Abstract
Aims
We examined whether distance from home to the nearest bar, i.e. alcohol outlet permitting consumption on the premises, is associated with risky alcohol behaviours.
Design
Cross-sectional and longitudinal study.
Setting and Participants
The cross-sectional data consisted of 78 858 and the longitudinal data of 54 778 Finnish Public Sector Study participants in between 2000 and 2009 [mean follow-up 6.8 years (SD=2.0)].
Measurements
Distances from home to the nearest bar were calculated using Global Positioning System-coordinates. The outcome variables were heavy alcohol use (drinking above the weekly guidelines) and extreme drinking occasions (passing out due to alcohol use). We used binomial logistic regression in cross-sectional analyses and in longitudinal mixed effects (between-individual) analyses. Conditional logistic regression was used in longitudinal fixed effects (within-individual) analyses.
Findings
Cross-sectionally, the likelihood of an extreme drinking occasion and heavy use was higher among those who resided <1 vs. ≥1 km from a bar. Longitudinally, between individuals, a decrease from >1 km to ≤1 km in distance was weakly associated with an extreme drinking occasion (1.18, 95% CI 0.98–1.41), and heavy use (1.12, 95% CI 0.97–1.29). Within-individual, the odds ratio for becoming a heavy user was 1.17 (95% CI, 1.02–1.34), per 1 km decrease in log-transformed continuous distance, the corresponding odds ratio for an extreme drinking occasion was 1.03 (95% CI, 0.89–1.18).
Conclusions
Moving place of residence close to or far from a bar appears to be associated with a small corresponding increase or decrease in risky alcohol behaviour.
INTRODUCTION
Heavy alcohol consumption is a significant contributor to the burden of diseases worldwide [1]. Risk factors for the development of alcohol disorders; heavy alcohol use and binge drinking, are common among adults and the middle-aged [2, 3]. Easy availability of alcohol is hypothesized to increase heavy alcohol consumption [4, 5], however, almost all prior studies on availability have been limited by cross-sectional designs, and have thus been unable to draw inferences about the causality of the association [6–11]. The only longitudinal study we are aware of found a weak association between the number of bars in the neighbourhood and alcohol consumption [12].
In the current study, we examined with greater precision the associations of living in proximity of a ‘bar’ (i.e. a place permitting consumption of alcohol on its premises such as a bar and or restaurant) with risky alcohol behaviours. In addition to cross-sectional data from a sample of over 78 000 participants, we examined a longitudinal subset of almost 55 000 participants who responded to at least two surveys. Furthermore, we quasi-experimentally studied the effect of change in distance on change in alcohol use among 6000 participants. These longitudinal analyses, providing an extension to prior research, enable an improved opportunity to evaluate the causality of the association. We hypothesised that the proximity of a bar to home increases the likelihood of risky alcohol behaviours, and that a decrease in distance to the nearest bar would be associated with an increase in risky drinking behaviours.
METHOD
Study population
The data were from the Finnish Public Sector Study (FPSS) cohort, an ongoing prospective study of employees working in ten towns and six hospital districts. The eligible population included all employees who had been working for the target organizations for a minimum of six months between 1991 and 2005 (n=151 618). The residential addresses and Geographical Positioning System (GPS)-coordinates of 146 600 employees between Jan 1st, 2000 and Dec 31st, 2010 were obtained from the Population Register Center. Identifiable questionnaire surveys have been repeated every four years at the participating organizations, starting from 2000. In 2005 and 2009 the surveys were also mailed to those who had completed questionnaires earlier, but had later left the organizations. The ethics committee of the Hospital District of Helsinki and Uusimaa approved the study.
In this study, we included all cohort members who were employed by the target organizations in 2000, 2004, and 2008 and responded to the surveys in these years. These data were complemented by the responses of the leavers from the 2005 and 2009 surveys. A total of 80 312 participants responded to at least one of these surveys (mean response rate 69%). Home address coordinates were not available for 296 participants, and 1156 had not responded to questions related to alcohol use. These were thus excluded from the study. In the cross-sectional analytic sample of 78 858 participants, we used data from the first year that survey responses and the distance variable were available. This sample did not differ substantially from the eligible cohort population in regard to sex (78% vs. 75% women) and age (mean age 44 vs. 44 years) distributions. For the prospective analyses, we used data for those who had responded to at least two surveys (n=54 778).
Alcohol consumption
The dichotomized variable for an extreme drinking occasion was created by asking the participants whether they had passed out at least once due to heavy drinking during the past 12 months [13]. The respondents also reported their habitual frequency and amount of beer, wine, and spirits intake, which was converted into grams of alcohol per week. One unit of pure alcohol (12 g) was equal to a 12 cl glass of wine, a 4 cl measure of spirits, or a 33 cl bottle of beer. The Finnish Ministry of Health and Social Affairs has set the lower limit for risk use of alcohol at 24 units/wk (288g) for men and 16 units/wk (192g) for women [14]. Based on these national cut-off points, we determined heavy alcohol use as >288g and >192g of consumed alcohol per week for men and women, respectively. These limits correspond with the medium risk levels of daily consumption set by the World Health Organization [15]. Supporting the validity of these self-reported measures, they were strong predictors of mortality from alcohol-related causes [16] in the FPSS cohort: the hazard ratio for mortality in the high versus low alcohol intake group was 2.32 (95% confidence interval (CI), 1.25–4.31), and in the group reporting extreme drinking occasions versus no extreme drinking 2.96 (95% CI, 1.62–5.39).
