Updated Overview of the SEER-Medicare Data: Enhanced Content and Applications (2024)

Search for other works by this author on:

Article history

Abstract

In 1991, the National Cancer Institute (NCI) and the Centers for Medicare and Medicaid Services (CMS) collaborated to link NCI’s Surveillance, Epidemiology, and End Results (SEER) cancer registry data with Medicare enrollment and claims data. Although they initially planned for this to be a one-time data linkage, the resulting SEER-Medicare database stimulated such great interest within the cancer research community that the NCI, CMS, and SEER registries decided that NCI should maintain the database and make it available to other researchers, after complying with confidentiality requirements. Now, almost 30 years later, the SEER-Medicare database continues to be a major resource to assess cancer care and outcomes in the United States.

The continued utility of the SEER-Medicare database comes from the large number of included cancer cases, the availability of detailed tumor characteristics, the population-based nature of both data sources, the longitudinal nature of the Medicare data, the breadth of services included in the Medicare data, and the biennial linkage updates. The resulting, invaluable database has been used for an array of epidemiological and health services research studies, producing more than 1800 peer-reviewed publications between 1993 and 2018 (1).

In 2002, NCI and its collaborators published an initial overview of the SEER-Medicare database (2). The information available through the SEER-Medicare linkage has subsequently been enhanced with each update to include additional cancer cases and noncancer controls and their Medicare data as well as supplementary data files. This article provides a current snapshot of the SEER-Medicare database. With a focus on elderly Medicare beneficiaries (age 65 years or older), this article highlights the generalizability of the individuals included in the SEER-Medicare data and the available sample sizes by beneficiary demographics, cancer characteristics, location of service (eg, inpatient vs outpatient), and type of Medicare coverage (fee for service [FFS] vs managed care).

Materials and Methods

Data Sources

SEER Program

Cancer is a reportable disease in the United States (3); therefore, when health providers care for a patient with cancer they are required to report the cancer case to the state or regional cancer registry. Through the SEER Program, the NCI supports select population-based cancer registries in the United States to collect information about individuals with newly diagnosed (incident) cancer who reside within each registry’s catchment area. The SEER registries typically attain annual Gold Standard Certification from the North American Association of Central Cancer Registries, which is the highest certification level that requires a case ascertainment of at least 95% (4).

The SEER registries collect information on each reported incident cancer case within their catchment areas, including demographics (eg, age at diagnosis, sex, race, and ethnicity), date of cancer diagnosis (month and year), tumor characteristics (eg, histology, grade, and stage), number of primary tumors, vital status, and, if applicable, cause of death derived via linkage to the National Center for Health Statistics data (5). First-course therapy information is also collected. However, given that hospital medical records serve as the main source of information for the registry data, outpatient treatments (eg, systemic therapies such as chemotherapy and radiation) may be underascertained in some circ*mstances. The SEER registries do not currently collect or routinely release information about cancer screening, mode of cancer detection (eg, screening, symptomatic, or incidental), subsequent courses of therapy, disease progression, or recurrence.

The geographical coverage of the SEER Program has expanded over time (6). The SEER Program began January 1, 1973, and during the initial decade of operation expanded to include nine registries (SEER-9: Connecticut, Hawaii, Iowa, New Mexico, Utah, and the metropolitan areas of Atlanta, Detroit, San Francisco-Oakland, and Seattle-Puget Sound), which collectively covered approximately 10% of the US population (7). In 1992, the SEER Program expanded to SEER-13 with the inclusion of the Alaska Native Tumor Registry, rural Georgia, Los Angeles County, and San Jose-Monterey registries. Then, in 2000, five additional registries (Kentucky, Louisiana, New Jersey, and the remainder of California and Georgia) were added to create SEER-18, which increased the population coverage to approximately 28%. Finally, in 2018, SEER-21 was formed with the inclusion of the Idaho, Massachusetts, and New York registries, increasing coverage to approximately 35% of the US population. Compared with the total US population, the population covered by the SEER Program is more racially and ethnically diverse and includes more people who are economically disadvantaged (7,8).

