記住我
Medicare is a large federal health insurance program covering 65 million beneficiaries.1 In the United States, people ages 65 years and older or with certain disabilities or end-stage renal disease are eligible for Medicare coverage. Since these factors are also associated with health status and health outcomes, Medicare beneficiaries and their families frequently face decisions about the use of health care treatments and services.
The economic impacts of health care decisions on patients and their families were recently identified as a patient-centered outcome in the 2019 reauthorization of the Patient-Centered Outcomes Research Trust Fund. Patient-centered outcomes research (PCOR) is designed to generate evidence about the health and economic outcomes that matter to patients and the people who care for them, with a goal of making better-informed decisions. Linking and analyzing Medicare billing data is one approach for studying the outcomes and effectiveness of treatments and services. Given the large number of Medicare beneficiaries and their high use of health care services, studies that use Medicare data can quickly identify associations between health services use and measures of effectiveness, as well as some of the potential economic impacts of treatment on patients, families, and communities.
For over a decade, clinicians, policymakers, and researchers have begun to emphasize “patient-centered” care as a cornerstone in improving health outcomes and quality of care.2,3 Additional data on outcomes and measures important to patients could help them to make better health care decisions with their families and providers. However, realizing this potential requires evidence, research, and data to facilitate this focus. Although PCOR is growing, it remains limited in part due to available data sources. For the Medicare population of older adults, Medicare program data present a rich potential option for conducting studies, although health care claims are limited in scope when used as the sole source in research.
To explore how Medicare data may be used in certain types of PCOR research, this paper inventories and evaluates linkages of federally funded survey and administrative data to Medicare fee-for-service (FFS) claims, for conducting PCOR on economic outcomes. Linked Medicare data holds tremendous potential for researchers to conduct economic analyses that can be combined with evidence from other studies to support decision-making. While Medicare data can only be used to investigate a subset of economic outcomes questions relevant to patients, Medicare data serve as an available resource to quickly launch multiple economic studies that explore a wide range of conditions, populations, and interventions. When linked to external sources, studies using Medicare data can expand the evidence base where information is lacking and over time be followed up with additional studies using other methodologies and data sources.
Medicare FFS claims (hereafter, “claims”) hold several advantages for researchers seeking to study the health and health-related economic outcomes of Medicare beneficiaries.4–6 These data capture detailed information about the types of health care received; payments for that care; associated health conditions; and basic patient demographics.7 Although several types of patient-level Medicare claims files exist, we focus on Medicare FFS claims for several reasons. First, our PCOR interest emphasizes economic outcomes, including Medicare’s costs of treatment and selected costs incurred by patients. FFS data contain charged, allowed, and paid amounts and data on coinsurance and copayments. Medicare Advantage (Part C) “encounter” data do not capture service-level payments, although both FFS and Medicare Advantage data track utilization of specific health services and basic patient demographics.8 Second, Medicare FFS are widely available, with best practices and limitations established over >2 decades of research, compared with Medicare Advantage encounter data, which have only recently been released to the research community.9 Third, FFS was historically the largest share of Medicare, although 2022 enrollment in Medicare Advantage10 nearly equals FFS.
There are also limitations on the information available within all types of Medicare claims. These data contain limited clinical information and exclude noncovered services, lab values, informal caregiving, supplemental insurance payments, over-the-counter medications, social circumstances, and patient experiences. A complete patient-centered perspective on economic costs includes a wide range of economic outcome measures, many of which are not available in Medicare claims. To help address this, data linkages may enable researchers to add economic outcomes from multiple data sources, improving the responsiveness, relevancy, and breadth of patient-centered economic research that may be conducted.
ECONOMIC OUTCOMES RELEVANT FOR PCOREconomic outcomes may be directly or indirectly associated with health and health care and include a wide range of costs and financial impacts.11 In a previous report, 3 domains of economic outcomes relevant to PCOR were defined and related to different stakeholders12,13: direct medical costs [tied to the receipt of health care services for a patient (eg, cost-sharing for a doctor visit)]; direct nonmedical costs [associated with health care or illness, but not part of the treatment (eg, transportation to a clinic)]; and indirect costs [resources lost from illness or treatment (eg, missed work time for a doctor visit)].14 Patient-centered care decisions may be informed by economic outcomes, but noneconomic factors may also be important considerations for patients and families. Researchers also include quality of life or expected length of life as aspects of PCOR. While these outcomes are tremendously important, the valuation of such outcomes is controversial.14 Further, because these have been studied in other work and the expanded definition of PCOR in the 2019 reauthorization15 does not explicitly list such costs, the present study focuses on economic outcomes.