Alcohol availability
The addresses of all on-site alcohol outlets, ‘bars’, in Finland that allow alcohol consumption on the premises (like bars, restaurants, and hotels) with active liquor licences in 2004 or 2008 were derived from the Regional State Administrative Agency, the licensing authority in Finland. A geocoding service converted 80% of these addresses into GPS-coordinates. Combined with a manual internet map search we reached a total of 92% and 94% of successfully converted addresses for 2004 and 2008, respectively (total number of bars in all these years=11,157). Because the license information was incomplete for 2000, the 2004 bar locations were also used for 2000. We examined possible errors in the coordinates from the geo-coding service [17] using an internet map search for randomly selected 50 addresses. The coordinates matched for all of the 50 addresses, and 49 addresses were confirmed as businesses related to alcohol retail (e.g. bar or restaurant).
Covariates
Age, sex, and occupational status (the last used as a proxy for individual socioeconomic status, SES), were obtained from employers’ administrative registers. As in our earlier studies [16, 18], we used the Classification of Occupations by Statistics Finland [19] to classify individuals into three socioeconomic positions: high (e.g. teachers, physicians), intermediate (e.g. registered nurses, technicians), and low (e.g. cleaners, maintenance workers). Sub-optimal health (self-rated health poor or fairly poor) and marital status may also affect alcohol consumption [6, 20], thus, we obtained information for these covariates via the questionnaires.
Alcohol outlets have been found to concentrate in the deprived neighbourhoods [7], and a relationship between neighbourhood disadvantage and the co-occurrence of smoking, heavy alcohol use, and physical inactivity has been reported [21]. Thus, controlling for neighbourhood characteristics is essential. Neighbourhood-level data based on the total population in Finland were obtained from Statistics Finland’s grid database [22] in which socioeconomic characteristics are available for 250×250m map squares (a neighbourhood in this study). These data were linked to the survey responses using the GPS-coordinates of the participants’ home addresses. We calculated an index for neighbourhood disadvantage using grid database information on income, education attainment, and unemployment rate [21]. The population density of each neighbourhood (residents per 1 km2) was used as a proxy for the degree of urbanization. Statistics Finland does not release information for map squares with less than 10 cases at the time of data collection, thus some data were missing. When possible, missing data were replaced with the average of the eight surrounding squares. However, data on neighbourhood disadvantage remained missing for 727 participants and on population density for an additional 533 participants, so all these participants were excluded from the analyses.
Statistical analyses
As a preliminary analysis, we plotted the prevalence of risky alcohol use by quintiles of distance to a bar. The cross-sectional associations were examined using binomial logistic regression with the generalized estimating equations (GEE) method with exchangeable correlation structure (GENMOD procedure of SAS 9.2) [23, 24]. This method takes into account the clustering of the participants within the neighbourhoods. Due to gender differences in alcohol use and its health effects [2, 25] interactions between distance and sex were first tested by entering the term “sex × distance” to the models.
In the cross-sectional data, the mean distance from home to the nearest bar was 0.83 (SD =1.2) km, and 90% of the participants lived within 1.7 km from a bar. For this reason the cross-sectional results are presented as odds ratios (OR) and 95% confidence intervals (CI) for living in proximity to (<1 km) compared to living far from (≥ 1 km) a bar. Distance was fitted first without adjustments, and then with adjustments for age, sex, socioeconomic status, marital status, sub-optimal health, extreme drinking occasion/heavy alcohol use, and neighbourhood disadvantage and population density.
As a sensitivity analysis, we used the average distance of the five closest bars, <1 km vs. ≥ 1 km, as an alternative measure of availability [mean distance to the five nearest bars 1.54 (SD =1.8) km]. We also ran the adjusted models, excluding those who had resided in their neighbourhood for less than three years, to examine if short residence time (i.e. being less exposed to the current distance to a bar) affected the results. Additionally, analyses were stratified by socioeconomic status because alcohol consumption may vary between socioeconomic groups [26].
In the longitudinal data the mean distance from home to the nearest bar was 0.90 (SD=1.3) km. For the between-individual analyses, change in distance to the nearest bar was categorized as: 1) “Remained far” = distance remained ≥ 1 km (reference), 2) “Increased” = distance increased from <1 km to ≥ 1 km, 3) “Remained close” = distance remained <1 km and 4) “Decreased” = distance decreased from ≥ 1 km to <1 km. For these analyses we used the same binomial logistic regression models as in the cross-sectional analyses.
