Article Text
Background Earlier research has shown that residents of Dutch deprived neighbourhoods drink less alcohol than people in other areas. We aimed to assess the role of individual and neighbourhood characteristics in a cross-sectional, nationwide, multilevel study.
Methods Individual data of 30 117 Dutch adults, living in 1722 neighbourhoods across the Netherlands, were obtained from the 2004 to 2009 national health survey (POLS). Chronic heavy alcohol consumption was measured as ≥14 drinks/week for women and ≥21 for men, and episodic heavy drinking as ≥6 drinks/day at least once a week. Neighbourhood deprivation was dichotomous; deprived districts as selected by the Dutch government versus other areas. Multilevel logistic regression models of the association between deprivation and heavy drinking were corrected for age, gender, household composition, population density and potential predictors ethnicity, socioeconomic status (education, income), neighbourhood-level social cohesion and percentage Muslims.
Results The prevalence of heavy drinking was lower in deprived neighbourhoods than in the rest of the Netherlands. This association was found for both chronic and episodic heavy drinking (OR=0.58 (0.47 to 0.72) and OR=0.57 (0.45 to 0.72), respectively). Adding ethnicity to the model reduced these associations by approximately one half. Socioeconomic composition did not contribute to the relationship. The proportion of Muslims explained a small part, while social cohesion explained even less of the association. Stronger associations were observed for women and older adults than for men and younger adults.
Conclusions The lower prevalence of heavy drinking occurring in deprived areas is largely explained by the ethnicity of neighbourhood residents.
Statistics from Altmetric.com
Introduction
Health inequalities between disadvantaged and more privileged groups in society have been described for decades.1–3 The role of the deprivation of the residential neighbourhood has also been recognised. Residents of more deprived neighbourhoods are often in poorer health and practice unhealthier behaviours (such as smoking, inactivity and poor diet) than residents of more affluent neighbourhoods.4 ,5
Neighbourhood-level inequalities in alcohol consumption have been studied less extensively and the results are less clear. Residents of more deprived neighbourhoods are more likely to abstain from alcohol.6 ,7 Studies on heavy drinking outcomes are much more inconsistent. Three North American studies suggest that binge drinking is more common in deprived areas,8–10 while Australian studies found no association between neighbourhood deprivation and binge drinking.11 ,12 Other North American studies, investigating a more chronic type of heavy drinking, report conflicting results. There are more studies reporting heavier drinking in deprived areas8 ,9 ,13 than studies reporting the opposite.6 ,1
European research is scarce. Lakshman et al4 found that even though other health behaviours like smoking and poor diet were more frequent in English deprived neighbourhoods, chronic heavy alcohol use was less common. Monden et al15 reported that neighbourhood socioeconomic status (SES) is not associated with chronic heavy drinking in the Dutch town of Eindhoven. A recent study from the Netherlands found that residents of deprived areas are more likely to abstain from alcohol and are less likely to engage in excessive drinking.16
Studies on general health outcomes imply that both demographic composition of the neighbourhood and the environmental context may affect health.17 ,18 Therefore, the association between deprivation and alcohol use can be explained by both individual characteristics of residents and features of the neighbourhood environment. In this explanatory study we focus on three types of possible predictors: ethnicity and neighbourhood religion, SES, and social environment of the neighbourhood.
First, there is a large compositional difference in ethnicity between deprived and non-deprived areas in the Netherlands. Ethnic minorities are often concentrated in the more deprived neighbourhoods.19 In the Netherlands, a large part of the immigrant population originates from Turkey and Morocco, countries with an Islamic culture. Since these groups traditionally have low levels of alcohol consumption, their presence might explain part of the lower heavy drinking rates in deprived neighbourhoods.20 We furthermore hypothesise that the presence of Turkish and Moroccan residents in the neighbourhood could influence the local social norm for alcohol use.20 A previous Dutch study among adolescents found that the presence of Muslim students in schools was associated with lower alcohol consumption of other students.21 We expect a similar phenomenon to occur among adults in the neighbourhood context.
