- Email: [email protected]
Neighbourhood safety and smoking in population subgroups: The HELIUS study
Preventive Medicine 112 (2018) 111–118
Contents lists available at ScienceDirect
Preventive Medicine journal homepage: www.elsevier.com/locate/ypmed
Neighbourhood safety and smoking in population subgroups: The HELIUS study
T
⁎
Erik J. Timmermansa, , Eleonore M. Veldhuizenb, Marieke B. Snijdera,c, Martijn Huismand,e, Anton E. Kunsta a
Academic Medical Center, University of Amsterdam, Department: Public Health, Amsterdam Public Health research institute, Amsterdam, The Netherlands Department of Geography, Planning & International Development Studies, University of Amsterdam, Amsterdam, The Netherlands c Academic Medical Center, University of Amsterdam, Department: Clinical Epidemiology, Biostatistics and Bioinformatics, Amsterdam Public Health research institute, Amsterdam, The Netherlands d Department of Epidemiology and Biostatistics, Amsterdam Public Health research institute, VU University Medical Center, Amsterdam, The Netherlands e Department of Sociology, Faculty of Social Sciences, VU University, Amsterdam, The Netherlands b
A R T I C LE I N FO
A B S T R A C T
Keywords: Environmental epidemiology Geographic Information Systems HELIUS study Neighbourhood safety Population subgroups Smoking
This study examines the associations between neighbourhood safety and three types of smoking behaviour, and whether these associations differ by sex, age, ethnicity and individual-level socio-economic position. Baseline data (2011–2015) from the The HEalthy LIfe in an Urban Setting (HELIUS) study (Amsterdam, the Netherlands) were used. Smoking behaviour was based on self-report. Heavy smoking was defined as smoking ≥10 cigarettes per day. Nicotine dependence was assessed using the Fagerström questionnaire. Geographic Information System techniques were used to construct local residential areas and to examine neighbourhood safety for these areas using micro-scale environmental data. Multilevel logistic regression analyses with 6-digit zip code area as a second level were used to assess the association between neighbourhood safety and smoking. In our study sample of 22,728 participants (18–70 years), 24.0% were current smokers, 13.7% were heavy smokers and 8.1% were nicotine dependent individuals. Higher levels of neighbourhood safety were significantly associated with less heavy smoking (OR = 0.88, 95% CI = 0.78–0.99) and less nicotine dependence (OR = 0.81, 95% CI = 0.69–0.95), but not with less current smoking (OR = 1.01, 95% CI = 0.91–1.11). The associations between neighbourhood safety and the three types of smoking behaviour varied by ethnicity. For instance, higher levels of neighbourhood safety were associated with less current smoking in participants of African Surinamese origin (OR = 0.71, 95% CI = 0.57–0.89), but not in those of Dutch (OR = 1.13, 95% CI = 0.91–1.39), South-Asian Surinamese (OR = 1.22, 95% CI = 0.95–1.55), Turkish (OR = 1.08, 95% CI = 0.84–1.38), Moroccan (OR = 1.53, 95% CI = 1.12–2.10) or Ghanaian (OR = 1.18, 95% CI = 0.47–2.94) origin. Policies that improve neighbourhood safety potentially contribute to less heavy smoking and nicotine dependence.
1. Introduction Tobacco smoking is the primary preventable cause of disability and death worldwide (United States Department of Health Human Services, 2014). It is increasingly recognized that the neighbourhood where people live is associated with their health behaviours, including smoking (Goenka and Andersen, 2016; Pearce et al., 2012; Shareck and Ellaway, 2011). To create neighbourhood-level interventions that discourage smoking, it is important to understand which neighbourhood characteristics affect smoking behaviour and which specific population subgroups are particularly susceptible for these neighbourhood effects
(Pearce et al., 2012). Neighbourhood safety represents socio-cultural characteristics of the neighbourhood and includes various components, such as physical/ social disorder, crime-related safety and traffic-related safety (McGinn et al., 2008). It has been suggested that unsafe neighbourhood conditions may act as chronic stressors and as such, may influence smoking through pathways, such as stress and psychological well-being (Nielsen et al., 2008; Pearce et al., 2012; Shareck and Ellaway, 2011; Weden et al., 2008). Furthermore, it has been suggested that pro-smoking norms are operating more strongly in deprived and unsafe neighbourhoods than in affluent and safe neighbourhoods. In addition, it has been
⁎ Corresponding author at: Academic Medical Center, University of Amsterdam, Department: Public Health, Amsterdam Public Health research institute, PO Box 22660, 1100 DD Amsterdam, The Netherlands. E-mail address: [email protected] (E.J. Timmermans).
https://doi.org/10.1016/j.ypmed.2018.04.012 Received 5 October 2017; Received in revised form 20 March 2018; Accepted 6 April 2018 Available online 12 April 2018 0091-7435/ © 2018 Elsevier Inc. All rights reserved.
Preventive Medicine 112 (2018) 111–118
E.J. Timmermans et al.
et al., 2017; Stronks et al., 2013). Briefly, baseline data were collected from January 2011 to December 2015. Participants aged 18–70 years were randomly sampled, stratified by ethnicity, through the municipality registry of Amsterdam. In total, 90,019 individuals were invited to participate in the HELIUS study. We were able to contact and get a response from 49,952 (55%) invited persons, either by response card or after a home visit by an ethnically matched interviewer. Of those, 24,789 individuals (50%) agreed to participate. Non-response analyses showed that socio-economic differences between participants and nonparticipants were very small (Snijder et al., 2017). Participants received a confirmation letter of the appointment for the physical examination, including a digital or paper version of the questionnaire (depending on the preference of the subject). Of the 24,789 participants, 23,942 (96.6%) participants completed the questionnaire. The HELIUS study has been approved by the Institutional Review Board of the Academic Medical Center, University of Amsterdam. For the current study, from the total sample of participants who completed the questionnaire (n = 23,942), we excluded participants who were aged over 70 year (n = 11) and those who had no full data available on all outcome measures (n = 395), neighbourhood safety (n = 236) or both (n = 3). Furthermore, we excluded participants of Javanese Surinamese (n = 243) and unknown Surinamese origin (n = 276) from the remaining sample because of the small sample sizes. We also excluded participants with an unknown or other ethnic origin (n = 50). Consequently, the analytical sample consisted of 22,728 individuals who lived across 8851 6-digit zip code areas. The average number of participants per 6-digit zip code area was 7, with a range of 1–92.
suggested that tobacco products are more readily available in deprived and unsafe neighbourhoods than in affluent and safe residential areas This may make it harder for individuals to quit smoking in deprived and unsafe neighbourhoods than in affluent and safe neighbourhoods (Ellaway and Macintyre, 2009; Giskes et al., 2006; Stead et al., 2001; Wiltshire et al., 2001). Previous studies investigating the association between neighbourhood safety and smoking behaviour are inconclusive (Diez Roux and Mair, 2010). Some studies showed that individuals who are exposed to higher levels of objectively measured crime (Shareck and Ellaway, 2011; Tsjeng et al., 2001) or perceive lower levels of neighbourhood safety (Ellaway and Macintyre, 2009; Ganz, 2000; Miles, 2006; Patterson et al., 2012) are more likely to be current smokers, but other studies found no associations (Diez Roux and Mair, 2010; Jitnarin et al., 2015; Rachelle et al., 2016; Virtanen et al., 2007). Several methodological limitations are presented among the current body of evidence. Particular problems include small study samples (e.g., Ganz, 2000; Tsjeng et al., 2001), ignorance of data-clustering at the environmentallevel (e.g., Ellaway and Macintyre, 2009; Patterson et al., 2012), and no adjustment for environmental-level characteristics, such as socio-economic status (e.g., Ellaway and Macintyre, 2009; Shareck and Ellaway, 2011). It remains undocumented whether the association between neighbourhood safety and smoking is stronger in some specific population subgroups than in others. Women, older adults, ethnic minorities and individuals with a low socio-economic position more often report poor neighbourhood safety than their counterparts (Skogan and Maxfield, 1981). It has been suggested that these specific groups are more exposed to their local living environment and are more aware, or sensitive to, what happens in their neighbourhood (Ellaway and Macintyre, 2009; Miles, 2006; Shareck and Ellaway, 2011; Stafford et al., 2005; Yen et al., 2009). These groups are suggested to be more dependent upon their immediate – “local” – environment and to be more concerned about safety because of their limited mobility and higher physical and social vulnerability (Skogan and Maxfield, 1981; Stafford and Marmot, 2003). A limited number of studies suggest that lower levels of neighbourhood safety are more strongly associated with current smoking in women and older adults (Dyck et al., 2001; Ellaway and Macintyre, 2009; Miles, 2006; Shareck and Ellaway, 2011). Studies examining the differential effects of neighbourhood safety on smoking behaviour across ethnic groups and socio-economic positions are lacking. This large-scale population-based study extends previous research and aimed to examine whether lower levels of neighbourhood safety are associated with three types of smoking behaviour, including current smoking, heavy smoking, and nicotine dependence, while adjusting for a range of individual- and environmental-level characteristics and for the clustering of participants within neighbourhoods. Furthermore, this study aimed to examine whether the associations between neighbourhood safety and smoking behaviour differ by sex, age, ethnicity and individual-level socio-economic position. We expected that higher levels of safety are associated with less current/heavy smoking and less nicotine dependence. Furthermore, we expected that higher levels of safety are more strongly associated with less smoking in women, older adults, ethnic minority groups and individuals with a low socio-economic position than in their counterparts.
2.2. Dependent variables 2.2.1. Smoking status Three measures were used to assess different smoking behaviours. Current smoking was assessed as currently smoking one or more tobacco products (i.e., cigarettes, cigars and/or pipe tobacco) (0 = no, 1 = yes). Heavy smoking was defined as smoking ≥10 cigarettes daily (0 = no, 1 = yes). For this measure, the number of cigars and packages of pipe tobacco were converted to cigarettes based on the tobacco content (i.e., 1 cigar is similar to 3 cigarettes, and 1 package of pipe tobacco is similar to 62.5 cigarettes). Often, studies use a cut-off of ≥20 cigarettes per day for heavy smoking (Neumann et al., 2013), which may represent nicotine dependence. To clearly distinguish between the smoking behaviours, we decided to use the cut-off ≥10 cigarettes per day (Visser et al., 2017). Non-smokers were classified as ‘no heavy smokers’. Nicotine dependence was determined by the Fagerström scale (Heatherton et al., 1991) consisting of six questions (e.g., ‘Do you find it hard not to smoke in places where it is not allowed?’). The sum score varied from 0 to 10, with a cut-off of ≥4 considered nicotine dependence (0 = no, 1 = yes). If one of the items was missing, the Fagerström sum score was calculated with a score of 0 for the missing item. If more than one item was missing the Fagerström sum score was coded as missing. For non-smokers the sum score is 0. This approach corresponds with previous research (Visser et al., 2017). 2.3. Independent variables
2. Methods 2.3.1. Neighbourhood safety A neighbourhood safety score was assessed by using data from the Amsterdam Safety Monitor 2013/2014 (ASM) maintained by the Municipality of Amsterdam (Municipality of Amsterdam, 2017; Smeets, 2015). Participants of the ASM were asked how they perceive safety in their neighbourhood on a scale from 1 to 10, with higher scores indicating higher levels of neighbourhood safety. More details about the assessment of neighbourhood safety in the ASM are described
2.1. Design and study sample Baseline data from the HEalthy LIfe in an Urban Setting (HELIUS) study were used. The HELIUS study is a multi-ethnic cohort study conducted in Amsterdam, the Netherlands, including Dutch, Surinamese, Turkish, Moroccan, and Ghanaian origin ethnic groups. The study protocol has been described in detail elsewhere (Snijder 112
Preventive Medicine 112 (2018) 111–118
E.J. Timmermans et al.
household financial difficulties (1). All four indicators of individuallevel socio-economic position were considered as potential effect modifier separately.
elsewhere (Municipality of Amsterdam, 2017; Smeets, 2015). Geographic Information Systems were used to assign a neighbourhood safety score to each included HELIUS participant. First, the centroid of the 6-digit zip code area in which the HELIUS participants were living were geocoded by using ArcGIS 10.1 (ESRI Inc., Redlands, California, United States of America (USA)). In the Netherlands, 6-digit zip code areas are the smallest geographical units available and are approximately sized 50 × 50 m and include 10 to 20 households (Veldhuizen et al., 2013). A buffer zone with a radius of 300 m was generated around each individual centroid. Subsequently, 300-m buffer zones were also created around the residential addresses of all ASM participants. Based on the extent of overlap (i.e., the percentage of overlap) between the buffer zones of HELIUS participants and ASM participants, a weighted average neighbourhood safety score was assigned to the buffer zones of the HELIUS participants. For this process, an overlay operation was performed, which joins data layers based on common geographical location. The average number of overlaying ASM buffer zones that were used to assess the neighbourhood safety score for each HELIUS-participant was 234.3 (SD = 143.1), with a range of 2.0–737.0.