For the within-individual analyses we used the fixed effects method (also known as quasi-experimental case-control method) with conditional logistic regression models (LOGISTIC procedure of SAS) [27]. These models utilized information from those who reported risky use in one survey (case) and no risky use in another survey (control), and for whom the distance to the nearest bar changed (either the participants moved, or the bar moved) between two surveys [27]. Thus, there was no need to control for time-invariant confounders (e.g. sex). Importantly, some unobserved covariates such as personality and genetic background could also be controlled via this design, which is the main advantage of this method [27, 28]. If three individual drinking points were available, each “case” and “control” point accompanied by different distance measure, were compared in the analysis [27]. In these analyses we used the continuous negative log-transformed distance variable, because the binary variable would have only accounted for changes over the cut-off point distance. The models first included only the measure of time-dependent distance (or time-dependent mean distance) variable, and then adjustments for time-dependent covariates: age, marital status, sub-optimal health, employment status (stayed in vs. left the target organization between surveys), extreme drinking occasion/ heavy use, neighbourhood disadvantage, and population density. These models take into account any changes in the exposure and the covariates. To examine whether living in proximity of a bar is exogenous (i.e. the person does not choose to move closer to, or further from a bar) we ran analyses for people who remained at the same address between two surveys, but whose drinking pattern and distance to a bar had changed. Because one unit change in the log-transformed continuous distance variable has no relevance to real life (and because most participants lived within 1 km from a bar), these results are provided as odds ratios for a 1 km decrease (from 1 to 0 km) in distance (or mean distance) to the nearest bar with a 95% CI.
RESULTS
The total study population consisted of 62 864 women (mean age 44.2 years, standard deviation 10.0) and 15 995 men (43.5, 10.0). Other descriptive statistics of the cross-sectional and longitudinal samples are presented in Table 1. Unadjusted prevalence of an extreme drinking occasion and heavy alcohol use by quintiles of distance to a bar is shown in Figure 1.
Table 1.
Description of study participant characteristics in cross-sectional and longitudinal data.
| Variable | n missing | Men n=15 995 | Women n=62 864 | p-value a | ||
|---|---|---|---|---|---|---|
| Cross sectional data | n | (%) | n | (%) | ||
| Married/cohabiting | 705 | <0.01 | ||||
| Yes | 12 624 | (79.9) | 46 402 | (74.5) | ||
| Socioeconomic status by occupation | 833 | <0.01 | ||||
| High | 6162 | (39.1) | 17 274 | (27.7) | ||
| Intermediate | 4220 | (26.8) | 36 297 | (58.3) | ||
| Low | 5372 | (34.1) | 8701 | (14.0) | ||
| Sub-optimal health b | 752 | <0.01 | ||||
| Yes | 4450 | (28.0) | 15 072 | (24.2) | ||
| Extreme drinking occasion | 334 | <0.01 | ||||
| Yes | 2483 | (15.6) | 3332 | (5.3) | ||
| Heavy alcohol use | - | <0.01 | ||||
| Yes | 1567 | (9.8) | 4776 | (7.6) | ||
| Living in proximity (<1 km) of a bar | 12 210 | (76.3) | 47 821 | (76.1) | 0.48 | |
| Longitudinal data | n=10 223 | n=44 555 | ||||
| Married/cohabiting | 594 | <0.01 | ||||
| Yes | 8344 | (82.2) | 32 893 | (74.2) | ||
| SES, by occupation | 501 | <0.01 | ||||
| High | 4266 | (42.5) | 12 986 | (29.6) | ||
| Intermediate | 2745 | (27.3) | 25 086 | (57.3) | ||
| Low | 3031 | (30.2) | 5744 | (13.1) | ||
| Sub-optimal healthb | 646 | <0.01 | ||||
| Yes | 3097 | (30.4) | 12 589 | (28.4) | ||
| Extreme drinking occasion | 526 | <0.01 | ||||
| Yes | 1218 | (12.9) | 1852 | (4.2) | ||
| Heavy alcohol use | 321 | <0.01 | ||||
| Yes | 1093 | (10.7) | 4004 | (9.0) | ||
| Change in distance to nearest bar | 0.11 | |||||
| Remained farc | 4214 | (41.2) | 18 594 | (41.7) | ||
| Increasedd | 1041 | (10.2) | 4549 | (10.2) | ||
| Remained closee | 4164 | (40.7) | 17 612 | (39.5) | ||
| Decreasedf | 804 | (7.86) | 3800 | (8.53) | ||
p-value for sex differences in covariates
Self-rated health poor or fairly poor
Distance remained ≥ 1 km
Distance increased from <1 km to ≥1 km
Distance decreased from ≥1 km to < 1 km
Distance remained <1 km
Figure 1.
Prevalence of heavy alcohol use and extreme drinking occasion by quintiles of distance from home to nearest bar. Cross-sectional data. Error bars represent 95% confidence intervals.
In the cross-sectional analyses, the interaction between distance to a bar and sex was not significant for heavy alcohol use or an extreme drinking occasion (p-values 0.74 and 0.35, respectively), thus the analyses were performed for all participants. Living in proximity (<1 vs. ≥ 1 km) to a bar was associated with risky alcohol use (Table 2). The results remained the same when using the mean distance to the five nearest bars. Sensitivity analyses including only those who had lived in their neighbourhood for a minimum of three years replicated these results (OR 1.10, 95% CI: 1.00–1.21 for an extreme drinking occasion, and 1.03, 95% CI: 0.95–1.12 for heavy use in the adjusted model). In the analyses stratified by socioeconomic status, the odds ratios for an extreme drinking occasion and heavy alcohol use were elevated in the group of intermediate SES: 1.14 (95% CI: 1.04–1.26), and 1.10 (95% CI: 1.00–1.22), respectively.
Table 2.
