, Turku, Finland
  1. Correspondence to Dr Jaana I Halonen, Finnish Institute of Occupational Health, Neulaniementie 4, Kuopio 70101, Finland; jaana.halonen{at}ttl.fi

Abstract

Background It is not clear whether the availability of tobacco affects the likelihood of smoking cessation. We examined whether the proximity to a tobacco store and the number of stores were associated with smoking cessation, and compared results for proximity variables based on walking and straight-line (as the crow flies) distance.

Methods The study population consisted of 8751 baseline smokers from the Finnish Public Sector study in 1997–2005. Smoking intensity (cigarettes/day) was determined at baseline and smoking cessation was determined from a follow-up survey in 2008–2009. Proximity was measured using straight-line and walking distance from home to the nearest tobacco store, and another exposure variable was the number of stores within 0.50 km from home. We calculated associations with log-binomial regression models, adjusting for individual-level and area-level confounders.

Results Of the participants, 3482 (39.8%) quit smoking during the follow-up (mean follow-up 5.5 years, SD 2.3 years). Among men who were moderate/heavy smokers at baseline and lived <0.50 km walking distance from the nearest tobacco store, the likelihood of smoking cessation was 27% (95% CI 12% to 40%) lower compared with those living ≥0.50 km from a store. Having even one store within 0.50 km walking distance from home decreased cessation in men who were moderate/heavy smokers by 37% (95% CI 19% to 51%). No decrease was found for men who were light smokers at baseline or for women.

Conclusions Living within walking distance of a tobacco store reduced the likelihood of smoking cessation among men who were moderate/heavy smokers.

  • Cessation
  • Environment
  • Public policy

https://doi.org/10.1136/tobaccocontrol-2012-050726

Statistics from Altmetric.com

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

    Introduction

    Smoking is one of the leading risk factors for premature mortality and disability in high-income and low-income countries.1 Estimates for yearly smoking-attributable deaths in the USA, for instance, are 200 000 for men and 180 000 for women.2 Smoking cessation could significantly decrease tobacco-related morbidity and mortality3 and therefore reliable information on modifiable factors that influence the likelihood of smoking cessation is of great public health relevance. Recent public health policies have aimed to reduce exposure to second-hand smoke by banning smoking in public places and workplaces, and these measures have been successful in decreasing tobacco-related morbidity.4 However, these benefits may be greater for non-smokers than for smokers themselves.5 Other preventive strategies for smoking have included tobacco advertising restrictions and higher taxation,6 but the role of availability of tobacco products has rarely been studied.

    Only a few studies have examined how the availability of tobacco products in the neighbourhood affect residents’ smoking,7 ,8 and even less is known about the effects of tobacco availability on smoking cessation.9 Another scarcely studied topic is the possibly different role of availability in cessation for female and male smokers.10 ,11 The results from tobacco availability studies and those on the availability of alcohol and alcohol consumption12 ,13 suggest that the physical availability of products that are harmful for health may be an independent factor influencing adverse health behaviours. However, further research is needed to reach firm conclusions. Moreover, although both straight-line (as the crow flies)13 and road12 distances have been used in studies examining the effects of service availability, comparison between these measures is lacking.

    This study examined whether the proximity of a tobacco store to home, and the number of stores within 0.50 km from home were associated with smoking cessation among adult men and women and whether these associations depended on baseline smoking intensity. We also compared the results for different measures of proximity using both straight-line and walking distances.

    Materials and methods

    Study setting and population

    The study data were obtained from the participants of the Finnish Public Sector study, an ongoing prospective cohort study of employees working in 10 towns and 6 hospital districts. The eligible population of the Finnish Public Sector study included all employees who had been working for the target organisations for a minimum of 6 months between 1991 and 2005 (n=151 618). The gender and age distribution of the cohort members is representative of Finnish public sector employees (75% vs 77% women; mean age 45 years). We obtained latitude and longitude of residences for 146 600 cohort members from 2000 to 2010 from the Finnish Population Register Centre. The first survey was completed in a nested cohort of current employees in 1997 after which surveys were repeated every 4 years, from 2000 to 2008, targeting those working at the participating organisations in the year of the survey. In 2005 and 2009, surveys were also mailed to those who had responded to a survey while employed but had later left the organisation. The response rate was on average 70%.