2.5. Potential confounders at the individual-level If applicable, age, sex, ethnicity and individual-level socio-economic position were considered as confounders in the analyses. Partner status (0 = no partner, 1 = partner) was additionally considered as a potential individual-level confounder.
2.6. Potential confounders at the environmental-level At the environmental-level, the following potential confounders were considered: population and household density, and socio-economic status. These data were derived from integral demographic and socio-economic registries maintained by the Municipality of Amsterdam. Population density was calculated as the number of residents in the 300-meter buffer zones and categorized into: (1) Low (Tertile 1 (≤2662) (reference category)), (2) Intermediate (Tertile 2 (2663–3942)), and (3) High (Tertile 3 (≥3943)). Household density was calculated as the number of households in the 300-meter buffer zones and categorized into: (1) Low (Tertile 1 (≤1156) (reference category)), (2) Intermediate (Tertile 2 (1157–1812)), and (3) High (Tertile 3 (≥1813)). Environmental-level socio-economic status was based on the proportion of households living on a minimum income and the average property value of dwellings in buffer zones (Veldhuizen et al., 2013). The proportion of households living on a minimum income was categorized into: (1) Low (Tertile 1 (≤16.9%), (2) Intermediate (Tertile 2 (17.0%–24.6%)), and (3) High (Tertile 3 (≥24.7%) (reference category)). The average property value of dwellings was categorized into: (1) Low (Tertile 1 (≤€166,660) (reference category)), (2) Intermediate (Tertile 2 (€166,661–€212,270)), and (3) High (Tertile 3 (≥€212,271)).
2.4. Potential effect modifiers The following potential effect modifiers were considered: age, sex (0 = men, 1 = women), ethnicity, and individual-level socio-economic position. Age was dummy-coded (0 = no, 1 = yes) into: (1) 18–30 years (reference category), (2) 31–40 years, (3) 41–50 years, (4) 51–60 years, and (5) 61–70 years. Participant's ethnicity was defined according to the country of birth of the participant as well as that of the parents, which is currently the most widely accepted and most valid assessment of ethnicity in the Netherlands (Stronks et al., 2009). Participants were considered of Dutch origin if the participant and both parents were born in the Netherlands. Participants were considered of non-Dutch ethnic origin if either they themselves were born outside the Netherlands and at least one of their parents (first-generation), or they themselves were born in the Netherlands but at least one of their parents was born outside the Netherlands (second-generation). Of the Surinamese immigrants in the Netherlands, approximately 80% are either African or South-Asian origin. Surinamese subgroups were classified according to self-reported ethnic origin. Ethnicity groups were dummy-coded (0 = no, 1 = yes) into: (1) Dutch (reference category), (2) South-Asian Surinamese, (3) African Surinamese, (4) Turkish, (5) Moroccan, and (6) Ghanaian origin. Individual-level socio-economic position was assessed using data on educational level, employment status, occupational level and household financial difficulties. Educational level was dummy-coded (0 = no, 1 = yes) into: (1) Low (no education or elementary education (reference category)), (2) Low-intermediate (lower vocational education or lower secondary education), (3) High-intermediate (intermediate vocational education or intermediate/higher secondary education), and (4) High (higher vocational education or university education). Employment status was dummy-coded (0 = no, 1 = yes) into: (1) incapacitated for work (reference category), (2) not in labour force (e.g., retired), (3) not employed, and (4) employed. Occupational level was classified according to the Dutch Standard Occupational Classification system for 2010 (Statistics Netherlands, 2010). Based on this classification system, occupational level was dummy-coded (0 = no, 1 = yes) into: (1) Low (elementary and lower (reference category)), (2) Intermediate (intermediate), and (3) High (higher and academic). Household financial difficulties were assessed by asking whether participants had problems managing their household income. The response categories ‘no, no problems at all’ and ‘no problems, but I have to pay attention to what I spend’ were considered as having no household financial difficulties (0), whereas the response categories ‘Yes, some problems’ and ‘Yes, a lot of problems’ were considered as having
2.7. Statistical analyses Multilevel logistic regression analyses with 6-digit zip code area as a second level were performed to examine the associations between neighbourhood safety and the three individual outcome measures on smoking. By conducting a multi-level study, the clustering of observations (level-1 unit) within the same 6-digit zip code area (level-2 unit) has been taken into account. To examine whether the associations differ by age, sex, ethnicity and individual-level socio-economic position, the analyses were stratified by subgroups of HELIUS participants. Interaction terms between neighbourhood safety and these variables were assessed in fully adjusted models to test for statistical significance. The interaction terms were considered as significant at a p-value below 0.10 (Aiken and West, 1991). The associations between neighbourhood safety and the individual outcome measures on smoking were examined in models constructed step by step. Model 1 adjusted the effects of neighbourhood safety on smoking status for age, sex and ethnicity. Model 2 additionally adjusted for all other individual-level characteristics (i.e., all four indicators of individual-level socio-economic position, and partner status). Model 3 additionally adjusted for all environmental-level characteristics (i.e., population and household density, and environmental-level socio-economic status). In all models, a probability level of ≤5.0% was regarded as statistically significant. Statistical analyses were performed in IBM SPSS Statistics (Version 24; IBM Corp, Armonk, New York, USA). 113
Preventive Medicine 112 (2018) 111–118
E.J. Timmermans et al.
Table 1 Characteristics of the study sample of the HELIUS study, Amsterdam, the Netherlands, 2011–2015.a
All participants Sex Men Women Age-group 18–30 years 31–40 years 41–50 years 51–60 years 61–70 years Ethnicity (%) Dutch South-Asian Surinamese African Surinamese Turkish Moroccan Ghanaian Individual-level socio-economic position (%) Educational level Low Low-intermediate High-intermediate High Employment status Incapacitated for work Not in the labour force Not employed Employed Occupational level Low Intermediate High Household financial difficulties No financial difficulties Financial difficulties Partner status (%) No partner Partner Environmental-level population density (%) Low Intermediate High Environmental-level household density (%) Low Intermediate High Environmental-level socio-economic status – proportion of minimum income households (%) Low Intermediate High Environmental-level socio-economic status – average property value of dwellings (%) Low Intermediate High a
All participants
Current smokers
Heavy smokers
Nicotine dependent individuals
N
(%)
(%)
(%)
24.0
13.7
8.1
42.3 57.7
32.5 17.8
20.9 8.5
12.0 5.2
20.9 18.4 25.5 24.4 10.8
25.6 26.8 23.8 23.2 18.5
12.0 14.2 14.9 14.5 11.9
6.7 8.9 9.5 8.3 5.8
20.1 14.4 18.9 17.3 18.7 10.6
24.7 29.6 31.7 34.4 13.2 3.6
12.2 17.3 16.0 22.7 8.9 1.4
6.5 12.0 8.1 14.2 5.2 1.0
17.8 26.5 29.6 26.1
19.9 30.1 25.1 19.6
13.0 19.8 13.6 8.2
8.7 12.1 7.6 4.2
7.7 18.1 14.5 59.7
26.5 16.4 31.2 24.4
17.4 8.0 20.3 13.4
13.2 4.1 13.3 7.4
46.6 26.6 26.8
27.5 27.2 20.4
17.3 15.9 8.7
10.6 9.4 4.5
61.3 38.7
20.5 30.0
10.7 18.7
5.7 12.0
50.3 49.7
27.8 20.3
15.5 12.0
9.9 6.4
33.3 33.4 33.3
22.5 23.0 26.6
12.2 13.4 15.5
7.0 7.9 9.4
33.4 33.3 33.3
22.5 22.7 26.9
12.6 13.0 15.6
7.0 7.9 9.4
33.3 33.3 33.4
23.1 25.5 23.5
13.7 15.2 12.3
8.5 9.2 6.6
33.4 33.3 33.3
23.4 24.0 24.6
13.7 14.1 13.3
8.6 8.0 7.7
(%)
22,728 22,728 9616 13,112 22,728 4754 4182 5776 5552 2464 22,728 4556 3279 4296 3940 4246 2411 22,544 4003 5979 6682 5880 22,478 1726 4062 3249 13,441 19,054 8882 5074 5098 22,286 13,656 8630 22,608 11,374 11,234 22,726 7576 7600 7550 22,727 7577 7575 7575 22,708 7566 7563 7579 22,728 7578 7577 7573
Abbreviation: N = number of valid observations.
3. Results
current smokers, heavy smokers and nicotine dependent individuals differed across various population subgroups (Table 1). In the full sample, the average neighbourhood safety score was 6.2 (SD = 0.5), with a range of 5.0–7.8 (Fig. 1).
The mean age of all 22,728 participants was 43.8 (SD = 13.4), with an age-range of 18–70. In the full sample, 57.7% were women. Of all participants, 20.1% were Dutch, 14.4% were South-Asian Surinamese, 18.9% African Surinamese, 17.3% were Turkish, 18.7% were Moroccan, and 10.6% were Ghanaian. Most participants had a highintermediate educational level (29.6%), were employed (59.7%), and experienced no household financial difficulties (61.3%). Furthermore, most participants had a low occupational level (46.6%) (Table 1). The study sample consisted of 24.0% current smokers, 13.7% heavy smokers and 8.1% nicotine dependent individuals. The proportions of
3.1. Neighbourhood safety and current smoking After adjustment for individual- and environmental-level confounders, neighbourhood safety was not statistically significantly associated with current smoking in the full sample (OR = 1.01, 95% CI = 0.91–1.11) (Table 2; Model 3). The association of neighbourhood safety with current smoking 114
Preventive Medicine 112 (2018) 111–118
E.J. Timmermans et al.
Fig. 1. Prevalence rate of current smokers, heavy smokers and nicotine dependent individuals by deciles of neighbourhood safety in the HELIUS study, Amsterdam, the Netherlands, 2011–2015.a
in those with a low educational level (OR = 1.14, 95% CI = 0.79–1.63). There was no statistical evidence that the association between neighbourhood safety and heavy smoking was modified by sex, age or the other individual-level socio-economic factors (Table 3; Model 3).
varied by ethnicity. Higher levels of neighbourhood safety were associated with less smoking in participants of African Surinamese origin (OR = 0.71, 95% CI = 0.57–0.89), but not in those of Dutch (OR = 1.13, 95% CI = 0.91–1.39), South-Asian Surinamese (OR = 1.22, 95% CI = 0.95–1.55), Turkish (OR = 1.08, 95% CI = 0.84–1.38), Moroccan (OR = 1.53, 95% CI = 1.12–2.10) or Ghanaian (OR = 1.18, 95% CI = 0.47–2.94) origin (Table 2; Model 3). There was no statistical evidence that the association between neighbourhood safety and current smoking was modified by sex, age or individual-level socio-economic position (Table 2; Model 3).
3.3. Neighbourhood safety and nicotine dependence After adjustment for individual- and environmental-level confounders, higher levels of neighbourhood safety were statistically significantly associated with less nicotine dependence in the full sample (OR = 0.81, 95% CI = 0.69–0.95) (Table 4; Model 3). The association of neighbourhood safety with nicotine dependence varied by age. Higher levels of safety were associated with less nicotine dependence in participants aged 18–30 years (OR = 0.54, 95% CI = 0.36–0.81), 31–40 years (OR = 0.92, 95% CI = 0.64–1.33), 41–50 years (OR = 0.84, 95% CI = 0.62–1.13), and 51–60 years (OR = 0.77, 95% CI = 0.56–1.05), but not in those aged 61–70 years (OR = 1.11, 95% CI = 0.64–1.90). The association of neighbourhood safety with nicotine dependence also varied by ethnicity. Higher levels of neighbourhood safety were associated with less nicotine dependence in participants of Dutch (OR = 0.91, 95% CI = 0.62–1.33), South-Asian Surinamese (OR = 0.74, 95% CI = 0.53–1.04), and African Surinamese origin (OR = 0.51, 95% CI = 0.35–0.73), and Ghanaian (OR = 0.69, 95% CI = 0.09–5.30) origin, but not in those of Turkish (OR = 1.07, 95% CI = 0.76–1.49) or Moroccan (OR = 1.58, 95% CI = 0.99–2.53) origin (Table 4; Model 3). There was no statistical evidence that the association between neighbourhood safety and heavy smoking was modified by sex or individual-level socio-economic position (Table 4; Model 3).