Second, residents living in deprived neighbourhoods are often characterised by a lower SES. Studies show that a lower SES is associated with abstinence and moderate alcohol use in the USA22 and UK.23 Differences in alcohol use between deprived areas and other areas might be attributed to the larger proportion of residents with lower SES living in deprived neighbourhoods.
Finally, social cohesion has been suggested as a potential explanatory factor for the association between neighbourhood deprivation and alcohol use.16 Social cohesion can be defined as the “extent of connectedness and solidarity among groups in society.”24 Stronger social cohesion is likely to encourage alcohol use,25–28 due to the social character of drinking behaviour.26 Social cohesion is expected to be even more important for binge drinking than for chronic drinking behaviour, because binge drinking is more of a social activity.29 Deprived neighbourhoods are characterised by weaker social cohesion.30 We therefore expect that a lower prevalence of heavy drinking in deprived areas might in part be attributable to lower levels of social cohesion.
The objective of this paper was to assess possible explanations for the association between heavy drinking and neighbourhood deprivation in the Netherlands. We investigate whether the three previously mentioned predictors contribute to differences in heavy alcohol use between deprived neighbourhoods and other areas. Analyses will be performed on a large national survey database covering the entire Dutch adult population.
Methods
Study population
Variables were measured at the individual and neighbourhood level. Individual-level data of Dutch adults, age 18 and over, were derived from the 2004–2009 National Integrated Survey on Living Conditions (POLS). The POLS is an extended questionnaire conducted annually among a population-based representative sample of the Dutch population.31 ,32 Municipalities were selected, from which individuals were derived using the Municipal Population Register. Computer-Assisted Personal Interviewing was used for the basic questionnaire, while questions on alcohol use were asked in supplementary, self-administered internet or paper and pencil questionnaires.
Neighbourhood-level variables were measured using information from the 57 405 respondents of the 2006 Netherlands Housing Research (WoON) survey.33 Neighbourhoods contained an average of 33 WoON respondents. As neighbourhood measurements (see ‘neighbourhood-level variables’) were aggregated mean scores, measurements based on fewer than 10 WoON respondents per postcode area were considered unreliable and these areas were thus excluded. POLS respondents living in these areas were therefore excluded as well, which reduced the initial sample size of 43 726 to 35 418. Exclusion of individuals without information on alcohol use resulted in a total study population of 30 117 POLS respondents divided over 1722 postcode areas. Non-response to alcohol-related items was mainly due to these questions being included in a follow-up questionnaire, which was not completed by all initial survey respondents. Demographic variables had <1% missing values.
The availability of the four-digit postcode for individuals in the POLS survey allowed us to link WoON-survey neighbourhood-level information to the individuals in the POLS survey.
Individual-level variables
We measured both chronic and episodic heavy alcohol consumption. Episodic heavy drinking (ie, binge drinking) was defined as having six drinks or more in 1 day at least once a week, derived from a single POLS question which was based on definitions by Garretsen.34 Chronic heavy drinking was defined using a quantity-frequency measurement, for a typical week. For both weekdays (Monday to Thursday) and weekend days (Friday to Sunday), the following questions were asked: (1) Do you drink alcohol? (2) On how many days do you drink alcohol? (3) How many drinks do you consume on such a drinking day? The sum of the number of drinks on weekdays and weekend days yielded the total score of drinks per week. Heavy alcohol use was dichotomised so that ≥14 drinks/week for women and ≥21 for men defined heavy drinking. This distinction corresponds with the recommendations for alcohol use by the British Medical Association.35
Two aspects of SES were measured: self-reported completed educational level in five categories (primary, low-level vocational, secondary, junior college/high-level vocational, college/university) and household equivalent income (net income in euros). Income information was obtained from linkage at the level of individual respondents to the national tax registry. Income was categorised into quintiles.
Ethnicity was derived from linkage to the national continuous population registry and was measured in three categories: ethnic Dutch origin, Western minorities and non-Western minorities. Other individual-level variables included age (in single years), gender and household composition (couple without children, couple with children, single without children, single with children, other). These variables were measured in the POLS survey.