Odds ratios (OR) and 95% confidence intervals (CI) for extreme drinking occasion and heavy alcohol use when living in proximity(<1 vs. ≥1 km) of a bar, cross-sectional data.
| Variable | Extreme drinking occasion | Heavy alcohol use | ||||
|---|---|---|---|---|---|---|
|
| ||||||
| OR | 95% | CI | OR | 95% | CI | |
| Crude modela | ||||||
| Proximity of a barb | 1.17 | 1.10 | 1.26 | 1.09 | 1.06 | 1.12 |
| Proximity of barsc | 1.24 | 1.17 | 1.31 | 1.23 | 1.16 | 1.30 |
| Adjusted modeld | ||||||
| Proximity of a barb | 1.09 | 1.01 | 1.17 | 1.04 | 0.97 | 1.11 |
| Sex (male) | 3.28 | 3.08 | 3.48 | 1.05 | 0.99 | 1.13 |
| Age (per 5 years) | 0.80 | 0.78 | 0.81 | 1.05 | 1.04 | 1.07 |
| Socioeconomic status (medium vs. high) | 1.24 | 1.15 | 1.33 | 0.65 | 0.61 | 0.70 |
| Socioeconomic status (low vs. high) | 1.72 | 1.59 | 1.87 | 0.61 | 0.56 | 0.66 |
| Marital status (single) | 1.39 | 1.31 | 1.49 | 1.00 | 0.94 | 1.07 |
| Sub-optimal healthe | 1.37 | 1.28 | 1.46 | 1.25 | 1.17 | 1.33 |
| Population density (per 1 SD) | 1.07 | 1.04 | 1.10 | 1.06 | 1.03 | 1.09 |
| Neighbourhood disadvantage (per 1 SD) | 1.03 | 1.00 | 1.06 | 0.96 | 0.93 | 1.00 |
| Heavy alcohol use | 3.67 | 3.41 | 3.95 | - | ||
| Extreme drinking occasion | - | 3.67 | 3.41 | 3.95 | ||
| Adjusted modeld | ||||||
| Proximity of barsc | 1.10 | 1.03 | 1.17 | 1.13 | 1.06 | 1.20 |
| Sex (male) | 3.27 | 3.08 | 3.48 | 1.05 | 0.99 | 1.12 |
| Age (per 5 years) | 0.80 | 0.78 | 0.81 | 0.66 | 0.62 | 0.70 |
| Socioeconomic status (medium vs. high) | 1.24 | 1.15 | 1.33 | 0.61 | 0.57 | 0.66 |
| Socioeconomic status (low vs. high) | 1.73 | 1.59 | 1.88 | 0.99 | 0.93 | 1.06 |
| Marital status (single) | 1.39 | 1.30 | 1.48 | 1.25 | 1.17 | 1.33 |
| Sub-optimal healthe | 1.37 | 1.28 | 1.46 | 0.66 | 0.62 | 0.70 |
| Population density (per 1 SD) | 1.06 | 1.02 | 1.09 | 1.04 | 1.01 | 1.07 |
| Neighbourhood disadvantage (per 1 SD) | 1.03 | 1.00 | 1.06 | 0.96 | 0.92 | 0.99 |
| Heavy alcohol use | 3.66 | 3.40 | 3.94 | - | ||
| Extreme drinking occasion | - | 3.67 | 3.41 | 3.95 | ||
No covariates in the model
Living <1 vs. ≥1 km away from the nearest bar
Mean distance to the five nearest bars <1 vs. ≥1 km
Model adjusted for age, sex, socioeconomic status, marital status, sub-optimal health, extreme drinking occasion/ heavy use, neighbourhood disadvantage and population density
self-rated health poor or fairly poor
SD= standard deviation
In the longitudinal dataset, interactions between sex and the time-dependent distance were non-significant (p-value 0.47 for an extreme drinking occasion, and 0.99 for heavy use). Thus, these analyses were also performed for all participants only. In the between-individual analyses, the likelihood of an extreme drinking occasion was higher if a bar remained close vs. remained far (Table 3). If distance to bar decreased the likelihood of risky use was slightly higher than if a bar remained far (Table 3).
Table 3.
Odds ratios (OR) and 95% confidence intervals (CI) for extreme drinking occasion and heavy alcohol use in association with change in distance to the nearest bar, longitudinal (between-individual) data.