    This study included participants who had reported being current smokers in any of the surveys between 1997 and 2005, of which the latest survey was selected, who responded to a follow-up survey in 2008 or 2009, and whose residential address was available at the follow-up, a total of 8751 adults (figure 1). Due to non-response and deaths, 4664 participants were lost from the follow-up. The mean age in the lost group was lower (42.8 years) than in the follow-up group (44.7 years), there were more men in the lost group (30%) than in the follow-up group (24%), more manual workers (36% vs 25%) and more baseline moderate/heavy smokers (59% vs 54%). However, the proportion of those whose distance to the nearest tobacco store was <0.50 km was similar in both groups: 68% in the lost group and 67% in the follow-up group, assuming that tobacco stores were at the same locations at the beginning and end of the follow-up. The analytic sample differed slightly from the eligible population in terms of age (mean age 50 years in this sample vs 45 years in the eligible cohort) and the portion of those with low socioeconomic position (24% vs 18%), but not in terms of gender (76% vs 75% women). The Ethics Committee of the Hospital District of Helsinki and Uusimaa approved the study.

    Figure 1
    Figure 1

    Selection of study participants.

    Smoking cessation

    In each survey smoking status was measured using the following questions: ‘Do you smoke or have you previously smoked regularly, that is, daily or nearly daily?’ ‘If you have smoked, do you still smoke regularly?’ Smoking status was categorised as current smoker, ex-smoker or never smoker. Smoking intensity (‘How many cigarettes a day do you smoke (or did you smoke) on average?’) was also requested. Participants were defined as ‘a quitter’ if they reported being a current smoker in at least one of the surveys requesting smoking status between 1997 and 2005, but not in the follow-up in 2008 or 2009. Smoking 10 or more cigarettes per day was used as the definition for moderate/heavy smoking.14

    Availability of tobacco products

    From April 2009 onwards, tobacco retail has been subject to license in Finland. The addresses of all outlets with a tobacco retail license in 2010 were obtained from the Regional State Administrative Agency, the authority that maintains the license register. The register contained a total of 9772 addresses of licensed retail places, and the latitude and longitude coordinates for 92% of them were obtained through a geocoding service. Using the coordinates of the participants’ residences and the tobacco stores, the straight-line, ‘as the crow flies’ distance from each participant's home to the nearest tobacco store was calculated in SAS 9.2. Walking distance was determined using the closest facility tool in the ArcGIS 10.0 Network Analyst (Environmental Systems Research Institute), for which we used the road network database for the whole of Finland. This database included sidewalks for pedestrians, but also freeways that are restricted from pedestrians, which could have resulted in misclassification of walking distance for those living next to a freeway. However, as only 1% of the participants lived within 0.50 km of a freeway junction (ie access to a freeway), any effect on the results is likely to be minor. The number of stores within 0.50 km from home was calculated in two phases: first we calculated 0.50 km straight-line and walking distance buffers from participants’ home addresses and then linked the stores on these buffers.

    The mean straight-line distance from home to the nearest tobacco store was 0.76 (SD=1.8) km, and mean walking distance was 1.1 (SD=1.8) km. We categorised both distance variables into five classes: 0–0.20 km, 0.21–0.40 km, 0.41–0.60 km, 0.61–1.0 km and >1.0 km, based on the distribution of distance measures (as the two distributions varied we chose even numbers for the categories) and also dichotomised them as <0.50 km and ≥0.50 km, as in a prior study.9 The distances were further used to calculate indexes of availability using a formula that follows Gaussian dispersion, exp (−6(distance)1.2), and that gives weight to the shorter distances. The index value is 1 when the distance is 0, 0.5 when the distance is 0.15 km, and approaches 0 when the distance is 1 km or greater.15 As the distributions of the number of tobacco stores within 0.50 km from home (mean 1.4, range 0–33 for walking distance; and mean 2.9, range 0–51 for straight-line distance) were skewed, these variables were categorised as 0, 1, >1.