Medicare Data

Medicare is a federally funded health insurance program administered by the CMS and insures the vast majority of the elderly population; approximately 94% of the US population age 65 years or older was enrolled Medicare in 2010 (9). Individuals are eligible for Medicare if they are 65 years or older and have sufficiently contributed to the system through payroll taxes either directly or indirectly (eg, via a spouse). Individuals younger than 65 years are eligible if they are long-term disabled, have end-stage renal disease, or have amyotrophic lateral sclerosis. Approximately 15% of Medicare beneficiaries are younger than 65 years (10).

Medicare-eligible individuals are entitled to Part A (inpatient hospital stays, skilled nursing facility stays, some home health visits, and hospice care) and can opt to add Part B coverage (physician visits, outpatient services, preventive services, and some home health visits) by paying an annual premium. More than 90% of Medicare beneficiaries enroll in both Parts A and B (10). Original Medicare reimburses providers for care provided on an FFS basis. Medicare Part C, also known as Medicare Advantage or managed care, is an alternative to FFS that allows individuals enrolled in both Parts A and B to join a privately managed care plan such as a health maintenance organization or preferred provider organization. Reimbursem*nt of care provided to beneficiaries enrolled in managed care plans is commonly based on a monthly capitation fee per beneficiary. In 2018, roughly 34% of all Medicare beneficiaries were enrolled in managed care (11). Medicare Part D, the prescription drug benefit, began in 2006 and is an additional optional coverage that beneficiaries can elect to receive. In 2018, approximately 72% of eligible Medicare beneficiaries were enrolled in Part D (11). Although the percentage of beneficiaries enrolled in managed care and/or Part D varies geographically, in general these percentages are increasing.

SEER-Medicare Linkage

Study Populations

Medicare Beneficiaries

Cancer Patients. Individuals included in the SEER cancer registry data are linked to their Medicare enrollment data through a deterministic algorithm based on social security number (SSN), name, sex, date of birth, and, if applicable, date of death (12). These variables are reported in the SEER and Medicare data, and SSN and name are removed from all files on completion of the linkage. Historically, during each linkage update, approximately 96% of individuals aged 65 years or older in the SEER data have been matched to Medicare enrollment data. Depending on age at cancer diagnosis and enrollment in Medicare, cancer cases included in the linked data may have Medicare data that predate their cancer diagnosis (eg, enrolled in Medicare at age 65 years and diagnosed with cancer at age 70 years) or may only have Medicare data several years after diagnosis (eg, diagnosed with cancer at age 60 years and enrolled in Medicare at age 65 years).

The initial year of cancer diagnosis data included in the SEER-Medicare database varies by registry, given the SEER Program expansions. The SEER-9 registries contribute data on cancer cases diagnosed as far back as 1973. The Alaska Native Tumor Registry, rural Georgia, Los Angeles County, and San Jose-Monterey registries contribute data on individuals diagnosed with cancer since 1992; the registries that cover Kentucky, Louisiana, New Jersey, and the remainder of California and Georgia contribute data on individuals diagnosed with cancer since 2000. Finally, the Idaho, Massachusetts, and New York registries were added after the most recent linkage update in 2018 and will be included in future linkages.

The most recent SEER-Medicare database update, which was released at the end of 2018, included cancer cases diagnosed through 2015. The 3-year delay between cancer diagnosis and inclusion in the SEER-Medicare database is due to a 2-year lag time for complete case reporting to the SEER Program plus one additional year to complete the linkage process.

Noncancer Controls. The SEER-Medicare files also include enrollment and Medicare claims for a random sample of Medicare beneficiaries who are not included in the SEER data. These beneficiaries serve as noncancer controls. To be included as a noncancer control, beneficiaries must reside in a SEER area at the time of a linkage update and must be included in the Medicare 5% random sample (13). Once a beneficiary is included as a SEER-Medicare noncancer control, they are retained even if they subsequently move to a non-SEER area. If a noncancer control is subsequently reported to a SEER registry as having been diagnosed with cancer, their information is updated and included with the cancer case data; there is an indicator variable to identify individuals diagnosed with cancer who are also included in the Medicare 5% random sample.

