Introduction

From a public health perspective, it is important to reduce social inequalities in health [1]. Previous research has shown that socioeconomic health inequalities have widened in recent decades [2,3]. In the last 20 years, there has been increasing interest in neighbourhoods’ influence on individual health-risk behaviours such as smoking, excessive alcohol consumption, physical inactivity and poor diet [4]. Numerous studies have shown that residents in deprived neighbourhood have higher rates of mortality and morbidity [5–7] than residents of more affluent neighbourhoods, even after taking into account individual-level characteristics such as sex, age, ethnicity and socioeconomic status (SES). Studies have also shown an association between neighbourhood deprivation and health-risk behaviour [6,8]. Most results are from surveys conducted in the USA, the United Kingdom, Canada and the Netherlands [5–8]. Health-risk behaviour is an important factor that increases the risk of morbidity [2] and can explain some of the socioeconomic inequalities in morbidity and mortality [8]. Individuals who engage in four healthy behaviours (fruit and vegetable intake of at least five servings per day, current non-smoker, moderate alcohol intake (1–14 units per week), and physical active) have a life expectancy that is, on average, 14 years longer than that of individuals who do not engage in any of these healthy behaviours [9].

Focusing on health-risk behaviour can be a part of the solution that reduces health inequalities because health behaviour can mediate the effect of SES on the risk of morbidity and mortality [2]. Health-risk behaviour is modifiable through health promotion and intervention programmes; therefore, it is possible to reduce the prevalence and development of these risk behaviours by, for example, developing policies to decrease tobacco smoking, alcohol consumption etc. Through such programmes, it is possible to reduce the gap in life expectancy over the long term [2].

Residents in deprived neighbourhoods are therefore an important target group in efforts to promote healthy behaviour and improve population health in general. “Deprived neighbourhood” is defined here as a geographically bounded area with a high proportion of adults with low SES, as characterised by indicators such as unemployment, low income, low education and low-paying jobs [10]. Knowledge of health-risk behaviour in socially deprived neighbourhoods can contribute to a deeper understanding of the complex interactions between social context, social determinants and health behaviour and to a greater understanding of the development of social inequalities in health behaviour [11]. Through such knowledge, it is possible to develop more targeted health promotion in the process of reducing social inequalities. Previous researchers have investigated how neighbourhood context affects the health of residents by adopting an overall conceptual model in which individual health outcomes are affected by the social and the physical environment of the neighbourhood [4,12–15]. In a review, Diez Roux and Mair have summarised that the social environment of neighbourhoods can affect residents’ health through factors related to safety/violence, social connections/cohesion, local institutions and norms [4]. In addition, they showed that the physical environment can affect health behaviour through environmental exposures, food and recreational resources, the built environment, aesthetic quality/natural spaces, services and quality of housing [4].

The increased interest in neighbourhood effects on individual health is due to, among other factors, multilevel statistical methods, which allow researchers to include both the individual level and the neighbourhood level in one regression model and thereby separate effects related to residents living in the neighbourhood from those related to the neighbourhoods themselves [4,6,16,17].

Previous reviews have primarily focused on examining associations between neighbourhood deprivation and health in general, and these reviews have investigated self-rated health, diseases and health behaviour [6,7,18]. To our knowledge, no systematic review has compared health-risk behaviour among adults in deprived neighbourhoods with that among adults in non-deprived neighbourhoods. By examining the differences in health-risk behaviour in deprived neighbourhoods compared to non-deprived neighbourhoods, it is possible to support future health promotion interventions in deprived neighbourhoods based on which health-risk behaviours warrant the greatest attention.

The aim of this review was to systematically identify and review recent studies on health-risk behaviour among adults who live in deprived neighbourhoods compared with those in non-deprived neighbourhoods. The following research questions were addressed in the present review: 1) What are the differences in health-risk behaviour (no or low consumption of fruits and vegetables, smoking, binge drinking or high-risk alcohol consumption, and physical inactivity) between adults living in deprived neighbourhoods and those living in non-deprived neighbourhoods based on quantitative observational studies and 2) what kind of operationalisations of neighbourhood deprivation were used in the studies?

