Assessment of economic status in trauma registries: A new algorithm for generating population-specific clustering-based models of economic status for time-constrained low-resource settings
Assessment of economic status in trauma registries: A new algorithm for generating population-specific clustering-based models of economic status for time-constrained low-resource settings
复制标题
DOI：
10.1016/j.ijmedinf.2016.05.004
复制
发表时间：
2016-10-01
影响因子：
4.9
通讯作者：
Juillard, Catherine
中科院分区：
文献类型：
--
作者：
Eyler, Lauren;Hubbard, Alan;Juillard, Catherine
目标：低收入和中等收入国家以及世界上的穷人在全球伤害负担中所占的份额不成比例。关于伤害差异的数据对于伤害预防和创伤系统加强针对弱势群体的干预措施至关重要，但在中低收入国家却很有限。我们的目标是通过制定一个标准化的方法来评估资源有限的国家创伤登记处的经济状况，从而促进伤害差异研究，这些国家的创伤登记处无法评估收入、支出和财富指数等复杂指标。为了满足这一需求，我们开发了一种基于聚类分析的算法，利用具有全国代表性的人口与健康调查（DHS），生成简单的特定人群经济状况指标家庭资产数据。对于有限数量的变量g，我们的算法使用g资产变量的所有组合对种群执行加权k-中心点聚类，并选择使平均轮廓宽度（ASW）最大化的变量组合和聚类数量。在包含随机分布变量和由相关分类变量定义的“真实”人口聚类的模拟数据集中，该算法选择正确的变量组合和适当的聚类数，除非变量相关性非常弱。当使用2011年爱沙尼亚国土安全部数据，我们的算法确定了20个经济集群与ASW 0.80，表明定义明确的人口clusters.Conclusions：这种经济模型评估健康差距将被用于新的爱沙尼亚六医院集中创伤登记。通过描述我们的标准化方法和算法生成经济聚类模型，我们的目标是促进在资源有限的国家其他创伤登记处的健康差距的测量。(C)2016爱思唯尔爱尔兰有限公司版权所有。
Objectives: Low and middle-income countries (LMICs) and the world's poor bear a disproportionate share of the global burden of injury. Data regarding disparities in injury are vital to inform injury prevention and trauma systems strengthening interventions targeted towards vulnerable populations, but are limited in LMICs. We aim to facilitate injury disparities research by generating a standardized methodology for assessing economic status in resource-limited country trauma registries where complex metrics such as income, expenditures, and wealth index are infeasible to assess.Methods: To address this need, we developed a cluster analysis-based algorithm for generating simple population-specific metrics of economic status using nationally representative Demographic and Health Surveys (DHS) household assets data. For a limited number of variables, g, our algorithm performs weighted k-medoids clustering of the population using all combinations of g asset variables and selects the combination of variables and number of clusters that maximize average silhouette width (ASW).Results: In simulated datasets containing both randomly distributed variables and "true" population clusters defined by correlated categorical variables, the algorithm selected the correct variable combination and appropriate cluster numbers unless variable correlation was very weak. When used with 2011 Cameroonian DHS data, our algorithm identified twenty economic clusters with ASW 0.80, indicating well-defined population clusters.Conclusions: This economic model for assessing health disparities will be used in the new Cameroonian six-hospital centralized trauma registry. By describing our standardized methodology and algorithm for generating economic clustering models, we aim to facilitate measurement of health disparities in other trauma registries in resource-limited countries. (C) 2016 Elsevier Ireland Ltd. All rights reserved.