Effects of Missing Data Strategies on Disparities Research Results in HCUP SID

描述（由申请人提供）：消除医疗保健差距，以确保服务不足的社区（例如少数民族、老年人、低收入者）和其他AHRQ优先人群获得优质医疗保健仍然是国家优先事项。解决医疗保健差异所需的大型、基于人口的研究可能成本高昂且难以执行，并且可能受到抽样策略和患者选择偏差的影响，使用医疗保健成本和利用项目（HCUP）国家住院患者数据库（SID）是一种越来越有吸引力的有效替代方案。SID和其他大型数据库的一个重要限制是缺失数据的数量。特别是，作为健康差异研究的一个关键指标，“患者种族”缺失的比例很高。本研究的目的是使SID成为研究种族差异更有用和可靠的资源。针对SID数据缺失问题，本文提出了顺序回归多元数据缺失和潜正态多元数据缺失两种多元数据缺失方法。这些方法将通过一个全面的模拟研究进行比较。它们相对于三种常用的缺失数据方法（即完整案例分析，缺失指标法，热甲板imputation）的优势也将通过模拟研究来说明。在仿真的基础上，选择最优的MI方法进行插补。因此，将生成多个输入数据集作为SID的伴侣，这将允许用户使用现有软件对广泛的实质性研究问题的完整数据进行分析。我们将使用输入的SID数据进行肌肉骨骼保健差异研究。该应用旨在确定种族是否是全膝关节置换术（TKR）后一系列不良后果的危险因素，以及在全膝关节置换术的使用中是否存在种族差异。

DESCRIPTION (provided by applicant): Eliminating healthcare disparities so underserved communities (e.g., minorities, elderly, low income) and other AHRQ priority populations are assured access to quality medical care remains a national priority. Large, population based studies necessary to address healthcare disparities can be costly and difficult to perform, and may be compromised by sampling strategies and patient selection biases, an efficient alternative that is becoming increasingly attractive is the use of the Healthcare Cost & Utilizatio Project (HCUP) State Inpatient Databases (SID). A significant limitation of SID and other large databases is the quantity of missing data. In particular, "patient race", a key indicator for healt disparities research, has a high proportion of missingness. The goal of this study is to make SID a more useful and reliable resource for the study of racial disparity. Accordingly, two multiple imputation (MI) methods (1) the sequential regression multivariate imputation, and (2) the latent normal multivariate imputation are proposed for addressing the missing data issue in the SID. These approaches will be compared through a comprehensive simulation study. Their advantages over the three commonly used missing data approaches (i.e. complete case analysis, missing indicator method, hot deck imputation) will also be illustrated through the simulation study. Based on the simulation, we will select the optimal MI method for imputation. As a result, multiply imputed datasets will be generated as a companion to the SID that will allow users to perform analysis using existing software for complete data for a wide range of substantive research questions. We will use imputed SID data to conduct musculoskeletal healthcare disparities research. The application is to determine whether race is a risk factor for set of adverse outcomes after total knee replacement (TKR) and whether racial disparities exist in utilization of TKR.