Zilber School of Public Health, University of Wisconsin-Milwaukee, 1240 N 10th St, Milwaukee, WI, 53205, USA.
Department of Epidemiology & Biostatistics, Dornsife School of Public Health, Drexel University, Philadelphia, PA, USA.
Int J Obes (Lond). 2019 Aug;43(8):1601-1610. doi: 10.1038/s41366-019-0322-3. Epub 2019 Jan 22.
Current knowledge regarding the relationship between segregation and body weight is derived mainly from cross-sectional data. Longitudinal studies are needed to provide stronger causal inference.
We use longitudinal data from the Multi-Ethnic Study of Atherosclerosis and apply an econometric fixed-effect strategy, which accounts for all time-invariant confounders, and compare results to conventional cross-sectional analyses. We examine the relationship between neighborhood-level racial/ethnic segregation, neighborhood poverty, and body mass index (BMI) separately for blacks, Hispanics, and whites. Segregation*gender interactions are included in all models. Neighborhood segregation was operationalized by the local G statistic, which assesses the extent to which a neighborhood's racial/ethnic composition is under (G statistic < 0) or over (G statistic > 0) represented, given the composition in the broader (e.g., county) area. For black, Hispanic, and white stratified models, the G statistic reflects the level of black, Hispanic, and white segregation, respectively. The G statistic was scaled such that a unit change represents a 1.96 difference in the score.
Cross-sectional models indicated higher segregation to be negatively associated with BMI for white females and positively associated for Hispanic females. No association was found for black females or males in general. In contrast, fixed-effect models adjusting for neighborhood poverty, higher segregation was positively associated with BMI for black females (coeff = 0.25 kg/m; 95% CI = [0.03, 0.46]; p-value = 0.03) but negatively associated for Hispanic females (coeff = -0.17 kg/m; 95% CI = [-0.33, -0.01]; p-value = 0.04) and Hispanic males (coeff = -0.20; 95% CI = [-0.39, -0.01]; p-value = 0.04). Further controls for socioeconomic factors fully explained the associations for Hispanics but not for black females.
Fixed-effect results suggest that segregation's impacts might not be universally harmful, with possible null or beneficial impacts, depending on race/ethnicity. The persistent associations after accounting for neighborhood poverty indicate that the segregation-BMI link may operate through different pathways other than neighborhood poverty.
目前关于隔离与体重之间关系的知识主要来自于横断面数据。需要进行纵向研究以提供更强的因果推理。
我们使用动脉粥样硬化多民族研究的纵向数据，并应用计量经济学固定效应策略，该策略考虑了所有时间不变的混杂因素，并将结果与传统的横断面分析进行比较。我们分别为黑人、西班牙裔和白人检查了社区层面的种族/族裔隔离、社区贫困与体重指数（BMI）之间的关系。所有模型均包含隔离*性别交互项。社区隔离由局部 G 统计量来操作，该统计量评估了一个社区的种族/族裔组成在多大程度上低于（G 统计量<0）或高于（G 统计量>0）其在更广泛（例如县）地区的组成。对于黑人、西班牙裔和白人分层模型，G 统计量分别反映了黑人、西班牙裔和白人的隔离程度。G 统计量进行了缩放，以便单位变化表示得分相差 1.96。
横断面模型表明，较高的隔离度与白人女性的 BMI 呈负相关，与西班牙裔女性的 BMI 呈正相关。一般来说，黑人和男性黑人没有关联。相比之下，调整了社区贫困的固定效应模型表明，较高的隔离度与黑人女性的 BMI 呈正相关（系数=0.25kg/m；95%CI=[0.03, 0.46]；p 值=0.03），与西班牙裔女性的 BMI 呈负相关（系数=-0.17kg/m；95%CI=[-0.33, -0.01]；p 值=0.04），与西班牙裔男性的 BMI 呈负相关（系数=-0.20；95%CI=[-0.39, -0.01]；p 值=0.04）。进一步控制社会经济因素完全解释了西班牙裔的关联，但不能完全解释黑人女性的关联。
固定效应结果表明，隔离的影响可能不是普遍有害的，而是可能具有零影响或有益影响，具体取决于种族/族裔。在考虑到社区贫困之后，仍存在关联表明，隔离与 BMI 之间的联系可能通过与社区贫困不同的途径发挥作用。
