Medicaid at 60: Addressing Data Gaps
本文讨论了医疗补助计划（Medicaid）在实施60年后仍面临的数据缺口问题，重点分析了家庭和社区服务、补充支付和定向支付以及受益人体验三个优先领域的数据挑战，并呼吁改进数据收集和透明度以支持政策研究。
Medicaid, a joint federal and state program, is a major provider of health insurance, covering 94 million beneficiaries with federal and state spending totaling $824 billion in fiscal year 2022 [1]. Sixty years after the program was enacted, Medicaid is a vital source of health coverage for nearly 3 out of 10 Americans and covers over 40% of births [2, 3]. Researchers, policymakers, and federal and state officials require complete and robust data to better understand how Medicaid serves beneficiaries, especially given the variation in state approaches to design and delivery. Yet, in spite of the maturity of the program, the limitations of Medicaid data have hindered research and created challenges for the Medicaid and CHIP Payment and Access Commission (MACPAC) in advancing our mission of providing timely, relevant analyses for Congress, the states, and the health policy community. The Transformed Medicaid Statistical Information System (T-MSIS) is the only national, uniform data set that includes person-level enrollment and service-level claims information on the Medicaid program. While T-MSIS has increased the amount of information available to researchers and the policy community to analyze Medicaid utilization and spending on particular services in fee-for-service (FFS) and managed care delivery systems compared to the predecessor Medicaid Statistical Information System (MSIS) data set, there are still concerns with data completeness and reliability as demonstrated by the data quality (DQ) information the Centers for Medicare & Medicaid Services (CMS) has published on its DQ Atlas website [4]. The state-level information in the DQ Atlas has shown that data are not consistently of low- or high-quality across all variables or even the same variable over time. As such, it is a challenge for researchers to assess the quality and completeness of data for the specific services, subpopulations, and time period they are seeking to analyze, and researchers must develop workarounds for these Medicaid data issues. Even with improvements to the data completeness and quality in T-MSIS, this resource can only tell us part of the story, and there is much left to know about this complex program and how the program is meeting federal and state goals of access, quality, and efficiency in program spending. In some cases, new data sources and methods, as well as greater transparency of these data, are needed. Below we highlight data challenges and gaps in three of MACPAC's priority areas for research: home- and community-based services (HCBS), FFS supplemental and managed care directed payments, and beneficiary experience. There is no straight-forward method to identify HCBS services in T-MSIS due to the level of state variation in types of services provided and how these services are billed and reported. Supplemental and directed payments can make up a significant amount of the total payments to certain providers (e.g., hospitals, nursing facilities) and these payments are frequently paid on a lump sum basis that are not tied to specific claims reported in T-MSIS. Finally, the information reported to T-MSIS is derived from state eligibility and claims systems and does not include information from all of the domains needed to evaluate access to care such as the beneficiary's experience in obtaining services. We must close the knowledge gap in these areas to understand how the Medicaid program provides access to necessary care, where there may be disparities in providing services in the most appropriate setting, and whether payment policies achieve the intended goals of improved access, quality, and efficiency in program spending. Medicaid HCBS have grown into the predominant way of delivering long-term services and supports (LTSS) to Medicaid beneficiaries who need them. HCBS are optional Medicaid benefits, but all states cover some HCBS benefits. HCBS encompass a wide range of services provided in a community setting, such as personal care services, supported employment, nonmedical transportation, and home-delivered meals. These services allow people with LTSS needs to live in their homes or in a homelike setting in the community. People with LTSS needs typically qualify for Medicaid based on their age (65 and older) or disability. Unlike other Medicaid beneficiaries, their income is not determined using modified adjusted gross income, and they must meet certain asset thresholds. Federal and state policymakers, beneficiaries, and other stakeholders have focused over the last two decades on moving the delivery of LTSS away from institutional settings and toward home- and community-based settings to be responsive to beneficiary preferences, foster greater community participation, and address program sustainability. Significant progress has been made. Starting in 2013, total federal and state Medicaid spending on HCBS exceeded total spending on institutional care. In fiscal year 2021, HCBS accounted for 63.2% of total