Epidemiology of Care Teams: Network Analysis of Providers and Shared Patients

描述（由申请人提供）：在一项关于照顾医疗保险受益人的初级保健医生的调查中，一名普通医生在117个实践中与229名其他医生协调护理。标准化后，这相当于每100名医疗保险受益人有99名医生和53名诊所。美国医疗保健系统的庞大规模为3.5亿多名患者提供服务，这使得这成为一个“大数据”问题。每个病人都是一群提供者的中心。我们试图在一个系统中衡量医疗服务提供者之间的相互联系，在这个系统中，数据往往是由付款人孤立的，因此不能反映许多提供者之间的联系。我们构建了一个独特的数据集，由选定的地理上定义的市场组成，由单一付款人超渗透，并通过它定义卫生保健团队的流行病学和团队结构。基于我们在保险索赔数据网络建模方面的经验，我们采用来自社会网络分析的工具来表征美国卫生系统中分散团队护理的性质。我们开发了计算方法，使政策制定者能够理解医疗保健社会网络基础设施的“大数据”问题，这些基础设施由相关提供者和患者组成。除了对提供者关系的社会网络分析以及通过大型网络的信息流和影响之外，我们还开发了以患者为中心的团队（本质上是围绕单个患者的提供者网络）为主要单位的卫生系统数据的新型社会网络方法。在构建提供者网络时，每个节点都是一个单独的提供者。提供者对之间的联系是由他们共享的病人数量加权的。我们引入了网络衍生的“团队稳定性”概念，并开发了新的量化方法。这些创新的指标允许通过在美国医疗保健系统的团队结构的第一次大规模系统测量团队的基线流行病学的阐明。我们假设，全国的医疗服务提供者不断地组合和重新组合成数量惊人的不同护理团队（可能有数百万）。作为团队结构有效性的初步测试，我们对团队特征与医疗质量指标之间的关系进行了形成性研究。

DESCRIPTION (provided by applicant): In a survey of primary care physicians taking care of Medicare beneficiaries, an average physician coordinates care with 229 other physicians working in 117 practices. Normalized, this is equivalent to 99 physicians and 53 practices for every 100 Medicare beneficiaries. The sheer magnitude of the US Healthcare system, providing services to more than 350 million patients, makes this a "big data" problem. Each patient is at the center of a cluster of providers. We seek to measure interconnectedness among health care providers in a system where data tend to be siloed by payor, and therefore do not reflect many inter-provider connections. We construct a unique dataset of selected geographically-defined markets, ultra-penetrated by a single payor and with it, define the epidemiology of health care teams and team structure. Building on our experience in network modeling with insurance claims data, we adapt tools from social network analysis to characterize the nature of fragmented team care in the US health system. We develop computational methods to allow policy makers to understand the "big data" problem of the healthcare social network infrastructure of linked providers and patients. Beyond social network analysis of provider relationships and the flow of information and influence through large networks, we develop novel social networking methods for health system data using the patient-centered team (essentially a network of providers around a single patient) as the primary unit. A provider network is built where each node is an individual provider. Links between provider pairs are weighted by the number of patients they share. We introduce the network-derived concept of "team stability" and develop novel methods to quantify it. These innovative metrics allow elucidation of the baseline epidemiology of teams through the first broad-scale system measurements of team structure in the US Healthcare system. We hypothesize that that the national population of providers constantly assorts and re-assorts into a staggeringly large number of different care teams (perhaps millions). As an initial test of the validity of our team constructs, we conduct a formative study of the association between team characteristics and health care quality metrics.