Developing a Prediction Model to Improve End‐of‐Life Prognostication and Hospice Referral in Parkinson's Disease

项目总结/摘要 帕金森病（PD）是第二常见的神经退行性疾病，影响约1.5 它是美国第14大死亡原因。目前没有医疗保险。 针对PD的临终关怀转诊指南，我们认为缺乏适当的模式或指南， 临床医生是导致住院死亡率不成比例高和低的主要因素之一， 这一人群的临终关怀转诊率。我们的长期目标是改善人们的临终结局， 与PD及其家庭照顾者一起生活，包括提高临终关怀转诊率和持续时间。 临终关怀使用我们假设，目前可用的临终关怀指南未得到充分利用，即使当 使用，显著低估了6个月死亡率，错过了重要的死亡率预测因素， 人口我们进一步假设，目前在最小数据集（MDS -一个联邦政府 养老院居民的强制性临床数据库）和国家死亡指数（NDI）可用于开发 更准确的预测模型。值得注意的是，Mitchell等人在痴呆症中使用了类似的方法来开发 高级痴呆症预后工具（ADEPT），风险评分证明比医疗保险更准确 老年痴呆症指南。为了开发有效的验证模型和风险评分，我们将遵循 Steyerberg概述的模型构建和验证结构，包括 缺失值的考虑，模型规格，参数估计，性能测量， 可重复性、可推广性和临床实用性。我们将实现这一建议的目标 通过两个具体的目标：1）开发一个预测模型的6个月死亡率的人生活在护理 使用最小数据集和国家死亡指数的PD家庭;以及2）比较 我们的PD特异性预测模型与当前可用的Medicare临终关怀指南相结合。该方法是 作为第一项将这种数据驱动方法应用于PD寿命终止确认和使用状态的创新研究 最先进的机器学习和建模方法，以及评估和解释工具，以实现 临床实用模型拟议的研究是重要的，因为这项建议的结果将创造第一个 基于证据和PD的临终关怀转诊风险评分，并将提供第一个系统性分析， 准确性和目前可用的临终关怀指南在这一人群中的使用。我们预计， 这项工作将对病人护理产生快速和直接的影响，也将为未来的医疗保健奠定基础。 研究，以改善临终关怀转介和在这一人群中，以及未来的基础 在养老院和社区居住队列中进行的前瞻性研究。

Project Summary/Abstract Parkinson's disease (PD) is the second most common neurodegenerative illness, affecting approximately 1.5 million Americans, and is the 14th leading cause of death in the United States. There are currently no Medicare Guidelines for hospice referral specific for PD and we believe the lack of appropriate models or guidelines for clinicians is one of the prime factors leading to the disproportionately high rates of in-hospital deaths and low rates of hospice referrals in this population. Our long-term goal is to improve end-of-life outcomes for persons living with PD and their family caregivers, including improving the rates of hospice referrals and duration of hospice use. We hypothesize that currently available hospice guidelines are underutilized and that, even when used, significantly underestimate 6-month mortality and miss important predictors of mortality in this population. We further hypothesize that data currently available in the Minimal Data Set (MDS – a federally mandated clinical database of nursing home residents) and National Death Index (NDI) can be used to develop a more accurate prediction model. Notably, Mitchell et al. utilized a similar approach in dementia to develop the Advanced Dementia Prognostic Tool (ADEPT), a risk score demonstrated to be more accurate than Medicare Guidelines for dementia. Toward developing a valid prognostication model and risk score, we will follow a model building and validation structure outlined by Steyerberg that includes statistical methodology for the consideration of missing values, model specification, parameter estimation, performance measurement, reproducibility, generalizability, and clinical usefulness. We will accomplish the objectives of this proposal through two Specific Aims: 1) Develop a predictive model of 6-month mortality in persons living in nursing homes with PD using the Minimum Data Set and the National Death Index; and 2) Compare the accuracy of our PD-specific prediction model to currently available Medicare Hospice Guidelines. The approach is innovative as the first study to apply this data-driven approach to PD end-of-life prognostication and uses state of the art machine learning and modeling approaches, and evaluation and interpretation tools, to achieve a clinically useful model. The proposed research is significant as the results of this proposal will create the first evidence-based and PD-specific risk scores for hospice referral and will provide the first systematic analysis of the accuracy and use of currently available hospice guidelines in this population. We anticipate that the results of this work will have a rapid and direct impact on patient care and will also create a foundation for future studies to improve hospice referrals and prognostication in this population as well as a foundation for future prospective studies in both nursing home and community-dwelling cohorts.