CAREER: A Unified Methodology for Optimizing the Management of Chronic Diseases

这项教师早期职业发展（Career）资助的目标是创造新的方法，使慢性病患者的操作和疾病管理决策能够整合。该项目侧重于(1)确定何时以及如何筛查、监测和治疗慢性疾病患者，因为每个患者的疾病进展参数值都是未知的（但随着时间的推移而学习），以及(2)在患者一生中提供最佳筛查、监测和治疗决策，从而改善资源分配。通过与临床医生和管理人员的多学科合作，该奖项创造的新方法将首先应用于三种慢性病的管理：青光眼、膀胱癌和高血压。但是，预计这些方法也将适用于其他慢性疾病。这项工作将支持研究生，本科生和大学预科学生，关注代表性不足的学生。该项目的成果将通过用户友好的工具、会议报告和期刊出版物广泛传播给工程界和医学界。研究小组将创建一个新的建模框架，将生存分析结果与疾病进展的表示相结合，并使用连续的、部分可观察的人群中患者的状态空间模型进行干预。这将通过(1)动态地将每个患者的病史纳入疾病状态演化随机模型的参数化，而不是使用静态的全民参数值；(2)从每个患者的角度比较导致改善的政策，而不是基于使用有限资源来改善人群健康结果的政策。(3)利用状态空间疾病演化模型的结构特性，提高在统一方法中使用这些模型的问题的最优解的计算能力。

The objective of this Faculty Early Career Development (CAREER) grant is to create new methodology that enables the integration of operational and disease management decisions for patients with chronic diseases. This project focuses on (1) determining when and how to screen, monitor and treat chronic-disease patients given that each patient's disease progression parameter values are not known, (but learned over time), and (2) improving resource allocation given optimal screening, monitoring and treatment decisions over the patient's lifetime. Through a multidisciplinary collaboration with clinicians and managers, the new methodologies created in this award will first be applied to the management of three chronic diseases: glaucoma, bladder cancer, and hypertension. However, it is anticipated that these methodologies will be applicable to other chronic diseases as well. This work will support graduate, undergraduate, and precollege students with attention to underrepresented students. The results of the project will be widely disseminated to engineering and medical communities through user-friendly tools, conference presentations, and journal publications.The research team will create a novel modeling framework that combines results in survival analysis with a representation of disease progression and intervention using continuous, partially observable state space models of patients within the population. This will be done by (1) dynamically incorporating each patient's disease history into the parameterization of the stochastic model of disease-state evolution, rather than using static population-wide parameter values, (2) comparing policies that lead to improvements from each patient's perspective, rather than those that are based on using constrained resources to improve the health outcomes of the population, and (3) exploiting structural properties of the state-space disease-evolution models to enhance the computation of optimal solutions to problems that use these models in the unified methodology.