Individual, household and neighborhood socioeconomic status and mortality: a study of absolute and relative deprivation

Abstract

Our objectives were twofold (1) to examine the effects of neighborhood socioeconomic status (SES) on mortality by gender and age and (2) to assess the impact of individual's household SES relative to one's neighborhood on mortality. Data were taken from the Israel Longitudinal Mortality Study, which linked a 20% sample of the 1983 census to mortality records through 1992. Multilevel modeling was performed on 131,156 men and women aged 45–89 years living in approximately 880 neighborhoods. Overall, 27,334 deaths were reported during the 9.5 year study period. Independent of individual characteristics, mortality risks increased 1–2% ($p<0.05$) per unit increase in area deprivation. Results did not vary by age or gender. Household deprivation relative to that of one's neighborhood and adjusted for absolute SES affected mortality in men only. Specifically, men living in relative disadvantage to their neighbors had lower risks of mortality than those living in concordance with their area (OR 0.86, 95% CI 0.80–0.92).

Introduction

Modern epidemiological inquiry began with the ecological association of disease and community (Lillienfeld & Stolley, 1994). Over the past several decades this science progressed toward the ‘individualization of risk’ (Diez-Roux, 1998). Recent advancements in epidemiology have synthesized these two approaches by addressing morbidity and mortality within and between geographical areas while accounting for individual-level variability (Diez-Roux et al., 2001). Many recent mortality studies have demonstrated consistent associations with area-level socioeconomic factors after adjusting for personal socioeconomic characteristics whereby areas with greater deprivation experience higher mortality rates (Diez-Roux et al., 2001; Barnett, Williams, Moore, & Chen, 2002; Bosma, van de Mheen, Borsboom, & Mackenbach, 2001; Borrell et al., 2002; Singh & Siahpush, 2002; Yen & Kaplan, 1999; Ben-Shlomo, White, & Marmot, 1996; Pickett & Pearl, 2001; Anderson, Sorlie, Backlund, Johnson, & Kaplan, 1997). This relationship has been observed using assorted contextual area-level socioeconomic variables, such as composite scores based on individual, household and neighborhood characteristics (Sloggett & Joshi, 1998; Borrell et al., 2002; Singh & Siahpush, 2002; Yen & Kaplan, 1999; Lynch et al., 1998). Other studies, however, failed to find an area-level socioeconomic status (SES) association with mortality and postulate that sufficient characterization of individual-level variables or methodological approaches may account for the discrepant findings (Malmstrom, Johansson, & Sundquist, 2001; Veugelers, Yip, & Kephart, 2001; Sloggett & Joshi, 1994).

By and large, individual-level social inequalities in health vary by gender and persist into older age (Mustard, Derksen, Berthelot, Wolfson, & Roos, 1997; Grundy & Sloggett, 2003; Manor, Eisenbach, Israeli, & Friedlander, 2000; Elo & Preston, 1996; Koskinen & Martelin, 1994). A number of studies reported larger inequalities for men than women (Elo & Preston, 1996; Koskinen & Martelin, 1994), while others observed similar or larger SES gradients for women (Manor et al., 2000; Manor, Eisenbach, Friedlander, & Kark, 2004). In community-level studies, this relationship is complicated by environmental and psychosocial constructs that may exert differential effects in the different gender and age groups (Picket & Pearl, 2001; Manor et al., 2000; Wiggins et al., 1998, Wiggins et al., 2002; Krieger, 2001). While most studies evaluating community-level influences adjust for the effects of gender and age, few have examined their distinctive role (Malmstrom et al., 2001; Wiggins et al., 1998, Wiggins et al., 2002).

Notwithstanding the deleterious effects of individual and community-level deprivation, it has long been argued that an individual's relative social position within a community influences his or her risk of death (Pearce & Davey-Smith, 2003; Elstad, 1998; Wilkinson, 2002). Psychosocial factors such as stress have been put forth as the main explanatory pathways linking relative deprivation and health (Elstad, 1998; Wilkinson, 2002). Evidence supporting this link includes biological factors (Wilkinson, 2002) and a negative association between income inequality and health that has been shown in a number of studies (Kennedy, Kawachi, & Prothrow-Stith, 1996; Kawachi, Kennedy, Lochner, & Prothrow-Stith, 1997). Others outside of the United States, however, fail to find such a relationship (Pearce & Davey-Smith, 2003; Ross et al., 2000). Furthermore, only limited information is available regarding the influence of discordance between the social position of the individual and that of his or her community above and beyond the influence of the individual's absolute status (Yen & Kaplan, 1999; Veugelers et al., 2001; Stafford & Marmot, 2003). In a recently published study (Stafford & Marmot, 2003) there was no evidence that personal deprivation combined with prosperous neighborhood adversely affects health. It may well be that alternative mechanisms operate by which, for example, relatively deprived persons living in wealthier areas benefit from their neighborhood's infrastructure and material circumstances (Pearce & Davey-Smith, 2003; Macleod & Davey-Smith, 2003).

In the present study we use the 1983 census records linked to mortality data through 1992 to assess area-level health inequalities in Israel. Our objectives were to (1) examine the area-level SES influence on mortality independent of individual-level and household-level characteristics by gender and age groups and (2) assess the impact of the individual's relative household-SES on mortality.