Neighbourhood-level variables
The definition of neighbourhood deprivation was adopted from the selection of neighbourhoods made by the Dutch national government in 2007. It determined which neighbourhoods required interventions to reduce deprivation. All Dutch neighbourhoods were assigned a sum score for deprivation based on 18 items. Items include neighbourhood problems like vandalism, social nuisance, poor housing, lack of safety and low neighbourhood SES. The full list of deprivation indicators is displayed in appendix A.36 Forty areas, containing 83 four-digit postcode neighbourhoods with the highest scores were defined as most deprived.19 The remaining 1639 neighbourhoods are ‘other’ neighbourhoods (the rest of the Netherlands).
Nine WoON 2006 items were selected to measure social cohesion (see appendix B),37 ,38 and mean scores for individual respondents were calculated to form a construct score. Cronbach's α was 0.83, indicating high internal reliability. The mean of all individual construct scores within a postcode area formed the aggregated neighbourhood-level scores. Preliminary analysis in which social cohesion was modelled with splines, showed that the association between chronic heavy drinking and social cohesion was not linear (χ²=10.8, p=0.005). Social cohesion was therefore analysed in quintiles.
To determine the extent to which a neighbourhood has a cultural norm of abstinence, the proportion of Muslims in the neighbourhood was measured. The variable was derived from response to a WoON 2006 item, asking whether the respondent adhered to a religion, and to which one. The percentage of neighbourhood residents reporting to be Muslims formed the neighbourhood score, which was linked to POLS data. We consider this measurement a neighbourhood cultural factor, additional to the concepts covered in social cohesion.
Population density was measured in addresses per square kilometres (km2) in the municipality. Data were obtained from a geographical database of Statistics Netherlands (CBS).
Statistical analyses
Logistic multilevel regression analyses were applied to investigate the relationship of both chronic and episodic heavy alcohol use with neighbourhood deprivation. Multilevel models were created by adding a random intercept at the neighbourhood level. In a series of nested models, we added variables to control for, respectively, basic demographic characteristics (age, gender and household composition), measures of the respondent's SES (education and income) and neighbourhood variables population density, social cohesion and percentage of Muslims. All models furthermore controlled for the survey year to correct for the secular trend in heavy alcohol use. All analyses were performed in Stata V.11.0.
Additionally, relationships were estimated separately for subgroups of gender, age (under and over 50), educational level (higher and lower) and income (higher and lower). These estimations were derived by adding interaction terms (eg, gender × deprivation) to regression models, fitted for the entire population. Lower income and lower education were defined, respectively, as the second and third lowest groups as shown in table 1.
Nationally representative estimates of the prevalence of heavy drinking were obtained by weighting for age, gender, marital status, household size, urbanity and county. When these weighs were included in regression analyses, results were virtually identical to the regression-based results presented in this paper.
Results
Table 1 shows differences in characteristics of residents of deprived districts and the rest of the Netherlands. In deprived areas, residents more often belong to minority groups and have both lower educational level and lower incomes than in other Dutch neighbourhoods. Furthermore, deprived neighbourhoods have higher population density, weaker social cohesion and a proportion of residents with an Islamic background six times larger than other Dutch neighbourhoods.
Figure 1 presents alcohol consumed in deprived neighbourhoods and other neighbourhoods in the Netherlands. In deprived neighbourhoods, 41% of residents do not drink alcohol, or only drink up to three drinks per week. Alcohol consumption of 4–22 drinks/week is less frequent in deprived neighbourhoods (53.2% vs 61.5%). The proportion of residents consuming more than 22 drinks/week is approximately equal for deprived neighbourhoods and the rest of the Netherlands (6.1% and 7.5%, respectively).
Table 1 shows the associations of covariates with chronic and episodic heavy drinking. Heavy drinking is least prevalent among non-Western minority groups. Chronic heavy drinking is barely related to education, and positively associated with income. Binge drinking is inversely associated with education, and has only a weak relationship with income. Furthermore, there is no linear association between social cohesion and chronic heavy drinking, while binge drinking occurs more in the most cohesive neighbourhoods.