LTSS expenditures. Analyzing spending by state in that same year, 37 states and the District of Columbia spent 50% or more of their total LTSS spending on HCBS. Among all states, rebalancing ratios, measures of progress toward moving LTSS spending away from institutional care and toward HCBS, ranged from 27.9% to 93.5% [5]. Despite the increasing role of HCBS in Medicaid, policymakers do not have a comprehensive understanding of its population, service utilization, and program spending. The data gap is exacerbated by the heterogeneity of HCBS users. HCBS users are made up of many different subpopulations such as seniors, people with physical disabilities, people with intellectual and developmental disabilities, and people with mental illness or serious emotional disturbance. In prior work reviewing demographic characteristics of HCBS users in 2013, we found that HCBS users were primarily persons with disabilities (64.3%) and individuals over the age of 65 (31.3%), although some adults did not have disabilities (1.6%) and some children (2.7%) used the services as well. Most were dually eligible for Medicare and Medicaid, and more than half were between the ages of 19 and 64 [6]. Spending and utilization patterns are likely different among these varied groups, but research at this level of granularity is limited. With the availability of and improvements in T-MSIS, researchers now have a comprehensive data source to analyze HCBS use in both FFS and managed care, but the lack of a standard way of identifying HCBS users and services is a barrier to research in this area. Even though there are numerous variables that researchers can use to identify a claim as HCBS in T-MSIS, these variables may not be consistently reported or states may vary in how they code the claims [7, 8]. Further, there are DQ issues. The Assistant Secretary for Planning and Evaluation (ASPE) and DQ Atlas assessments have found that the data element that states are supposed to populate for the HCBS taxonomy category is unusable. Instead, individual claims for services must be identified separately to classify them into one of the taxonomy categories [7, 9]. MACPAC recently published an issue brief comparing how CMS, ASPE, and KFF have started to use T-MSIS data to analyze LTSS utilization and spending; each analysis employs a distinct methodological approach [10]. For example, both CMS and KFF examine expenditures and user counts for HCBS and institutional LTSS, while the ASPE and DQ Atlas briefs provide a methodological approach for identifying HCBS but do not contain any statistics on use or spending. These approaches, including in the definition of LTSS, lead to differences in the count of users in each analysis [11, 12]. These different analyses from CMS, ASPE, and KFF are good starting points for establishing a common methodology, but collaboration across the research community is needed to identify the pros and cons of various methods and to work toward standard definitions to the extent possible so that findings are more comparable and generalizable [9, 13]. Additional research is also needed on the factors, including those unrelated to payment, that can address persistent workforce shortages. MACPAC has ongoing work on HCBS payment policy and how it might affect the HCBS workforce [14, 15]. The use of supplemental and directed payments makes it challenging to calculate how much was spent for particular services. For example, hospital services represent the largest share of spending in Medicaid, accounting for about one-third of Medicaid spending in 2022 [16]. Approximately 40% of Medicaid hospital payments in 2022 were made through FFS supplemental and managed care directed payments [17]. The increasing role of supplemental and managed care directed payments complicates the analysis of spending and how payments relate to statutory goals of efficiency, economy, quality, and access (§ 1902(a)(30)(A) of the Social Security Act) because these payments may not be tied to a specific beneficiary or service. Additionally, some of these payments support additional goals outside of Medicaid such as uncompensated care for the uninsured or delivery system reform and raise questions as to how to account for these dollars when assessing Medicaid payment. Because FFS supplemental payments are generally made on a lump-sum basis and are not tied to claims, the paid amounts reported in sources such as T-MSIS do not represent the total amount Medicaid paid to certain providers, such as hospitals. For managed care directed payments, the picture is less clear. Certain types of directed payments, such as minimum fee schedules, may have been paid as part of the claim, while other types, such as uniform rate increases, are frequently paid on a lump-sum basis similar to FFS supplemental payments. As such, we still do not have a comprehensive understanding of the full amount that Medicaid paid for hospital services, such as deliveries. MACPAC has previously recommended that CMS collect additional data on these supplemental and directed payments and make them publicly available in a standard