    Covariates

    Individual-level data from the employers’ registers included information on the participants’ age, gender and occupational status. Occupational status was used as an indicator of socioeconomic status (SES) and was split into three categories (high, intermediate and low) according to the Classification of Occupations by Statistics Finland, as in previous studies.14 Another indicator for SES was housing tenure (owner occupier vs other) that was obtained from the Population Register Center. In addition to smoking status and smoking intensity, we obtained marital status (married or cohabiting vs not) and habitual alcohol intake from the surveys. Heavy alcohol use was determined as 24 and 16 units of alcohol per week for men and women, respectively.16 Because chronic disease may influence smoking behaviour,17 we measured baseline diseases (yes/no) from the Finnish Drug Reimbursement Register and from the Finnish Cancer Registry.18

    Neighbourhood SES has been linked to differences in tobacco availability.19 The index for neighbourhood disadvantage was calculated using small-area data on median household income (coded as additive inverse), education attainment (percentage of people over 18 years old whose highest education level is elementary school) and unemployment rate (unemployed people belonging to the labour force/total labour force). As an additional area-level covariate, population density (inhabitants/km2) was used as a proxy for the degree of urbanisation. These data were obtained from Statistics Finland, which calculates socio-demographic information by map grid squares of 250×250 m (‘a neighbourhood’) covering the whole of Finland.20 All area-level data were based on the total population living in Finland at the end of 2008, and linked to individual-level data using the latitude/longitude of the participants’ residences and the coordinates on the 250 m map squares. On average, 2.6 participants lived in the same neighbourhood (range 1–29).

    Statistical analyses

    The associations between tobacco availability and smoking cessation were calculated using log-binomial regression with the generalised estimating equations (GEE) method with the participants’ neighbourhood as the clustering variable (GENMOD procedure of SAS 9.2). Because smoking habits and cessation may differ between women and men,10 ,11 and because smoking intensity is likely to modify the ability to quit smoking,21 we first examined the interactions of proximity measures with gender and smoking intensity.

    In the main analyses, the models for straight-line and walking distance as well as the number of stores were adjusted for age, SES, marital status, housing tenure, neighbourhood disadvantage and population density. As sensitivity analyses, models were further adjusted for chronic disease and heavy alcohol consumption, and the likelihood of loss of follow-up was examined with binomial logistic regression models (GENMOD procedure of SAS 9.2). Any cases with missing data were excluded from the analyses.

    To compare the GEE models’ goodness of fit, we examined the quasilikelihood under the independence model criterion (QICu) statistic that corresponds to Akaike's information criterion used for comparing models with likelihood-based methods. In the comparison of models’ QICu, the model with the smaller statistic is considered to have a better fit.22

    The results are presented as prevalence ratio (PR) with 95% CI of smoking cessation by the categorised distances (with >1.0 km as the reference); for living within <0.50 vs ≥0.50 km from the nearest store; and per increase from 0 to 1 in the index of availability, that is, when the distance decreases from 1.0 km to 0 m, but with more weight on the short distances. The results for the number of stores are presented as PR (95% CI) for 1 vs 0 and >1 vs 0 stores within a 0.50 km buffer.

    Results

    Of the participants, 6663 (76.1%) were women and the mean age at the last survey was 50.0 (SD=9.5) (table 1). A total of 451 (55.6%) men who quit had been moderate/heavy smokers prior to cessation, whereas among women who quit the corresponding figure was 1017 (38.1%).

    View this table:
    Table 1

    Descriptive statistics of the study population

    Interactions for gender with the categorised walking and straight-line distance to a tobacco store were not significant. However, interactions with gender were significant for the binary distance measures, for the indexes of availability and for categorised number of tobacco stores (p interactions <0.01 for each). Thus, all analyses were stratified by gender. As the interaction for baseline smoking intensity was also significant with <0.50 km walking distance (p<0.01), with the indexes of availability (p<0.05) and with the categorised number of stores (p<0.05), analyses were further stratified by baseline smoking intensity.

