Utilizing Technology and AI Approaches to Facilitate Independence and Resilience in Older Adults

翻译后摘要：姑息治疗（PC）是一个跨学科的概念，旨在改善福祉的 在整个病程中患有严重疾病的人，包括临终者。PC使 护理决策符合患者和护理人员的偏好。对于老年痴呆症患者 疾病和相关疾病（ADRD），PC是特别具有挑战性的，因为确定患者的 ADRD病程的现状通常是困难的。为患者提供成功的PC服务 ADRD的重点是将治疗方案与及时识别和缓解相结合 患者及其家属的生理、心理和决策需求。PC可以 促进道德，公平和有效的人口健康原则，实现最佳的 避免过度使用和使用不足。许多挑战阻碍了适当的 ADRD患者在患者和人群水平的PC整合水平。那里 正在讨论PC交付的最佳时间，以及 训练有素的团队和资源来提供PC。一个普遍存在的挑战是护理系统的 识别那些将从PC中受益最多的人的能力。此外，有文件证明， PC输送的差异，例如少数种族/民族患者接受的太少、太晚 与大多数同行相比，人工智能（AI）预测建模 这些技术可以准确和及时地识别可能患有ADRD的人， 从PC评估中获益。为了实现我们的目标，使用先进的人工智能分析工具， 该项目的目标是：1）开发和 验证高级预测模型（PM），以识别可能受益的ADRD患者 从PC评估; 2）评价基于PM的姑息治疗干预措施对 人口水平的医疗保健利用结果; 3）评估PC服务的差异 非裔美国人和其他少数民族人口的分娩和医疗保健利用率， ADRD; 4）启动我们开发的先进分析工具的第一阶段技术转让 通过开展初步试点，开发可公开访问的软件， JHU“ACG”基于人口的平台。Chintan Pandya博士（PI）和团队将开发 机器学习预测模型识别可能从PC中获益的ADRD患者 考核这些模型将使用电子健康记录（EHR）中捕获的数据开发 以及其它大型电子数据库（例如，ADRD患者的保险索赔） 在大量的患者/消费者群体中。除了共享开放架构的免费访问 工具在这个项目的结论，我们计划整合基于软件的算法，我们 发展成为我们广泛使用的（覆盖20多个国家的2.5亿多患者）约翰霍普金斯ACG 预测建模和疾病分层软件。这将有助于快速传播新的 美国和全球大型医疗保健组织的姑息治疗分析技术 客户群在患者层面，我们的ADRD PC决策支持框架将促进 患者/护理人员和提供者之间的护理讨论的互动目标。人口我们开发的基于分析的工具将有助于促进资源规划、质量和公平 在卫生服务系统、卫生计划和社区内对ADRD队列进行评估。

Abstract: Palliative care (PC) is an interdisciplinary concept aimed at improving the wellbeing of persons with serious illness throughout their course of illness including end-of-life. PC enables care decisions that align with patient and caregiver preferences. For persons with Alzheimer’s Disease and Related Disorders (ADRD), PC is particularly challenging, as determining a patient’s current status on the ADRD disease course is often difficult. Successful PC services in patients with ADRD focuses on integrating therapeutic regimens with timely identification and alleviation of physical, psychosocial, and decision-making needs of patients and their families. PC can promote ethical, equitable, and efficient population health principles by achieving optimal healthcare utilization by avoiding overuse and underuse. Many challenges hinder appropriate levels of PC integration for persons with ADRD at both the patient- and population-levels. There is ongoing discussion on optimal timing of PC delivery in addition to variation in availability of well-trained teams and resources to deliver PC. One ubiquitous challenge is the care system’s ability to identify those persons that will benefit most from PC. Moreover, there are documented disparities in delivery of PC, such that minority race/ethnicity patients receive too little, too late care compared to their majority counterparts. Artificial intelligence (AI) predictive modeling techniques may enable accurate and timely identification of persons with ADRD who are likely to benefit from PC assessment. To achieve our goal of using advanced AI analytic tools to improve PC received by persons with ADRD, the project has the following objectives:1) To develop and validate advanced predictive models (PM) to identify persons with ADRD who are likely to benefit from PC assessment; 2) To evaluate the impact of PM based palliative care interventions on population-level healthcare utilization outcomes; 3) To assess the disparities in PC services delivery and healthcare utilization in African American and other minority populations with ADRD; 4) To initiate first stage of technology transfer of the advanced analytic tools we develop by undertaking initial pilots and developing both publicly accessible software and integration into the JHU “ACG” population-based platform. Dr. Chintan Pandya (PI) and team will develop machine learning prediction models to identify ADRD patients likely to benefit from PC assessment. These models will be developed using data captured in electronic health record (EHR) and other large electronic databases (e.g., insurance claims) of patients with ADRD cared for within a large patient/consumer population. In addition to sharing open architecture free-access tools at the conclusion of this project, we plan on integrating the software-based algorithms we develop into our widely used (reaching 250+ million patients in 20+ nations) Johns Hopkins ACG predictive modeling and disease stratification software. This will allow for rapid diffusion of new palliative care analytic technologies among a very large U.S. and global healthcare organizations client base. At the patient level, our ADRD PC decision support framework will promote interactive goals of care discussions between patients/caregivers and providers. The populationbased analytic tools we develop will help promote resource planning, quality, and equity assessment among ADRD cohorts within health delivery systems, health plans, and communities.