Methods

The PRISMA guidelines for systematic reviews were followed as the reporting guidelines for this review [19,20]. There was no protocol for this review.

First, studies were identified by systematically searching electronic databases (PubMed, Embase, Web of Science and Sociological Abstracts) using relevant search terms related to deprived neighbourhoods and health-risk behaviours, Boolean operators, and truncations (see S1 File for a complete list of the search terms used in PubMed). A search strategy was developed and adapted for each database using appropriate subject headings and keywords and was restricted to studies that had been published between 1 January 1996 and 1 July 2014. This period was selected because it was considered to cover the most recent research in health-risk behaviour in deprived neighbourhoods. The search strategies for the Embase, Web of Science, and Sociological Abstracts databases are available upon request (please contact the first author). Second, reference lists in articles for which the full text was assessed were scanned to detect articles that were not found in the database search. The literature search was carried out by the first author (MHA).

Results

The database searches provided a total of 7,909 citations, and three additional citations were identified through the manual reference search. After duplicates were removed, 4,361 citations remained. Among these, 4,291 citations were excluded because they did not meet the criteria following our review of the titles and abstracts. The full text of the remaining 70 citations was examined, and 46 articles were excluded for different reasons (see Fig 1). In total, 22 studies were included in the systematic review.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Flow diagram showing the literature search strategy.

Flow diagram of the study illustrating identification, screening, eligibility and inclusion processes in the systematic review of health-risk behaviour in deprived neighbourhoods compared with non-deprived neighbourhoods (from the PRISMA statement [19]).

https://doi.org/10.1371/journal.pone.0139297.g001

The overall quality of the reviewed studies was low. According to the EPHPP criteria, most of the studies had global scores of either “moderate” (n = 10) or “weak” (n = 12). No studies were scored as “strong” (see S2 File).

All included studies had cross-sectional designs (Table 1). Eleven of the 22 studies used structured interviews [8,22–31], and eight studies used self-administered questionnaires [32–39]. Three studies used both methods [40–42]. One study also used clinic biomedical assessment as a supplement [40]. Studies varied widely both in sample size and in the characteristics of the populations studied. The size of the study populations varied from 655 [32] to 58,282 respondents [34]. Nearly all of the studies focused on both men and women, and the sex distribution across most studies was relatively evenly balanced. Most of the results are from studies in the Netherlands (n = 6) [23,24,26,30,32,42], Australia (n = 5) [25,35,38–40], the USA (n = 3) [27,28,33], and the United Kingdom (n = 3) [8,34,37]. Data were collected between 1988 and 2010.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Characteristics of the reviewed cross-sectional studies.

https://doi.org/10.1371/journal.pone.0139297.t001

The statistical methods of the 22 studies are described in Table 2. Data sources for measuring neighbourhood deprivation consisted of census-defined neighbourhoods (n = 13) [25,27–29,31,33–35,37–41], aggregated self-reported neighbourhood deprivation (n = 2) [23,30] and neighbourhood deprivation based on different public data sources (n = 7) [8,22,24,26,32,36,42]. Most of the studies (n = 18) used population-based respondent selection [8,22–30,32–34,36,37,40–44]; four studies selected their respondents by selecting the neighbourhoods in advance of the surveys [31,35,38,39]. Nineteen studies had multilevel designs [8,22,24–29,32–44], i.e., they included individual-level data nested within the neighbourhood level. Two studies linked individual data with aggregated self-reported neighbourhood characteristics [23,30]. One study used only census data to select the neighbourhoods to survey [31].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Operationalisations of neighbourhood deprivation and statistical methods in the reviewed cross-sectional studies.

