| Sign In to gain access to subscriptions and/or personal tools. |
Affluent Neighborhoods Reduce Excess Risk of Tooth Loss among the Poor
1 School of Dentistry, University of Michigan, Ann Arbor, MI, USA; Correspondence: * corresponding author, anne.sanders{at}adelaide.edu.au
The effect of neighborhood on health may vary according to the characteristics of the residents. We tested the hypothesis that, in affluent neighborhoods, low-income adults retain more teeth than their income-equivalent peers in poor neighborhoods. In 2003, the Adelaide Small Area Dental Study collected sociodemographic and tooth retention information from 2860 adults in 60 neighborhoods. Neighborhood socio-economic position was a census-based composite measure. Using multilevel modelling, we fitted a series of two-level random intercept variance component models. Findings revealed significant main effects for individual and neighborhood predictors and a significant interaction between neighborhood disadvantage and low income. In affluent areas, disparities in tooth retention were negligible, but in poor neighborhoods, substantial variation in tooth retention between individuals was found based on their level of income. Low-income adults appeared to benefit from living in affluent areas, while wealthier adults living in poor neighborhoods did not lose their oral health advantage.
Key Words: cross-level effect modification • health inequalities • multi-level analysis • neighborhood • tooth retention
An established body of evidence confirms that the locality in which people live has an influence on their health. Initial explanations attributed the effect to characteristics of the individuals who lived in those localities—that is, to compositional effects. Typically, these explanations argued that poor health in disadvantaged areas arose from the greater propensity of residents for risk behavior. Yet longitudinal studies have shown that behavioral explanations for socio-economic differences in health are overstated (Lantz et al., 1998; van Oort et al., 2005)—a finding replicated in dental research (Sanders et al., 2006a, 2007; Slade et al., 2006). Moreover, these compositional explanations overlook the possibility that the local environment itself may contribute to the health outcome of residents. Effects pertaining to areas, as opposed to individuals, are known as contextual effects. For example, environmental attributes such as water quality can be hazardous, while lower rates of crime may be protective of health. The introduction of multilevel statistical techniques has facilitated empirical testing of the simultaneous effects of compositional- and contextual-level influences and the interactions between these. These studies confirm that place does influence the health status of residents, independently of the characteristics of the residents themselves (Robert, 1999; Pickett and Pearl, 2001; Riva et al., 2007). Multilevel studies in the dental literature are sparse, with some finding an independent effect of place on dental decay (Pattussi et al., 2006a) and tooth loss (Tellez et al., 2006; Turrell et al., 2007), while others have not (Bower et al., 2007). A fundamental concept that is relatively unexplored is the potential for an interaction between compositional- and contextual-level influences. In this study, a cross-level interaction occurs when the relationship between individuals’ level of household income and oral health status varies across levels of neighborhood socio-economic position. Arguably, an affluent neighborhood promotes health for poor individuals through greater availability of health-enhancing resources in those neighborhoods. Conversely, if those resources are costly, then poor people in affluent neighborhoods will be no better off than poor people in disadvantaged areas. One criterion used to detect the presence of effect modification is through statistical evaluation for the presence of interaction between compositional and contextual influences. To explore whether neighborhood contextual characteristics may alter compositional influences on oral health, this study aimed to investigate the presence, direction, and magnitude of a cross-level interaction of neighborhood socio-economic position on tooth retention across strata of individual household income. We hypothesized that tooth retention would be positively associated with an individual’s household income, but that the magnitude of the association would be more pronounced in poor neighborhoods than in affluent neighborhoods.
The Human Research Ethics Committee of the University of Adelaide gave ethical approval for the study.