| Change in distance to the nearest bar (n=54,778) | Extreme drinking occasion | Heavy alcohol use | ||||
|---|---|---|---|---|---|---|
|
| ||||||
| OR | 95% | CI | OR | 95% | CI | |
| Crude modela | ||||||
| Remained farb | 1 | 1 | ||||
| Increasedc | 1.29 | 1.10 | 1.52 | 0.94 | 0.82 | 1.07 |
| Remained closed | 1.22 | 1.10 | 1.35 | 1.17 | 1.08 | 1.27 |
| Decreasede | 1.35 | 1.14 | 1.59 | 1.18 | 1.03 | 1.36 |
| Adjusted modelf | ||||||
| Remained farb | 1 | 1 | ||||
| Increasedc | 1.14 | 0.96 | 1.35 | 0.91 | 0.79 | 1.05 |
| Remained closed | 1.13 | 1.00 | 1.27 | 1.04 | 0.95 | 1.14 |
| Decreasede | 1.18 | 0.98 | 1.41 | 1.12 | 0.97 | 1.29 |
| Change in mean distance to the 5 nearest bars | ||||||
| Crude modela | ||||||
| Remained farb | 1 | 1 | ||||
| Increasedc | 1.15 | 1.06 | 1.25 | 1.12 | 1.05 | 1.20 |
| Remained closed | 1.33 | 1.19 | 1.48 | 1.40 | 1.29 | 1.52 |
| Decreasede | 1.39 | 1.09 | 1.76 | 1.51 | 1.25 | 1.82 |
| Adjusted modelf | ||||||
| Remained farb | 1 | 1 | ||||
| Increasedc | 1.09 | 1.00 | 1.20 | 1.07 | 1.00 | 1.15 |
| Remained closed | 1.18 | 1.02 | 1.36 | 1.23 | 1.10 | 1.37 |
| Decreasede | 1.15 | 0.88 | 1.50 | 1.39 | 1.13 | 1.71 |
No covariates in the model
Distance (mean distance) remained ≥1 km;
Distance (mean distance) increased from <1 km to ≥1 km
Distance (mean distance) remained <1 km
Distance (mean distance) decreased from ≥1 km to <1 km
Model adjusted for age, sex, socioeconomic status, marital status, sub-optimal health, extreme drinking occasion /heavy use, neighbourhood disadvantage and population density
Figure 2 shows that the shortening of distance to a bar during the follow-up was associated with an increase in likelihood of heavy alcohol use. When the log-transformed distance to the nearest bar decreased by 1 km, the odds ratio of heavy alcohol use was 1.17 (95% CI: 1.02–1.34, adjusted model). In relation to changes in extreme drinking occasions, the association was weaker (OR 1.03, 95% CI: 0.89–1.18, adjusted model). Results were similar when using the measure for a 1 km decrease in the mean log-transformed distance to the five nearest bars (OR for heavy drinking in the adjusted model 1.14, 95% CI: 1.01–1.30) (Figure 2). Results by all covariates of the adjusted within-individual models are presented in supplemental eTable 1.
Figure 2.
Odds ratios (OR) and 95% confidence intervals (CI) for heavy alcohol use and extreme drinking occasion in association with 1 km decrease in log-transformed distance from home to the nearest bar. Results from longitudinal within-individual analyses.
a Model adjusted for time-dependent confounders: age, socioeconomic status, marital status, suboptimal health, employment status, extreme drinking occasion/ heavy use, and neighbourhood disadvantage and population density
We studied the exogeneity of these associations using information on non-movers. The addresses of 2508 (53%) of those reporting change in extreme drinking occasions, and of 3576 (60%) of those reporting change in heavy use remained the same between surveys. If the log-transformed distance to the nearest bar decreased by 1 km because the bar moved, the odds of heavy use was elevated (1.36, 95% CI: 0.96–1.94), although non-significant. For an extreme drinking occasion the corresponding odds ratio was 1.07 (95% CI: 0.70–1.65).
DISCUSSION
In the present study, living in proximity of ‘a bar’ (i.e. an on-site alcohol outlet) was associated with increased odds of extreme drinking occasions and heavy alcohol use among adults. Importantly, results from the longitudinal analyses suggested that a reduction in distance to a bar may be causally related to an increase in risky alcohol behaviours.
We are not aware of any previous longitudinal studies that have used distance from home to an alcohol outlet as the measure of availability. Overall, most prior studies have used outlet density as the measure of availability[5]. However, outlet density may not fully capture all these associations, as even in the low density areas the distance to an outlet may be short, and the prevalence of exposure in the neighbourhood may not represent the etiologically meaningful exposure [29]. Furthermore, density measures may not equally effectively describe the exposure of those living in the centre of the area versus those at the periphery[30].
Few prior cross-sectional studies have examined the association between distance to on-site alcohol outlet and alcohol consumption [7, 8, 12]. In contrast to our findings, one study reported no association between high alcohol consumption among adults and living “near” or “far” from on-site outlets [7]. That study, however, did not study risk-level alcohol consumption as did ours. Another study, investigating suburban youth, found no association between distance from home to an outlet and alcohol use [8]. However, since adolescents are legally not permitted to buy alcohol, it is not clear whether distance to the nearest outlet is a relevant exposure for that age group. In addition, a rather long mean distance (>2 km) from home to an outlet suggests a different study context to that of our study (mean distance <1 km). A third study found that among adolescents, binge drinking (having consumed five subsequent drinks during the previous 30 days) was associated with on-site alcohol outlets within 0.5 miles (0.8 km) from home [31], which is in agreement with our findings.
In the only previous longitudinal study on alcohol availability and alcohol use we are aware of, the measures for availability were the number of bars and bar density within 0.5 km from home [12]. In line with the current findings, the authors concluded that adding bars to the individuals’ local neighbourhoods raises alcohol consumption by small amounts. They found evidence for exogeneity in areas of greater outlet density, which agrees with our findings. Besides the smaller sample size compared to our study, one drawback of that study was that they used the Yellow Pages as the source of alcohol outlet data, which may not include all outlets, thus biasing their findings towards the null. Longitudinal studies on alcohol-related harms, other than heavy consumption, have suggested an association between the availability of bars and off-site outlets with violence rates [32, 33] and alcohol-related mortality [34].
