School of Epidemiology and Public Health, University of Ottawa, Ottawa, Ontario, Canada.
Department of Microbiology and Immunology, Kilimanjaro Christian Medical University College, Moshi, Tanzania.
J Epidemiol Community Health. 2017 Nov;71(11):1046-1051. doi: 10.1136/jech-2017-209119. Epub 2017 Aug 19.
Principal component analysis (PCA) is frequently adopted for creating socioeconomic proxies in order to investigate the independent effects of wealth on disease status. The guidelines and methods for the creation of these proxies are well described and validated. The Demographic and Health Survey, World Health Survey and the Living Standards Measurement Survey are examples of large data sets that use PCA to create wealth indices particularly in low and middle-income countries (LMIC), where quantifying wealth-disease associations is problematic due to the unavailability of reliable income and expenditure data. However, the application of this method to smaller survey data sets, especially in rural LMIC settings, is less rigorously studied.In this paper, we aimed to highlight some of these issues by investigating the association of derived wealth indices using PCA on risk of vector-borne disease infection in Tanzania focusing on malaria and key arboviruses (ie, dengue and chikungunya). We demonstrated that indices consisting of subsets of socioeconomic indicators provided the least methodologically flawed representations of household wealth compared with an index that combined all socioeconomic variables. These results suggest that the choice of the socioeconomic indicators included in a wealth proxy can influence the relative position of households in the overall wealth hierarchy, and subsequently the strength of disease associations. This can, therefore, influence future resource planning activities and should be considered among investigators who use a PCA-derived wealth index based on community-level survey data to influence programme or policy decisions in rural LMIC settings.
主成分分析（PCA）常用于创建社会经济代理变量，以研究财富对疾病状况的独立影响。创建这些代理变量的指南和方法已经得到了很好的描述和验证。人口与健康调查、世界卫生调查和生活水平衡量调查是使用 PCA 创建财富指数的大型数据集的示例，特别是在中低收入国家（LMIC），由于缺乏可靠的收入和支出数据，量化财富与疾病之间的关联存在问题。然而，这种方法在较小的调查数据集上的应用，特别是在农村 LMIC 环境中，研究得较少。在本文中，我们旨在通过研究坦桑尼亚基于 PCA 的衍生财富指数与蚊媒疾病感染风险之间的关联来强调其中的一些问题，重点关注疟疾和主要虫媒病毒（即登革热和基孔肯雅热）。我们表明，与结合所有社会经济变量的指数相比，由社会经济指标子集组成的指数对家庭财富的代表性最差。这些结果表明，在财富代理中包含的社会经济指标的选择会影响家庭在整体财富层次结构中的相对位置，进而影响疾病关联的强度。因此，这可能会影响未来的资源规划活动，并且应该在使用基于社区层面调查数据的 PCA 衍生财富指数来影响农村 LMIC 环境中的方案或政策决策的研究人员中进行考虑。
