Comparing Quality of Care between VA-Delivered and VA-Purchased Care: The Need for Better Risk Adjustment

背景：2018年的《使命法案》进一步扩大了退伍军人获得护理的机会， 社会各界虽然这可能改善了获得护理的机会，但也可能导致减少了 退伍军人得到的医疗服务质量。 意义：本研究将首次改进VA使用的现有风险调整方法， VA/社区护理（CC）质量比较。随着越来越多的退伍军人登记使用 CC，迫切需要知道VA购买的社区护理质量是否 至少相当于VA提供的护理质量。我们还将研究是否添加社交 健康的决定因素（SDOH）的风险调整方法的影响评估的质量和健康 差距。 创新和影响：更好地了解哪些新数据源和SDOH变量得到改善 需要风险调整方法，以便更公平和更准确地比较VA/CC质量， 国家和地方两级。我们将获得一个更丰富的图片的条件和其他个人- 以及影响退伍军人风险或疾病负担的社区因素，还包括来自 多系统使用（医疗保险和所有付款人索赔数据库）。 具体目标：对于2020-2022年联邦财政年度，我们的具体目标是：1）审查是否增加 容易获得的VA，CC和药房数据以及个人和社区水平的SDOH变量改善了 Gagne（一种易于修改的comorbance测量）的鉴别和校准; 2）检查是否 增加非VA系统使用，提高了Gagne的辨别力和校准; 3）总体比较 使用“Gagne 1”（包括额外的VA和SDOH）在国家和地方区域级别的VA/CC质量 数据）、“Gagne 2”（包括非VA系统使用数据集）和Nosos风险评分（“现成”VA 风险调整方法）。我们选择了两个退伍军人亚组进行研究： (SMI)和接受全髋关节或全膝关节置换术（THA/TKA）的退伍军人。这两个小组都在 高成本、高容量的外包CC类别。护理质量由4个健康结果定义： 患有SMI的退伍军人的急诊科（艾德）就诊和再入院，以及并发症和 THA/TKA后再次入院。 方法：对于目标1，我们将检查Gagne合并症的系数 在将其他数据源（CC和药房数据）添加到VA利用率数据时以及在 增加SDOH变量以预测结果。我们还将比较 Gagne，Gagne1，Nosos.对于目标2，我们将检查将非VA系统使用数据添加到 Gagne 1的模式识别和校准的Gagne 2。目标3： 使用Gagne 1、Gagne 2和Nosos比较国家和地方层面的整体VA/CC质量。 下一步/执行：通过与主要业务伙伴（社区办公室）合作， 护理[OCC]、分析和性能集成[API]、健康公平办公室[OHE]、访问办公室和 严重精神疾病研究和评估中心[SMITREC]），我们将提供VA领导人和 政策制定者对质量进行公平比较，可将其纳入正在进行的工具开发 以及正在全国范围内实施或用于帮助促进实践的举措（风险调整人 使用）和政策（与是否分配额外资源给VA提供的护理或扩大 使用VA购买的护理）。我们的研究结果也将使更多知情的工作人员，临床医生和退伍军人共享 决定在哪里接受护理。

Background: The MISSION Act of 2018 further expanded the opportunity for Veterans to receive care in the community. While this may have led to improved access to care, it may also have resulted in decreases in the quality of care that Veterans receive. Significance: This study will be the first to improve existing risk adjustment methods used by VA with an application for VA/Community Care (CC) quality comparisons. As increasing numbers of Veteran enrollees use CC, there is an urgent need to know if the quality of care delivered in the community that is purchased by VA is at least equivalent to the quality of care delivered in VA. We will also examine whether adding social determinants of health (SDOH) to the risk adjustment methods impacts assessments of quality and health disparities. Innovation and Impact: A better understanding of which new data sources and SDOH variables improve risk adjustment methods is needed to enable fairer and more accurate comparisons of VA/CC quality at both the national and local area levels. We will obtain a much richer picture of the conditions and other individual- and community-level factors that affect the risk or disease burden of Veterans by also including data from multi-system use (Medicare and All-Payer Claims Databases). Specific Aims: For federal fiscal years 2020-2022, our specific aims are to: 1) Examine whether adding readily available VA, CC, and pharmacy data and individual- and community-level SDOH variables improve the discrimination and calibration of Gagne (an easily modifiable comorbidity measure); 2) Examine whether adding non-VA system use improves the discrimination and calibration of Gagne; and 3) Compare overall VA/CC quality at the national and local area levels using “Gagne1” (which includes additional VA and SDOH data), “Gagne2” (which includes non-VA system use datasets), and the Nosos risk score (an “off-the-shelf” VA risk adjustment method). We selected two Veteran subgroups to study: Veterans with serious mental illness (SMI) and Veterans undergoing total hip or total knee arthroplasty (THA/TKA). Both of these subgroups are in high-cost, high-volume categories of outsourced CC. Quality of care is defined by 4 health outcomes: emergency department (ED) visits and readmissions for Veterans with SMI, and complications and readmissions after THA/TKA. Methodology: For Aim 1, we will examine the extent to which the coefficients on the Gagne comorbidities change when additional data sources (CC and pharmacy data) are added to VA utilization data and then when SDOH variables are added to predict outcomes. We will also compare discrimination and calibration between Gagne, Gagne1, and Nosos. For Aim 2, we will examine the effect of adding non-VA system use data to Gagne1 and evaluate their effect on model discrimination and calibration of Gagne2. For Aim 3, we will compare overall VA/CC quality at the national and local levels using Gagne1, Gagne2, and Nosos. Next Steps/Implementation: Through collaboration with key operational partners (the Office of Community Care [OCC], Analytics and Performance Integration [API], Office of Health Equity [OHE], Access Office, and Serious Mental Illness Research and Evaluation Center [SMITREC]), we will provide VA leaders and policymakers with equitable comparisons of quality that can be integrated into ongoing development of tools and initiatives that are being implemented nationwide or used to help facilitate practice (which risk adjuster to use) and policy (decisions related to whether to allocate additional resources to VA-provided care or expand use of VA-purchased care). Our findings will also enable more informed staff, clinician, and Veteran shared decision-making about where to receive care.