Friday, April 07, 2006

A multilevel analysis of race, community disadvantage, and body mass index among adults.

Understanding the factors that contribute to overweight and obesity in the US, and to disparities between black and non-black women in particular, may help us better address the rising epidemic of obesity in the US. This study is the first we know of to examine how racial differences in BMI are explained by both individual SES and community disadvantage. Using multilevel modeling, we first demonstrated that overall variation in BMI among adults in the US is due to both between-community and within-community variation, suggesting that future studies should continue to address how both individual and community factors are related to BMI. However, most of the variation in BMI was due to individual-level variation rather than systematic differences between communities.




Consistent with previous research, we showed that black women had higher BMI than non-black women (Flegal, Carroll, Kuczmarski, & Johnson (1998) and Flegal, Carroll, Ogden, & Johnson (2002)), and that race (black) and low individual SES were each independent risk factors for higher BMI among women ( Lakdawalla & Philipson, 2002; Harrell & Gore, 1998; Sundquist & Johansson, 1998). Individual SES slightly reduced racial differences in BMI among women, with large racial differences in BMI persisting.




Community socioeconomic disadvantage and community income inequality (Gini) were each independently associated with BMI among women, net of demographic and individual SES measures, whereas community black concentration (% black) was not. When considered simultaneously, individual SES and community socioeconomic disadvantage each explained some of the association between race and BMI among women. However, strong race effects remained.





Among women, physical activity and current smoking were independent predictors of (low) BMI while indices of chronic financial stress, stressful life events, and social support were not. However, physical activity and current smoking only slightly altered the associations between race and BMI, individual SES and BMI, and community SDI and BMI. Although we hypothesized that race, individual SES, and community SDI would be associated with BMI partly due to smoking, physical activity, stress, and social support, we found little evidence of this. Similarly, Rucker (2000) controlled for multiple measures of individual stress, social support, and psychological well-being and found that strong racial differentials in BMI among women ages 24–60 persisted.




Future studies should examine the mediating role of alternate or improved measures of health behaviors, stress, and social support. In particular, measures of physical activity should examine nuanced distinctions between sedentary, light, moderate, and high physical activities in work and leisure activities. Similarly, stress could be examined in several social, psychological, environmental, or physical domains. In addition, there are clearly other unmeasured individual-level variables that should be included, such as those regarding nutrition and diet. Moreover, studies should include potential community-level mediators such as access to supermarkets, community crime levels, community access to parks and recreation, and proxy measures of community norms, such as percentage of the community that is overweight.




Regarding men, our results showed that black and non-black men had similar average BMI, which is consistent with recent data from NHANES (1999–2000) that showed similar prevalence of obesity among black, white, and Hispanic men (Flegal et al., 2002). But unlike Lakdawalla and Philipson's (2002) analysis of NHIS data, we did not find that black men had significantly higher BMI than non-black men. This may be a function of smaller sample size in the ACL, since our individual-level models (Models 1 and 2) present race coefficients that are practically identical to those reported by Lakdawalla and Philipson. In contrast to the results for women, none of the individual SES measures or the community disadvantage measures was associated with BMI for men. But consistent with results for women, physical activity and current smoking were significantly associated with BMI for men, whereas chronic financial stress, stressful life events, and social support were not.




Although there were no racial differences in BMI among men, these findings should not distract us from recognizing that overweight and obesity remain problems among men in general. In our sample, black women had the highest mean BMI, followed by black men, non-black men, and non-black women. Although we focused on explaining why there are such large differences in BMI between black women and non-black women, another interesting question is why there are not large racial differences in BMI among men. In the US, black men and white men have very different social and economic contexts, and they vary significantly on most social, economic, behavioral, and health characteristics and experiences. The fact that they do not vary significantly in BMI can be seen as an anomaly. However, the fact that there are clear gender differences in the relation between race and BMI does provide important clues regarding etiology. Indeed, any explanations for variations in BMI must account for the fact that there are significant differences by race and gender and potential complex interactions with individual SES. For example, if community access to healthy food is a mechanism affecting overweight and obesity, why is there no effect of community socioeconomic disadvantage index for men? In fact, we found marginally significant interactions such that black men in more socioeconomically disadvantaged communities have lower BMI, suggesting opposite effects of community environments on black men and women.




