Although it is clear that data collection and dissemination within the Centers for Medicare and Medicaid Services (CMS) will accelerate rapidly in the coming years, CMS already holds vast amounts of information, much of which can be accessed by appropriate outside organizations. This appendix briefly reviews some of the current sources of data within CMS and some of the details of data related to improvement of the quality of care, along with discussion of some of the external, secondary uses of these data. The emphasis is largely on the data collected with respect to Medicare.
DATA SOURCES AND STRUCTURE
The data held currently by CMS come from a variety of sources, of which the most important is claims for all types of services provided. All claims contain basic diagnostic information, as well as information on date of service, the type of service provided, and the identity of the prescribing physician. Some types of data, such as hospital discharges, include multiple diagnoses, as well as a record of procedures performed, diagnosis-related group (DRG) assigned, and other information on the hospital stay. Managed-care plans serving Medicare beneficiaries (Part C of Medicare) are required to submit extensive “benefit utilization” reports, which provide encounter data for these beneficiaries very similar to data from claims submitted on behalf of those in fee-for-service Medicare.
This information is merged into the fee-for-service data sets to generate a comprehensive view of facts such as hospital discharges.
Part D of Medicare is administered by pharmacy benefit organizations that pay the claims rather than CMS’s doing so directly. Part D providers are required to submit detailed reports of the drugs prescribed as well as identifying the prescriber and the pharmacy where each prescription is filled.
Special supplemental information aimed at both monitoring quality and assigning patients to payment groups is collected on nursing home patients (the Minimum Data Set, or MDS), home health patients (the Outcome and Assessment Information Set, or OASIS), patients in rehabilitation facilities (Inpatient Rehabilitation Facility-Patient Assessment Instrument, or IRF-PAI), and those in psychiatric facilities (Inpatient Psychiatric Facility Prospective Payment System, or IPF PPS).
Quality information is collected from surveys done by the Joint Commission and state agencies and entered into the Online Survey Certification and Reporting (OSCAR) database. Since 2007, hospitals are required to report “never events,” adverse outcomes that ought not to have happened (such as wrong-site surgery), as defined by the National Quality Forum. Payment is withheld for the hospital stays during which such events occur. Additional quality reporting is also required. The amount of reporting is expected to increase as electronic medical records (EMRs) come into common use; ease of reporting should improve as well. Physicians are not required to submit quality information but have been encouraged to do so under a voluntary plan that can lead to incentive payments. Physicians’ quality reports are based on measures developed by the National Quality Forum. End-stage renal disease facilities report patient outcome measures into a system known as CROWNWeb, for Consolidated Renal Operations in a Web-enabled Network.
CMS also conducts a number of beneficiary surveys. The Medicare Current Beneficiary Survey (MCBS) is a rolling survey of beneficiaries that includes questions on out-of-pocket costs, services used, and the experience of care. The Health Outcomes Survey (HOS) measures outcomes for individuals enrolled in Medicare managed care. A survey specific to patients’ experience of hospital care—Hospital Consumer Assessment of Healthcare Providers and Systems (HCAHPS)—is also conducted.
Demographic information on beneficiaries, including race and ethnicity data, is provided to CMS by the Social Security Administration (SSA). According to recent reports, race and ethnicity information are now being collected using OMB standards when new Social Security and Supplemental Security Income claims are filed, and when applications are made
for Social Security numbers and replacement cards. However, until 19801 very limited data on race and ethnicity (white, black, or “other”) were collected at the time of enrollment. While the categories have expanded since then, data available for Hispanic/Latino and Asian beneficiaries remain of limited accuracy despite efforts to repopulate SSA data received through focused outreach efforts, arrangements with the Indian Health Service and the collection of self-reported data through the Medicare Current Beneficiary Survey.2
As a result, accurate demographic information is often lacking on Medicare beneficiaries, posing a challenge in terms of identifying and reducing racial disparities in health and treatment. Demographic information on Medicaid beneficiaries is collected by states under a variety of rules.
Information on providers and institutions participating in the Medicare program, including information on the ethnicity of providers, is collected at the time of request for a Medicare number.
The new reporting rules for meaningful use of electronic health records (EHRs) should yield substantial new data that can easily be accessed, but the exact specifications for the information to be generated are not yet determined.
CMS also collects and retains extensive cost information based on regular reports submitted by participating facilities.
The Office of Research, Development, and Information (ORDI) in CMS is responsible for coordinating the agency’s research. ORDI conducts research projects such as those that have served as the basis for the design of the new accountable care organizations. The data sets collected in the process of research are usually held by ORDI until the demonstration is complete and analysis has been concluded.
