Predictive Analytics in Hemodialysis: Enabling Precision Care for Patient with ESKD

摘要 终末期肾病(ESKD)是一种复杂的疾病，患者的生命周期各不相同。 预期寿命，25%的人在1年内死亡，41%的人至少存活5年。而提供程序 认识到患者是不同的--也应该是不同的--没有工具可以可靠地 预测个人预期寿命和帮助个体化治疗。相反，提供商是 对于如何最好地管理病人，往往没有明确或不完整的指导方针。为了 为血液透析(HD)患者提供精准护理，迫切需要能够(1) 动态评估医疗决策的预期寿命；以及(2)确定不同的临床 表型，以加强临床监测和护理规划。我们的中心假设是 在患者存活率和疾病轨迹中存在异质性，如果知道的话，可能是 用于提供更个性化和更有效的护理。通过耦合新的机器学习 利用血液透析患者的细粒临床数据预测生存的方法，我们将能够 开发必要的分析工具来支持精准护理。在完成这项工作后 我们将拥有工具来动态评估患者的预期寿命和对 ESKD患者的异质性疾病表型。这些工具将允许提供商 做出明智的治疗决定，并为进一步的精确研究奠定基础 转化为优化的病人护理。

ABSTRACT End stage kidney disease (ESKD) is a complex disease with individuals having variable life- expectancies, with 25% dying within 1 year and 41% surviving at least 5 years. While providers recognize that patients are different – and ought to be differently – there are no tools to reliably forecast individual life expectancy and aid in treatment individualization. Instead, providers are left with often unclear or incomplete guidelines on how best to manage patients. In order to provide precision care for patients on hemodialysis (HD), there is a critical need to be able to (1) dynamically assess life expectancy for medical decision-making; and (2) identify distinct clinical phenotypes to enhance clinical monitoring and care planning. Our central hypothesis is that there is heterogeneity in patient survivorship and disease trajectory that, when known, can be used to provide more personalized and effective care. By coupling novel machine learning approaches for survival prediction with granular clinical data on HD patients, we will be able to develop the analytic tools necessary to support precision care. At the completion of this proposal we will have tools to dynamically assess a patient's life expectancy and insights into heterogeneous disease phenotypes for patients with ESKD. These tools will allow providers to make informed treatment decisions as well as lay the groundwork for further precision research into optimized patient care.