SCOPE AND OBJECTIVESThis paper reviews Medicare claims-based data linkages for conducting PCOR on economic outcomes, taking traditional FFS claims in the CMS Research Identifiable File (RIF) format as a starting point. The richness of linked data sources in general is well known to health services researchers,16–19 and a previous review12,13 identified linkages as a key strategy for future research. Medicare claims linkages are salient for several reasons. First, CMS has an established research data infrastructure: the Research Data Assistance Center (ResDAC), the Chronic Conditions Data Warehouse (CCW), and the CCW Virtual Research Data Center (VRDC). Second, Medicare is the largest single payer of health care in the United States, and its expenditures are projected to continue growing.20 Third, Medicare covers an older population with more health care needs, chronic conditions, and higher costs than in younger populations.20 Given these factors, research using linked Medicare claims can facilitate meaningful PCOR and economic evaluations for the Medicare FFS population.
This paper has 3 primary objectives:
To guide researchers on data linkages between Medicare FFS claims and other federally funded administrative and survey data sources that capture information on 1 or more economic outcome domains. To identify the specific economic outcome measures with each data linkage. To illustrate gaps that remain for economic outcomes for the Medicare population to inform priorities into additional Medicare data linkage efforts. METHODSWe followed 3 main steps during April–August 2022 to develop and extract the catalog of eligible data sources capturing economic outcomes relevant to PCOR. Data inclusion criteria were: (1) federally funded; (2) captures at least 1 economic outcome domain relevant for PCOR; (3) linked or linkable to Medicare claims at the individual level; and (4) available to external researchers. See Appendix Table A-1 (Supplemental Digital Content 1, https://links.lww.com/MLR/C686) for the relevant perspectives and examples of economic outcome measures which may be captured under each domain. (Given the large number of datasets surveyed, we do not list all specific variables from each dataset.) The study was nonhuman subjects research, and institutional review board review did not apply.
First, we assessed all data sources in a previous federal report on economic outcomes,12,13 retaining those which are linked or linkable to Medicare claims. Next, we reviewed the websites of all divisions of the US Department of Health and Human Services (HHS) and of the Medicare and Medicaid Resource Information Center (MedRIC).21 We then performed a targeted search (PubMed, Google Scholar) for additional data sources or relevant peer-reviewed literature. Search terms are provided in Appendix Table A-2 (Supplemental Digital Content 2, https://links.lww.com/MLR/C687).
Second, we abstracted relevant information from each data source: major characteristics (name, steward, URL, source), sample size, economic outcome domains and measures, time frame, and access to Medicare linkage. We also rated the potential of each source for economic evaluations as: high (multiple economic outcome domains and multiple perspectives); medium (at least 1 of: multiple economic outcome domains, multiple perspectives, or multiple economic outcome measures within a domain); or low (all others).
Third, results were reviewed with a 9-member expert panel 3 times during May–September 2022. We included a diverse array of federal, academic, and nonfederal experts in PCOR, health economics, health services research, and health equity. We also conducted 9 individual interviews with additional experts (researchers, data stewards, and Medicare policymakers) to review the findings.
RESULTSWe identified 18 federally funded data sources (Table 1) which are linked or linkable to Medicare claims and which capture economic outcomes relevant to PCOR. Table 2 shows select characteristics of these data sources. Survey data sources (12 sources) are more prevalent than administrative data (6 sources).
TABLE 1 - Federally Funded Data Sources Currently Linked or Linkable to Medicare FFS Claims Data source name (Abbreviation), steward URL for linked data Source of data Sample size Available economic outcome categories Available economic outcome measures, by category Available time frame for linked data Outside researchers’ access to linked data Rating of linked data for PCOR economic evaluations Rationale for rating Medicare Part D claims (PDE), CMS https://resdac.org/cms-data/files/pde Administrative 48 million Direct medical costs Direct medical costs: Insurance premium*; OOP health care costs, Paid/reimbursed amount*With the exception of the SEER-Medicare data which are already in linked form, the economic outcome measures field for all other data sources represents the specific outcomes that are available independently in the data source versus via the linkage.