    Among men who were moderate/heavy smokers, an inverse association was observed between walking distance to the nearest tobacco store and smoking cessation for distances no longer than 0.6 km (PR 0.70–0.80), although it was significant for the category of 0.21–0.4 vs >1.0 km only (table 2). We found lower cessation prevalence for living within a <0.50 vs ≥0.50 km walking distance from a tobacco store (PR 0.73, 95% CI 0.60 to 0.88) (table 2). Proximity measured with the index of availability also showed lower cessation prevalence among moderate/heavy male smokers, although the effect estimate for walking distance index was not statistically significant (table 2). Having one versus no stores within 0.50 km from home decreased the likelihood of smoking cessation among men who were moderate/heavy smokers (PR 0.63, 95% CI 0.49 to 0.81). The association was negative, although slightly weaker, for having more than one store (vs no store) within 0.50 km from home (table 2). Adjustment for heavy alcohol use and chronic disease did not change the effect estimates (data not shown). OR for loss of follow-up was 0.95 (95% CI 0.80 to 1.14) among men who were moderate/heavy smokers at baseline and 1.18 (95% CI 0.89 to 1.59) among men who were light smokers at baseline, indicating no differences according to baseline smoking intensity. No associations were found between proximity and cessation in women who were moderate/heavy smokers at baseline (table 2).

    View this table:
    Table 2

    Prevalence ratios for smoking cessation among baseline moderate/heavy smokers in association with tobacco store proximity and availability

    The results using measures for straight-line distance were mainly similar to the results obtained for walking distance. Differences were that the men's effect estimate for the index of availability reached statistical significance, and the estimate for having one versus no stores within 0.5 km from home did not (see online supplementary appendix e-table 1). Based on the goodness of fit statistics (QICu), the walking (vs straight-line) distance models’ fit was slightly better; for example, <0.50 vs ≥0.50 km model for men: QICu for walking distance was 1649 and for straight-line distance 1661 (table 2; and see online supplementary e-table 1).

    No inverse association between proximity and smoking cessation were observed among male and female baseline light smokers. On the contrary, women living <0.50 km from a tobacco store had 8% higher likelihood of smoking cessation, and a similar association was observed for the index of availability (table 3). Having one or more tobacco stores within 0.50 km from home also gave positive although non-significant effect estimates in association with cessation in female light smokers. The results for women were in the same direction when using the straight-line distance measure, but did not reach statistical significance (see online supplementary appendix e-table 2). However, the goodness of fit statistics suggested better fit for the walking distance models (<0.50 vs ≥0.50 km model for women: QICu for walking distance 4459 and for straight-line distance 4473).

    View this table:
    Table 3

    Prevalence ratios for smoking cessation among baseline light smokers in association with tobacco store proximity and availability

    Discussion

    In this study, living in close proximity to a store selling tobacco products and having one or more tobacco stores within 0.50 km from home were associated with a decreased likelihood of smoking cessation among men who were moderate or heavy smokers at baseline. For women who were light smokers, we found some associations which were in the opposite direction. Measures based on walking versus straight-line distance had slightly better model fit. The results with different proximity variables were mainly comparable, although all the effect estimates for the five-class distance variables did not reach statistical significance, possibly due to the lower number of cases within each category.