Confidentiality Considerations

Protecting the identities of the cancer patients and noncancer controls included in the data is paramount to NCI, CMS, and the SEER registries. As a result, personally identifiable information that is used to link the SEER and Medicare data (eg, SSN, names, and full date of birth) is not included in the SEER-Medicare database. Instead, each beneficiary is assigned a unique, nonidentifiable number, which allows the same individual to be tracked across SEER-Medicare data files and time. Linkage to additional data resources on an individual beneficiary level by database users is not allowed. Additionally, all requests for SEER-Medicare data must adequately describe the research question and must detail how the data will be stored and protected (14). Only the minimum data needed to conduct the outlined research question are provided. All requests must be approved by NCI and a representative of the SEER registries. Knowledge of beneficiary residence can also increase the possibilities of reidentification; therefore, residential ZIP codes and census tracts are encrypted in the SEER-Medicare data. Access to the unencrypted ZIP codes and census tracts is restricted and only provided if there is strong justification and increased data security measures are in place. All restricted-variable requests must be approved by NCI and all SEER registries.

As described in more detail below, Medicare claims data also include unique identifiers for both institutional (eg, hospitals and practices) and individual (eg, physicians, physician assistants, and nurse practitioners) health-care providers and their ZIP codes. To similarly protect the identities of the included providers, all provider identifiers and their ZIP codes are encrypted in the SEER-Medicare data. Encryption is completed in the same manner across files, which allows linkage of providers across different SEER-Medicare files. Access to unencrypted institutional provider identifiers and their ZIP codes must be approved by NCI and each SEER registry. Access to unencrypted individual provider identifiers (eg, physicians, physician assistants, and nurse practitioners) is not permissible.

Data Included

The SEER-Medicare database consists of multiple different sets of files that are linkable via the unique, nonidentifiable number assigned to each included Medicare beneficiary (eg, cancer patient and noncancer control). The various file types are summarized below and in Table1.

Denominator Files

SEER-Medicare provides researchers with three denominator files that may be used alone or in combination to identify a study cohort(s) of interest:

1. Patient Entitlement and Diagnosis Summary File (PEDSF) is a person-level file that contains all individuals found in both the SEER and Medicare data (eg, cancer patients). The PEDSF includes SEER demographic and clinical information for up to 10 primary cancer diagnoses as well as Medicare demographic, entitlement, and enrollment information for each included individual.

2. Summarized Denominator (SUMDENOM) file includes the Medicare demographic, entitlement, and enrollment information for the noncancer controls (ie, individuals who are included in the Medicare 5% random sample who resided in the SEER catchment areas but do not have a cancer diagnosis and, therefore, are not included in the SEER data). The individuals included in the SUMDENOM file are mutually exclusive from those included in the PEDSF file.

3. 5% All Cancer Diagnosis File is a subset of the PEDSF that only includes cancer patients who are included in the Medicare 5% sample (15). Researchers can use information from this file to recreate a complete 5% sample by adding these individuals back into the SUMDENOM file.

Medicare Claim Files

From the beginning of the SEER-Medicare linkage, five types of Medicare files have been included in the SEER-Medicare database to capture FFS claims from hospitals (MEDPAR), outpatient facilities, physicians and suppliers (Carrier), hospice care, and home health agencies (HHA). Over the years, in response to changes in CMS data structure and benefits, the SEER-Medicare database has been enhanced to include two additional file types capturing claims from: 1) durable medical equipment providers, starting in 1994, and 2) Part D Prescription Drug Event (PDE) claims, starting in 2007, which allowed 1 year of onboarding to the prescription drug coverage following the introduction of Part D in 2006.