format that enables analysis, and some progress has been made. T-MSIS contains substantially more information on the services provided through managed care than was previously available under the predecessor system. Under the provisions of the Consolidated Appropriations Act, 2021 (P.L. 116-220), CMS has started to collect provider-level information on supplemental payments (other than disproportionate share hospital payments). However, this supplemental payment information is collected in a different data system, and additional work is needed to link these data to T-MSIS. Additionally, CMS has started to publish managed care directed payment preprints that provide some information on the amounts projected to be paid to providers, but this information does not represent actual payment amounts. Furthermore, the spending estimates are reported for a class of providers and not for individual providers. The 2024 managed care rule includes a requirement to report managed care directed payments at the provider level into T-MSIS; however, it is not clear how long it may take for the data to be complete and reliable [18]. MACPAC is actively engaged in combining information from these various data sources, documenting the use of these payments, and calculating overall payment rates, starting with hospital services. However, this will only provide part of the picture needed to assess the efficiency of these payments in achieving policy goals. Additional research is needed to identify and develop meaningful measures of quality and outcomes. As some states have started to pay up to average commercial rates for hospital services through managed care directed payments, it is important to understand whether these additional payments are achieving their intended goals of improving access, quality, and efficiency in program spending. Further complicating analyses of hospital services is the fact that many hospitals are a source of financing for a portion of the nonfederal share. Medicaid financing is a shared responsibility between states and the federal government, and states are permitted to raise the nonfederal share of Medicaid expenditures through multiple sources, including state general revenue, health care-related taxes, and funds from other state agencies and local governments (including government-owned providers such as public hospitals) through intergovernmental transfers and certified public expenditures. States increasingly use funds that come from providers to finance the nonfederal share. From state fiscal year 2008 to 2018, the Government Accountability Office found that the use of state general funds declined from 75% to 68% of the nonfederal share, and the use of health care-related taxes more than doubled, from 7% to 17% of the nonfederal share [19, 20]. The nonfederal share financed by providers is commonly used to make FFS supplemental payments (e.g., disproportionate share hospital and upper payment limit payments) and managed care directed payments. The amount providers pay in health care-related taxes, intergovernmental transfers, and certified public expenditures can be seen as a cost that effectively reduces the initial payment (i.e., gross payment) these providers receive. MACPAC has recommended that Congress require collecting and reporting data on the sources of the nonfederal share necessary to determine net Medicaid payment at the provider level [21]. Without collecting and making these data publicly available, researchers will not be able to fully understand how Medicaid payment and financing policies affect net payment to providers and are linked to goals of access, quality, and efficiency in program spending. Medicaid provides coverage for health care and other related services to the nation's most economically disadvantaged people, including low-income children and their families, pregnant women, people over the age of 65, people with disabilities, and in some states, low-income adults. A key measure of the effectiveness of any health coverage program is whether it provides timely access to high-quality services. Given the effects of poverty and related socioeconomic factors on the health and ability of many Medicaid beneficiaries to obtain services, ensuring access to needed services is particularly important. In its June 2022 Report to Congress on Medicaid and CHIP, the Commission discussed a new Medicaid access monitoring framework with three key domains of access: (1) provider availability and accessibility, (2) use of services, and (3) beneficiary perceptions and experiences of care [22]. Current systems to monitor access have been insufficient due to their design, data availability, and state capacity constraints and do not capture all of the domains of access, most notably beneficiary experience. For example, many states rely on administrative data to monitor access and utilization of services, but this approach does not capture unmet health needs, barriers to care, beneficiary perceptions of care, or self-reported health status. Additionally, these data do not always capture all relevant demographic or socioeconomic