3.2. Neighbourhood safety and heavy smoking After adjustment for individual- and environmental-level confounders, higher levels of neighbourhood safety were statistically significantly associated with less heavy smoking (OR = 0.88, 95% CI = 0.78–0.99) (Table 3; Model 3). Ethnicity was found to be an effect modifier in the association between neighbourhood safety and heavy smoking. Higher levels of neighbourhood safety were associated with less heavy smoking in participants of South-Asian Surinamese (OR = 0.77, 95% CI = 0.58–1.04), African Surinamese (OR = 0.61, 95% CI = 0.46–0.81), and Ghanaian (OR = 0.35, 95% CI = 0.09–1.42) origin, but not in those of Dutch (OR = 1.13, 95% CI = 0.85–1.50), Turkish (OR = 1.16, 95% CI = 0.88–1.54) or Moroccan (OR = 1.56, 95% CI = 1.07–2.27) origin (Table 3; Model 3). Educational level was also found to be an effect modifier in the association between neighbourhood safety and heavy smoking. Higher levels of neighbourhood safety were associated with less heavy smoking in participants with a low-intermediate (OR = 0.83, 95% CI = 0.67–1.03), high-intermediate (OR = 0.86, 95% CI = 0.68–1.08), and high educational level (OR = 0.86, 95% CI = 0.65–1.14), but not
4. Discussion This study aimed to examine the associations between neighbourhood safety and three types of smoking behaviour in a large sample of adults
a The neighbourhood safety score ranges from 1 to 10, with higher scores representing higher levels of neighbourhood safety.
115
Preventive Medicine 112 (2018) 111–118
E.J. Timmermans et al.
Table 2 Associations between neighbourhood safety score (range = 1–10) and current smoking in the HELIUS study, Amsterdam, the Netherlands, 2011–2015a,b.
Full sample All participants Sex Men Women Age 18–30 years 31–40 years 41–50 years 51–60 years 61–70 years Ethnicity Dutch South-Asian Surinamese African Surinamese Turkish Moroccan Ghanaian Individual-level socioeconomic position Educational level Low Low-intermediate High-intermediate High Employment status Incapacitated for work Not in the labour force Not employed Employed Occupational level Low Intermediate High Household financial difficulties No financial difficulties Financial difficulties
Table 3 Associations between neighbourhood safety score (range = 1–10) and heavy smoking in the HELIUS study, Amsterdam, the Netherlands, 2011–2015a,b.
Model 1
Model 2
Model 3
Model 1
Model 2
Model 3
OR
95% CI
OR
95% CI
OR
95% CI
OR
95% CI
OR
95% CI
OR
95% CI
0.96
0.90–1.03
1.05
0.97–1.13
1.01
0.91–1.11
0.87
0.80–0.94
0.96
0.88–1.06
0.88
0.78–0.99
0.96 0.97
0.88–1.06 0.88–1.08
1.01 1.10
0.91–1.12 0.98–1.23
1.00 1.05
0.87–1.15 0.90–1.22
0.90 0.83
0.81–1.01 0.73–0.95
0.97 0.98
0.86–1.09 0.84–1.13
0.85 0.99
0.72–0.99 0.81–1.22
1.16 1.08 0.86 0.94 0.77
1.02–1.37 0.93–1.26 0.75–0.99 0.81–1.08 0.62–0.95
1.28 1.13 0.96 1.02 0.88
1.08–1.51 0.96–1.34 0.82–1.12 0.87–1.19 0.69–1.12
1.17 0.98 0.92 1.04 0.99
0.92–1.48 0.78–1.23 0.75–1.13 0.84–1.28 0.71–1.38
0.94 0.98 0.85 0.88 0.71
0.77–1.14 0.81–1.19 0.72–1.00 0.75–1.05 0.55–0.92
0.98 1.07 0.95 0.99 0.82
0.78–1.22 0.87–1.33 0.79–1.13 0.83–1.19 0.62–1.09
0.79 0.91 0.79 1.01 0.90
0.58–1.08 0.68–1.22 0.62–1.01 0.78–1.30 0.61–1.34
0.84 1.11
0.74–0.96 0.93–1.33
1.05 1.19
0.91–1.21 0.98–1.46
1.13 1.22
0.91–1.39 0.95–1.55
0.64 0.88
0.55–0.76 0.71–1.10
0.90 0.96
0.74–1.08 0.75–1.22
1.13 0.77
0.85–1.50 0.58–1.04
0.79 0.99 1.54 1.02
0.67–0.92 0.86–1.14 1.28–1.85 0.53–1.94
0.85 1.03 1.49 0.86
0.71–1.00 0.87–1.22 1.22–1.83 0.41–1.84
0.71 1.08 1.53 1.18
0.57–0.89 0.84–1.38 1.12–2.10 0.47–2.94
0.77 1.03 1.39 0.55
0.63–0.95 0.88–1.20 1.12–1.73 0.20–1.57
0.83 1.10 1.37 0.41
0.67–1.03 0.92–1.33 1.08–1.73 0.12–1.41
0.61 1.16 1.56 0.35
0.46–0.81 0.88–1.54 1.07–2.27 0.09–1.42
0.97 0.98 1.26 0.97
0.80–1.18 0.85–1.11 1.11–1.42 0.85–1.11
0.96 0.98 1.25 0.99
0.77–1.21 0.85–1.14 1.09–1.43 0.85–1.14
1.05 0.90 1.17 0.95
0.76–1.45 0.74–1.09 0.97–1.40 0.78–1.15
1.10 0.96 1.03 0.85
0.88–1.37 0.83–1.12 0.88–1.20 0.70–1.03
1.17 0.96 1.02 0.86
0.91–1.51 0.82–1.13 0.86–1.20 0.70–1.05
1.14 0.83 0.86 0.86
0.79–1.63 0.67–1.03 0.68–1.08 0.65–1.14
1.15
0.90–1.47
1.06
0.81–1.39
1.13
0.77–1.67
1.08
0.81–1.42
1.02
0.75–1.40
0.98
0.63–1.53
0.95
0.80–1.14
1.12
0.89–1.41
1.12
0.82–1.54
0.75
0.59–0.96
0.91
0.67–1.22
0.86
0.58–1.29
1.05 0.94
0.88–1.25 0.86–1.03
1.13 1.02
0.93–1.39 0.93–1.12
1.20 0.94
0.90–1.60 0.83–1.07
1.05 0.83
0.86–1.28 0.74–0.93
1.20 0.91
0.96–1.50 0.81–1.02
1.21 0.80
0.87–1.68 0.69–0.94
1.02 1.13 0.99
0.91–1.14 0.98–1.29 0.85–1.14
1.02 1.15 1.02
0.91–1.15 1.00–1.32 0.88–1.18
1.02 1.02 0.99
0.87–1.19 0.84–1.23 0.81–1.21
1.05 0.95 0.81
0.92–1.20 0.81–1.12 0.66–0.99
1.05 0.97 0.86
0.92–1.20 0.82–1.14 0.70–1.06
0.92 0.87 0.84
0.77–1.11 0.69–1.09 0.64–1.12
0.95
0.87–1.04
1.01
0.91–1.11
0.95
0.83–1.08
0.80
0.71–0.90
0.88
0.77–1.00
0.82
0.70–0.98
1.04
0.93–1.15
1.11
0.99–1.25
1.10
0.93–1.30
0.99
0.87–1.12
1.08
0.94–1.24
0.96
0.79–1.16
Full sample All participants Sex Men Women Age 18–30 years 31–40 years 41–50 years 51–60 years 61–70 years Ethnicity Dutch South-Asian Surinamese African Surinamese Turkish Moroccan Ghanaian Individual-level socioeconomic position Educational level Low Low-intermediate High-intermediate High Employment status Incapacitated for work Not in the labour force Not employed Employed Occupational level Low Intermediate High Household financial difficulties No financial difficulties Financial difficulties a
Abbreviations: CI = confidence interval, OR = odds ratio. The associations in Model 1 are adjusted for sex, age and ethnicity (if applicable). The associations in Model 2 are additionally adjusted for individuallevel socio-economic position (i.e., educational level, employment status, occupational level, household financial difficulties) (if applicable) and partner status. The associations in Model 3 are additionally adjusted for all environmental-level characteristics (i.e., population density, household density, proportion of minimum income households, and average property value of dwellings).
a
b
Abbreviations: CI = confidence interval, OR = odds ratio. b The associations in Model 1 are adjusted for sex, age and ethnicity (if applicable). The associations in Model 2 are additionally adjusted for individuallevel socio-economic position (i.e., educational level, employment status, occupational level, household financial difficulties) (if applicable) and partner status. The associations in Model 3 are additionally adjusted for all environmental-level characteristics (i.e., population density, household density, proportion of minimum income households, and average property value of dwellings).
between neighbourhood safety and the three types of smoking behaviour. Second, we lacked information about perceived stress in this study. As a consequence, we could not draw conclusions regarding the assumption that lower levels of neighbourhood safety leads to higher levels of stress, which in turn leads to (more) smoking among residents. Third, although we adjusted for relevant confounders in our analyses, there still might be residual confounding factors that we did not take into account, such as the availability/accessibility of tobacco retailers in residential areas. However, we did adjust for population density and environmental-level socio-economic status, which are strongly related to the availability/accessibility of tobacco retailers (Monshouwer et al., 2014; Rodriguez et al., 2013). Fourth, we used Geographic Information System techniques to construct residential areas on a relatively small scale and to calculate a neighbourhood safety score for these areas
from various backgrounds, and assessed whether these associations differ by sex, age, ethnicity and individual-level socio-economic position. The results of this study showed that higher levels of neighbourhood safety were significantly associated with less heavy smoking and less nicotine dependence, but not with less current smoking. The findings further showed that the associations between neighbourhood safety and the three types of smoking behaviour varied by ethnicity. A strength of this study is that the associations could be adjusted for a wide range of relevant confounders at the individual- as well as environmental-level. Another strength of the present study is the focus on three types of smoking behaviour. Some limitations have to be acknowledged as well. First, the cross-sectional design of the current study does not make it possible to determine causal relationships 116
Preventive Medicine 112 (2018) 111–118
E.J. Timmermans et al.
areas were more likely to be heavy smokers and nicotine dependent. A possible explanation for this finding could be that excessive smoking is a coping mechanism to reduce stress that is associated with living in an unpleasant or threatening residential environment (Diez Roux et al., 2003; Ellaway and Macintyre, 2009; Nielsen et al., 2008; Pearce et al., 2012; Shareck and Ellaway, 2011; Weden et al., 2008). Furthermore, smoking is described as a socially and culturally ingrained behaviour in deprived and unsafe neighbourhoods (Giskes et al., 2006). It has been suggested that pro-smoking norms are operating more strongly in such neighbourhoods, and that tobacco products are more readily available in these areas (Ellaway and Macintyre, 2009; Giskes et al., 2006; Stead et al., 2001; Wiltshire et al., 2001). It might be harder to decrease smoking habits by residents of deprived and unsafe neighbourhoods compared to those of more affluent and safe residential areas, due to a lack of social support, strong prevailing community norms and social reinforcement for smoking (Giskes et al., 2006; Stead et al., 2001). The results of the subgroup analyses show that the associations of neighbourhood safety with the three types of smoking behaviour vary by ethnicity. The observed variations in these associations might be due to differences in how ethnic groups experience neighbourhood safety (e.g., in terms of presence of ethnic discrimination) and/or to differences in the meaning they attach to it (Essed, 1991; Visser et al., 2017). In addition, the ethnic variations in these associations might be related to the differences in the coping resources (e.g., differences in social support networks through family systems and ethnic institutions) as used by the various ethnic group to effectively deal with lower levels of neighbourhood safety (Dhami and Skeikh, 2000; Visser et al., 2017). It has been suggested that women, older adults, ethnic minorities, and individuals with a low socio-economic position are more exposed to their immediate environment, and are more aware of, or sensitive to neighbourhood conditions than their counterparts (Dyck et al., 2001; Ellaway and Macintyre, 2009; Miles, 2006; Shareck and Ellaway, 2011; Stafford et al., 2005; Yen et al., 2009). In addition, these groups are suggested to be more dependent upon their local living environment and to be more concerned about safety because of their limited mobility, higher physical and/or social vulnerability (Skogan and Maxfield, 1981; Stafford and Marmot, 2003). In contrast with our expectations, findings of the subgroup-analyses suggest that lower levels of neighbourhood safety are associated with less heavy smoking in low educated individuals. Furthermore, the results of the subgroup-analyses suggest that lower levels of neighbourhood safety are associated with less nicotine dependence in the oldest age-group. These findings may need replication in other studies. The current findings provide empirical support that policies that improve neighbourhood safety potentially contribute to less heavy smoking and nicotine dependence. Future studies should focus on the mediating role of stress in the association between neighbourhood safety and smoking behaviour. In addition, future research should make a distinction between the various components of neighbourhoods safety (i.e., physical/social disorder, crime-related safety, and traffic-related safety) to obtain more detailed information about the association between neighbourhood safety and smoking.
Table 4 Associations between neighbourhood safety score (range = 1–10) and nicotine dependence in the HELIUS study, Amsterdam, the Netherlands, 2011–2015a,b.