Although the specific variables included vary across these seven claim file types, each Medicare claim file includes the unique patient identifier and typically includes date(s) of service, unique health-care provider identifiers (eg, institutional and individual National Provider Identifiers [NPIs]), diagnosis codes, procedure codes, amount charged, and amount reimbursed. All health-care providers, defined by the Health Insurance Portability and Accountability Act of 1996 (16), submitting medical claims are required to obtain an NPI. All provider numbers included in the SEER-Medicare database are encrypted. As noted above, a restricted-variable permission process is available whereby researchers can request access to unencrypted institutional provider NPIs; access is granted with sufficient justification and appropriate approvals. Release of unencrypted individual provider NPIs is not allowed because individual providers are considered protected entities in the database.

As a general strategy, SEER-Medicare users typically limit analysis to people “likely to have complete claims,” that is, individuals with equal months of Part A and Part B coverage and no managed care coverage. In contrast to FFS coverage, managed care providers are reimbursed via a per beneficiary capitation fee and are only required to submit patient-level claims for hospice and Part D services. Thus, all other care provided to beneficiaries enrolled in managed care plans will be incompletely captured in the Medicare claims data. Depending on the research question, some researchers additionally require equal months of Part D coverage.

Medicare Patient Assessment Files

Beginning with the 2018 linkage update, for cancer patients diagnosed in 2004 or later, the SEER-Medicare data include the Minimum Data Set (MDS) nursing home assessments back to 2011 and Home Health Outcome and Assessment Information Set (OASIS) data back to 2010. The inclusion of the MDS and OASIS data in the SEER-Medicare database is discussed in more detail elsewhere (17,18). Briefly, the MDS data include information from federally mandated, clinical assessments for all CMS-certified nursing home residents. The OASIS data include information from federally mandated, clinical assessments of adult Medicare patients receiving skilled home health care. Both assessments are conducted at mandated time points (eg, at admission, discharge, when health status changes, and, if applicable, at set time intervals while the patient is using the respective services) and gauge patients’ comorbidities and physical, psychological, and psychosocial functioning. The OASIS assessment further gauges patients’ health-care needs when using home health services (eg, hospice care, oxygen therapy, and physical therapy) as well as their living arrangements.

Ancillary Files

Since the original SEER-Medicare linkage, NCI has also created a set of ancillary files to allow SEER-Medicare database users to explore the associations between socioeconomic, geographic, and provider characteristics and health-care outcomes. These files are related to:

1. Ecologic measures of socioeconomic status. Neither the SEER data nor the Medicare files include individual-level socioeconomic characteristics, which have been demonstrated to be strongly associated with health-care use and outcomes (19,20). However, both data sources do include geographic information (eg, SEER: residential census tract and ZIP code at cancer diagnosis; Medicare: annual residential ZIP code and provider billing ZIP code). Therefore, through linkages with US Census Bureau data (1990 and 2000 Census and the 2008–2012 American Community Survey) (21), NCI has created SEER-Medicare linkable files that include aggregate socioeconomic measures (eg, median household income, education level) at the ZIP code and census tract levels without the need to request actual, unencrypted geographic variables. As noted above, ZIP codes and census tracts included in the SEER-Medicare database are encrypted for confidentiality concerns; access to the unencrypted data can only be obtained if sufficient justification is provided and approval is granted from the NCI and each SEER registry. The NCI has also created a crosswalk between encrypted ZIP codes and the Dartmouth Atlas Project hospital referral regions, which represent regional health-care markets for tertiary medical care that generally require the services of a major referral center (22).