https://doi.org/10.1371/journal.pone.0139297.t002

To investigate the association between health-risk behaviour and neighbourhood deprivation, 14 studies used multilevel logistic regression (also called hierarchical modelling) [22–26,29,32–36,38,39,42–44], and the remaining eight studies used binary logistic regression [8,27,28,30,31,37,40,41]. Only hierarchical modelling allows the consideration of data on the individual and contextual levels simultaneously while accounting for the potential dependency of individual observations that share the same characteristics as higher-level variables. Therefore, the results from studies that applied standard regression techniques could be biased. All studies controlled for individual-level confounders; most commonly, these were sex, age, marital status, ethnicity, education, employment status and income.

The operationalisation of neighbourhood deprivation varied widely across studies. Ten studies used different predefined indexes to operationalise neighbourhood deprivation [8,22,25,28,29,34,37–40]. Two studies examined deprived neighbourhoods using SES indicators such as education, occupation, and unemployment, which were aggregates of individual-level variables that had been derived from census and survey data [23,30]. Two studies used a summary score and a composite index of different SES indicators [33,36]. Six studies used a number of different indicators to operationalise neighbourhood deprivation [24,26,27,31,41,42], and two studies used a single indicator [32,35].

Diverse measures were used to assess health-risk behaviour. We broadly categorised the studies according to the following health-risk behaviour outcomes: low fruit and vegetable consumption (n = 4), smoking (n = 16), alcohol consumption (n = 7) and physical inactivity (n = 12). Many studies presented results for multiple outcomes. For more specific definitions of each health-risk behaviour outcome, see Table 3.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Risk estimates of the reviewed studies for health-risk behaviour in deprived neighbourhoods compared with non-deprived neighbourhoods.

https://doi.org/10.1371/journal.pone.0139297.t003

As presented in Table 3, the studies were grouped by type of health-risk behaviour to allow for a more straightforward comparison of the different behaviours. Two out of four studies found a positive association between low consumption of fruit and vegetables and living in deprived neighbourhoods compared with non-deprived neighbourhoods [8,38]. The majority of the studies found a positive association between current smoking and living in deprived neighbourhoods [8,22,25–31,33,36,37,41]. Among the seven studies that examined alcohol consumption, the results were ambiguous. In regard to binge drinking, two studies found a positive association between binge drinking and living in deprived neighbourhoods [28,34]. The East of England Lifestyle Survey found a negative association of exceeding recommended alcohol consumption limits and living in deprived neighbourhoods [8]. There was a clear association between living in deprived neighbourhoods and physical inactivity; three-quarters of the studies found a positive association of physical inactivity in these neighbourhoods [22,28,29,31,36,39,41,43,44].

Discussion

This review, which compared health-risk behaviour among adults in deprived neighbourhoods and those in non-deprived neighbourhoods, found a clear pattern of increased smoking and physical inactivity in deprived neighbourhoods. These results are in line with the conclusions of previous systematic reviews [6,7,18]. Most of the reviewed studies did not specify any causal mechanisms that linked neighbourhood deprivation to health-risk behaviour. It has been stressed that the social and physical environment of neighbourhoods may be important in understanding how neighbourhoods can contribute to health inequalities [4]. A review by Pickett and Pearl noted that neighbourhoods might affect health-risk behaviour directly (i.e., via the effects of simply living in a deprived neighbourhood) or indirectly through mechanisms such as the availability of and access to healthy foods or recreational facilities, normative attitudes towards health behaviour, and social support [6]. The physical and social features of neighbourhoods may, for instance, affect health behaviour through mechanisms involving the experience of stress and the buffering effects of social support and social relations [4,6,28,45]. It appears likely that stress is associated with an increased health-risk behaviour and that access to social support can reduce these behaviours. Stress may lead persons to engage in coping behaviours related to, e.g., unhealthy diet and smoking, and living in a deprived neighbourhood may itself be a source of stress [6,46,47]. Van Lenthe and Mackenback found that neighbourhood stressors mediated neighbourhood and individual socioeconomic inequalities in smoking [30]. Furthermore, previous systematic reviews on how factors of the built environment affect health have concluded that the built environment can significantly influence individuals’ health [48–50]. To understand how the environment can affect health between neighbourhoods, it is relevant to mention the environmental justice framework. This is a conceptual model that hypothesises that environmental exposures are unequally distributed across social classes and that neighbourhoods or residents with low SES are more vulnerable to environmental exposures [48].

