Development and Validation of Computational Algorithms to Assess Kidney Health in Electronic Health Records

项目摘要 这项提议的一个关键目的是为候选人奥兹拉兹加特·巴斯兰蒂博士提供必要的保护时间和 额外的培训和资源，以发展其量化方法的技能和对潜在因素的理解 肾脏疾病的进展机制，并促进她向独立的翻译 医疗保健研究人员。长期的职业目标是成为一名独立的数据科学家，专注于 关于医院对急性疾病和由这种护理引起的并发症的护理。这样做的总体目标是 应用是建立分析方法的基础，以确定患者在 急性住院事件和量化健康状态的转变，可以应用于任何急性 生病了。我们的中心假设是，使用肾脏健康作为这种方法的范例，我们可以确定 使用纵向的、高颗粒的时间序列研究住院期间肾脏健康的个体状态变化 电子健康记录中的数据，确定过渡到更严重的急性和慢性阶段的可能性 肾脏疾病，并增进对影响这些转变的潜在过程的理解。当前 急性肾损伤(AKI)的诊断和风险评估的重点是确定AKI发作的严重程度 评估肾脏恢复情况的综合框架也不存在。显然缺乏对……的研究。 通过非线性和非正常时间估计肾脏健康不同状态之间的转移概率 使用纵向电子健康记录数据的依赖域。底层流程的复杂性 影响急性和慢性肾脏疾病从肾脏风险过渡到更严重阶段的概率 需要在足够大和粒度的数据集中应用高级计算模型。具体的 该提案的目的是：目标1--大规模扩展和验证肾脏健康的可计算表型 医疗数据。目的2-确定肾脏健康变化的流行病学和临床结果。目标3- 开发和验证概率图形模型，以预测肾脏健康状态和 确定疾病进展的风险因素。这项拟议的研究意义重大，因为我们将拥有表型算法 肾脏健康，在多中心研究中得到验证，可以增强机构间的共享，并使 研究肾脏健康变化的流行病学和结局。这种方法是创新的，因为它 以创新的步骤实施数据科学和统计方面的技术进步，以开发和验证 确定肾脏健康和图形变化的可计算表型的表型算法 通过非线性和非正常时间预测肾脏健康状态转换的模型- 使用高度细粒度的电子健康记录的依赖域。这将为以下方面的变化提供基础 AKI患者的护理，通过识别那些有发展AKI风险的患者并进展到 急性和慢性肾脏疾病。在拟议的调查完成后，交付成果将是新的 肾脏健康的知识和诊断和预测工具。

Project Summary A key aim of this proposal is to equip the candidate, Dr.Ozrazgat Baslanti, with the necessary protected time and additional training and resources to develop her skillset on quantitative methods and understanding of underlying mechanism of progression of kidney disease and facilitate her transition to an independent translational researcher in health care. The long-term career goal is to become an independent data scientist, with a focus on hospital care for acute disease and complications arising from that care. The overall objective of this application is to build the foundation of the analytical approach for identifying patients’ health trajectories during episode of acute hospitalization and quantifying the transitions in health states that can be applied to any acute illness. Our central hypothesis is that using kidney health as a paradigm for this approach we can determine individual states of change in kidney health during hospitalization using longitudinal, highly granular temporal data in electronic health records, determine transition probabilities to more severe stages of acute and chronic kidney disease, and improve understanding of the underlying processes influencing these transitions. Current diagnosis and risk evaluation for acute kidney injury (AKI) are focused on determination of severity of AKI episode and an integrated framework for assessing renal recovery does not exist. There is a clear lack of research on estimating transition probabilities among different states of kidney health through nonlinear and non-normal time- dependent domains using longitudinal electronic health records data. The complexity of underlying processes influencing the transition probabilities from renal risk to more severe stages of acute and chronic kidney disease requires application of advanced computational models in sufficiently large and granular datasets. The specific aims of the proposal are: Aim 1- Expand and validate computable phenotypes of kidney health in large-scale medical data. Aim 2- Determine the epidemiology and clinical outcomes of changes in kidney health. Aim 3- Develop and validate probabilistic graphical models to predict transition through the states of kidney health and identify risk factors for progression. The proposed research is significant as we will have phenotyping algorithms of kidney health, validated in multi-center study, that can enhance their inter-institutional sharing and that enable to study epidemiology and outcomes of changes in kidney health. The approach is innovative because it implements technological advances in data science and statistics in innovative steps to develop and validate a phenotyping algorithm that determines computable phenotypes of changes in kidney health and graphical models to predict transition through the states of kidney health through nonlinear and non-normal time- dependent domains using highly granular electronic health records. This will provide foundation for changes in the care of patients with AKI, through identification of those patients at risk of developing AKI and progressing to acute and chronic kidney disease. On completion of the proposed investigations the deliverables will be new knowledge and a diagnosis and prognostication tool for kidney health.