2. Information on health-care providers. Numerous studies have demonstrated that health-care provider characteristics can influence patient treatment and outcomes. Health-care provider characteristics can be determined directly from Medicare claims or through linkable ancillary SEER-Medicare files (23,24). The SEER-Medicare Hospital File, which compiles information from the CMS Healthcare Cost Report (25) and the Provider of Service (26) surveys, includes information on hospital characteristics (eg, location, accreditation, number of beds, type of ownership, and available services). The Hospital Mergers and Acquisitions File allows the longitudinal tracking of hospitals that have changed ownership over time. Additionally, the NCI has created a linkable version of the CMS Medicare Data on Provider Practice and Specialty file, which includes provider-level specialty and utilization data. Further, the NCI has created the NPI-UPIN Crosswalk file to facilitate longitudinal tracking of individual physicians during the transition from Unique Physician Identification Numbers (UPINs) to NPIs, which occurred in 2007. Finally, NCI facilitates a process by which users can obtain physician characteristics from the American Medical Association data without needing access to unencrypted physician identifiers (27).

3. Assessment of patient comorbidity. In 2000, the NCI created a comorbidity index for cancer patients based on the Charlson index. This index was updated in 2014. NCI has made the macros for these claims-based comorbidity index algorithms available to researchers (28). Currently, these macros only identify comorbidities based on International Classification of Diseases ninth revision (ICD-9) diagnosis codes, which were submitted on claims until September 30, 2015; thereafter, ICD-10 codes were submitted. The NCI is in the process of modifying the SAS macros to incorporate ICD-10 diagnosis codes. The NCI has also made CMS’s Chronic Condition Flags, a set of Medicare claims-based flags for common chronic and potentially disabling conditions that are linkable at the individual level (29).

Statistical Analyses

This article presents descriptive statistics on the clinical and sociodemographic characteristics of persons included in the data and their health-care utilization. We included a comparison of characteristics of cancer patients in the SEER-Medicare data with individuals from the Medicare 5% sample who resided in a SEER area who were not known to have a cancer diagnosis (eg, noncancer sample). We then presented the number of patients with cancer by cancer site and indicate the percentage enrolled in FFS during the month of cancer diagnosis to better estimate the proportion of patients who will have available claims at the time of cancer diagnosis. Third, to demonstrate the extent and type of Medicare services used by cancer patients, we calculated the annual average number of claims per cancer patient in the year before diagnosis (eg, 12 months before cancer month of diagnosis), the year after diagnosis (eg, 12 months after the cancer month of diagnosis), and the 12 months before death (eg, 11 months before and month of death).

Medicare beneficiaries who received coverage due to disability (eg, those who enrolled before age 65 years) are not typically considered to be representative of the younger than 65-year-old general or disabled populations (because not everyone with a disability qualifies); thus, like most published SEER-Medicare analyses, we excluded these individuals from our analyses. Additionally, although the SEER-Medicare data include cancer cases diagnosed as early as 1973 and Medicare claims starting in 1991, we restricted all analyses to persons who were diagnosed with a first primary cancer between 2011 and 2015 (ie, the most recent 5 years of cancer data). The NCI comorbidity index was calculated for individuals who had sufficient claims data, defined as 12 months of continuous enrollment in FFS Parts A and B in the year before cancer diagnosis (28).

Results

Table2 summarizes the characteristics of the cancer patients and noncancer controls who were 65 years or older at first inclusion in the SEER-Medicare data between 2011 and 2015. During this 5-year time period, 857 056 cancer patients and 469 591 eligible noncancer controls were added to the database. Overall, the included cancer patients were slightly more likely to be older, male, and white and had higher comorbidity levels than the noncancer controls. Variations by Medicare enrollment type (Parts A, B, C, and D) were also observed, likely because the controls were younger and healthier. The controls were slightly more likely to be full Medicare-Medicaid dual eligible, indicating lower socioeconomic status among the controls. The geographic distribution was very similar between the two groups. Notably, approximately 40% of the included beneficiaries were California residents. The vast majority of the included cancer patients had only one primary cancer diagnosed between 2011 and 2015; however, a sizable number, almost 60 000 individuals, had at least one additional new primary cancer diagnosed by December 31, 2015.