Full sample All participants Sex Men Women Age 18–30 years 31–40 years 41–50 years 51–60 years 61–70 years Ethnicity Dutch South-Asian Surinamese African Surinamese Turkish Moroccan Ghanaian Individual-level socioeconomic position Educational level Low Low-intermediate High-intermediate High Employment status Incapacitated for work Not in the labour force Not employed Employed Occupational level Low Intermediate High Household financial difficulties No financial difficulties Financial difficulties
Model 1
Model 2
Model 3
OR
95% CI
OR
95% CI
OR
95% CI
0.85
0.76–0.94
0.92
0.82–1.03
0.81
0.69–0.95
0.90 0.78
0.79–1.03 0.66–0.92
0.93 0.91
0.80–1.08 0.75–1.10
0.76 0.93
0.62–0.94 0.72–1.20
0.80 0.92 0.96 0.80 0.72
0.62–1.03 0.73–1.16 0.79–1.16 0.65–0.99 0.51–1.01
0.79 0.98 1.05 0.89 0.81
0.59–1.06 0.75–1.27 0.84–1.30 0.71–1.12 0.56–1.18
0.54 0.92 0.84 0.77 1.11
0.36–0.81 0.64–1.33 0.62–1.13 0.56–1.05 0.64–1.90
0.57 0.83
0.46–0.70 0.64–1.07
0.82 0.89
0.64–1.05 0.67–1.17
0.91 0.74
0.62–1.33 0.53–1.04
0.73 1.03 1.43 0.91
0.56–0.95 0.85–1.24 1.09–1.88 0.26–3.18
0.70 1.08 1.52 0.88
0.53–0.94 0.86–1.34 1.13–2.04 0.18–4.21
0.51 1.07 1.58 0.69
0.35–0.73 0.76–1.49 0.99–2.53 0.09–5.30
1.03 0.97 1.03 0.79
0.80–1.34 0.81–1.16 0.85–1.26 0.61–1.03
1.00 0.93 1.00 0.81
0.74–1.35 0.77–1.13 0.81–1.23 0.61–1.07
1.15 0.76 0.76 0.79
0.75–1.77 0.58–0.98 0.57–1.02 0.54–1.16
1.05
0.77–1.43
1.00
0.71–1.42
0.98
0.60–1.61
0.89
0.64–1.23
1.05
0.71–1.56
1.14
0.66–1.94
1.02 0.76
0.81–1.28 0.67–0.88
1.09 0.84
0.84–1.41 0.72–0.97
0.97 0.70
0.67–1.41 0.57–0.86
1.03 0.90 0.81
0.88–1.20 0.74–1.11 0.62–1.07
1.00 0.88 0.87
0.86–1.18 0.71–1.09 0.65–1.15
0.82 0.84 0.81
0.66–1.03 0.63–1.12 0.54–1.19
0.70
0.60–0.82
0.76
0.64–0.89
0.72
0.57–0.90
1.06
0.91–1.22
1.11
0.9–1.30
0.91
0.72–1.14
a
Abbreviations: CI = confidence interval, OR = odds ratio. The associations in Model 1 are adjusted for sex, age and ethnicity (if applicable). The associations in Model 2 are additionally adjusted for individuallevel socio-economic position (i.e., educational level, employment status, occupational level, household financial difficulties) (if applicable) and partner status. The associations in Model 3 are additionally adjusted for all environmental-level characteristics (i.e., population density, household density, proportion of minimum income households, and average property value of dwellings). b
using micro-scale environmental data that were independent of participants' perceptions. We assume that this safety score provides an appropriate measure of the actual safety levels in a neighbourhood, but it should also be acknowledged that this measure not necessarily matches with the participants' perceptions of neighbourhood safety. Finally, the current study only examines the association of smoking with safety in the residential environment and does not consider safety in other places where people spend substantial amounts of time (e.g., for work, shopping, and recreation). However, this focus on place of residence is common to most geographical studies, not only for practical reasons, but also because people's home is important to their identity, rest and recovery from daily stress. The findings suggest that individuals who live in unsafe residential
Authors' contributions AEK and EJT conceptualized the study. EJT, EMV and MBS prepared the data. EMV obtained the geodata and performed the spatial analyses. EJT performed the statistical analyses, interpreted the data, and drafted the manuscript. All authors revised intermediate manuscript versions critically for important intellectual content. All authors have read and approved the final version of the manuscript. Conflict of interest statement The authors declare that they have no competing interests. 117
Preventive Medicine 112 (2018) 111–118
E.J. Timmermans et al.
Acknowledgments
Safety Monitor 2013/2014, . www.ois.amsterdam.nl/visualisatie/veiligheidsindex. html (Accessed: May, 2017). Neumann, T., Rasmussen, M., Heitmann, B.L., et al., 2013. Gold standard program for heavy smokers in a real-life setting. Int. J. Environ. Res. Public Health 10, 4186–4199. Nielsen, L., Curtis, T., Kristensen, T.S., et al., 2008. What characterizes persons with high levels of perceived stress in Denmark? A national representative study. Scand. J. Public Health 36, 369–379. Patterson, F., Seravalli, L., Hanlon, A., et al., 2012. Neighbourhood safety as a correlate of tobacco use in a sample of urban, pregnant women. Addict. Behav. 37, 1132–1137. Pearce, J., Barnett, R., Moon, G., 2012. Sociospatial inequalities in health-related behaviours: pathways linking place and smoking. Prog. Hum. Geogr. 36, 3–24. Rachelle, J.N., Wood, L., Nathan, A., et al., 2016. Neighbourhood disadvantage and smoking: examining the role of neighbourhood-level psychosocial characteristics. Health Place 40, 98–105. Rodriguez, D., Carlos, H.A., Adachi-Mejia, A.M., Berke, E.M., Sargent, J.D., 2013. Predictors of tobacco outlet density nationwide: a geographic analysis. Tob. Control. 22, 349–355. Stead, M., MacAskill, S., Makintosh, A., et al., 2001. “It's as if you're locked in”: qualitative explanations for area effects on smoking disadvantaged communities. Health Place 7, 333–343. Shareck, M., Ellaway, A., 2011. Neighbourhood crime and smoking: the role of objective and perceived crime measures. BMC Public Health 11, 930. Skogan, W.G., Maxfield, M.G., 1981. Coping with Crime. Individual and Neighbourhood Reactions. Sage Publications, London. Smeets, H., 2015. Verantwoording Veiligheidsindex 2014 [in English: Justification of the Safety Index in 2014]. Municipality of Amsterdam, Department of Research and Statistics, Amsterdam. Snijder, M.B., Galenkamp, H., Prins, M., et al., 2017. Cohort profile: the Healthy Life in an Urban Setting (HELIUS) study. BMJ Open 7, e017873. Stafford, M., Marmot, M., 2003. Neighbourhood deprivation and health: does it affect us all equally? Int. J. Epidemiol. 32, 357–366. Stafford, M., Cummins, S., Macintyre, S., et al., 2005. Gender differences in the associations between health and neighbourhood environment. Soc. Sci. Med. 60, 1681–1692. Statistics Netherlands, 2010. Standaard Beroepenclassificatie 2010 - Editie 2010 (In English: Standard Occupational Classification system 2010 - Edition 2010). The Hague/Heerlen: Statistics Netherlands. Stronks, K., Kulu-Glasgow, I., Agyemang, C., 2009. The utility of ‘country of birth’ for the classification of ethnic groups in health research: the Dutch experience. Ethn. Health 14, 255–269. Stronks, K., Snijder, M.B., Peters, R.J.G., et al., 2013. Unravelling the impact of ethnicity on health in Europe: the HELIUS study. BMC Public Health 13, 402. Tsjeng, M., Yeatts, K., Millikan, R., et al., 2001. Environmental-level characteristics and smoking in women. Am. J. Public Health 91, 1847–1850. United States Department of Health Human Services, 2014. The Health Consequences of Smoking-50 Years of Progress: a Report of the Surgeon General. United States Department of Health and Human Services, Center for Diseases Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion, Office on Smoking and Health, Atlanta, GA. Veldhuizen, E.M., Stronks, K., Kunst, A.E., 2013. Assessing associations between socioeconomic environment and self-reported health in Amsterdam using bespoke environments. PLoS One 8, e68790. Virtanen, M., Kivimäki, M., Kouvonen, A., et al., 2007. Average household income, crime, and smoking behaviour in a local area: the Finnish 10-town study. Soc. Sci. Med. 64, 1904–1913. Visser, M.J., Ikram, U.Z., Derks, E.M., et al., 2017. Perceived ethnic discrimination in relation to smoking and alcohol consumption in ethnic minority groups in the Netherlands: the HELIUS study. Int. J. Public Health. http://dx.doi.org/10.1007/ s00038-017-0977-2. Weden, M.M., Carpiano, R.M., Robert, S.A., 2008. Subjective and objective neighbourhood characteristics and adult health. Soc. Sci. Med. 66, 1256–1270. Wiltshire, S., Bancroft, A., Amos, A., Parry, O., 2001. “They're doing people a service” qualitative study of smoking, smuggling, and social deprivation. BMJ 323, 203–207. Yen, I.H., Michael, Y.L., Pedue, L., 2009. Neighbourhood environment in studies of health of older adults: a systematic review. Am. J. Prev. Med. 37, 455–463.
This work was financially supported by the Amsterdam Public Health Research Institute in Amsterdam, the Netherlands. The HEalthy LIfe in an Urban Setting (HELIUS) study is conducted by the Academic Medical Center (AMC) Amsterdam and the Public Health Service (GGD) of Amsterdam. Both organisations provided core support for HELIUS. The HELIUS study is also funded by the Dutch Heart Foundation, the Netherlands Organisation for Health Research and Development (ZonMw), the European Union 7th Framework Programme, and the European Fund for the Integration of non-EU immigrants (EIF). We are most grateful to the participants of the HELIUS study and the management team, research nurses, interviewers, research assistants and other staff members who have taken part in gathering the data of this study. Furthermore, we would like to thank the Department of Research and Statistics of the municipality of Amsterdam for providing the data of the Amsterdam Safety Monitor 2013/2014 and the integral demographic and socio-economic registries. References Aiken, L.S., West, S.G., 1991. Multiple Regression: Testing and Interpreting Interactions. SAGE Publications, Inc., Thousand Oaks, CA. Dhami, S., Skeikh, A., 2000. The Muslim family: predicament and promise. West. J. Med. 137, 352–356. Diez Roux, A.V., Mair, C., 2010. Neighborhoods and health. Ann. N. Y. Acad. Sci. 1186, 125–145. Diez Roux, A.V., Merkin, S.S., Hannan, P., et al., 2003. Area characteristics, individuallevel socioeconomic indicators, and smoking in young adults - the coronary artery disease risk development in young adults study. Am. J. Epidemiol. 157, 315–326. Dyck, I., Davis Lewis, N., McLafferty, S., et al., 2001. Women in their place: gender and perceptions of neighbourhoods in the West of Scotland. In: Dyck, I., Davis Lewis, N., McLafferty, S. (Eds.), Geographies of Women's Health. Routledge, London. Ellaway, A., Macintyre, S., 2009. Are perceived neighbourhood problems associated with the likelihood of smoking. J. Epidemiol. Community Health 63, 78–80. Essed, P., 1991. Understanding Everyday Racism: An Interdisciplinary Theory. Sage Publications, Inc., Thousand Oaks, CA. Ganz, M., 2000. The relationship between external threats and smoking in central Harlem. Am. J. Public Health 90, 367–371. Giskes, K., Van Lenthe, F.J., Turell, G., et al., 2006. Smokers living in deprived areas are less likely to quit: a longitudinal follow-up. Tob. Control. 15, 485–488. Goenka, S., Andersen, L.B., 2016. Our health is a function of where we live. Lancet 387, 2168–2170. Heatherton, T.F., Kozlowski, L.T., Frecker, R.C., et al., 1991. The Fagerström Test for nicotine dependence: a revision of the Fagerström Tolerance Questionnaire. Br. J. Addict. 86, 1119–1127. Jitnarin, N., Heinrich, K.M., Haddock, C.K., 2015. Neighbourhood environment perceptions and the likelihood of smoking and alcohol use. Int. J. Environ. Res. Public Health 12, 784–799. McGinn, A.P., Evenson, K.R., Herrin, A.H., et al., 2008. The association of perceived and objectively measured crime with physical activity: a cross-sectional analysis. J. Phys. Act. Health 5, 117–131. Miles, R., 2006. Neighbourhood disorder and smoking: findings of a European urban survey. Soc. Sci. Med. 63, 2464–2475. Monshouwer, K., Verdurmen, J., Ketelaars, T., Van Laar, M., 2014. Points of Sale of Tobacco Products: Synthesis of Scientific and Practice-based Knowledge on the Impact of Reducing the Number of Points of Sale and Restrictions on Tobacco Product Displays. Trimbos-Instituut, Netherlands Institute of Mental Health and Addiction, Utrecht, The Netherlands. Municipality of Amsterdam, 2017. Department of Research and Statistics. In: Amsterdam
118
Contents lists available at ScienceDirect
Preventive Medicine journal homepage: www.elsevier.com/locate/ypmed
Neighbourhood safety and smoking in population subgroups: The HELIUS study
T
⁎
Erik J. Timmermansa, , Eleonore M. Veldhuizenb, Marieke B. Snijdera,c, Martijn Huismand,e, Anton E. Kunsta a
Academic Medical Center, University of Amsterdam, Department: Public Health, Amsterdam Public Health research institute, Amsterdam, The Netherlands Department of Geography, Planning & International Development Studies, University of Amsterdam, Amsterdam, The Netherlands c Academic Medical Center, University of Amsterdam, Department: Clinical Epidemiology, Biostatistics and Bioinformatics, Amsterdam Public Health research institute, Amsterdam, The Netherlands d Department of Epidemiology and Biostatistics, Amsterdam Public Health research institute, VU University Medical Center, Amsterdam, The Netherlands e Department of Sociology, Faculty of Social Sciences, VU University, Amsterdam, The Netherlands b
A R T I C LE I N FO
A B S T R A C T
Keywords: Environmental epidemiology Geographic Information Systems HELIUS study Neighbourhood safety Population subgroups Smoking
This study examines the associations between neighbourhood safety and three types of smoking behaviour, and whether these associations differ by sex, age, ethnicity and individual-level socio-economic position. Baseline data (2011–2015) from the The HEalthy LIfe in an Urban Setting (HELIUS) study (Amsterdam, the Netherlands) were used. Smoking behaviour was based on self-report. Heavy smoking was defined as smoking ≥10 cigarettes per day. Nicotine dependence was assessed using the Fagerström questionnaire. Geographic Information System techniques were used to construct local residential areas and to examine neighbourhood safety for these areas using micro-scale environmental data. Multilevel logistic regression analyses with 6-digit zip code area as a second level were used to assess the association between neighbourhood safety and smoking. In our study sample of 22,728 participants (18–70 years), 24.0% were current smokers, 13.7% were heavy smokers and 8.1% were nicotine dependent individuals. Higher levels of neighbourhood safety were significantly associated with less heavy smoking (OR = 0.88, 95% CI = 0.78–0.99) and less nicotine dependence (OR = 0.81, 95% CI = 0.69–0.95), but not with less current smoking (OR = 1.01, 95% CI = 0.91–1.11). The associations between neighbourhood safety and the three types of smoking behaviour varied by ethnicity. For instance, higher levels of neighbourhood safety were associated with less current smoking in participants of African Surinamese origin (OR = 0.71, 95% CI = 0.57–0.89), but not in those of Dutch (OR = 1.13, 95% CI = 0.91–1.39), South-Asian Surinamese (OR = 1.22, 95% CI = 0.95–1.55), Turkish (OR = 1.08, 95% CI = 0.84–1.38), Moroccan (OR = 1.53, 95% CI = 1.12–2.10) or Ghanaian (OR = 1.18, 95% CI = 0.47–2.94) origin. Policies that improve neighbourhood safety potentially contribute to less heavy smoking and nicotine dependence.