    To our knowledge, only one previous study examined the effects of tobacco availability on smoking cessation. In Houston, Texas, living within a road network distance of 0.25 km or 0.50 km from the nearest tobacco store was associated with a decreased likelihood to quit smoking during a 26-week follow-up after cessation of treatment.9 Partly in agreement with these results, our findings were derived from a large observational cohort followed over 5 years. Tobacco availability was associated with cessation among men who smoked over 10 cigarettes per day at baseline, but proximity was not associated with decreased likelihood of quitting among male light smokers or female smokers of any intensity. However, previous studies suggest that heavy smokers may face more difficulties in cessation also for reasons other than availability, such as nicotine dependence,23 and that heavy smokers may not benefit from ‘collective quitting’, that is, when their spouse or friends quit, as much as light smokers do.24

    Tobacco store proximity and availability were associated with a decreased likelihood of quitting among men, but only for those who were moderate/heavy smokers at baseline. Among women, some measures of tobacco proximity and availability were associated with a higher likelihood of cessation, but only among those who were light smokers at baseline. This gender difference might seem surprising, as prior scarce studies on the reasons and mechanisms of gender differences have suggested that women have more cessation difficulties than men.10 ,11 In these data on employed adults there are differences in the gender distributions between jobs, and therefore the availability of tobacco in relation to workplace may have influenced the findings. Women may also more often work in organisations (eg, hospitals) that are smoke free and encourage cessation, which may be a more important factor than tobacco availability for cessation in women. Moreover, the observed associations were not particularly strong, and it may be that the potential to find such associations was limited by the relative ubiquity of tobacco stores among the study participants.

    The abundant number of stores also explains the small difference in the means of straight-line and walking distance to a tobacco store, which may have diminished the differences between models using either straight-line or walking distances. However, based on the goodness of fit statistics in this study, walking distance may be a more appropriate measure than straight-line distance for availability studies as straight-line distance tends to underestimate the true distance and time needed for travel.

    The policy implications of these data are not straightforward. Limiting the availability of tobacco products is not easy, although our findings suggest that the licensing authorities should aim to decrease the number of sales outlets to increase the prevalence of smoking cessation. In Finland, local authorities grant the licenses and determine the license fees. Thus, differences in the numbers of outlets and licensing fees may occur between local authorities. Other high-level policy strategies suggested for increasing smoking cessation include restrictions on tobacco advertising as this may increase smoking initiation among young people,25 and cessation difficulties among adults.26 In Finland, however, retail and other advertising of tobacco or its price has long been banned by the Tobacco Act,27 which suggests that local availability of tobacco products affects smoking cessation regardless of the restrictions on tobacco advertising. Other, more recent legal restrictions such as the smoking ban ‘in the joint and public indoor premises of workplaces’ took place in 2010,27 and therefore does not affect the data used in this study.

    One of the limitations of this study is the use of self-reported data to assess smoking, with no biological measures to validate self-reported cessation. This may have led to under-estimation of tobacco use. However, earlier studies have shown that self-report may be a valid and reliable way to measure smoking habits among Finns.28 Another limitation is that the data on tobacco stores were from 2010, whereas smoking cessation and availability measures were determined based on individual data from 2008 and 2009. The granting of licenses from April 2009 onwards reduced the total number of tobacco-selling outlets, but it has been estimated that most of these places were restaurants, in which tobacco retail was minor in the first place. However, it is possible that a reduced number of outlets over the follow-up period biased the associations of this study towards null. Further, we had no information on the success and length of abstinence, and limited information on nicotine dependence. As many attempts to quit fail,23 we may have overestimated the cessation rate. A possible effect modifier that was omitted in this study is the use of nicotine replacement therapy or other cessation treatments that were not requested in the surveys. We also had no information on the locations of the study participants’ workplaces, which may have caused some exposure misclassification if tobacco was more readily available at the workplace than at home. Finally, because the study population consisted of Finnish adults employed in the public sector, the generalisability of these results to other populations (eg, the unemployed or adolescents) and study locations may be limited. However, we did have a large study population and made adjustments for a number of individual-level and area-level confounders. Moreover, to our knowledge, this is the first study to compare various proximity variables.

    According to our data, men with a history of moderate/heavy smoking were less likely to quit smoking if they lived in close proximity to a tobacco store. Although walking distance measures revealed the associations slightly more clearly than straight-line distance measures, straight-line distance was not substantially inferior to walking distance in examining the effect of proximity to a tobacco store on smoking cessation.

    What this paper adds

    • It is not clear whether the availability of tobacco affects the likelihood of smoking cessation.