Similarly, why is lower individual SES related to higher BMI in women but not in men? We found a marginally significant interaction between race and income among men, such that higher income is associated with higher BMI among black men, a finding consistent with an earlier study showing that SES is strongly inversely related to overweight in women, and slightly positively related to overweight in men (Ross & Mirowsky, 1983). Salmon, Owen, Bauman, Schmitz, and Booth (2000) found that male, less skilled workers may "make up" for their lack of leisure time physical activity through greater occupational strenuousness. Further research might therefore examine whether occupational strenuousness helps explain the smaller racial differentials in BMI among men.




Although a strength of our study was that we measured black racial concentration (% black) at the community level rather than at the larger county or MSA level as many previous studies on health outcomes have done, it may be that social norms related to acceptability of overweight among black people may be produced and maintained at larger levels of aggregation, such as cities, counties, or larger regions. Alternately, racial differences in social norms regarding weight may be transmitted and maintained not locally or regionally, but through national media and through family and social norms immune to regional boundaries, resulting in race-specific but not region-specific social norms.




There are, of course, limitations to our definition of community as census tracts. If community characteristics do matter to BMI, then crude measurement of the boundaries of community using census-defined areas may underestimate effects of community context. Moreover, it is likely that BMI is affected by numerous social, psychological, economic, biologic, physical, and environmental phenomena that operate at different levels. This study investigated only individual-level and community-level socioeconomic variables, ignoring other contextual characteristics that vary systematically by family, social networks, workplace, and other levels of socially or physically bounded measures of community or geography.




As this is the first study we know of to simultaneously examine individual-level and community-level explanations of race differences in BMI in a national sample, we chose to start by examining cross-sectional relationships. However, we expect that the causal relations between-individual SES, community disadvantage, and BMI are complex, requiring further longitudinal analyses. For example, examining independent associations between BMI and both individual SES and community disadvantage ignores pathways through which people of different races come to accumulate dissimilar levels of individual SES based on community experiences, and how they become residentially distributed and segregated in the first place. In addition, although we examine BMI as the dependent variable in this study, the causal direction also flows the other way. For example, people who are overweight may experience job discrimination, health problems, and missed workdays, which all may affect income, stress, physical activity, and smoking behavior.




We are also left with questions about BMI trajectories over the life course and how they vary by race and gender. Ferraro et al. (2003) demonstrated that childhood overweight is associated with severe obesity in adulthood. They also found that although women have a higher likelihood of severe obesity, the association between childhood overweight and severe obesity is stronger in men ( Ferraro et al., 2003). Future research should address not only BMI trajectories across the life course by gender and race, but time-varying exposures to multilevel risk factors as well.




Moreover, although we controlled for the non-linear relationship between age and BMI, we did not examine age variations in the relationship between race, individual SES, community disadvantage, and BMI. We expect that these relationships do vary by age, and recommend that future research use longitudinal data to examine age variations both cross-sectionally and over time.



There are also limitations to looking at BMI itself as the dependent variable. We did not present potentially meaningful (though empirically arbitrary) cut-offs representing underweight, normal weight, overweight, obesity, and severe obesity. In sensitivity analyses (not presented), removing underweight respondents at the bottom of the BMI scale did not significantly change our results. Though we found no racial differences in BMI for men, there may be racial differences in severe obesity among men (Ferraro et al., 2003).




In conclusion, one of the main contributions of this study is the finding that both individual and community socioeconomic context help explain black women's higher BMI, although racial differences persist. This suggests that reducing racial disparities in BMI among women may require interventions with both individuals and communities. A second main contribution is our finding that race, individual SES, community socioeconomic disadvantage, and individual health behaviors are each independent predictors of BMI among women in the US, net of each other. Moreover, significant between-individual and between-community variance in BMI remained unexplained even after considering these factors, though most unexplained variance is due to individual-level rather than community-level factors. Our conclusion that multiple measured and unmeasured individual-level and community-level factors are associated with BMI among women provides an important yet somewhat daunting message. If the determinants of obesity in women are multiple and multilevel, so must be the efforts to address them.

0 Comments:

Post a Comment

<< Home