1. Introduction Tobacco smoking is the primary preventable cause of disability and death worldwide (United States Department of Health Human Services, 2014). It is increasingly recognized that the neighbourhood where people live is associated with their health behaviours, including smoking (Goenka and Andersen, 2016; Pearce et al., 2012; Shareck and Ellaway, 2011). To create neighbourhood-level interventions that discourage smoking, it is important to understand which neighbourhood characteristics affect smoking behaviour and which specific population subgroups are particularly susceptible for these neighbourhood effects
(Pearce et al., 2012). Neighbourhood safety represents socio-cultural characteristics of the neighbourhood and includes various components, such as physical/ social disorder, crime-related safety and traffic-related safety (McGinn et al., 2008). It has been suggested that unsafe neighbourhood conditions may act as chronic stressors and as such, may influence smoking through pathways, such as stress and psychological well-being (Nielsen et al., 2008; Pearce et al., 2012; Shareck and Ellaway, 2011; Weden et al., 2008). Furthermore, it has been suggested that pro-smoking norms are operating more strongly in deprived and unsafe neighbourhoods than in affluent and safe neighbourhoods. In addition, it has been
⁎ Corresponding author at: Academic Medical Center, University of Amsterdam, Department: Public Health, Amsterdam Public Health research institute, PO Box 22660, 1100 DD Amsterdam, The Netherlands. E-mail address: [email protected] (E.J. Timmermans).
https://doi.org/10.1016/j.ypmed.2018.04.012 Received 5 October 2017; Received in revised form 20 March 2018; Accepted 6 April 2018 Available online 12 April 2018 0091-7435/ © 2018 Elsevier Inc. All rights reserved.
Preventive Medicine 112 (2018) 111–118
E.J. Timmermans et al.
et al., 2017; Stronks et al., 2013). Briefly, baseline data were collected from January 2011 to December 2015. Participants aged 18–70 years were randomly sampled, stratified by ethnicity, through the municipality registry of Amsterdam. In total, 90,019 individuals were invited to participate in the HELIUS study. We were able to contact and get a response from 49,952 (55%) invited persons, either by response card or after a home visit by an ethnically matched interviewer. Of those, 24,789 individuals (50%) agreed to participate. Non-response analyses showed that socio-economic differences between participants and nonparticipants were very small (Snijder et al., 2017). Participants received a confirmation letter of the appointment for the physical examination, including a digital or paper version of the questionnaire (depending on the preference of the subject). Of the 24,789 participants, 23,942 (96.6%) participants completed the questionnaire. The HELIUS study has been approved by the Institutional Review Board of the Academic Medical Center, University of Amsterdam. For the current study, from the total sample of participants who completed the questionnaire (n = 23,942), we excluded participants who were aged over 70 year (n = 11) and those who had no full data available on all outcome measures (n = 395), neighbourhood safety (n = 236) or both (n = 3). Furthermore, we excluded participants of Javanese Surinamese (n = 243) and unknown Surinamese origin (n = 276) from the remaining sample because of the small sample sizes. We also excluded participants with an unknown or other ethnic origin (n = 50). Consequently, the analytical sample consisted of 22,728 individuals who lived across 8851 6-digit zip code areas. The average number of participants per 6-digit zip code area was 7, with a range of 1–92.
suggested that tobacco products are more readily available in deprived and unsafe neighbourhoods than in affluent and safe residential areas This may make it harder for individuals to quit smoking in deprived and unsafe neighbourhoods than in affluent and safe neighbourhoods (Ellaway and Macintyre, 2009; Giskes et al., 2006; Stead et al., 2001; Wiltshire et al., 2001). Previous studies investigating the association between neighbourhood safety and smoking behaviour are inconclusive (Diez Roux and Mair, 2010). Some studies showed that individuals who are exposed to higher levels of objectively measured crime (Shareck and Ellaway, 2011; Tsjeng et al., 2001) or perceive lower levels of neighbourhood safety (Ellaway and Macintyre, 2009; Ganz, 2000; Miles, 2006; Patterson et al., 2012) are more likely to be current smokers, but other studies found no associations (Diez Roux and Mair, 2010; Jitnarin et al., 2015; Rachelle et al., 2016; Virtanen et al., 2007). Several methodological limitations are presented among the current body of evidence. Particular problems include small study samples (e.g., Ganz, 2000; Tsjeng et al., 2001), ignorance of data-clustering at the environmentallevel (e.g., Ellaway and Macintyre, 2009; Patterson et al., 2012), and no adjustment for environmental-level characteristics, such as socio-economic status (e.g., Ellaway and Macintyre, 2009; Shareck and Ellaway, 2011). It remains undocumented whether the association between neighbourhood safety and smoking is stronger in some specific population subgroups than in others. Women, older adults, ethnic minorities and individuals with a low socio-economic position more often report poor neighbourhood safety than their counterparts (Skogan and Maxfield, 1981). It has been suggested that these specific groups are more exposed to their local living environment and are more aware, or sensitive to, what happens in their neighbourhood (Ellaway and Macintyre, 2009; Miles, 2006; Shareck and Ellaway, 2011; Stafford et al., 2005; Yen et al., 2009). These groups are suggested to be more dependent upon their immediate – “local” – environment and to be more concerned about safety because of their limited mobility and higher physical and social vulnerability (Skogan and Maxfield, 1981; Stafford and Marmot, 2003). A limited number of studies suggest that lower levels of neighbourhood safety are more strongly associated with current smoking in women and older adults (Dyck et al., 2001; Ellaway and Macintyre, 2009; Miles, 2006; Shareck and Ellaway, 2011). Studies examining the differential effects of neighbourhood safety on smoking behaviour across ethnic groups and socio-economic positions are lacking. This large-scale population-based study extends previous research and aimed to examine whether lower levels of neighbourhood safety are associated with three types of smoking behaviour, including current smoking, heavy smoking, and nicotine dependence, while adjusting for a range of individual- and environmental-level characteristics and for the clustering of participants within neighbourhoods. Furthermore, this study aimed to examine whether the associations between neighbourhood safety and smoking behaviour differ by sex, age, ethnicity and individual-level socio-economic position. We expected that higher levels of safety are associated with less current/heavy smoking and less nicotine dependence. Furthermore, we expected that higher levels of safety are more strongly associated with less smoking in women, older adults, ethnic minority groups and individuals with a low socio-economic position than in their counterparts.
2.2. Dependent variables 2.2.1. Smoking status Three measures were used to assess different smoking behaviours. Current smoking was assessed as currently smoking one or more tobacco products (i.e., cigarettes, cigars and/or pipe tobacco) (0 = no, 1 = yes). Heavy smoking was defined as smoking ≥10 cigarettes daily (0 = no, 1 = yes). For this measure, the number of cigars and packages of pipe tobacco were converted to cigarettes based on the tobacco content (i.e., 1 cigar is similar to 3 cigarettes, and 1 package of pipe tobacco is similar to 62.5 cigarettes). Often, studies use a cut-off of ≥20 cigarettes per day for heavy smoking (Neumann et al., 2013), which may represent nicotine dependence. To clearly distinguish between the smoking behaviours, we decided to use the cut-off ≥10 cigarettes per day (Visser et al., 2017). Non-smokers were classified as ‘no heavy smokers’. Nicotine dependence was determined by the Fagerström scale (Heatherton et al., 1991) consisting of six questions (e.g., ‘Do you find it hard not to smoke in places where it is not allowed?’). The sum score varied from 0 to 10, with a cut-off of ≥4 considered nicotine dependence (0 = no, 1 = yes). If one of the items was missing, the Fagerström sum score was calculated with a score of 0 for the missing item. If more than one item was missing the Fagerström sum score was coded as missing. For non-smokers the sum score is 0. This approach corresponds with previous research (Visser et al., 2017). 2.3. Independent variables
2. Methods 2.3.1. Neighbourhood safety A neighbourhood safety score was assessed by using data from the Amsterdam Safety Monitor 2013/2014 (ASM) maintained by the Municipality of Amsterdam (Municipality of Amsterdam, 2017; Smeets, 2015). Participants of the ASM were asked how they perceive safety in their neighbourhood on a scale from 1 to 10, with higher scores indicating higher levels of neighbourhood safety. More details about the assessment of neighbourhood safety in the ASM are described
2.1. Design and study sample Baseline data from the HEalthy LIfe in an Urban Setting (HELIUS) study were used. The HELIUS study is a multi-ethnic cohort study conducted in Amsterdam, the Netherlands, including Dutch, Surinamese, Turkish, Moroccan, and Ghanaian origin ethnic groups. The study protocol has been described in detail elsewhere (Snijder 112
Preventive Medicine 112 (2018) 111–118
E.J. Timmermans et al.
household financial difficulties (1). All four indicators of individuallevel socio-economic position were considered as potential effect modifier separately.
elsewhere (Municipality of Amsterdam, 2017; Smeets, 2015). Geographic Information Systems were used to assign a neighbourhood safety score to each included HELIUS participant. First, the centroid of the 6-digit zip code area in which the HELIUS participants were living were geocoded by using ArcGIS 10.1 (ESRI Inc., Redlands, California, United States of America (USA)). In the Netherlands, 6-digit zip code areas are the smallest geographical units available and are approximately sized 50 × 50 m and include 10 to 20 households (Veldhuizen et al., 2013). A buffer zone with a radius of 300 m was generated around each individual centroid. Subsequently, 300-m buffer zones were also created around the residential addresses of all ASM participants. Based on the extent of overlap (i.e., the percentage of overlap) between the buffer zones of HELIUS participants and ASM participants, a weighted average neighbourhood safety score was assigned to the buffer zones of the HELIUS participants. For this process, an overlay operation was performed, which joins data layers based on common geographical location. The average number of overlaying ASM buffer zones that were used to assess the neighbourhood safety score for each HELIUS-participant was 234.3 (SD = 143.1), with a range of 2.0–737.0.
