Precision Approaches to Reduce Asthma Disparities with Electronic Health Record Data

哮喘是一种慢性疾病，表现为气道对特定环境刺激的高反应性， 影响了两千万美国成年人美国成人哮喘患病率、严重程度和死亡率的差异 是众所周知的，但很少有方法能显著降低它们。对现实世界人口的研究， 因为来自电子健康记录（EHR）的信息对于指导个性化护理的设计非常宝贵 战略，因为他们抓住了大量不同的和脆弱的人。此外，EHR数据可以 通过研究地理信息， 受影响患者的分布。我们已经确定了个人和地区层面的因素， 使用EHR的哮喘与丰富多样的社会、经济和环境变量来源有关， 我已经开发出了从异质可用的EHR数据中适当提取信息的方法 并减少这些不完美数据分析中的偏倚。本提案将确定下列小组： 具有共同的人口统计学、临床、社会和环境暴露模式的成人哮喘患者 使用EHR数据的特征与社会，经济和环境因素的数据增强， 能够设计有效的精确策略来减少哮喘急性发作。我们的目标是：1）发展 并验证自然语言处理（NLP）算法，以提取社交、职业和过敏信息 2）使用空间数据确定增加哮喘急性发作风险的地理空间区域 广义线性混合模型，并确定在这些领域的危险因素;和3）使用贝叶斯分层 用于识别哮喘亚表型的聚类技术，其将形成临床决策支持工具的基础 提供精准的护理策略。该项目将导致创建一个工具，可以帮助照顾 成人哮喘基于可采取行动的风险因素，并推动公共卫生战略，以减轻负担， 哮喘在特定地区我们新颖的数据整合方法、亚表型分析方法和软件 开发将有广泛的适用性，任何条件的研究使用EHR或其他现实世界的数据。

Asthma, a chronic disease that manifests as airway hyperresponsiveness to specific environmental stimuli, affects over 20 million American adults. Disparities in adult asthma prevalence, severity, and death in the U.S. are well known but few approaches have significantly decreased them. Studies of real-world populations such as those derived from Electronic Health Records (EHRs) are invaluable to guide the design of personalized care strategies because they capture a large number of diverse and vulnerable people. Additionally, EHR data can be leveraged to identify geospatial areas where people are at peak risk of a condition by studying the geographic distribution of affected patients. We have identified individual- and area-level factors that are associated with asthma using EHRs linked to rich and diverse sources of social, economic, and environmental variables, and we have developed methods to appropriately extract information from EHR data that are heterogeneously available across patients and reduce bias in the analysis of these imperfect data. This proposal will identify sub-groups of adults with asthma who share common patterns of demographic, clinical, social, and environmental exposure characteristics using EHR data augmented with data on social, economic, and environmental factors, which will enable the design of effective precision strategies to reduce asthma exacerbations. Our aims are to: 1) develop and validate natural language processing (NLP) algorithms to extract social, occupational and allergy information from EHR notes; 2) determine geospatial areas that have increased asthma exacerbation risk using spatial generalized linear mixed models and identify risk factors in these areas; and 3) use Bayesian hierarchical clustering techniques to identify asthma sub-phenotypes that will form the basis of a clinical decision support tool that offers precision care strategies. This project will result in the creation of a tool that can assist in the care of adults with asthma based on actionable risk factors and motivate public health strategies to mitigate the burden of asthma in specific regions. Our novel data integration approaches, sub-phenotyping methods, and software developed will have broad applicability for the study of any condition using EHR or other real-world data.