measures. Further, measures of provider availability have overly relied on structural measures such as network adequacy or time and distance standards and have not also assessed accessibility, such as the availability of translation and interpretation services for beneficiaries with limited English proficiency and accommodations for individuals with disabilities. Beneficiaries' perceptions of their needs, barriers to care, and care experiences are important components of access which are not available in administrative data. Beneficiary perspectives can inform connection to the health care system, timeliness of care, barriers to care, unmet needs, and culturally competent care [23]. It is also important to understand beneficiaries' perceptions of interactions with providers (including being treated with respect and without bias), beneficiary knowledge and understanding of benefits, stability of care, and perceived quality of care. Understanding beneficiary experience with their care is an area in need of further research. To date, there is no federal, coordinated data collection and analysis effort. While states and managed care organizations may have their own approaches for collecting information on beneficiary experience, data collected at the state and plan level lack generalizability do not permit state-to-state comparisons, and typically are not reported to the CMS. For example, a number of states use the Consumer Assessment of Healthcare Providers and Systems, a survey designed to assess patient experiences with health care, including access to care. However, these surveys only capture information about beneficiaries covered by managed care plans, and data are only available from plans that voluntarily submit to the Consumer Assessment of Healthcare Providers and Systems database. In 2022, MACPAC recommended that CMS field an annual federal Medicaid beneficiary survey to consistently collect information across all states to measure unmet need, barriers to care, knowledge of benefits, and how beneficiaries perceive they are being treated [22]. States engage in myriad activities to monitor beneficiary perceptions and experiences; however, the comparability of findings may be limited depending on the methods used because the collection of these data is not universal. A federal Medicaid survey would provide consistent data across states and relieve some state administrative burden. The 2024 managed care rule requires states to conduct an annual experience survey of enrollees in managed care and does not propose to collect similar information in FFS [18]. As a result, the Commission raised concerns that the lack of comparable information for beneficiaries enrolled in FFS would limit the ability of states, CMS, and other stakeholders to compare access across delivery systems. In the Commission's view, standardizing a core set of access measures would help ensure comparability across states, but states should have the flexibility to add additional measures to meet their priorities. To reduce administrative burden, it would be helpful for CMS to capitalize on existing efforts [24]. Based on rules in effect as of February 2025, states will be required to develop and conduct an annual enrollee experience survey for each managed care plan beginning in 2027, which will add new and important data to state and CMS efforts to monitor access and quality, and identify areas in need of improvement from the beneficiary perspective (42 CFR 438.66(b)(4) and (c)(5)). In establishing the enrollee experience survey requirement, CMS declined to establish a standard survey instrument or define minimum survey characteristics to provide states with flexibility to select the survey design that will best aid in their specific monitoring, oversight, and quality improvement activities. Although states will be required to submit summary results from the enrollee experience survey to CMS through the Managed Care Program Annual Report (42 CFR 438.66(e)(2)(vii)), it is unclear if those data will provide a national picture of beneficiary experience with the program. The knowledge gaps that we identify represent challenges for MACPAC and other researchers in providing timely, relevant analyses for the health policy community. As these gaps are addressed, data and information should be made broadly available to the research community. Given the breadth and complexity of the Medicaid program, MACPAC relies on empirical, scientific literature driven by a robust research community to develop the evidence needed to make policy recommendations. As we move forward into the next 60 years of Medicaid, we encourage CMS to work closely with the research community to find ways to reduce the burden of obtaining and analyzing data and further analyses of this vital program. Although the cost of accessing Medicare and Medicaid data may be comparable, the cost of obtaining a sufficiently large Medicaid sample for state-level analyses may limit the ability of many researchers to independently conduct research on the Medicaid program. The authors declare no conflicts of interest.