2.5. Potential confounders at the individual-level If applicable, age, sex, ethnicity and individual-level socio-economic position were considered as confounders in the analyses. Partner status (0 = no partner, 1 = partner) was additionally considered as a potential individual-level confounder.
2.6. Potential confounders at the environmental-level At the environmental-level, the following potential confounders were considered: population and household density, and socio-economic status. These data were derived from integral demographic and socio-economic registries maintained by the Municipality of Amsterdam. Population density was calculated as the number of residents in the 300-meter buffer zones and categorized into: (1) Low (Tertile 1 (≤2662) (reference category)), (2) Intermediate (Tertile 2 (2663–3942)), and (3) High (Tertile 3 (≥3943)). Household density was calculated as the number of households in the 300-meter buffer zones and categorized into: (1) Low (Tertile 1 (≤1156) (reference category)), (2) Intermediate (Tertile 2 (1157–1812)), and (3) High (Tertile 3 (≥1813)). Environmental-level socio-economic status was based on the proportion of households living on a minimum income and the average property value of dwellings in buffer zones (Veldhuizen et al., 2013). The proportion of households living on a minimum income was categorized into: (1) Low (Tertile 1 (≤16.9%), (2) Intermediate (Tertile 2 (17.0%–24.6%)), and (3) High (Tertile 3 (≥24.7%) (reference category)). The average property value of dwellings was categorized into: (1) Low (Tertile 1 (≤€166,660) (reference category)), (2) Intermediate (Tertile 2 (€166,661–€212,270)), and (3) High (Tertile 3 (≥€212,271)).
2.4. Potential effect modifiers The following potential effect modifiers were considered: age, sex (0 = men, 1 = women), ethnicity, and individual-level socio-economic position. Age was dummy-coded (0 = no, 1 = yes) into: (1) 18–30 years (reference category), (2) 31–40 years, (3) 41–50 years, (4) 51–60 years, and (5) 61–70 years. Participant's ethnicity was defined according to the country of birth of the participant as well as that of the parents, which is currently the most widely accepted and most valid assessment of ethnicity in the Netherlands (Stronks et al., 2009). Participants were considered of Dutch origin if the participant and both parents were born in the Netherlands. Participants were considered of non-Dutch ethnic origin if either they themselves were born outside the Netherlands and at least one of their parents (first-generation), or they themselves were born in the Netherlands but at least one of their parents was born outside the Netherlands (second-generation). Of the Surinamese immigrants in the Netherlands, approximately 80% are either African or South-Asian origin. Surinamese subgroups were classified according to self-reported ethnic origin. Ethnicity groups were dummy-coded (0 = no, 1 = yes) into: (1) Dutch (reference category), (2) South-Asian Surinamese, (3) African Surinamese, (4) Turkish, (5) Moroccan, and (6) Ghanaian origin. Individual-level socio-economic position was assessed using data on educational level, employment status, occupational level and household financial difficulties. Educational level was dummy-coded (0 = no, 1 = yes) into: (1) Low (no education or elementary education (reference category)), (2) Low-intermediate (lower vocational education or lower secondary education), (3) High-intermediate (intermediate vocational education or intermediate/higher secondary education), and (4) High (higher vocational education or university education). Employment status was dummy-coded (0 = no, 1 = yes) into: (1) incapacitated for work (reference category), (2) not in labour force (e.g., retired), (3) not employed, and (4) employed. Occupational level was classified according to the Dutch Standard Occupational Classification system for 2010 (Statistics Netherlands, 2010). Based on this classification system, occupational level was dummy-coded (0 = no, 1 = yes) into: (1) Low (elementary and lower (reference category)), (2) Intermediate (intermediate), and (3) High (higher and academic). Household financial difficulties were assessed by asking whether participants had problems managing their household income. The response categories ‘no, no problems at all’ and ‘no problems, but I have to pay attention to what I spend’ were considered as having no household financial difficulties (0), whereas the response categories ‘Yes, some problems’ and ‘Yes, a lot of problems’ were considered as having
2.7. Statistical analyses Multilevel logistic regression analyses with 6-digit zip code area as a second level were performed to examine the associations between neighbourhood safety and the three individual outcome measures on smoking. By conducting a multi-level study, the clustering of observations (level-1 unit) within the same 6-digit zip code area (level-2 unit) has been taken into account. To examine whether the associations differ by age, sex, ethnicity and individual-level socio-economic position, the analyses were stratified by subgroups of HELIUS participants. Interaction terms between neighbourhood safety and these variables were assessed in fully adjusted models to test for statistical significance. The interaction terms were considered as significant at a p-value below 0.10 (Aiken and West, 1991). The associations between neighbourhood safety and the individual outcome measures on smoking were examined in models constructed step by step. Model 1 adjusted the effects of neighbourhood safety on smoking status for age, sex and ethnicity. Model 2 additionally adjusted for all other individual-level characteristics (i.e., all four indicators of individual-level socio-economic position, and partner status). Model 3 additionally adjusted for all environmental-level characteristics (i.e., population and household density, and environmental-level socio-economic status). In all models, a probability level of ≤5.0% was regarded as statistically significant. Statistical analyses were performed in IBM SPSS Statistics (Version 24; IBM Corp, Armonk, New York, USA). 113
Preventive Medicine 112 (2018) 111–118
E.J. Timmermans et al.
Table 1 Characteristics of the study sample of the HELIUS study, Amsterdam, the Netherlands, 2011–2015.a
All participants Sex Men Women Age-group 18–30 years 31–40 years 41–50 years 51–60 years 61–70 years Ethnicity (%) Dutch South-Asian Surinamese African Surinamese Turkish Moroccan Ghanaian Individual-level socio-economic position (%) Educational level Low Low-intermediate High-intermediate High Employment status Incapacitated for work Not in the labour force Not employed Employed Occupational level Low Intermediate High Household financial difficulties No financial difficulties Financial difficulties Partner status (%) No partner Partner Environmental-level population density (%) Low Intermediate High Environmental-level household density (%) Low Intermediate High Environmental-level socio-economic status – proportion of minimum income households (%) Low Intermediate High Environmental-level socio-economic status – average property value of dwellings (%) Low Intermediate High a
All participants
Current smokers
Heavy smokers
Nicotine dependent individuals
N
(%)
(%)
(%)
24.0
13.7
8.1
42.3 57.7
32.5 17.8
20.9 8.5
12.0 5.2
20.9 18.4 25.5 24.4 10.8
25.6 26.8 23.8 23.2 18.5
12.0 14.2 14.9 14.5 11.9
6.7 8.9 9.5 8.3 5.8
20.1 14.4 18.9 17.3 18.7 10.6
24.7 29.6 31.7 34.4 13.2 3.6
12.2 17.3 16.0 22.7 8.9 1.4
6.5 12.0 8.1 14.2 5.2 1.0
17.8 26.5 29.6 26.1
19.9 30.1 25.1 19.6
13.0 19.8 13.6 8.2
8.7 12.1 7.6 4.2
7.7 18.1 14.5 59.7
26.5 16.4 31.2 24.4
17.4 8.0 20.3 13.4
13.2 4.1 13.3 7.4
46.6 26.6 26.8
27.5 27.2 20.4
17.3 15.9 8.7
10.6 9.4 4.5
61.3 38.7
20.5 30.0
10.7 18.7
5.7 12.0
50.3 49.7
27.8 20.3
15.5 12.0
9.9 6.4
33.3 33.4 33.3
22.5 23.0 26.6
12.2 13.4 15.5
7.0 7.9 9.4
33.4 33.3 33.3
22.5 22.7 26.9
12.6 13.0 15.6
7.0 7.9 9.4
33.3 33.3 33.4
23.1 25.5 23.5
13.7 15.2 12.3
8.5 9.2 6.6
33.4 33.3 33.3
23.4 24.0 24.6
13.7 14.1 13.3
8.6 8.0 7.7
(%)
22,728 22,728 9616 13,112 22,728 4754 4182 5776 5552 2464 22,728 4556 3279 4296 3940 4246 2411 22,544 4003 5979 6682 5880 22,478 1726 4062 3249 13,441 19,054 8882 5074 5098 22,286 13,656 8630 22,608 11,374 11,234 22,726 7576 7600 7550 22,727 7577 7575 7575 22,708 7566 7563 7579 22,728 7578 7577 7573
Abbreviation: N = number of valid observations.
3. Results
current smokers, heavy smokers and nicotine dependent individuals differed across various population subgroups (Table 1). In the full sample, the average neighbourhood safety score was 6.2 (SD = 0.5), with a range of 5.0–7.8 (Fig. 1).
The mean age of all 22,728 participants was 43.8 (SD = 13.4), with an age-range of 18–70. In the full sample, 57.7% were women. Of all participants, 20.1% were Dutch, 14.4% were South-Asian Surinamese, 18.9% African Surinamese, 17.3% were Turkish, 18.7% were Moroccan, and 10.6% were Ghanaian. Most participants had a highintermediate educational level (29.6%), were employed (59.7%), and experienced no household financial difficulties (61.3%). Furthermore, most participants had a low occupational level (46.6%) (Table 1). The study sample consisted of 24.0% current smokers, 13.7% heavy smokers and 8.1% nicotine dependent individuals. The proportions of
3.1. Neighbourhood safety and current smoking After adjustment for individual- and environmental-level confounders, neighbourhood safety was not statistically significantly associated with current smoking in the full sample (OR = 1.01, 95% CI = 0.91–1.11) (Table 2; Model 3). The association of neighbourhood safety with current smoking 114
Preventive Medicine 112 (2018) 111–118
E.J. Timmermans et al.
Fig. 1. Prevalence rate of current smokers, heavy smokers and nicotine dependent individuals by deciles of neighbourhood safety in the HELIUS study, Amsterdam, the Netherlands, 2011–2015.a
in those with a low educational level (OR = 1.14, 95% CI = 0.79–1.63). There was no statistical evidence that the association between neighbourhood safety and heavy smoking was modified by sex, age or the other individual-level socio-economic factors (Table 3; Model 3).
varied by ethnicity. Higher levels of neighbourhood safety were associated with less smoking in participants of African Surinamese origin (OR = 0.71, 95% CI = 0.57–0.89), but not in those of Dutch (OR = 1.13, 95% CI = 0.91–1.39), South-Asian Surinamese (OR = 1.22, 95% CI = 0.95–1.55), Turkish (OR = 1.08, 95% CI = 0.84–1.38), Moroccan (OR = 1.53, 95% CI = 1.12–2.10) or Ghanaian (OR = 1.18, 95% CI = 0.47–2.94) origin (Table 2; Model 3). There was no statistical evidence that the association between neighbourhood safety and current smoking was modified by sex, age or individual-level socio-economic position (Table 2; Model 3).
3.3. Neighbourhood safety and nicotine dependence After adjustment for individual- and environmental-level confounders, higher levels of neighbourhood safety were statistically significantly associated with less nicotine dependence in the full sample (OR = 0.81, 95% CI = 0.69–0.95) (Table 4; Model 3). The association of neighbourhood safety with nicotine dependence varied by age. Higher levels of safety were associated with less nicotine dependence in participants aged 18–30 years (OR = 0.54, 95% CI = 0.36–0.81), 31–40 years (OR = 0.92, 95% CI = 0.64–1.33), 41–50 years (OR = 0.84, 95% CI = 0.62–1.13), and 51–60 years (OR = 0.77, 95% CI = 0.56–1.05), but not in those aged 61–70 years (OR = 1.11, 95% CI = 0.64–1.90). The association of neighbourhood safety with nicotine dependence also varied by ethnicity. Higher levels of neighbourhood safety were associated with less nicotine dependence in participants of Dutch (OR = 0.91, 95% CI = 0.62–1.33), South-Asian Surinamese (OR = 0.74, 95% CI = 0.53–1.04), and African Surinamese origin (OR = 0.51, 95% CI = 0.35–0.73), and Ghanaian (OR = 0.69, 95% CI = 0.09–5.30) origin, but not in those of Turkish (OR = 1.07, 95% CI = 0.76–1.49) or Moroccan (OR = 1.58, 95% CI = 0.99–2.53) origin (Table 4; Model 3). There was no statistical evidence that the association between neighbourhood safety and heavy smoking was modified by sex or individual-level socio-economic position (Table 4; Model 3).
