Methods for Profiling Hospital Performance Based on Healthcare-AssociatedInfections

项目摘要/摘要 医疗保健相关感染(HAI)影响每31名住院患者中的1人，是一种显著的 潜在的可预防的病人伤害的原因。医疗保险和医疗补助服务中心 (CMS)在以下指标中纳入了结肠手术部位感染(SSI)和其他HAI比率 用于根据医院的护理质量对医院进行排名。国家政策对医院排名的依赖 强调需要稳健的方法，能够正确区分有意义的差异 在护理中，而不是患者群体的差异或随机变异。拟议中的工作 旨在开发改进的医院概况描述方法，并解决三个方法差距 利用来自189家社区医院网络的详细管理和临床数据。 评估医院绩效需要进行风险调整，这需要选择患者级别 预测SSI风险的特征，同时考虑医院内的聚集性。然而， 变量选择过程对于集群数据是有限的，因为在处理 复杂的依赖结构。目标1建议开发一个新的变量选择框架 对于高维聚集数据，容纳缺失的协变量。令人担忧的是 提出了对手术量较低的医院进行排名的可靠性。目标2建议 开发可用于确定特定环境下所需手术的分析工具 用户指定的误分类为最差性能阈值的音量 四分位数。旨在提高医院排名可靠性的方法也将由 汇集多年或多个指标的信息。目标3建议开发 比较不同排名系统和识别医院特征的有效方法 这是造成这种差异的原因之一。我们会开发方便易用的软件，以方便 实施新办法。方法的开发将由HCA冒号SSI指导 数据集和人力厅的国家信息系统数据库(2014-2016)。这些方法具有广泛的应用前景。 HAI和其他重要情况的结果，如败血症。拟议的研究是 意义重大，因为成功解决这些问题将提高区分 不同医院之间的HAI比率差异是真正有意义的还是不同的人工制品 患者群体，或者，否则可能被低手术量掩盖。创新 在于开发新的方法和工具，以更好地调整风险，增加 医院排名的可靠性，以及比较排名系统的可靠性。建议的研究结果 研究将有助于为正在进行的绩效工资计划的决策提供信息，以及 最终提高我们预防甲型HAIS的能力，提高护理质量。

Project Summary/Abstract Healthcare-associated infections (HAIs) affect 1 in 31 hospitalized patients and are a significant cause of potentially preventable patient harm. The Centers for Medicare and Medicaid Services (CMS) incorporates colon surgical site infections (SSIs) and other HAI rates in metrics that are used to rank hospitals on their quality of care. The reliance of national policy on hospital rankings underscores the need for robust methodology that can properly distinguish meaningful differences in care as opposed to differences in patient populations or random variation. The proposed work aims to develop improved methods for hospital profiling and addresses three methodological gaps leveraging detailed administrative and clinical data from a network of 189 community hospitals. Profiling hospital performance requires risk-adjustment, which entails selecting patient-level characteristics that predict SSI risks while accounting for clustering within hospitals. However, variable selection procedures are limited for clustered data due to challenges in handling the complex dependence structure. Aim 1 proposes to develop a new variable selection framework for high-dimensional clustered data, accommodating missing covariates. Concerns have been raised about the reliability of rankings for hospitals with a low surgical volume. Aim 2 proposes to develop analytic tools that can be used to determine, for a particular setting, the required surgical volume for a user-specified threshold of the rate of misclassifying into the worst-performing quartile. Methods that aim to improve the reliability of hospital rankings will also be developed by pooling information from multiple years or from multiple indicators. Aim 3 proposes to develop valid methods for comparing different ranking systems and for identifying hospital characteristics that contribute to the differences. User-friendly software will be developed to facilitate the implementation of new methods. The methods development will be guided by an HCA colon SSI dataset and the AHRQ HCUP’s NIS database (2014-2016). The methods can be applied broadly to HAIs and outcomes of other important conditions such as sepsis. The proposed research is significant, because success in addressing these issues will improve the ability to distinguish differences in HAI rates across hospitals that are truly meaningful versus an artifact of different patient populations, or that might otherwise be masked by the low surgical volume. Innovation lies in the development of new methods and tools for better risk-adjustment, to increase the reliability of hospital rankings, and for comparing ranking systems. The results of the proposed research will help inform decision-making on the ongoing pay-for-performance programs, and ultimately improve our capacity to prevent HAIs and improve quality of care.