Developing an autism-specific mortality risk index using data from Medicare-enrolled autistic older adults

使用参加医疗保险的自闭症老年人的数据制定特定于自闭症的死亡风险指数

• 批准号: 10716884
• 负责人: Lauren Bishop
• 金额: $ 67.73万
• 依托单位: OHIO STATE UNIVERSITY
• 依托单位国家: 美国
• 财政年份: 2023
• 资助国家: 美国
• 起止时间: 2023-08-01 至 2028-04-30
• 项目状态: 未结题
• 来源: https://reporter.nih.gov/project-details/10716884
• 关键词: Academic Medical Centers; Adult; Age; Aging; Algorithms; Area Under Curve; Characteristics; Chronic; Clinical; Communities; Community Practice; Complex; Data; Data Analyses; Disease; Disparity; Elderly; Electronic Health Record; Enrollment; Fatigue; General Population; Goals; Gold; Health; Health Status; Healthcare Systems; Intervention; Intervention Studies; Life Expectancy; Logic; Machine Learning; Measurement; Measures; Medicare; Medicare claim; Observational Study; Outcome; Participant; Pattern; Persons; Population; Premature Mortality; Prospective Studies; Public Health; Research; Research Personnel; Risk; Risk Reduction; Sample Size; Strategic Planning; Subgroup; Techniques; Testing; Time; Urinary Incontinence; Weight; adult with autism spectrum disorder; age related; aging population; autism spectrum disorder; autistic; billing data; clinical practice; comorbidity; comorbidity Index; falls; forest; frailty; high risk; human old age (65+); improved; indexing; innovation; member; mortality; mortality risk; novel; outcome prediction; phrases; practical application; predictive tools; preference; prospective; research study

High risk for premature mortality is one of the most pressing issues faced by the growing population of aging autistic adults. Autistic adults are disproportionately more likely to have chronic conditions, leading to increased risk for mortality compared to the general population. However, one major barrier to identifying those at greatest risk for mortality is the absence of accurate predictive tools for this population. Our objective is to establish a novel, machine-learning derived mortality risk index for autistic older adults. We will leverage our team’s unique expertise in autism aging research, population-level administrative data analysis, and machine learning to achieve our specific aims: (Aim 1) identify comorbidities and geriatric complaints that differentially influence time- to-mortality for autistic and non-autistic older adults; (Aim 2) compare existing mortality risk indices to a novel, autism-specific index for predicting autistic older adults’ risk of mortality; (Exploratory Aim 3) determine the distribution of mortality risk among autistic older adults in local healthcare systems as a precursor for prospective studies. We will achieve these aims through the synergistic use of national administrative billing data and local electronic health records data. In Aim 1, we will apply a machine learning technique called “logic forest” in an innovative way to identify specific comorbidities and age-related conditions, or combinations thereof, that differentially influence time-to-mortality among autistic and non-autistic older adults using the most recent nine years of national Medicare data. In Aim 2, we will apply a stochastic hill climbing optimization technique, a type of machine learning, to national Medicare data to develop an algorithm-based index that quantifies autistic older adults’ risk of mortality based on comorbidities and demographic characteristics. We will compare the predictive validity of our novel autism-specific algorithm-based index to the Charlson and Elixhauser comorbidity indices, the gold-standards of mortality risk measurement among the general population. Last, in Exploratory Aim 3 we will obtain sample size estimates for prospective studies by quantifying mortality risk among aging autistic adults in two large healthcare systems using the Charlson, Elixhauser, and our novel autism-specific mortality risk indices. Findings of this study will have practical applications for researchers to identify participants for prospective observational and intervention studies and clinicians to identify high-risk cases for special management and intervention. This study is responsive to NOT-AG-21-020 in that we will analyze existing Medicare claims data to examine “subgroups of older adults with special needs” and “health outcomes in complex multimorbid older adults”. Further, this study is aligned with the NIA’s Strategic Plan as we seek to “understand disparities related to aging and [inform] strategies to improve the health status of older adults” on the autism spectrum. This project will have a high public health impact yielding critical new information about mortality risk among a historically understudied aging population that can ultimately be used to improve life-expectancy among autistic people.

高风险的过早死亡是日益增长的老龄化人口所面临的最紧迫的问题之一 自闭症成年人自闭症成年人更有可能患有慢性疾病，导致增加 与一般人群相比，死亡风险。然而，一个主要的障碍，以确定那些在最大的 死亡率的风险是缺乏准确的预测工具。我们的目标是建立一个 新的，机器学习衍生的自闭症老年人死亡风险指数。我们将利用我们团队的独特性 在自闭症老龄化研究，人口水平的管理数据分析和机器学习方面的专业知识， 实现我们的具体目标：（目标1）确定对时间有不同影响的合并症和老年病投诉- 自闭症和非自闭症老年人的死亡率;（目的2）将现有的死亡率风险指数与一种新的， 预测孤独症老年人死亡风险的孤独症特异性指数;（探索性目标3）确定 当地医疗保健系统中自闭症老年人的死亡风险分布作为前瞻性研究的前兆 问题研究我们将通过协同使用国家行政计费数据和地方 电子健康记录数据。在目标1中，我们将在一个应用程序中应用一种称为“逻辑森林”的机器学习技术。 识别特定合并症和年龄相关疾病或其组合的创新方法， 使用最近的9项研究， 多年的国家医疗数据。在目标2中，我们将应用随机爬山优化技术， 机器学习，国家医疗保险数据，以开发一个基于算法的指数，量化自闭症老年人 基于合并症和人口统计学特征的成人死亡风险。我们将比较预测的 我们的新的自闭症特异性算法为基础的指数的有效性，以查尔森和Elixhauser共病指数， 一般人群死亡风险衡量的金标准。最后，在探索性目标3中，我们 将通过量化老年孤独症成人的死亡风险来获得前瞻性研究的样本量估计值 在两个大型医疗保健系统中使用Charlson，Elixhauser和我们的新自闭症特异性死亡风险， 指数。这项研究的结果将有实际的应用研究人员确定参与者， 前瞻性观察和干预研究以及临床医生， 管理和干预。本研究响应NOT-AG-21-020，我们将分析现有的 医疗保险声称数据检查“有特殊需要的老年人亚组”和“复杂的健康结果” 多病老年人”。此外，这项研究与NIA的战略计划保持一致，因为我们试图“了解 与老龄化有关的差异和[告知]改善老年人健康状况的战略”关于自闭症 江西篇章该项目将对公共卫生产生重大影响，产生关于死亡风险的重要新信息 在历史上未充分研究的老龄化人口中， 孤独症患者