Table 2 shows the prevalence of chronic and episodic heavy drinking for various subgroups of the studied population. In the total population, as well as in all specified gender, age, education and income groups, less heavy drinking occurs in deprived neighbourhoods. The difference is particularly large for the prevalence of chronic heavy drinking among residents over 50. This is caused by a very high chronic heavy drinking prevalence (18%) in this age group in other Dutch neighbourhoods.
Table 3 shows the results of the regression analyses for the associations between deprivation and chronic and episodic heavy drinking, respectively. Residents in deprived neighbourhoods are significantly less likely to be heavy drinkers than in the rest of the Netherlands (OR chronic=0.58 (0.47 to 0.72), OR episodic=0.57 (0.45 to 0.72)). These strong associations are largely reduced when adjusted for ethnicity (OR chronic=0.72 (0.57 to 0.90), OR episodic=0.77 (0.60 to 0.98)). Controlling for educational level and income has only very limited impact on the associations found, as ORs barely changed when both variables were added to the regression models.
Social cohesion slightly weakens the association between deprivation and binge drinking and chronic heavy use, respectively (table 3). Adding percentage of Muslims to the model further weakens associations to a small extent. The explanatory role of percentage of Muslims is slightly larger than that of social cohesion. The mediating effect of percentage of Muslims and social cohesion was the same when we added percentage Muslims first and then social cohesion.
Table 3 also contains results for a population from which all ethnic minorities were excluded. By doing this, we assessed whether associations for the ethnic Dutch-origin residents are comparable with results for the total population, which are corrected for ethnicity. This table shows that the results are indeed similar.
Table 4 shows the results for gender and age groups. It is notable that associations for women and respondents over 50 are stronger than for men and younger adults, although not significantly. For chronic heavy drinking, for example, the most adjusted OR for residents over 50 (OR=0.65 (0.45 to 0.94)) is comparable with the least adjusted OR for residents under 50 (OR=0.64 (0.48 to 0.84)). The predictors investigated seem to have a relatively low explanatory power for females and older adults, as ORs are modified less by adding ethnicity, SES, and neighbourhood factors. According to tests for interaction, the association between neighbourhood deprivation and heavy alcohol use was not significantly different for low and high education and income groups (results not shown).
Discussion
Key findings
The results of this study confirm the inverse association between neighbourhood deprivation and heavy alcohol use as found by Ruijsbroek et al.16 Despite social and environmental problems in deprived neighbourhoods, residents of these areas tend to be less likely to be heavy drinkers. A large proportion, but not all, of the differences found in heavy alcohol use were explained by the ethnic background of the neighbourhood population. To a small degree, the association was explained by the proportion of neighbourhood residents with an Islamic background. Social cohesion and residents’ individual SES barely contributed to the relationship between deprivation and heavy drinking.
Evaluation of potential data problems
The prevalence of heavy alcohol use is probably underestimated. Under-reporting of alcohol use may result from giving socially desirable answers. The literature shows that underreporting is more severe among heavy drinkers.39 However, no studies were found assessing the association between living in a deprived neighbourhood or having a low SES and underreporting alcohol use. Therefore, we do not expect our results to be biased by underreporting.
Total response rates to the POLS surveys are between 64% and 67%. Selective response may cause bias, depending on the extent to which the response rate is associated with both alcohol use and neighbourhood deprivation. Residents of deprived neighbourhoods are less likely to participate in research.40 Furthermore, late response to surveys is associated with higher alcohol use.41 However, we have corrected for several characteristics that could be associated with non-response, like ethnicity, age, education and social cohesion. Substantial bias of results could only occur if neighbourhood deprivation is independently related to non-response.
It is notable that the SES hardly contributes to the association between neighbourhood deprivation and heavy alcohol use. This is due to education and income being only weakly related to alcohol use and associations of education are opposite to those of income (table 1). Using data on other SES indicators, like occupational class or wealth, in addition to education and income, might have shown different results. We did have information on the wealth (sum of all monetary assets minus debts) of part of the population (survey years 2007–2009). When we controlled for wealth (in addition to education and income) we found that ORs for both chronic and episodic heavy drinking for the total population remained identical (0.76 and 0.77, respectively). There were only very minor changes in the CIs (0.61 to 0.96 and 0.60 to 0.98, respectively). Wealth, therefore, had no additional value in predicting heavy drinking.
