Eibner Christine, Sturm Roland
RAND Corporation, Arlington, VA 22202, USA.
Soc Sci Med. 2006 Jan;62(2):348-59. doi: 10.1016/j.socscimed.2005.06.017. Epub 2005 Jul 21.
In this analysis we create census-tract level indices of area deprivation for the US that parallel similar indices developed in Britain, and we determine whether these indices are related to physical and mental health outcomes. Data for the indices come from the 2000 Census Summary File Tapes and the 2001 Zip Code Business Patterns Files. These indices are then linked by census tract to cross-sectional data from the HealthCare for Communities (HCC) study, and equations are estimated using multi-level models with census-tract random effects. We find that area-level deprivation predicts poor mental and physical health outcomes, but different components of deprivation have different effects. When we measure deprivation using three factor scores that emerged from our analysis (rather than combining all measures of deprivation into a single index), we find that access to services has a more pronounced association with physical health, whereas racial composition and local language barriers are more strongly correlated with mental health. These findings suggest that grouping all variables into a single index may mask important heterogeneity in the ways in which area characteristics affect health outcomes.
在本分析中，我们创建了美国人口普查区层面的地区贫困指数，该指数与英国开发的类似指数相似，并且我们确定这些指数是否与身心健康结果相关。指数数据来自2000年人口普查摘要文件磁带和2001年邮政编码商业模式文件。然后，这些指数通过人口普查区与来自社区医疗保健（HCC）研究的横截面数据相链接，并使用具有人口普查区随机效应的多层次模型估计方程。我们发现，地区层面的贫困预示着不良的身心健康结果，但贫困的不同组成部分具有不同的影响。当我们使用分析中得出的三个因素得分来衡量贫困时（而不是将所有贫困衡量指标合并为一个单一指数），我们发现获得服务与身体健康的关联更为显著，而种族构成和当地语言障碍与心理健康的相关性更强。这些发现表明，将所有变量归为一个单一指数可能会掩盖地区特征影响健康结果方式中的重要异质性。