No clear differences between deprived and non-deprived neighbourhoods were found in relation to fruit and vegetable consumption or alcohol consumption. With regard to fruit and vegetable consumption, it should be mentioned that there were only four studies on the subject, which may explain the few significant results. Only one study showed that fruit and vegetable consumption is decreased in deprived neighbourhood. Another study found the same association for vegetable consumption but not for fruit consumption. Furthermore, two cross-sectional studies did not show any association. Therefore, these results should be interpreted with caution. There is a need for more research on fruit and vegetable consumption. In relation to alcohol consumption, we did not find any geographical pattern between countries that might explain the equivocal findings. All studies except for one analysed alcohol consumption adjusted for ethnicity or the proportion of Muslim residents; thus, these factors cannot explain the results. In a systematic review on disadvantaged areas and substance use outcomes, Karriker-Jaffe reported equivocal findings for alcohol consumption [18]. Furthermore, other studies showed that abstaining from alcohol and moderate drinking are more prevalent in deprived versus non-deprived neighbourhoods [8,51]. It is important for future research to consider that alcohol consumption is one of the leading risk factors for mortality and morbidity [52,53]. We highly advise that future interventions combat risky alcohol consumption in all populations. Our findings suggest that the situation in regard to alcohol consumption is not worse in deprived neighbourhoods, which could suggest that interventions can be population-based without considering the specific needs of deprived neighbourhoods. However, these suggestions need to be followed up and tested in future research. Additionally, future research should bear in mind that alcohol consumption is complex in terms of the interaction between age and education level. For example, in Denmark, in the younger age groups, high alcohol consumption is most evident among persons with elementary school as their highest education level; by contrast, in the 65 years or older age group, it is most evident among persons with higher education [54].

Neighbourhoods were operationalised using a variety of techniques. The majority of the studies operationalised neighbourhoods using statistical (e.g., census tracts) or administrative spatial units (e.g., city-defined neighbourhoods). Some studies used multiple neighbourhood socioeconomic characteristics to rank neighbourhood deprivation. The VICLANES [35] and FP7 EURO-URHIS2 [32] used only income and unemployment, respectively, as neighbourhood deprivation measures, and others used indexes. Among the studies using different indicators, the measures that were most often used were income, employment and education. It appears that the indicators used were mainly based on the availability of data rather than on conceptual considerations. Most studies were not explicit about why certain indicators were selected to measure neighbourhood deprivation. Moreover, most studies provided little or no information on the validity and reliability of the measures used.

Neighbourhood deprivation is a frequently used term, but it has no singular definition or operationalisation, as shown in this review. Future research should focus on how to define and operationalise neighbourhood deprivation, which will facilitate systematic review and allow for meta-analysis. However, despite an accepted definition, the problem of how neighbourhood deprivation is operationalised will remain because many researchers must rely on data availability. In the future, researchers should choose neighbourhood deprivation indicators that have been validated. None of the reviewed studies stated whether the measures of neighbourhood deprivation that they used had been validated. A search of the different measures of neighbourhood deprivation used in the reviewed studies did only reveal documentation in regard to the validation of the SEIFA Index [55], which was used in the HS [25] and SESAW [38] studies. In the reviewed studies, data on income, employment and education were frequently used, although there is no true consensus in the literature that these are the best measures of neighbourhood deprivation [56]. Researchers should provide their reasons, both practical and theoretical, for choosing specific measures of neighbourhood deprivation.

