Do neighborhoods affect individual mortality? A systematic review and meta-analysis of multilevel studies

Abstract

There has been increasing interest in investigating whether inhabitants in socially or physically deprived neighborhoods have higher mortality when individual socioeconomic status is adjusted for. Results so far appear ambiguous and the objective of this study was to conduct a systematic literature review of previous studies and to quantify the association between area-level socioeconomic status (ALSES) and all-cause mortality in a meta-analysis. Current guidelines for systematic reviews and meta-analyses were followed. Articles were retrieved from Medline, Embase, Social Sciences Citation Index and PsycInfo and individually evaluated by two researchers. Only peer-reviewed multilevel studies from high-income countries, which analyzed the influence of at least one area-level indicator and which controlled for individual SES, were included. The ALSES estimates in each study were first combined into a single estimate using weighted linear regression. In the meta-analysis we calculated combined estimates with random effects to account for heterogeneity between studies. Out of the 40 studies found eligible for the systematic review 18 studies were included in the meta-analysis. The systematic review suggests that there is an association between social cohesion and mortality but found no evidence for a clear association for area-level income inequality or for social capital. Studies including more than one area level suggest that characteristics on different area levels contribute to individual mortality. In the meta-analysis we found significantly higher mortality among inhabitants living in areas with low ALSES. Associations were stronger for men and younger age groups and in studies analyzing geographical units with fewer inhabitants.

Introduction

The number of studies attempting to demonstrate an association between individual mortality and the characteristics of areas in which individuals live has steadily increased since the mid 1990s. A recent review (Diez Roux & Mair, 2010) has summarized the processes through which, according to various authors, neighborhoods can contribute to health inequalities. First, residential segregation by race, ethnicity and socioeconomic position influences the inequality of resource distribution in areas. These processes influence on the one hand the physical environment in terms of environmental exposures, food and recreational resources, the built environment, aesthetic quality, natural spaces, services and quality of housing. For instance, inhabitants with high SES are more likely to attract healthy food stores or sports clubs to their local areas because they can afford these, which means that they have easier access to health promoting goods and activities and that even more resourceful inhabitants are attracted to the area. On the other hand the social environment is affected in terms of safety/violence, social connections/cohesion, local institutions and norms. As an example, people are probably more likely to begin exercising if they repeatedly observe fellow inhabitants exercising or if they are encouraged to do so by neighbors who are active in sports clubs. Thus the hypothesis is that physical and social environments have consequences for inhabitants’ morbidity and mortality by influencing individual health behavior.

In contrast to the traditional distinction between people and places, the current literature recognizes their mutually reinforcing and reciprocal relationships suggesting that neighborhoods affect people, by for example attracting certain socioeconomic groups, and that people affect neighborhoods, as for example by attracting certain workplaces or demanding goods from certain shops (Cummins et al., 2007, Macintyre and Ellaway, 2003, Macintyre et al., 2002). Cummins et al. (2007) also note that place and context vary in time and space because people engage in different contexts over a day, a week or over the life-course and argue for the importance of considering that contextual influences operates on different spatial scales, i.e. that influences from national and regional levels co-exist with influences from the neighborhood level.

Despite advances in this research, it remains unclear whether area conditions have an effect on mortality that is distinct from health determinants linked to individual SES. The present paper investigates this question in two steps: by systematically identifying and reviewing recent studies, and through a meta-analysis estimating the association between ALSES and all-cause mortality.

To isolate neighborhood influences from individual influences we included only multilevel studies that controlled for individual SES. This distinguishes our study from four existing reviews (Ellen et al., 2001, Pickett and Pearl, 2001, Riva et al., 2007, Yen et al., 2009). Three of these (Ellen et al., 2001, Pickett and Pearl, 2001, Yen et al., 2009) were not based exclusively on studies using multilevel modeling, which is essential if health variations associated with area conditions are to be isolated from relationships with individual risk factors (Subramanian et al., 2003, Subramanian and Kawachi, 2004). The small number of multilevel studies available in the past might have prevented such specific reviews from being conducted earlier. Secondly, in contrast to the previous reviews only publications that adjusted for at least one socioeconomic indicator at the individual level were included. Pickett and Pearl (2001) found that studies not adjusting for individual socioeconomic measures were more likely to observe area effects and that those studies including socioeconomic indicators on the individual level found a reduced area effect. It has been extensively demonstrated that individual socioeconomic position is associated with mortality (Mackenbach et al., 2003, Marmot, 2010) and since most area-effect studies use socioeconomic indicators on the area-level, individual-level SES should be taken into account to ensure the validity of ‘independent’ area-level factors.