    • We found that close proximity to a tobacco store may reduce the likelihood of smoking cessation, particularly among men who were moderate/heavy smokers.

    References

    1. ↵
      WHO. Global Health Risks: Mortality and Burden of Disease Attributable to Selected Major Risks. Geneva: World Health Organization, 2009.
    2. ↵
      1. Rostron B
      . Smoking-attributable mortality by cause in the United States: revising the CDC's data and estimates. Nicotine Tob Res 2013;15:238–46.
      OpenUrlAbstract/FREE Full Text
    3. ↵
      1. Godtfredsen NS,
      2. Prescott E
      . Benefits of smoking cessation with focus on cardiovascular and respiratory comorbidities. Clin Respir J 2011;5:187–94.
      OpenUrlCrossRefPubMed
    4. ↵
      1. Menzies D
      . The case for a worldwide ban on smoking in public places. Curr Opin Pulm Med 2011;17:116–22.
      OpenUrlCrossRefPubMed
    5. ↵
      1. Bajoga U,
      2. Lewis S,
      3. McNeill A,
      4. et al
      . Does the introduction of comprehensive smoke-free legislation lead to a decrease in population smoking prevalence? Addiction 2011;106:1346–54.
      OpenUrlCrossRefPubMedWeb of Science
    6. ↵
      1. McGoldrick DE,
      2. Boonn AV
      . Public policy to maximize tobacco cessation. Am J Prev Med 2010;38:S327–32.
      OpenUrlCrossRefPubMed
    7. ↵
      1. Peterson NA,
      2. Lowe JB,
      3. Reid RJ
      . Tobacco outlet density, cigarette smoking prevalence, and demographics at the county level of analysis. Subst Use Misuse 2005;40:1627–35.
      OpenUrlCrossRefPubMedWeb of Science
    8. ↵
      1. Paul CL,
      2. Mee KJ,
      3. Judd TM,
      4. et al
      . Anywhere, anytime: retail access to tobacco in New South Wales and its potential impact on consumption and quitting. Soc Sci Med 2010;71:799–806.
      OpenUrlCrossRefPubMed
    9. ↵
      1. Reitzel LR,
      2. Cromley EK,
      3. Li Y,
      4. et al
      . The effect of tobacco outlet density and proximity on smoking cessation. Am J Public Health 2011;101:315–20.
      OpenUrlCrossRefPubMedWeb of Science
    10. ↵
      1. Amos A,
      2. Greaves L,
      3. Nichter M,
      4. et al
      . Women and tobacco: a call for including gender in tobacco control research, policy and practice. Tob Control 2012;21:236–43.
      OpenUrlAbstract/FREE Full Text
    11. ↵
      1. Ward KD,
      2. Klesges RC,
      3. Zbikowski SM,
      4. et al
      . Gender differences in the outcome of an unaided smoking cessation attempt. Addict Behav 1997;22:521–33.
      OpenUrlCrossRefPubMedWeb of Science
    12. ↵
      1. Kavanagh AM,
      2. Kelly MT,
      3. Krnjacki L,
      4. et al
      . Access to alcohol outlets and harmful alcohol consumption: a multi-level study in Melbourne, Australia. Addiction 2011;106:1772–9.
      OpenUrlCrossRefPubMedWeb of Science
    13. ↵
      1. Truong KD,
      2. Sturm R
      . Alcohol environments and disparities in exposure associated with adolescent drinking in California. Am J Public Health 2009;99:264–70.
      OpenUrlCrossRefPubMedWeb of Science
    14. ↵
      1. Kouvonen A,
      2. Vahtera J,
      3. Vaananen A,
      4. et al
      . Relationship between job strain and smoking cessation: the Finnish Public Sector Study. Tob Control 2009;18:108–14.
      OpenUrlAbstract/FREE Full Text
    15. ↵
      1. Pasanen K,
      2. Tissari S,
      3. Jänis M
      . Ympäristöterveysinformatiikka, Loppuraportti [Environmental Health Informatics]. Kuopio: University of Kuopio, 2005: 41.
    16. ↵
      1. Halonen JI,
      2. Kivimaki M,
