Adverse Outcomes of Dual Use of Health Systems Among Older Male Veterans

描述（由申请人提供）：65岁或以上的退伍军人有资格使用退伍军人健康管理局（VHA）系统和通过医疗保险提供的私人医疗保健系统。这种“双重”系统的使用既有积极的影响，也有消极的影响。我们假设老年男性退伍军人的双重使用增加了他们因门诊护理敏感条件（ACSCs）住院的风险，并增加了他们的死亡风险。我们在老年男性退伍军人的资产和健康动态调查（AHEAD）中检验了这一假设，该调查包括2,911名年龄在70岁或以上的男性（1993年）。其中1,574人（54%）是退伍军人，281人（17.9%）报告患有与服役有关的残疾。我们将把AHEAD调查数据与医疗保险索赔和国家死亡指数联系起来，我们将对医疗保健市场结构的数据进行地理编码。由于VHA索赔数据不可用，我们将间接测量双重使用基于自我报告和医疗保险索赔之间的差异，通过三个步骤的过程：(1)通过生成二元住院和门诊报告差异标记来识别自我报告使用和医疗保险索赔之间的不一致；(2)使用多元逻辑回归和所有适当的基线协变量，包括退伍军人状态标记，为步骤1中每个标记的不一致可能性创建倾向分数；(3)利用第2步所得的倾向得分和退伍军人身份标记(a)首先对ACSCs的风险和基线后的死亡率进行加性建模，然后(b)为退伍军人身份标记和倾向得分之间的乘法交互作用添加一个标记，以挖掘它们的协同效应，最后(c)调整潜在的混杂因素，试图分解步骤3b中观察到的效应。我们将使用比例风险模型，我们的重点是退伍军人的状态标记和倾向得分之间的乘法交互作用。因为非退伍军人在退伍军人状态标记上被编码为零，所以这个乘法交互项的显著和积极的参数估计将是支持我们假设的初步证据。届时，我们将向AHEAD、VHA和CMS申请许可，将VHA索赔与现有的AHEAD综合数据文件联系起来，然后提出一项后续的NIH R01研究，以使用医疗保险和VHA索赔数据，为我们关于双重用途对医疗保健结果的不利影响的假设提供更明确和直接的检验。

DESCRIPTION (provided by applicant): Veterans aged 65 years old or older are eligible to use both the Veterans Health Administration (VHA) system, and the private health care delivery system via Medicare. Such "dual" system use can have both positive and negative effects. We hypothesize that dual use among older male veterans increases their risk of being hospitalized for ambulatory care sensitive conditions (ACSCs), and increases their risk of mortality. We examine this hypothesis among older male veterans in the Survey of Assets and Health Dynamics Among the Oldest Old (AHEAD), which included 2,911 men aged 70 years old or older at baseline (1993). 1,574 (54%) of the AHEAD men are veterans, and 281 (17.9%) reported having a service-related disability. We will link the AHEAD survey data to Medicare claims and the National Death Index, and we will geocode data on health care market structure. Because VHA claims data are not available, we will indirectly measure dual use based on the discrepancy between self-report and Medicare claims through a 3-step process: (1) identify discordance between self-reported use and Medicare claims by generating binary inpatient and outpatient reporting discrepancy markers; (2) create propensity scores for the likelihood of discordance on each of the markers in Step 1 using multiple logistic regression and all appropriate baseline covariates including the veterans' status marker; and, (3) use the propensity scores obtained in Step 2 and the veterans' status marker to (a) first additively model the risk of ACSCs and mortality after baseline, then (b) add a marker for the multiplicative interaction between the veterans' status marker and the propensity scores to tap their synergistic effect, and finally (c) adjust for potential confounders in an attempt to decompose the effect observed in Step 3b. We will use proportional hazards models and our focus is on the multiplicative interaction between the veterans' status marker and the propensity score. Because non-veterans are coded zero on the veterans' status marker, significant and positive parameter estimates for this multiplicative interaction term will be prima facie evidence supporting our hypothesis. At that point we will petition AHEAD, VHA, and CMS for permission to link VHA claims to the existing AHEAD comprehensive data files, and then propose a follow-on NIH R01 study to provide a more definitive and direct examination of our hypothesis about the adverse effect of dual use on health care outcomes using both Medicare and VHA claims data.