Overall, approximately 60% of the included cancer patients were enrolled in FFS Parts A and B during the month of cancer diagnosis and, thus, are likely to have claims data available for study (Table3). Enrollment in FFS plans decreased during the study period from 63.2% to 58.3% and was relatively uniform across cancer sites (range: 57.6% [prostate]–63.2% [lung and bronchus]).

Health-care services provided to the included cancer patients enrolled in FFS Parts A and B varied over the cancer continuum (Table4). For example, in the year before cancer diagnosis, less than one-third of cancer patients had durable medical equipment, home health services, or in-patient hospitalization and only 1% had a hospice claim. Almost all cancer patients (95%) had at least one claim from an individual provider, and the vast majority had at least one outpatient clinic claim (74%) and a prescription drug claim (59%). Health-care use was higher during the 12 months after cancer diagnosis compared with the 12 months before diagnosis. In particular, the frequency of hospitalizations and durable medical equipment claims doubled (from 21% to 44% and 10% to 21%, respectively), and hospice utilization increased exponentially from 1% to 17%. The annual average number of service claims per cancer case was also higher postdiagnosis for hospitalizations (from 1.97 to 2.19), outpatient clinics (from 5.93 to 11.56), and individual providers (from 29.88 to 55.45). Health-care use during the last year of life increased further, as expected.

Table5 presents an overview of characteristics of individuals enrolled in Part D by managed care status. Enrollment in the Part D low-income subsidy (LIS), which is a government assistance program that reduces prescription drug cost-sharing responsibilities for qualifying beneficiaries, was similarly more common among older persons, females, nonwhites, and persons with higher comorbidities regardless of FFS status (Table5). Receipt of LIS decreased over time among cancer patients enrolled in FFS plans but increased with time among those enrolled in managed care.

Discussion

The SEER-Medicare database has many strengths. Namely, it includes a large population-based sample size, which provides an established denominator. The included patient demographics (eg, age, race-ethnicity, sex), extensive clinical cancer characteristics (eg, diagnosis date, site, histology, stage), and detailed health-care use and cost information also allow for the identification and comparison of cancer treatment and outcomes across populations included in the data. Additionally, the longitudinal nature of the data, often available both before and after cancer diagnosis, allows for the calculation of time-dependent measures such as comorbidity indices, detailed multi-course treatments (eg, surgery, radiation, and systemic therapy), and outcomes such as time to subsequent event or death. Moreover, the recent addition of Medicare nursing home and home health care assessments allows for more comprehensive insights into the long-term health-care needs among cancer survivors. Finally, the ancillary files provide opportunities to study how provider characteristics affect cancer care.

There are also limitations to the SEER-Medicare database that warrant mentioning. First, as with most claims-based data sources, there is a time lag between an individual’s receipt of services and when the files become available for research (on average 2–3 years). The most recent data released in 2018 included cancer cases through 2015 and claims through 2016. In part, the time lag is driven by a SEER Program mandate to have complete case ascertainment in the catchment areas and CMS requiring that claims have sufficient maturity before a final research file can be created. Although the field of cancer treatment often moves more quickly than the implementation of new data collection requirements, the NCI consistently works to make high-quality data available as soon as it is feasible.

Second, SEER-Medicare analyses may not be generalizable to the entire cancer population. By nature of linking to the Medicare data and mostly including individuals 65 years and older, findings from SEER-Medicare analyses may not be generalizable to younger populations. Additionally, results based on the FFS population may not be generalizable to those enrolled in managed care, given there are demonstrated demographic differences between these two groups. Beneficiaries enrolled in managed care are often excluded from analyses due to their largely absent claims data. However, CMS has recently begun releasing managed care encounter data; thus, the encounter data may prove insightful to understanding how health-care use and outcomes vary by managed care enrollment status. Additionally, most investigators elect to restrict their analysis to the first primary cancer diagnosis; however, given cancer treatments are prolonging cancer survivorship and the US population is aging, a growing proportion of cancer patients are being excluded. NCI staff and investigators alike have begun exploring methods to better incorporate individuals who have been diagnosed with more than one cancer. For example, among patients who have been diagnosed with two cancers, NCI staff have developed a method to assign claims-based receipt of chemotherapy to a specific tumor (31) and have assessed the impact of having multiple tumors on the cost of cancer care (32). Others have considered whether the treatment of one cancer would affect treatment decision-making for a second cancer and have attempted to include as many people as possible in their analysis.