†Although publicly available information about the MEPS-Medicare data linkage is not available, a technical expert panel (TEP) panel member at AHRQ confirmed with the research team that external researchers can request for MEPS-Medicare linked data through the NCHS RDC.
DUA indicates data use agreement; FFS, fee-for-service; NA, not available; OOP, out-of-pocket; PCOR, patient-centered outcomes research; SEER, Surveillance, Epidemiology, and End Results.
Note: Findings are based on the 18 federal and federally funded data sources identified using the search process described in the Methods section. (Although a comprehensive scan was conducted, the exhibit may not represent all datasets that are linked or linkable to Medicare claims.)
*Totals may not add up to 100% due to rounding or nonmutual exclusivity of categories.
†Represents agencies at the United States Department of Health and Human Services (HHS).
‡Categories are not mutually exclusive. Medicaid and Part D claims files include both individual-level enrollment data and encounter/claim-level data.
§Categories are not mutually exclusive. The National Study of Caregiving has historically been a source of cross-sectional data. However, longitudinal data collection began in 2017, and will continue to be implemented going forward. Therefore, this data source was counted as both cross-sectional and longitudinal data.
∥Categories are not mutually exclusive. The Panel Study of Income Dynamics provided annual data from 1968 to 1997, and then changed to biennial after 1997. Therefore, this data source was counted as both annual and biennial data.
¶Values in “Other” include: (1) every 5 years (National Long Term Care Survey) and (2) 1973–74, 1977, 1985, 1995, 1997, 1999, and 2004 (National Nursing Home Survey).
#This field was populated based on 5 social determinants of health (SDOH) categories in a framework from the Agency for Healthcare Research and Quality (AHRQ) (https://www.ahrq.gov/sdoh/about.html).**These characteristics have been identified as priority populations in the Executive Order On Advancing Racial Equity and Support for Underserved Communities Through the Federal Government.
LGBTQ indicates lesbian, gay, bisexual, transgender, queer or questioning.
Within Medicare FFS claims, only direct medical costs are available, as paid amounts. Among the sources linked or linkable to Medicare claims, most (15 sources) also contain direct medical costs. Two data sources contain information on all 3 economic outcome domains—Panel Study of Income Dynamics (PSID)22 and Medical Expenditure Panel Survey (MEPS).23 Both have relatively few economic outcome measures in total within a broader domain, although dozens of individual variables are included to capture detail within an economic outcome measure (eg, out-of-pocket amount for physician services, out-of-pocket amount for inpatient care). Survey data sources also capture a wider range of economic outcomes than administrative data sources.
We also examined the specific economic outcome measures in each of the 3 domains (Table 3). Out-of-pocket costs and paid/reimbursed amounts are the most commonly available economic outcome measures both within the direct medical cost category and across all economic outcome measures, available in 11 and 10 data sources, respectively. The value of informal caregiving was the most common direct nonmedical cost (3 sources), and the value of absenteeism was the most common indirect medical cost (5 sources).
TABLE 3 - Economic Outcome Measures in Federally Funded Data Sources Linked or Linkable to Medicare Claims Economic outcome measure No. data sources % of data sources* Domain: Direct medical costs Paid/reimbursed amount 10 56 Out-of-pocket health care costs and/or medical expenses 11 61 Insurance premiums 8 44 Domain: Direct nonmedical costs Transportation and/or travel costs associated with seeking medical care 2 11 Vehicle modification expenses 0 0 Paid professional care (child care expenses, senior care expenses, and/or housekeeping expenses) 1 6 Relocation/moving costs 0 0 Specialized clothing/laundry costs 0 0 Value of time spent in seeking health care 2 11 Value of informal caregiving 3 17 Special food 2 11 Home modifications 2 11 Domain: Indirect medical costs Value of absenteeism 5 28 Value of presenteeism 2 11 Value of time spent in home production and leisure 4
留言 (0)