Design and Sample
Neighborhood Definition
Data Collection
Dependent Variable
Compositional Variables
Contextual Socio-economic Indicator
Analytic Approach In the baseline model (Model 1), we fitted the random intercept term for neighborhoods and fixed effects for sex and of age centered at mean age. Model 2 additionally estimated main effects of fixed compositional socioeconomic predictors. Model 3 additionally estimated main effect of neighborhood tertiles. Finally, in Model 4, we tested the study hypothesis of a cross-level interaction by adding ‘dummy’ variables that cross-classified 3 neighborhood socio-economic tertiles with 4 categories of household income, nominating as reference groups the advantaged neighborhood that had the highest income category, adjusted for sex, age, and education.
From the 2915 participants who returned a completed questionnaire (response rate = 69.4%), analysis was limited to those who reported number of retained teeth (n = 2860). Males contributed 45.6%, and age ranged from 43 to 58 yrs (mean = 50.1). The compositional and contextual socio-economic gradients in crude estimates of tooth retention persisted after adjustment for age, sex, and clustering (Table 1  ).
The variance components model showed statistically significant variation between neighborhoods in numbers of retained teeth (Table 2 ). In the baseline model (Table 2, Model 1), the mean intercept of 24.9 represents the mean number of retained teeth among females of mean age. The intercepts for the 60 neighborhoods varied around this mean intercept, with a variance of 2.89 [standard error (SE), 0.68], indicating statistically significant variation between and among neighborhoods. The intraclass correlation of 6.77% indicates the proportion of variation in retained teeth attributable to neighborhoods. Adjustment for compositional fixed effects of education and household income attenuated the level 2 random variance to 1.27 (SE. 0.38), but the effect of neighborhood remained statistically significant (Table 2, Model 2). In this model, adults with no post-school qualification or vocational training retained fewer teeth than those with a Bachelor’s or higher degree. With the inclusion of neighborhood socio-economic tertiles (Table 2, Model 3), the model estimated that adults with lowest household income had 3.6 fewer teeth than adults with highest income (95% CI = –4.4 to –2.8), and predicted an independent neighborhood effect such that adults in poor neighborhoods retained approximately 2.0 fewer teeth (95% CI = –2.8 to –1.2) than those in affluent neighborhoods, regardless of their age, sex, education, and household income.
There was a statistically significant cross-level interaction between compositional and contextual socio-economic levels (P = 0.005), and predicted estimates revealed a larger income-related difference in tooth retention in poor neighborhoods compared with affluent neighborhoods (Fig. , Table 3). In poor neighborhoods, people in lowest-income households had retained an adjusted mean of 21.7 teeth per person, compared with 26.0 teeth for people in highest-income households—a difference of 4.3 teeth per person that was statistically significant as indicated by a distinct lack of overlap of standard errors (Fig.). Yet in affluent neighborhoods, the adjusted mean numbers of retained teeth were 26.6 and 27.8 for poor and wealthy individuals, respectively—a difference of less than 2 teeth per person that displayed considerable overlap of standard errors. We separately tested for an effect modification of education and neighborhood advantage on retained teeth, but this interaction was non-significant (see Appendix for results).