The association between risky alcohol use and distance to a bar supports the hypothesis that alcohol availability is an important determinant of the development of alcohol disorders. Limiting on-premises alcohol sales has been found to be effective in reducing homicides, violence, and other alcohol-related problems [5, 35]. Restricting opening hours and/or times for selling alcohol could be a universal way of reducing risky alcohol consumption [4]. Higher taxation and pricing of alcohol are also seen as efficient and generalizable ways to reduce alcohol dependence and alcohol-related harms [4, 36]. Recent findings from a natural experiment in Finland reported an increase in alcohol-related mortality after an increase in availability of alcohol, indicated by price reduction [6]. Although alcohol prices in Finland are already relatively high, higher taxation could affect on-site alcohol consumption.
Our study has several limitations. One is that we used self-reported data on alcohol consumption although people tend to under-report their drinking, and they may not correctly recall the amounts consumed [37]. Therefore, our findings may underestimate the actual alcohol use potentially attenuating the observed associations. However, although self-reporting is sensitive to response style, it is less likely to vary within an individual. In the within-individual analyses, there was no need to control for time invariant variables such as response style, thus it is unlikely that the observed within-individual variance was due to cross-sectional differences between individuals [27]. We had no data on the number of drinks consumed consecutively. Therefore we could not assess episodes of heavy drinking as in most studies: “more than five or four drinks consumed on one occasion” [38]. However, passing out during the prior 12 months, which we used instead, has been acknowledged as a proxy for an at-risk drinking pattern [13].
Although we were able to use longitudinal data, we cannot rule out the possibility of self-selection bias or reverse causation. The first means that those with a high level of alcohol consumption may choose to live close to bars. However, due to economic circumstances not everyone can choose their area of residence, and self-selection may therefore be more associated with people with a high SES. Reverse causation occurs when bars are established in areas where the entrepreneurs know people are likely to consume more alcohol (e.g. areas inhabited by people with lower SES, who, in Finland, are most likely to be heavy users [39]). To control for these biases we adjusted the models for individual- and area-level SES, and ran individual SES-stratified analyses. These associations were the strongest among those with intermediate SES, possibly because this group was the largest in this study. Although we were able to show that individuals changed their risky alcohol use when distance to an outlet changed, further longitudinal studies are needed to examine whether any biases drive these associations. We also recognize that because the distance from the participants’ residence to a bar was calculated “as the crow flies”, it may not provide accurate information about the distance needed to travel to a bar. Finally, the study population consisted of Finnish adults, the majority of whom were employed. Because the long-term unemployed may drink more alcohol than the employed [40], our results may be inaccurate estimations for the unemployed. Moreover, the drinking behaviours and cultural norms vary between countries, for instance, in the UK [41] and Australia [42] heavy drinking is more common, and in the U.S. less common [43], than in our sample, which is why the generalizability of our results to other cultures may be limited.
In summary, the proximity of a bar increased the odds of risky alcohol behaviours among working-aged men and women, and the results from the longitudinal analyses suggest the association may be causal. Thus, availability of alcohol seems to be an important determinant in risky alcohol use.
Supplementary Material
Acknowledgments
Sources of Funding
This study was supported by the Academy of Finland (projects 124271, 124322, 129262 and 126602) and the participating organizations. Prof. Mika Kivimäki is supported by the BUPA Foundation, UK, the UK Medical Research Council, the National Institute on Aging (R01AG034454), and the National Heart, Lung, and Blood Institute (R01HL036310), NIH, USA. The funders have imposed no contractual constraints on publishing.
Footnotes
Declaration of Interest
All authors declare that they have no financial interests or connections to the tobacco, alcohol, pharmaceutical, or gaming industries.