3.2. Neighbourhood safety and heavy smoking After adjustment for individual- and environmental-level confounders, higher levels of neighbourhood safety were statistically significantly associated with less heavy smoking (OR = 0.88, 95% CI = 0.78–0.99) (Table 3; Model 3). Ethnicity was found to be an effect modifier in the association between neighbourhood safety and heavy smoking. Higher levels of neighbourhood safety were associated with less heavy smoking in participants of South-Asian Surinamese (OR = 0.77, 95% CI = 0.58–1.04), African Surinamese (OR = 0.61, 95% CI = 0.46–0.81), and Ghanaian (OR = 0.35, 95% CI = 0.09–1.42) origin, but not in those of Dutch (OR = 1.13, 95% CI = 0.85–1.50), Turkish (OR = 1.16, 95% CI = 0.88–1.54) or Moroccan (OR = 1.56, 95% CI = 1.07–2.27) origin (Table 3; Model 3). Educational level was also found to be an effect modifier in the association between neighbourhood safety and heavy smoking. Higher levels of neighbourhood safety were associated with less heavy smoking in participants with a low-intermediate (OR = 0.83, 95% CI = 0.67–1.03), high-intermediate (OR = 0.86, 95% CI = 0.68–1.08), and high educational level (OR = 0.86, 95% CI = 0.65–1.14), but not
4. Discussion This study aimed to examine the associations between neighbourhood safety and three types of smoking behaviour in a large sample of adults
a The neighbourhood safety score ranges from 1 to 10, with higher scores representing higher levels of neighbourhood safety.
115
Preventive Medicine 112 (2018) 111–118
E.J. Timmermans et al.
Table 2 Associations between neighbourhood safety score (range = 1–10) and current smoking in the HELIUS study, Amsterdam, the Netherlands, 2011–2015a,b.
Full sample All participants Sex Men Women Age 18–30 years 31–40 years 41–50 years 51–60 years 61–70 years Ethnicity Dutch South-Asian Surinamese African Surinamese Turkish Moroccan Ghanaian Individual-level socioeconomic position Educational level Low Low-intermediate High-intermediate High Employment status Incapacitated for work Not in the labour force Not employed Employed Occupational level Low Intermediate High Household financial difficulties No financial difficulties Financial difficulties
Table 3 Associations between neighbourhood safety score (range = 1–10) and heavy smoking in the HELIUS study, Amsterdam, the Netherlands, 2011–2015a,b.
Model 1
Model 2
Model 3
Model 1
Model 2
Model 3
OR
95% CI
OR
95% CI
OR
95% CI
OR
95% CI
OR
95% CI
OR
95% CI
0.96
0.90–1.03
1.05
0.97–1.13
1.01
0.91–1.11
0.87
0.80–0.94
0.96
0.88–1.06
0.88
0.78–0.99
0.96 0.97
0.88–1.06 0.88–1.08
1.01 1.10
0.91–1.12 0.98–1.23
1.00 1.05
0.87–1.15 0.90–1.22
0.90 0.83
0.81–1.01 0.73–0.95
0.97 0.98
0.86–1.09 0.84–1.13
0.85 0.99
0.72–0.99 0.81–1.22
1.16 1.08 0.86 0.94 0.77
1.02–1.37 0.93–1.26 0.75–0.99 0.81–1.08 0.62–0.95
1.28 1.13 0.96 1.02 0.88
1.08–1.51 0.96–1.34 0.82–1.12 0.87–1.19 0.69–1.12
1.17 0.98 0.92 1.04 0.99
0.92–1.48 0.78–1.23 0.75–1.13 0.84–1.28 0.71–1.38
0.94 0.98 0.85 0.88 0.71
0.77–1.14 0.81–1.19 0.72–1.00 0.75–1.05 0.55–0.92
0.98 1.07 0.95 0.99 0.82
0.78–1.22 0.87–1.33 0.79–1.13 0.83–1.19 0.62–1.09
0.79 0.91 0.79 1.01 0.90
0.58–1.08 0.68–1.22 0.62–1.01 0.78–1.30 0.61–1.34
0.84 1.11
0.74–0.96 0.93–1.33
1.05 1.19
0.91–1.21 0.98–1.46
1.13 1.22
0.91–1.39 0.95–1.55
0.64 0.88
0.55–0.76 0.71–1.10
0.90 0.96
0.74–1.08 0.75–1.22
1.13 0.77
0.85–1.50 0.58–1.04
0.79 0.99 1.54 1.02
0.67–0.92 0.86–1.14 1.28–1.85 0.53–1.94
0.85 1.03 1.49 0.86
0.71–1.00 0.87–1.22 1.22–1.83 0.41–1.84
0.71 1.08 1.53 1.18
0.57–0.89 0.84–1.38 1.12–2.10 0.47–2.94
0.77 1.03 1.39 0.55
0.63–0.95 0.88–1.20 1.12–1.73 0.20–1.57
0.83 1.10 1.37 0.41
0.67–1.03 0.92–1.33 1.08–1.73 0.12–1.41
0.61 1.16 1.56 0.35
0.46–0.81 0.88–1.54 1.07–2.27 0.09–1.42
0.97 0.98 1.26 0.97
0.80–1.18 0.85–1.11 1.11–1.42 0.85–1.11
0.96 0.98 1.25 0.99
0.77–1.21 0.85–1.14 1.09–1.43 0.85–1.14
1.05 0.90 1.17 0.95
0.76–1.45 0.74–1.09 0.97–1.40 0.78–1.15
1.10 0.96 1.03 0.85
0.88–1.37 0.83–1.12 0.88–1.20 0.70–1.03
1.17 0.96 1.02 0.86
0.91–1.51 0.82–1.13 0.86–1.20 0.70–1.05
1.14 0.83 0.86 0.86
0.79–1.63 0.67–1.03 0.68–1.08 0.65–1.14
1.15
0.90–1.47
1.06
0.81–1.39
1.13
0.77–1.67
1.08
0.81–1.42
1.02
0.75–1.40
0.98
0.63–1.53
0.95
0.80–1.14
1.12
0.89–1.41
1.12
0.82–1.54
0.75
0.59–0.96
0.91
0.67–1.22
0.86
0.58–1.29
1.05 0.94
0.88–1.25 0.86–1.03
1.13 1.02
0.93–1.39 0.93–1.12
1.20 0.94
0.90–1.60 0.83–1.07
1.05 0.83
0.86–1.28 0.74–0.93
1.20 0.91
0.96–1.50 0.81–1.02
1.21 0.80
0.87–1.68 0.69–0.94
1.02 1.13 0.99
0.91–1.14 0.98–1.29 0.85–1.14
1.02 1.15 1.02
0.91–1.15 1.00–1.32 0.88–1.18
1.02 1.02 0.99
0.87–1.19 0.84–1.23 0.81–1.21
1.05 0.95 0.81
0.92–1.20 0.81–1.12 0.66–0.99
1.05 0.97 0.86
0.92–1.20 0.82–1.14 0.70–1.06
0.92 0.87 0.84
0.77–1.11 0.69–1.09 0.64–1.12
0.95
0.87–1.04
1.01
0.91–1.11
0.95
0.83–1.08
0.80
0.71–0.90
0.88
0.77–1.00
0.82
0.70–0.98
1.04
0.93–1.15
1.11
0.99–1.25
1.10
0.93–1.30
0.99
0.87–1.12
1.08
0.94–1.24
0.96
0.79–1.16
Full sample All participants Sex Men Women Age 18–30 years 31–40 years 41–50 years 51–60 years 61–70 years Ethnicity Dutch South-Asian Surinamese African Surinamese Turkish Moroccan Ghanaian Individual-level socioeconomic position Educational level Low Low-intermediate High-intermediate High Employment status Incapacitated for work Not in the labour force Not employed Employed Occupational level Low Intermediate High Household financial difficulties No financial difficulties Financial difficulties a
Abbreviations: CI = confidence interval, OR = odds ratio. The associations in Model 1 are adjusted for sex, age and ethnicity (if applicable). The associations in Model 2 are additionally adjusted for individuallevel socio-economic position (i.e., educational level, employment status, occupational level, household financial difficulties) (if applicable) and partner status. The associations in Model 3 are additionally adjusted for all environmental-level characteristics (i.e., population density, household density, proportion of minimum income households, and average property value of dwellings).
a
b
Abbreviations: CI = confidence interval, OR = odds ratio. b The associations in Model 1 are adjusted for sex, age and ethnicity (if applicable). The associations in Model 2 are additionally adjusted for individuallevel socio-economic position (i.e., educational level, employment status, occupational level, household financial difficulties) (if applicable) and partner status. The associations in Model 3 are additionally adjusted for all environmental-level characteristics (i.e., population density, household density, proportion of minimum income households, and average property value of dwellings).
between neighbourhood safety and the three types of smoking behaviour. Second, we lacked information about perceived stress in this study. As a consequence, we could not draw conclusions regarding the assumption that lower levels of neighbourhood safety leads to higher levels of stress, which in turn leads to (more) smoking among residents. Third, although we adjusted for relevant confounders in our analyses, there still might be residual confounding factors that we did not take into account, such as the availability/accessibility of tobacco retailers in residential areas. However, we did adjust for population density and environmental-level socio-economic status, which are strongly related to the availability/accessibility of tobacco retailers (Monshouwer et al., 2014; Rodriguez et al., 2013). Fourth, we used Geographic Information System techniques to construct residential areas on a relatively small scale and to calculate a neighbourhood safety score for these areas
from various backgrounds, and assessed whether these associations differ by sex, age, ethnicity and individual-level socio-economic position. The results of this study showed that higher levels of neighbourhood safety were significantly associated with less heavy smoking and less nicotine dependence, but not with less current smoking. The findings further showed that the associations between neighbourhood safety and the three types of smoking behaviour varied by ethnicity. A strength of this study is that the associations could be adjusted for a wide range of relevant confounders at the individual- as well as environmental-level. Another strength of the present study is the focus on three types of smoking behaviour. Some limitations have to be acknowledged as well. First, the cross-sectional design of the current study does not make it possible to determine causal relationships 116
Preventive Medicine 112 (2018) 111–118
E.J. Timmermans et al.
areas were more likely to be heavy smokers and nicotine dependent. A possible explanation for this finding could be that excessive smoking is a coping mechanism to reduce stress that is associated with living in an unpleasant or threatening residential environment (Diez Roux et al., 2003; Ellaway and Macintyre, 2009; Nielsen et al., 2008; Pearce et al., 2012; Shareck and Ellaway, 2011; Weden et al., 2008). Furthermore, smoking is described as a socially and culturally ingrained behaviour in deprived and unsafe neighbourhoods (Giskes et al., 2006). It has been suggested that pro-smoking norms are operating more strongly in such neighbourhoods, and that tobacco products are more readily available in these areas (Ellaway and Macintyre, 2009; Giskes et al., 2006; Stead et al., 2001; Wiltshire et al., 2001). It might be harder to decrease smoking habits by residents of deprived and unsafe neighbourhoods compared to those of more affluent and safe residential areas, due to a lack of social support, strong prevailing community norms and social reinforcement for smoking (Giskes et al., 2006; Stead et al., 2001). The results of the subgroup analyses show that the associations of neighbourhood safety with the three types of smoking behaviour vary by ethnicity. The observed variations in these associations might be due to differences in how ethnic groups experience neighbourhood safety (e.g., in terms of presence of ethnic discrimination) and/or to differences in the meaning they attach to it (Essed, 1991; Visser et al., 2017). In addition, the ethnic variations in these associations might be related to the differences in the coping resources (e.g., differences in social support networks through family systems and ethnic institutions) as used by the various ethnic group to effectively deal with lower levels of neighbourhood safety (Dhami and Skeikh, 2000; Visser et al., 2017). It has been suggested that women, older adults, ethnic minorities, and individuals with a low socio-economic position are more exposed to their immediate environment, and are more aware of, or sensitive to neighbourhood conditions than their counterparts (Dyck et al., 2001; Ellaway and Macintyre, 2009; Miles, 2006; Shareck and Ellaway, 2011; Stafford et al., 2005; Yen et al., 2009). In addition, these groups are suggested to be more dependent upon their local living environment and to be more concerned about safety because of their limited mobility, higher physical and/or social vulnerability (Skogan and Maxfield, 1981; Stafford and Marmot, 2003). In contrast with our expectations, findings of the subgroup-analyses suggest that lower levels of neighbourhood safety are associated with less heavy smoking in low educated individuals. Furthermore, the results of the subgroup-analyses suggest that lower levels of neighbourhood safety are associated with less nicotine dependence in the oldest age-group. These findings may need replication in other studies. The current findings provide empirical support that policies that improve neighbourhood safety potentially contribute to less heavy smoking and nicotine dependence. Future studies should focus on the mediating role of stress in the association between neighbourhood safety and smoking behaviour. In addition, future research should make a distinction between the various components of neighbourhoods safety (i.e., physical/social disorder, crime-related safety, and traffic-related safety) to obtain more detailed information about the association between neighbourhood safety and smoking.
Table 4 Associations between neighbourhood safety score (range = 1–10) and nicotine dependence in the HELIUS study, Amsterdam, the Netherlands, 2011–2015a,b.