CMS also maintains a number of data sets related to Medicaid. The Medicaid Analytic Extract (MAX) files track utilization and enrollment data at the person level on an annual basis. This data set, according to CMS, is used to support research and policy analysis for Medicaid and other low-income populations. The Medicaid Drug Rebate Product Description file is a catalog listing of all pharmaceuticals that qualify for drug manufacturer rebates to states under the Medicaid Drug Rebate Initiative (MDRI) (part of the Omnibus Budget Reconciliation Act of 1990
1 Joel S. Weissman and Romana Hasnain-Wynia, 2011, “Advancing Health Care Equity Through Improved Data Collection,” New England Journal of Medicine 364:2276-2277.
2 RTI International, 2008, “Creation of New Race-Ethnicity Codes and Socioeconomic Status (SES) Indicators for Medicare Beneficiaries” available at http://www.ahrq.gov/qual/medicareindicators/, last accessed October 24, 2011; AHRQ “Race, Ethnicity, and Language Data: Standardization for Health Care Quality Improvement,” available at http://www.ahrq.gov/research/iomracereport/reldata3b.htm, last accessed October 25, 2011.
[Public Law 101-508]). This file is updated and maintained by CMS on a quarterly basis. Supplementing the MDRI is the Medicaid Drug Rebate Utilization file, which captures drug utilization and vendor payments data submitted to CMS by the states in order to calculate state reimbursement amounts.
The Payment Error Rate Measurement is an annual compilation of error rates in payments to states for Medicaid and the Children’s Health Insurance Program (CHIP) services.
CMS’s Central Data Administration team is the primary custodian of most CMS data. Quality reporting and support of the quality databases are the responsibility of the Office of Clinical Standards and Quality. According to CMS, the process of data administration itself consists of the following:
• Guiding the creation and monitoring the usage of data and information as vital agency assets;
• Promulgating agency standards, procedures, and guidelines related to data names and definitions;
• Maintaining the inventory of agency data assets;
• Facilitating understanding of the meaning, accuracy, and timeliness of data assets; and
• Promoting the reuse of standardized data names, definitions, elements, and values.
A few CMS databases are readily available to the public; many are available to researchers, with appropriate restrictions related to the privacy of individuals; and a few, most notably those collected by quality improvement organizations (QIOs) for their improvement work, are never available for secondary use. In order to facilitate effective outside use of CMS data, CMS has contracted with the University of Minnesota to create the Research Data Assistance Center (ResDAC). ResDAC provides free assistance to researchers interested in using Medicare data for their research; it maintains a comprehensive list of the data sets available and when the next update is expected. All requests must be reviewed by ResDAC prior to submission to CMS; this requirement reduces rework by inexperienced researchers and ensures an efficient process for review.
For users who require more consistent access to CMS data, such as government and state agencies and providers, an active CMS Data Use Agreement can be established, stipulating the manner and time frame in which the data are to be used. Interested external data customers and stakeholders that may wish access to data include the following: academic institutions and the private sector, congressional entities, Department of Health and Human Services (HHS) federal agencies and contractors, non-
HHS federal agencies, providers, state government agencies, and state Medicaid agencies.
CMS organizes its data sets with different levels of specificity and beneficiary personal information in order to facilitate research. Data are maintained in identifiable data files, which contain actual beneficiaryspecific and physician-specific information, such as per year person-level enrollment and utilization. Accessing this class of data requires authorization from CMS and is subject to Health Insurance Portability and Accountability Act (HIPAA; Public Law 104-191) protections. Limited data files are files that have been stripped of data elements that might permit the identification of beneficiaries but which include beneficiary-level health information. Accessing this class of data also requires authorization from CMS and is also subject to HIPAA protections. Non-identifiable data files and public use files are accessible to the public and are not subject to CMS authorization or HIPAA protections, as they have been stripped of all individual-identifying information. Cost report data coming from all Medicare program providers is provided on an annual basis and contain information on costs, statistics, and facility characteristics. Medical review data refer to a number of quality-of-care assessment reports by facility such as the MDS. This information includes personal-level specific data on facility residents and other topics, and accessing it is subject to CMS authorization and HIPAA protections.
Demonstration data on “the likely impact of new methods of service delivery, coverage of new types of service, and new payment approaches on beneficiaries, providers, health plans, states, and the Medicare Trust Funds,”3 as well as CMS’s evaluation projects data validating research and providing useful information for monitoring CMS’s various programs, may contain identifiable, limited-information, and/or non-identifying information. Accessing demonstration data may or may not warrant CMS approval or HIPAA protections, depending on the data collected. Consumer assessment data such as those collected in the MCBS and the HOS are in the form of survey responses from beneficiaries as consumers on the interpersonal aspects of health care. Some are available with CMS approval, and some are fully de-identified and are available without special approval.
