Effects of Chronic Kidney Disease on Cardiovascular Disease and Dementia Among People with Diabetes: Causal Modeling with Machine Learning Approach

项目摘要/摘要 糖尿病一直是给人们带来巨大健康和经济负担的主要公共卫生问题 个人和社会。鉴于联合国在可持续发展目标(SDGs)中 决心到2030年将非传染性疾病的发病率和死亡率降低三分之一， 了解心血管疾病等长期不良健康后果的主要风险因素 在糖尿病患者中进行心血管疾病的治疗势在必行。而慢性肾病(CKD)和抑郁症 与糖尿病和心血管疾病密切相关，这是这些非传染性疾病之间的因果联系 还没有得到充分的证实。这可能是由于(1)不明确的时间性(即不清楚的时间- 疾病发生的顺序)和(2)其复杂的多因素和高维相互作用 潜在的混杂因素，如人口特征、社会经济地位和共病。 本应用程序的总体目标是调查糖尿病与其 并发症包括慢性肾功能不全。我的具体目标如下：目标1(F99阶段)评估原因 糖尿病患者抑郁与心血管疾病的关系在总结了前面的 文献，我将利用纵向数据来研究糖尿病和抑郁症对心血管疾病的联合影响 充分考虑依赖时间的暴露和混杂因素。目标2(K00阶段)检查原因 从糖尿病到慢性肾脏病和心血管疾病死亡的途径。我将开发基于机器学习的预测 糖尿病患者的慢性肾脏病模型，然后利用 得到了因果推理结构内的预测模型。我还将调查CKD在多大程度上 介导了从糖尿病到心血管疾病死亡的途径。 这项研究提供了一个及时的机会，有助于增加关于这些非传染性疾病如何 疾病(即糖尿病、抑郁症、慢性肾脏病和心血管疾病)相互作用。此外，它的应用还包括 因果推理结构中的机器学习将通过目标化的方式来促进精确健康的概念 并设计有效的干预措施，以预防未来的非传染性疾病及其 并发症。

PROJECT SUMMARY/ABSTRACT Diabetes has been the major public health issues imposing substantial health and economic burden on individuals and society. Given the Sustainable Development Goals (SDGs) in which United Nations has resolved to reduce morbidity and mortality from non-communicable diseases by one-third by year 2030, understanding the major risk factors of long-term adverse health outcomes such as cardiovascular disease (CVD) among patients with diabetes are imperative. While chronic kidney disease (CKD) and depression are closely interrelated with both diabetes and CVD, the causal link between these non-communicable diseases have not been sufficiently established. This is possibly due to (1) ill-defined temporality (i.e. unclear time- ordering of disease occurrence) and (2) their complex multifactorial and high-dimensional interaction with potential confounders such as demographic characteristics, socio-economic status, and comorbidities. The overall objective of this application is to investigate the causal relationship between diabetes and its complications including CKD. My specific aims are as follows: Aim 1 (F99 phase) assesses the causal relationship between depression and CVD among people with diabetes. After summarizing the previous literature, I will utilize longitudinal data to examine the joint effect of diabetes and depression on CVD sufficiently considering time-dependent exposure and confounders. Aim 2 (K00 phase) examines the causal pathway from diabetes to CKD, and to CVD mortality. I will develop the machine learning-based prediction model of CKD among people with diabetes, and then estimate the effect of CKD on CVD mortality using the obtained prediction model within causal inference structure. I will also investigate the extent to which CKD mediates the pathway from diabetes to CVD mortality. This study presents a timely opportunity to contribute to growing literature on how these non-communicable diseases (i.e. diabetes, depression, CKD, and CVD) interact with each other. Moreover, applications of machine learning in causal inference structure will contribute to the “precision health” concept by targeting high-risk populations and design effective interventions to prevent future non-communicable diseases and their complications.