Full sample All participants Sex Men Women Age 18–30 years 31–40 years 41–50 years 51–60 years 61–70 years Ethnicity Dutch South-Asian Surinamese African Surinamese Turkish Moroccan Ghanaian Individual-level socioeconomic position Educational level Low Low-intermediate High-intermediate High Employment status Incapacitated for work Not in the labour force Not employed Employed Occupational level Low Intermediate High Household financial difficulties No financial difficulties Financial difficulties
Model 1
Model 2
Model 3
OR
95% CI
OR
95% CI
OR
95% CI
0.85
0.76–0.94
0.92
0.82–1.03
0.81
0.69–0.95
0.90 0.78
0.79–1.03 0.66–0.92
0.93 0.91
0.80–1.08 0.75–1.10
0.76 0.93
0.62–0.94 0.72–1.20
0.80 0.92 0.96 0.80 0.72
0.62–1.03 0.73–1.16 0.79–1.16 0.65–0.99 0.51–1.01
0.79 0.98 1.05 0.89 0.81
0.59–1.06 0.75–1.27 0.84–1.30 0.71–1.12 0.56–1.18
0.54 0.92 0.84 0.77 1.11
0.36–0.81 0.64–1.33 0.62–1.13 0.56–1.05 0.64–1.90
0.57 0.83
0.46–0.70 0.64–1.07
0.82 0.89
0.64–1.05 0.67–1.17
0.91 0.74
0.62–1.33 0.53–1.04
0.73 1.03 1.43 0.91
0.56–0.95 0.85–1.24 1.09–1.88 0.26–3.18
0.70 1.08 1.52 0.88
0.53–0.94 0.86–1.34 1.13–2.04 0.18–4.21
0.51 1.07 1.58 0.69
0.35–0.73 0.76–1.49 0.99–2.53 0.09–5.30
1.03 0.97 1.03 0.79
0.80–1.34 0.81–1.16 0.85–1.26 0.61–1.03
1.00 0.93 1.00 0.81
0.74–1.35 0.77–1.13 0.81–1.23 0.61–1.07
1.15 0.76 0.76 0.79
0.75–1.77 0.58–0.98 0.57–1.02 0.54–1.16
1.05
0.77–1.43
1.00
0.71–1.42
0.98
0.60–1.61
0.89
0.64–1.23
1.05
0.71–1.56
1.14
0.66–1.94
1.02 0.76
0.81–1.28 0.67–0.88
1.09 0.84
0.84–1.41 0.72–0.97
0.97 0.70
0.67–1.41 0.57–0.86
1.03 0.90 0.81
0.88–1.20 0.74–1.11 0.62–1.07
1.00 0.88 0.87
0.86–1.18 0.71–1.09 0.65–1.15
0.82 0.84 0.81
0.66–1.03 0.63–1.12 0.54–1.19
0.70
0.60–0.82
0.76
0.64–0.89
0.72
0.57–0.90
1.06
0.91–1.22
1.11
0.9–1.30
0.91
0.72–1.14
a
Abbreviations: CI = confidence interval, OR = odds ratio. The associations in Model 1 are adjusted for sex, age and ethnicity (if applicable). The associations in Model 2 are additionally adjusted for individuallevel socio-economic position (i.e., educational level, employment status, occupational level, household financial difficulties) (if applicable) and partner status. The associations in Model 3 are additionally adjusted for all environmental-level characteristics (i.e., population density, household density, proportion of minimum income households, and average property value of dwellings). b
using micro-scale environmental data that were independent of participants' perceptions. We assume that this safety score provides an appropriate measure of the actual safety levels in a neighbourhood, but it should also be acknowledged that this measure not necessarily matches with the participants' perceptions of neighbourhood safety. Finally, the current study only examines the association of smoking with safety in the residential environment and does not consider safety in other places where people spend substantial amounts of time (e.g., for work, shopping, and recreation). However, this focus on place of residence is common to most geographical studies, not only for practical reasons, but also because people's home is important to their identity, rest and recovery from daily stress. The findings suggest that individuals who live in unsafe residential
Authors' contributions AEK and EJT conceptualized the study. EJT, EMV and MBS prepared the data. EMV obtained the geodata and performed the spatial analyses. EJT performed the statistical analyses, interpreted the data, and drafted the manuscript. All authors revised intermediate manuscript versions critically for important intellectual content. All authors have read and approved the final version of the manuscript. Conflict of interest statement The authors declare that they have no competing interests. 117
Preventive Medicine 112 (2018) 111–118
E.J. Timmermans et al.
Acknowledgments
Safety Monitor 2013/2014, . www.ois.amsterdam.nl/visualisatie/veiligheidsindex. html (Accessed: May, 2017). Neumann, T., Rasmussen, M., Heitmann, B.L., et al., 2013. Gold standard program for heavy smokers in a real-life setting. Int. J. Environ. Res. Public Health 10, 4186–4199. Nielsen, L., Curtis, T., Kristensen, T.S., et al., 2008. What characterizes persons with high levels of perceived stress in Denmark? A national representative study. Scand. J. Public Health 36, 369–379. Patterson, F., Seravalli, L., Hanlon, A., et al., 2012. Neighbourhood safety as a correlate of tobacco use in a sample of urban, pregnant women. Addict. Behav. 37, 1132–1137. Pearce, J., Barnett, R., Moon, G., 2012. Sociospatial inequalities in health-related behaviours: pathways linking place and smoking. Prog. Hum. Geogr. 36, 3–24. Rachelle, J.N., Wood, L., Nathan, A., et al., 2016. Neighbourhood disadvantage and smoking: examining the role of neighbourhood-level psychosocial characteristics. Health Place 40, 98–105. Rodriguez, D., Carlos, H.A., Adachi-Mejia, A.M., Berke, E.M., Sargent, J.D., 2013. Predictors of tobacco outlet density nationwide: a geographic analysis. Tob. Control. 22, 349–355. Stead, M., MacAskill, S., Makintosh, A., et al., 2001. “It's as if you're locked in”: qualitative explanations for area effects on smoking disadvantaged communities. Health Place 7, 333–343. Shareck, M., Ellaway, A., 2011. Neighbourhood crime and smoking: the role of objective and perceived crime measures. BMC Public Health 11, 930. Skogan, W.G., Maxfield, M.G., 1981. Coping with Crime. Individual and Neighbourhood Reactions. Sage Publications, London. Smeets, H., 2015. Verantwoording Veiligheidsindex 2014 [in English: Justification of the Safety Index in 2014]. Municipality of Amsterdam, Department of Research and Statistics, Amsterdam. Snijder, M.B., Galenkamp, H., Prins, M., et al., 2017. Cohort profile: the Healthy Life in an Urban Setting (HELIUS) study. BMJ Open 7, e017873. Stafford, M., Marmot, M., 2003. Neighbourhood deprivation and health: does it affect us all equally? Int. J. Epidemiol. 32, 357–366. Stafford, M., Cummins, S., Macintyre, S., et al., 2005. Gender differences in the associations between health and neighbourhood environment. Soc. Sci. Med. 60, 1681–1692. Statistics Netherlands, 2010. Standaard Beroepenclassificatie 2010 - Editie 2010 (In English: Standard Occupational Classification system 2010 - Edition 2010). The Hague/Heerlen: Statistics Netherlands. Stronks, K., Kulu-Glasgow, I., Agyemang, C., 2009. The utility of ‘country of birth’ for the classification of ethnic groups in health research: the Dutch experience. Ethn. Health 14, 255–269. Stronks, K., Snijder, M.B., Peters, R.J.G., et al., 2013. Unravelling the impact of ethnicity on health in Europe: the HELIUS study. BMC Public Health 13, 402. Tsjeng, M., Yeatts, K., Millikan, R., et al., 2001. Environmental-level characteristics and smoking in women. Am. J. Public Health 91, 1847–1850. United States Department of Health Human Services, 2014. The Health Consequences of Smoking-50 Years of Progress: a Report of the Surgeon General. United States Department of Health and Human Services, Center for Diseases Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion, Office on Smoking and Health, Atlanta, GA. Veldhuizen, E.M., Stronks, K., Kunst, A.E., 2013. Assessing associations between socioeconomic environment and self-reported health in Amsterdam using bespoke environments. PLoS One 8, e68790. Virtanen, M., Kivimäki, M., Kouvonen, A., et al., 2007. Average household income, crime, and smoking behaviour in a local area: the Finnish 10-town study. Soc. Sci. Med. 64, 1904–1913. Visser, M.J., Ikram, U.Z., Derks, E.M., et al., 2017. Perceived ethnic discrimination in relation to smoking and alcohol consumption in ethnic minority groups in the Netherlands: the HELIUS study. Int. J. Public Health. http://dx.doi.org/10.1007/ s00038-017-0977-2. Weden, M.M., Carpiano, R.M., Robert, S.A., 2008. Subjective and objective neighbourhood characteristics and adult health. Soc. Sci. Med. 66, 1256–1270. Wiltshire, S., Bancroft, A., Amos, A., Parry, O., 2001. “They're doing people a service” qualitative study of smoking, smuggling, and social deprivation. BMJ 323, 203–207. Yen, I.H., Michael, Y.L., Pedue, L., 2009. Neighbourhood environment in studies of health of older adults: a systematic review. Am. J. Prev. Med. 37, 455–463.
This work was financially supported by the Amsterdam Public Health Research Institute in Amsterdam, the Netherlands. The HEalthy LIfe in an Urban Setting (HELIUS) study is conducted by the Academic Medical Center (AMC) Amsterdam and the Public Health Service (GGD) of Amsterdam. Both organisations provided core support for HELIUS. The HELIUS study is also funded by the Dutch Heart Foundation, the Netherlands Organisation for Health Research and Development (ZonMw), the European Union 7th Framework Programme, and the European Fund for the Integration of non-EU immigrants (EIF). We are most grateful to the participants of the HELIUS study and the management team, research nurses, interviewers, research assistants and other staff members who have taken part in gathering the data of this study. Furthermore, we would like to thank the Department of Research and Statistics of the municipality of Amsterdam for providing the data of the Amsterdam Safety Monitor 2013/2014 and the integral demographic and socio-economic registries. References Aiken, L.S., West, S.G., 1991. Multiple Regression: Testing and Interpreting Interactions. SAGE Publications, Inc., Thousand Oaks, CA. Dhami, S., Skeikh, A., 2000. The Muslim family: predicament and promise. West. J. Med. 137, 352–356. Diez Roux, A.V., Mair, C., 2010. Neighborhoods and health. Ann. N. Y. Acad. Sci. 1186, 125–145. Diez Roux, A.V., Merkin, S.S., Hannan, P., et al., 2003. Area characteristics, individuallevel socioeconomic indicators, and smoking in young adults - the coronary artery disease risk development in young adults study. Am. J. Epidemiol. 157, 315–326. Dyck, I., Davis Lewis, N., McLafferty, S., et al., 2001. Women in their place: gender and perceptions of neighbourhoods in the West of Scotland. In: Dyck, I., Davis Lewis, N., McLafferty, S. (Eds.), Geographies of Women's Health. Routledge, London. Ellaway, A., Macintyre, S., 2009. Are perceived neighbourhood problems associated with the likelihood of smoking. J. Epidemiol. Community Health 63, 78–80. Essed, P., 1991. Understanding Everyday Racism: An Interdisciplinary Theory. Sage Publications, Inc., Thousand Oaks, CA. Ganz, M., 2000. The relationship between external threats and smoking in central Harlem. Am. J. Public Health 90, 367–371. Giskes, K., Van Lenthe, F.J., Turell, G., et al., 2006. Smokers living in deprived areas are less likely to quit: a longitudinal follow-up. Tob. Control. 15, 485–488. Goenka, S., Andersen, L.B., 2016. Our health is a function of where we live. Lancet 387, 2168–2170. Heatherton, T.F., Kozlowski, L.T., Frecker, R.C., et al., 1991. The Fagerström Test for nicotine dependence: a revision of the Fagerström Tolerance Questionnaire. Br. J. Addict. 86, 1119–1127. Jitnarin, N., Heinrich, K.M., Haddock, C.K., 2015. Neighbourhood environment perceptions and the likelihood of smoking and alcohol use. Int. J. Environ. Res. Public Health 12, 784–799. McGinn, A.P., Evenson, K.R., Herrin, A.H., et al., 2008. The association of perceived and objectively measured crime with physical activity: a cross-sectional analysis. J. Phys. Act. Health 5, 117–131. Miles, R., 2006. Neighbourhood disorder and smoking: findings of a European urban survey. Soc. Sci. Med. 63, 2464–2475. Monshouwer, K., Verdurmen, J., Ketelaars, T., Van Laar, M., 2014. Points of Sale of Tobacco Products: Synthesis of Scientific and Practice-based Knowledge on the Impact of Reducing the Number of Points of Sale and Restrictions on Tobacco Product Displays. Trimbos-Instituut, Netherlands Institute of Mental Health and Addiction, Utrecht, The Netherlands. Municipality of Amsterdam, 2017. Department of Research and Statistics. In: Amsterdam
118