Furthermore, the reviewed studies used different labels for neighbourhoods, such as “community”, “area”, and “place”. In general, there is no clear distinction between these terms, and the concepts of neighbourhood and community are not precise [57].

In the current review, we included only studies that at least adjusted for one confounder besides sex and age to control for sociodemographic and socioeconomic differences between respondents living in deprived and non-deprived neighbourhoods. We considered ethnicity and educational level as two of the main potential confounders that could influence the results of this review because these factors have been found to constitute important determinants of health behaviours [58].

We used the EPHPP risk of bias tool to assess the quality of the reviewed studies. Most studies were scored with a global rating of either “moderate” or “weak”, partly because many of the studies (n = 8) had response rates below 60% and all had cross-sectional designs. Accordingly, no causal pathway can be interpreted, and only assumptions about associations are possible. Future studies should, when possible, use cohort designs to capture the long-term effects of neighbourhood deprivation on health-risk behaviour to explicitly examine causal processes over time.

We cannot be certain that all of the reviewed studies included the most deprived neighbourhoods because the data used in most of the studies were based on general health surveys merged with census data, and we know that residents of deprived neighbourhoods are less likely to participate in research [59]. Only four of the reviewed studies [31,35,38,39] selected the deprived neighbourhoods in advance. It is important for future research on deprived neighbourhoods to make a greater effort (e.g., using interpreters in interviews and making multiple contact attempts) to obtain higher resident response rates [60].

A limitation of this review is that it includes only peer-reviewed, English-language articles that could be found in the four selected databases. However, we believe that our search ensured robust data collection because we checked references. Another weakness of this study is that the screening process for selecting and excluding studies was performed by one researcher only due to resource constraints, and this could have potentially reduced the objectivity of study inclusion. In addition, it was not possible for us to conduct a meta-analysis because of the different operationalisations of neighbourhood deprivation and different definitions of health-risk behaviour, which also prevented us from performing a specific check of publication bias such as a funnel plot. However, publication bias cannot be excluded.

The strength of this study is that it gives an update of the research in the field of neighbourhood deprivation and health-risk behaviour from economically developed Western countries in the period between 1996 and 2014. Furthermore, the study emphasises the lack of a definition and operationalisation of neighbourhood deprivation. In addition, we used a very broad search string, which made our searches sensitive and ensured the identification of as many relevant studies as possible.

Conclusions

Based on the studies that were included in this review, there is consistent evidence that smoking and physical inactivity are more prevalent among adult residents in deprived neighbourhoods than among residents in non-deprived neighbourhoods. No clear differences between deprived and non-deprived neighbourhoods were found in relation to low fruit and vegetable consumption or alcohol consumption, and the results were equivocal. The reviewed studies used different operationalisations of neighbourhood deprivation.

Future health prevention interventions in deprived neighbourhoods should specifically focus on smoking and physical inactivity. We suggest that alcohol interventions should be population-based without considering the specific needs of deprived neighbourhoods.

Future research in this area should address fruit and vegetable consumption. The lack of a uniform definition and operationalisation of neighbourhood deprivation should also be addressed. An understanding of the mechanisms by which neighbourhood deprivation in general affects health-risk behaviour is still lacking. Future research is therefore needed to identify mediators of the association between neighbourhood deprivation and health-risk behaviour. A better theoretical and empirical understanding of these mechanisms or environmental justice will be important for developing and designing more targeted and prioritised health promotion interventions in the process of reducing social inequalities in health. Furthermore, examinations of which groups are most affected by neighbourhood deprivation would be valuable for developing interventions for the most at-risk residents. Additionally, there is a need for more Scandinavian research in this area, as Scandinavia was underrepresented among the reviewed studies.

Supporting Information

Acknowledgments

We thank Ola Ekholm from the National Institute of Public Health, University of Southern Denmark, for critically reading the article.