In this population, residence in affluent neighborhoods attenuated the excess risk of tooth loss associated with low income, supporting the study hypothesis. One manifestation of this effect was that, in affluent neighborhoods, the average number of retained teeth among the poor was indistinguishable, statistically, from that in all but the wealthiest residents. A related observation was that residence in a poor neighborhood compared with an affluent neighborhood was associated with only a modest reduction in number of retained teeth for high-income adults (1.2 teeth per person, on average), while a much larger reduction was seen for low-income adults (4.0 teeth per person, on average). Viewed in this way, the findings illustrate that the magnitude of the contextual neighborhood effect was significantly stronger for people whose household income was low compared with people whose household income was high. This suggests a heightened sensitivity to adverse oral health effects of neighborhoods among the poor, while the affluent were relatively impervious to neighborhood effects. These findings build on earlier evidence that the area effect on tooth loss cannot be accounted for solely by the fact that poorer people tended to live in poor areas (Turrell et al., 2007). Several characteristics of areas may account for these observed effects. One explanation is that the finding is an artifact of residential mobility—for example, that poor people with poor oral health preferentially move to disadvantaged areas to be close to publicly funded dental care. This seems unlikely in Adelaide, where public dental clinics are distributed throughout the metropolitan area for low-income groups. A more plausible interpretation is that wealthy neighborhoods create better opportunities for people to retain teeth through greater availability of private-sector, specialty dental services (e.g., endodontics). Yet low-income Australians find private-sector dental care unaffordable (Harford and Spencer, 2004), so this explanation appears unlikely. A further possibility is that certain features of advantaged neighborhoods buffer the oral health hazard posed by low income. In the United States, nutritious food is more readily accessible in affluent neighborhoods than in poorer ones (Lewis et al., 2005; Jetter and Cassady, 2006). Yet in Australia, the accessibility of fruit and vegetables is not systematically related to area socio-economic characteristics (Winkler et al., 2006). Moreover, there is little evidence that a nutritious diet is protective of tooth retention. An alternative explanation is that area disadvantage undermines social capital, since deprivation generates disorder, mistrust, and social exclusion, all of which are correlated with health deficits (Cattell, 2001)—an explanation with some theoretical and empirical support in the dental literature (Pattussi et al., 2001, 2006b; Watt, 2002). Our finding, that the poor living in poor areas experienced the worst health outcomes, is consistent with previous investigations of all-cause mortality (Yen and Kaplan, 1999) and mental health (Fone et al., 2007). Our results confirmed previous findings that health risk among the affluent was equal regardless of where they lived (Yen and Kaplan, 1999). Yet there are inconsistencies. Some studies found a protective effect of affluent areas among disadvantaged adults (Fone et al., 2007), and others found excess mortality among poor adults in affluent areas (Winkleby et al., 2006). While we found a flatter gradient in affluent areas, others reported a steeper gradient in affluent areas (Veugelers et al., 2001; Roos et al., 2004). The principal strengths of this study are the multilevel analytic approach from a random sample of Adelaide adults and the high sampling fraction for neighborhoods. In our previous analyses (Sanders et al., 2006b, 2007), weighting left estimates virtually unchanged, leading us to conclude that findings are generalizable to Adelaide residents of this age range. Our models were intentionally parsimonious, adjusting only for key covariates. While many other factors (e.g., levels of dental decay and periodontitis) are associated with tooth loss, they represent intermediaries that lie on the causal pathway between exposure and outcome, and their statistical adjustment can bias results (Robins and Greenland, 1992). Evidence that where one lives has a substantive effect on oral health, especially if economically disadvantaged, is important to clinicians concerned about population oral health. The topic of cross-level effect modification has public health significance, since socio-economic diversity may minimize the adverse oral health consequences of concentrated disadvantage. The notion that mixed-income neighborhoods reduce the health and social problems of concentrated poverty underpins policy debate in many countries (Ostendorf et al., 2001). Further research is recommended to identify those characteristics of affluent neighborhoods that are protective against tooth loss, recognizing that such characteristics will be of greatest benefit to individuals with low levels of household income.
The National Health and Medical Research Council (NHMRC) funded this study: Project Grant #250315. Dr. Sanders is supported by a NHMRC Sidney Sax Public Health Fellowship (#399222). Dr. Turrell is supported by a NHMRC Senior Research Fellowship (#390109).
A supplemental appendix to this article is published electronically only at http://jdr.iadrjournals.org/cgi/content/full/87/10/969/DC1. Received for publication November 9, 2007. Revision received May 28, 2008. Accepted for publication July 6, 2008.
Journal of Dental Research, Vol. 87, No. 10,
969-973 (2008)
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
|
|
 |
|
Sign In to gain access to subscriptions and/or personal tools.
TOP
ABSTRACT
INTRODUCTION
$20,799; $20,800 36,399; $36,400 51,999; $52,000+; missing/don’t know). Covariates were sex and age. On average, participants had lived in their current neighborhood for 17.5 yrs, and only 8.7% had lived there fewer than 10 yrs. Since this variable differed by less than one yr between neighborhood tertiles, residential mobility was not included in analyses. 