References
- 1.Rehm J, Mathers C, Popova S, Thavorncharoensap M, Teerawattananon Y, Patra J. Global burden of disease and injury and economic cost attributable to alcohol use and alcohol-use disorders. Lancet. 2009 Jun 27;373(9682):2223–33. doi: 10.1016/S0140-6736(09)60746-7. [DOI] [PubMed] [Google Scholar]
- 2.Parna K, Rahu K, Helakorpi S, Tekkel M. Alcohol consumption in Estonia and Finland: Finbalt survey 1994–2006. BMC Public Health. 2010;10(261):261. doi: 10.1186/1471-2458-10-261. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Blazer DG, Wu LT. The Epidemiology of At-Risk and Binge Drinking Among Middle-Aged and Elderly Community Adults National Survey on Drug Use and Health. Am J Psychiatry. 2009 Oct;166(10):1162–9. doi: 10.1176/appi.ajp.2009.09010016. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Kendler KS. Levels of explanation in psychiatric and substance use disorders: implications for the development of an etiologically based nosology. Mol Psychiatry. 2012;17(1):11–21. doi: 10.1038/mp.2011.70. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Popova S, Giesbrecht N, Bekmuradov D, Patra J. Hours and days of sale and density of alcohol outlets: impacts on alcohol consumption and damage: a systematic review. Alcohol Alcohol. 2009 Sep-Oct;44(5):500–16. doi: 10.1093/alcalc/agp054. [DOI] [PubMed] [Google Scholar]
- 6.Herttua K, Martikainen P, Vahtera J, Kivimaki M. Living alone and alcohol-related mortality: a population-based cohort study from Finland. PLoS Med. 2011 Sep;8(9):e1001094. doi: 10.1371/journal.pmed.1001094. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Pollack CE, Cubbin C, Ahn D, Winkleby M. Neighbourhood deprivation and alcohol consumption: does the availability of alcohol play a role? Int J Epidemiol. 2005 Aug;34(4):772–80. doi: 10.1093/ije/dyi026. [DOI] [PubMed] [Google Scholar]
- 8.Pasch KE, Hearst MO, Nelson MC, Forsyth A, Lytle LA. Alcohol outlets and youth alcohol use: exposure in suburban areas. Health Place. 2009 Jun;15(2):642–6. doi: 10.1016/j.healthplace.2008.10.002. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Schonlau M, Scribner R, Farley TA, Theall K, Bluthenthal RN, Scott M, et al. Alcohol outlet density and alcohol consumption in Los Angeles county and southern Louisiana. Geospat Health. 2008 Nov;3(1):91–101. doi: 10.4081/gh.2008.235. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Scribner RA, Cohen DA, Fisher W. Evidence of a structural effect for alcohol outlet density: a multilevel analysis. Alcohol Clin Exp Res. 2000 Feb;24(2):188–95. [PubMed] [Google Scholar]
- 11.Connor JL, Kypri K, Bell ML, Cousins K. Alcohol outlet density, levels of drinking and alcohol-related harm in New Zealand: a national study. J Epidemiol Community Health. 2011 Oct;65(10):841–6. doi: 10.1136/jech.2009.104935. [DOI] [PubMed] [Google Scholar]
- 12.Picone G, MacDougald J, Sloan F, Platt A, Kertesz S. The effects of residential proximity to bars on alcohol consumption. Int J Health Care Finance Econ. 2010 Dec;10(4):347–67. doi: 10.1007/s10754-010-9084-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Paljarvi T, Makela P, Poikolainen K, Suominen S, Car J, Koskenvuo M. Subjective measures of binge drinking and alcohol-specific adverse health outcomes: a prospective cohort study. Addiction. 2012 Feb;107(2):323–30. doi: 10.1111/j.1360-0443.2011.03596.x. [DOI] [PubMed] [Google Scholar]
- 14.Työterveyslaitos ja Sosiaali- ja terveysministeriö, [Finnish Institute of Occupatinal Health and Finnish Ministry of Social Affairs and Health]. Riskikulutuksen varhainen tunnistaminen ja mini-interventio -hoitosuosituksen yhteenveto [Adapted translation into Finnish based on “Alcohol and Primary Health Care: Clinical Guidelines on Identification and Brief Interventions. Department of Health of the Government of Catalonia: Barcelona.” by Anderson, P., Gual, A., Colom, J. (2005).]; 2006.
- 15.WHO. International Guide for Monitoring Alcohol Consumption and Related Harm Department of Mental Health and Substance Dependence. Noncommunicable Diseases and Mental Health Cluster, WHO; 2000. [Google Scholar]
- 16.Vahtera J, Pentti J, Kivimaki M. Sickness absence as a predictor of mortality among male and female employees. J Epidemiol Community Health. 2004 Apr;58(4):321–6. doi: 10.1136/jech.2003.011817. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Krieger N, Waterman P, Lemieux K, Zierler S, Hogan JW. On the wrong side of the tracts? Evaluating the accuracy of geocoding in public health research. Am J Public Health. 2001 Jul;91(7):1114–6. doi: 10.2105/ajph.91.7.1114. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Kivimaki M, Lawlor DA, Davey Smith G, Kouvonen A, Virtanen M, Elovainio M, et al. Socioeconomic position, co-occurrence of behavior-related risk factors, and coronary heart disease: the Finnish Public Sector study. Am J Public Health. 2007 May;97(5):874–9. doi: 10.2105/AJPH.2005.078691. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Statistics Finland. Classification of Occupations. Helsinki: Statistics Finland; 1987. [Google Scholar]
- 20.Tran TV, Nguyen D, Chan K, Nguyen TN. The association of self-rated health and lifestyle behaviors among foreign-born Chinese, Korean, and Vietnamese Americans. Qual Life Res. 2012 Mar 16; doi: 10.1007/s11136-012-0155-1. [DOI] [PubMed] [Google Scholar]
- 21.Halonen JI, Kivimaki M, Pentti J, Kawachi I, Virtanen M, Martikainen P, et al. Quantifying neighbourhood socioeconomic effects in clustering of behaviour-related risk factors: a multilevel analysis. PLoS One. 2012;7(3):e32937. doi: 10.1371/journal.pone.0032937. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Statistics Finland. Grid Database. 2007 [cited January 11th, 2012]; Available from: http://www.tilastokeskus.fi/tup/ruututietokanta/index_en.html.