      3. Virtanen M,
      4. et al
      . Living in proximity of a bar and risky alcohol behaviours: a longitudinal study. Addiction 2013;108:320–8.
      OpenUrlCrossRefPubMedWeb of Science
    17. ↵
      1. Bednarek M,
      2. Gorecka D,
      3. Wielgomas J,
      4. et al
      . Smokers with airway obstruction are more likely to quit smoking. Thorax 2006;61:869–73.
      OpenUrlAbstract/FREE Full Text
    18. ↵
      1. Salo P,
      2. Oksanen T,
      3. Sivertsen B,
      4. et al
      . Sleep disturbances as a predictor of cause-specific work disability and delayed return to work. Sleep 2010;33:1323–31.
      OpenUrlPubMedWeb of Science
    19. ↵
      1. Siahpush M,
      2. Jones PR,
      3. Singh GK,
      4. et al
      . Association of availability of tobacco products with socio-economic and racial/ethnic characteristics of neighbourhoods. Public Health 2010;124:525–9.
      OpenUrlCrossRefPubMedWeb of Science
    20. ↵
      Statistics Finland. Grid Database 2007. http://www.tilastokeskus.fi/tup/ruututietokanta/index_en.html (accessed 14 Feb 2013).
    21. ↵
      1. Klech HH
      . Reduced smoking—an acceptable goal for the hopeless heavy smoker? Eur Respir J 1998;11:263–4.
      OpenUrlCrossRefPubMedWeb of Science
    22. ↵
      1. Pan W
      . Akaike's information criterion in generalized estimating equations. Biometrics 2001;57:120–5.
      OpenUrlCrossRefPubMedWeb of Science
    23. ↵
      1. Aveyard P,
      2. West R
      . Managing smoking cessation. BMJ 2007;335:37–41.
      OpenUrlFREE Full Text
    24. ↵
      1. Christakis NA,
      2. Fowler JH
      . The collective dynamics of smoking in a large social network. N Engl J Med 2008;358:2249–58.
      OpenUrlCrossRefPubMedWeb of Science
    25. ↵
      1. Henriksen L,
      2. Schleicher NC,
      3. Feighery EC,
      4. et al
      . A longitudinal study of exposure to retail cigarette advertising and smoking initiation. Pediatrics 2010;126:232–8.
      OpenUrlAbstract/FREE Full Text
    26. ↵
      1. Paynter J,
      2. Edwards R
      . The impact of tobacco promotion at the point of sale: a systematic review. Nicotine Tob Res 2009;11:25–35.
      OpenUrlAbstract/FREE Full Text
    27. ↵
      Tobacco Act. Section 8, 1976. http://www.finlex.fi/en/laki/kaannokset/1976/en19760693.pdf (accessed 14 Feb 2013).
    28. ↵
      1. Vartiainen E,
      2. Seppala T,
      3. Lillsunde P,
      4. et al
      . Validation of self reported smoking by serum cotinine measurement in a community-based study. J Epidemiol Community Health 2002;56:167–70.
      OpenUrlAbstract/FREE Full Text
    View Abstract

    Supplementary materials

    • Supplementary Data

      This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

      Files in this Data Supplement:

    Footnotes

    • Contributors JIH, MK, AK, JP, IK, SVS and JV conceived and designed the experiments; JIH and JP analysed the data; MK, JP and JV contributed to materials and analysis tools; JIH, AK, JP and JV wrote the paper; MK, IK and SVS critically reviewed the article. All authors reviewed and approved the manuscript prior to submission.

    • Funding This study was supported by the Academy of Finland (projects 124271, 124322, 129262 and 126602). Professor Mika Kivimäki is supported by a professorial fellowship from the Economic and Social Research Council, UK.

    • Competing interests None.

    • Ethics approval The Ethics Committee of the Hospital District of Helsinki and Uusimaa has approved the study.

    • Provenance and peer review Not commissioned; externally peer reviewed.