DATA-CENTRIC EFFORTS TOWARD QUALITY IMPROVEMENT
A regulatory and payment agency like CMS has two potential approaches to improving the quality of care: (1) it can establish standards aimed at eliminating the worst care, and (2) it can encourage, in some
manner, overall improvement. If quality is viewed as a normal curve, the first can be seen as limiting the length of the left tail, whereas the second involves shifting the entire curve to the right. Despite a popular focus on “getting the bad guys,” overall improvement can have a much greater effect on more individuals.
Conditions of Participation
Since passage in 1965 of the legislation creating Medicare (Public Law 89-97), CMS, or its predecessor agencies, have had responsibility for monitoring the quality of care in part of the health care system. The focus of the original law was exclusively on establishing minimum standards for institutional providers, particularly hospitals and nursing homes. Regulatory standards, known as “Conditions of Participation,” were developed, and physical surveys were conducted to ensure that standards were met. These surveys are conducted by state agencies and, in many instances, by private accrediting agencies whose standards are deemed to be equal to or better than those of the Conditions of Participation. The only accrediting agency mentioned in the original law and used in the early years was the Joint Commission on Accreditation of Hospitals (now the Joint Commission on Accreditation of Healthcare Organizations); more recently other agencies have also qualified for “deemed” status.
Whoever conducts the survey, the core question is straightforward: Is care good enough for CMS beneficiaries? In theory, CMS can act to withdraw all payments if care is substandard; in practice, it has proved very difficult to de-certify a whole institution, even one with serious problems. Although the Conditions of Participation have attempted to improve overall care by means such as requiring quality improvement committees, there is little evidence that this approach makes a difference.
In recent years, CMS has moved, even within the context of the Conditions of Participation, to deal with substandard care in a more precise manner on the basis of objective data. Facilities are now required to report to CMS a list of “never events,” and payment is denied whenever these occur.
Quality Improvement Organizations and Predecessor Organizations
The first effort to use data sets to improve care came with the creation of professional standards review organizations (PSROs) in 1972; these physician-run organizations, each of which covered a state or smaller area, had access to Medicare claims data and were expected both to reduce overuse of services and to improve quality. Evaluations of the program, and interest from Congress, focused almost exclusively on
whether or not the PSROs were saving money, with little attention to the matter of quality. In the 1980s, the PRSO program was eliminated, and new entities—professional review organizations (PROs)—were created, with more of a focus on quality. The new law allowed more flexibility in terms of what kinds of organizations could qualify to perform reviews and what areas they would cover. Tasks to be carried out included not only data-based efforts to improve care but also a number of less relevant activities such as the investigation of patient complaints. The name PRO was changed to QIO in 2002 to emphasize further the focus on improving population-based measures of health, but the conflicting tasks remain. A study by the Institute of Medicine (IOM) completed in 2006 argued for even more focused tasks for the QIOs. It is not clear how fully those recommendations have been followed. The conclusions of the IOM study were as follows:
• The quality of the health care received by Medicare beneficiaries has improved over time.
• The existing evidence is inadequate to determine the extent to which the QIO program has contributed directly to those improvements.
• The QIO program provides a potentially valuable nationwide infrastructure dedicated to promoting quality health care.
• The value of the program could be enhanced through the use of strategies designed to focus the QIOs’ attention on the provision of technical assistance in support of quality improvement, to broaden their governance base and structure, and to improve CMS’'s management of related data systems and program evaluations.4
EXTERNAL SECONDARY USES OF CMS DATA
The various data sets described here, particularly those produced by Medicare, have served as a rich base for health services research in the United States. Among current researchers, the best known is probably the Dartmouth Atlas of Health Care,5 which began with studies of variation in hospital use and now reports on a wide variety of issues such as the number of individuals who see 10 or more doctors in the last 6 months of life. The Dartmouth files have been maintained continuously for 20 years, and so patterns of use can be traced back over time. The original research that led to the development of DRGs was also conducted by
4 Institute of Medicine, 2006, Medicare’s Quality Improvement Organization Program: Maximizing Potential, Washington, D.C.: The National Academies Press, p. 4.
outside researchers using Medicare Provider Analysis and Review (MEDPAR) data. The DRG system, which was originally intended to facilitate utilization reviews by identifying aberrant cases, or outliers, was first reported 35 years ago6 and first used as a payment mechanism in New Jersey in 1980.
The information available today is much richer than that available when the Dartmouth Atlas first appeared, and research interest continues to grow. The value of CMS data is limited by the fact that they reflect only care delivered to Medicare and to a certain extent Medicaid beneficiaries. The availability of more universal data will serve to enhance greatly the understanding of the functioning of the U.S. health care system.
6 John D. Thompson, Robert B. Fetter, and Charles D. Mross, 1975, “Case Mix and Resource Use,” Inquiry 12(4):300-312.