- 23.Lipsitz S, Kim K, Zhao L. Analysis of repeated categorical data using generalized estimating equations. Stat Med. 1994;13:1149–63. doi: 10.1002/sim.4780131106. [DOI] [PubMed] [Google Scholar]
- 24.SAS. software release 9.2. Cary, NC: SAS Institute; 2001. [Google Scholar]
- 25.Matheson FI, White HL, Moineddin R, Dunn JR, Glazier RH. Drinking in context: the influence of gender and neighbourhood deprivation on alcohol consumption. J Epidemiol Community Health. 2011 doi: 10.1136/jech.2010.112441. Epub Feb 17:17. [DOI] [PubMed] [Google Scholar]
- 26.Giskes K, Turrell G, Bentley R, Kavanagh A. Individual and household-level socioeconomic position is associated with harmful alcohol consumption behaviours among adults. Aust N Z J Public Health. 2011 Jun;35(3):270–7. doi: 10.1111/j.1753-6405.2011.00683.x. [DOI] [PubMed] [Google Scholar]
- 27.Allison PD. Fixed Effects Regression Methods for Longitudinal Data Using SAS: SAS. Cary, NC: SAS Institute Inc; 2005. [Google Scholar]
- 28.Gilman SE, Martin LT, Abrams DB, Kawachi I, Kubzansky L, Loucks EB, et al. Educational attainment and cigarette smoking: a causal association? Int J Epidemiol. 2008 Jun;37(3):615–24. doi: 10.1093/ije/dym250. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Basta LA, Richmond TS, Wiebe DJ. Neighborhoods, daily activities, and measuring health risks experienced in urban environments. Soc Sci Med. 2010 Dec;71(11):1943–50. doi: 10.1016/j.socscimed.2010.09.008. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Matisziv T, Crubesic T, Wei H. Downscaling spatial structure for the analysis of epidemiological data. Computers, Environment and Urban Systems. 2008;32(1):81–93. [Google Scholar]
- 31.Truong KD, Sturm R. Alcohol environments and disparities in exposure associated with adolescent drinking in California. Am J Public Health. 2009 Feb;99(2):264–70. doi: 10.2105/AJPH.2007.122077. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Livingston M. A longitudinal analysis of alcohol outlet density and domestic violence. Addiction. 2011 May;106(5):919–25. doi: 10.1111/j.1360-0443.2010.03333.x. [DOI] [PubMed] [Google Scholar]
- 33.Gruenewald PJ, Remer L. Changes in outlet densities affect violence rates. Alcohol Clin Exp Res. 2006 Jul;30(7):1184–93. doi: 10.1111/j.1530-0277.2006.00141.x. [DOI] [PubMed] [Google Scholar]
- 34.Stockwell T, Zhao J, Macdonald S, Vallance K, Gruenewald P, Ponicki W, et al. Impact on alcohol-related mortality of a rapid rise in the density of private liquor outlets in British Columbia: a local area multi-level analysis. Addiction. 2011 Apr;106(4):768–76. doi: 10.1111/j.1360-0443.2010.03331.x. [DOI] [PubMed] [Google Scholar]
- 35.Duailibi S, Ponicki W, Grube J, Pinsky I, Laranjeira R, Raw M. The effect of restricting opening hours on alcohol-related violence. Am J Public Health. 2007 Dec;97(12):2276–80. doi: 10.2105/AJPH.2006.092684. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Elder RW, Lawrence B, Ferguson A, Naimi TS, Brewer RD, Chattopadhyay SK, et al. The effectiveness of tax policy interventions for reducing excessive alcohol consumption and related harms. Am J Prev Med. 2010 Feb;38(2):217–29. doi: 10.1016/j.amepre.2009.11.005. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Ekholm O, Strandberg-Larsen K, Gronbaek M. Influence of the recall period on a beverage-specific weekly drinking measure for alcohol intake. Eur J Clin Nutr. 2011 Apr;65(4):520–5. doi: 10.1038/ejcn.2011.1. [DOI] [PubMed] [Google Scholar]
- 38.Jackson KM. Heavy episodic drinking: determining the predictive utility of five or more drinks. Psychol Addict Behav. 2008 Mar;22(1):68–77. doi: 10.1037/0893-164X.22.1.68. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Mäkelä P, Mustonen H, Tigerstedt C, Juo Suomi. Alcohol consumption and its changes among Finns in 1968–2008] Helsinki: National Institute for Health and Welfare; 2010. Suomalaisten alkoholinkäyttö ja sen muutokset 1968–2008 [Finland Drinks. [Google Scholar]
- 40.Henkel D. Unemployment and substance use: a review of the literature (1990–2010) Curr Drug Abuse Rev. 2011 Mar 1;4(1):4–27. doi: 10.2174/1874473711104010004. [DOI] [PubMed] [Google Scholar]
- 41.Robinson S, Harris H. Smoking and drinking among adults, 2009. A report on the 2009 General Lifestyle Survey. London: The Office for National Statistics; 2011. [Google Scholar]
- 42.Australian Institute of Health and Welfare. Drug statistics series no 25 Cat no PHE 145. Canberra: 2011. 2010 National Drug Strategy Household Survey report. [Google Scholar]
- 43.Centers for Disease Control and Prevention, Office of Surveillance Epidemiology and Laboratory Services. Behavioral Risk Factor Surveillance System. Prevalence and Trends Data. Nationwide (States and DC) - 2008 Alcohol Consumption. 2008 [cited June 27th, 2012]; Available from: http://apps.nccd.cdc.gov/BRFSS/sex.asp?cat=AC&yr=2008&qkey=4413&state=UB.
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.