Vable Anusha M, Gilsanz Paola, Nguyen Thu T, Kawachi Ichiro, Glymour M Maria
Center for Population Health Sciences, Department of Medicine, Stanford University, Palo Alto, California, United States of America.
Center for Primary Care and Outcomes Research, Department of Health Research and Policy, Stanford University, Palo Alto, California, United States of America.
PLoS One. 2017 Oct 13;12(10):e0185898. doi: 10.1371/journal.pone.0185898. eCollection 2017.
Childhood socioeconomic status (cSES) is a powerful predictor of adult health, but its operationalization and measurement varies across studies. Using Health and Retirement Study data (HRS, which is nationally representative of community-residing United States adults aged 50+ years), we specified theoretically-motivated cSES measures, evaluated their reliability and validity, and compared their performance to other cSES indices. HRS respondent data (N = 31,169, interviewed 1992-2010) were used to construct a cSES index reflecting childhood social capital (cSC), childhood financial capital (cFC), and childhood human capital (cHC), using retrospective reports from when the respondent was <16 years (at least 34 years prior). We assessed internal consistency reliability (Cronbach's alpha) for the scales (cSC and cFC), and construct validity, and predictive validity for all measures. Validity was assessed with hypothesized correlates of cSES (educational attainment, measured adult height, self-reported childhood health, childhood learning problems, childhood drug and alcohol problems). We then compared the performance of our validated measures with other indices used in HRS in predicting self-rated health and number of depressive symptoms, measured in 2010. Internal consistency reliability was acceptable (cSC = 0.63, cFC = 0.61). Most measures were associated with hypothesized correlates (for example, the association between educational attainment and cSC was 0.01, p < 0.0001), with the exception that measured height was not associated with cFC (p = 0.19) and childhood drug and alcohol problems (p = 0.41), and childhood learning problems (p = 0.12) were not associated with cHC. Our measures explained slightly more variability in self-rated health (adjusted R2 = 0.07 vs. <0.06) and number of depressive symptoms (adjusted R2 > 0.05 vs. < 0.04) than alternative indices. Our cSES measures use latent variable models to handle item-missingness, thereby increasing the sample size available for analysis compared to complete case approaches (N = 15,345 vs. 8,248). Adopting this type of theoretically motivated operationalization of cSES may strengthen the quality of research on the effects of cSES on health outcomes.
儿童社会经济地位（cSES）是成人健康状况的有力预测指标，但其操作化和测量方法在不同研究中存在差异。利用健康与退休研究数据（HRS，该数据具有全国代表性，涵盖年龄在50岁及以上的美国社区居民），我们制定了基于理论的cSES测量方法，评估了其信度和效度，并将其表现与其他cSES指数进行了比较。HRS受访者数据（N = 31169，于1992年至2010年进行访谈）被用于构建一个cSES指数，该指数反映儿童社会资本（cSC）、儿童金融资本（cFC）和儿童人力资本（cHC），使用受访者16岁之前（至少34年前）的回顾性报告。我们评估了量表（cSC和cFC）的内部一致性信度（克朗巴哈系数），以及所有测量方法的结构效度和预测效度。效度通过与cSES的假设相关因素（教育程度、测量的成人身高、自我报告的儿童健康状况、儿童学习问题、儿童药物和酒精问题）进行评估。然后，我们将经过验证的测量方法的表现与HRS中用于预测2010年自我评定健康状况和抑郁症状数量的其他指数进行了比较。内部一致性信度是可以接受的（cSC = 0.63，cFC = 0.61）。大多数测量方法与假设的相关因素相关（例如，教育程度与cSC之间的关联为0.01，p < 0.0001），但测量的身高与cFC（p = 0.19）、儿童药物和酒精问题（p = 0.41）以及儿童学习问题（p = 0.12）与cHC不相关。与替代指数相比，我们的测量方法在自我评定健康状况（调整后的R2 = 0.07对<0.06）和抑郁症状数量（调整后的R2 > 0.05对< 0.04）方面解释的变异性略多。我们的cSES测量方法使用潜在变量模型来处理项目缺失问题，因此与完全病例法相比，可用于分析的样本量有所增加（N = 15345对8248）。采用这种基于理论的cSES操作化方法可能会提高关于cSES对健康结果影响的研究质量。