Explaining results
The prevalence of heavy drinking is low in deprived areas, due to the larger proportion of abstainers and moderate drinkers in these neighbourhoods. This is in accordance with Ruijsbroek et al16 and other studies that show abstaining and moderate drinking is more prevalent in deprived than in non-deprived neighbourhoods.4 ,6 ,7 ,14 However, such differences do not seem to exist for very high levels of alcohol use, that is, above 22 glasses per week. A previous study has shown that the prevalence of very high alcohol intake was associated with the occurrence of neighbourhood problems.42 People with a higher alcohol intake and a problematic type of alcohol use might selectively migrate into more deprived neighbourhoods. This could explain why differences between deprived and other areas are not observed for very high alcohol intake.
The ethnic composition of the neighbourhood, with large minority groups in deprived areas, explained about half of the association found between area deprivation and heavy alcohol use. Minority groups in the Netherlands are largely of Turkish and Moroccan origins,19 and therefore descending from a culture of abstaining or very moderate alcohol use.20 The contextual influence of the presence of neighbourhood residents with an Islamic background adds another explanatory value to the ethnic composition, but is a much weaker predictor.
The association between neighbourhood deprivation and heavy drinking differs between younger and older adults. Associations remain stronger for older adults when corrected for individual and contextual confounders. Residents of Dutch deprived neighbourhoods report being in poor health more often.16 Elderly people are more likely to suffer from health problems,43 which may have induced part of the older residents to cut back drinking to improve health.44 Therefore, the unexplained difference in alcohol use might be partly attributable to the relatively poor health status of residents in deprived neighbourhoods.
Neighbourhood social cohesion does not seem to be strongly associated with individual heavy alcohol use. Our results suggest that the social environment does not play a major role in drinking differences between deprived and affluent neighbourhoods. In contrast, Ahern et al45 found that neighbourhood norms about getting drunk regularly were related to individual drinking patterns, independent of friend and family norms and individual norms. Including a more direct measurement of neighbourhood social norms for alcohol use might therefore provide useful information for further research on neighbourhood drinking differences.
Conclusion
This study shows that the prevalence of heavy alcohol consumption is lower in deprived neighbourhoods than in the rest of the Netherlands. We found that the roles of SES and the contextual factors social cohesion and percentage of Muslims in the neighbourhood are limited. Discrepancies in the prevalence of heavy alcohol use between deprived and more affluent neighbourhoods were mainly determined by differences in ethnicity. We cannot conclude that heavy drinking is not a problem in deprived neighbourhoods. However, according to our results, policy and interventions that target reducing heavy alcohol use do not need to prioritise deprived neighbourhoods.
What is already known on this subject
-
Earlier research from the Netherlands has shown that the prevalence of heavy drinking is lower in Dutch deprived neighbourhoods than in other areas.
-
Studies from other countries show conflicting results, and more clarity on possible explanations for found differences is therefore needed.
What this study adds
-
Residents of deprived neighbourhoods were up to 43% less likely to be involved in either chronic or episodic heavy drinking.
-
The lower prevalence of heavy drinking in deprived areas is largely explained by the ethnicity of residents in these neighbourhoods and for a small part by the percentage of residents with an Islamic background.
-
Differences were not caused by the socio-economic status of residents or the social cohesion of the neighbourhood.
Acknowledgments
We gratefully acknowledge Statistics Netherlands for preparing microlevel data of the POLS survey, and providing access to these data. We would also like to thank the Data Archiving and Networked Services (DANS) for making the free use of WoON 2006 data possible.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online appendix
Footnotes
-
Contributors MAGK and AEK designed the study. MAGK performed all statistical analyses. MW prepared survey data files and checked the execution of the analyses. MAGK wrote the manuscript. Authors AEK, BJ-G, MD and KS critically reviewed several versions of the manuscript. All authors contributed to and have approved of the final manuscript.
-
Funding This study is funded by a Grant of The Netherlands Organisation for Health Research and Development, which had no further role in the study design, analysis, interpretation, writing or publication.
-
Competing interests None.
-
Provenance and peer review Not commissioned; externally peer reviewed.