Third, not all health data are captured in the claims. Although diagnoses are included, information such as health-related behaviors, anthropomorphic data, and nonprescription medication use that may be found in medical records but are not included as diagnoses or procedures are not captured in the SEER-Medicare database. Likewise, if Medicare is not the primary insurance, a full claim for reimbursem*nt of services may not be submitted to Medicare. There also will not be claims for services recommended and not received. Additionally, although the SEER-Medicare database includes codes to indicate a specific test was performed, in general the results of the test (eg, hematological malignancy phase or severity, cytogenetic results, biomarker data, and laboratory values) will not be known. However, some test results are being incorporated into the data through data linkages and/or changes to Medicare coverage policies. For example, the NCI has augmented the SEER data through linkages with private entities. Most notably, the SEER data have been linked on the individual level to Genomic Health Inc. data to identify Oncotype Dx Breast Cancer Recurrence Score test results (33). In the future, results from this and/or similar tests may be included in the SEER-Medicare data. Once a test result becomes instrumental to characterizing a cancer subtype (eg, breast cancer hormone receptor or Human Epidermal Growth Factor Receptor 2 status) and/or to treatment reimbursem*nt (eg, epidermal growth factor receptor mutation status and erlotinib use), such results may become routinely captured in the SEER and/or Medicare claims data. Administrative claims data alone also lack information about the cancer care decision-making process (eg, who made the decisions, how or why the decisions were made, what was the correlation between planned and received treatment, and why was treatment altered or discontinued) and other patient-reported outcomes. In response, NCI has created separate linkages between the SEER data and the Medicare Health Outcome Survey and the Medicare Consumer Assessment of Healthcare Providers and Systems survey (34–36).

Fourth, researchers need to be prepared for the complexity of the SEER-Medicare database. Medicare data comprise multiple different file types with unusual one-to-many linkage relationships. Often, specialized knowledge about Medicare billing and coding is essential to properly manipulate and interpret the data (eg, billing rules and codes change over time). To increase reproducibility and to hopefully improve project time horizons, the NCI is developing coding databases, such as the Cancer Medication Enquiry Database, which consists of two searchable databases based on the Healthcare Common Procedure Coding System (HCPCS) and National Drug Code nomenclatures that will allow for quick identification of relevant cancer drug codes (37,38). Additionally, researchers are often faced with logistical and/or financial delays when they want to link SEER-Medicare data to additional data sources. For example, obtaining linked provider-specialty data from the American Medical Association requires additional time and funding. Therefore, NCI has explored the utility of the CMS Medicare Data on Provider Practice and Specialty data and the National Plan and Provider Enumeration System as alternatives to the American Medical Association data (39,40).

In conclusion, the SEER-Medicaredatabase remains a unique and powerful database for a diverse array of studies that track health-care delivery longitudinally among individuals over the course of the cancer continuum from cancer screening, diagnosis, treatment, and long-term follow-up (41–54). Analyses of the data have also provided insights into cancer care costs and health-care preparedness in the United States (55–57). Collaborators at NCI, CMS, and the SEER registries are committed to the continued maintenance and improvement of the SEER-Medicare database, thus ensuring future insights into cancer health services in the United States.

Funding

This work was supported in part by NIH P30 CA77598 (Masonic Cancer Center).

Notes

Affiliations of authors: National Cancer Institute, Division of Cancer Control and Population Sciences, Healthcare Delivery Research Program, Bethesda, MD (LE, JLW); Division of Health Policy and Management, School of Public Health, University of Minnesota, Minneapolis, MN (HP, BAV); Research and Rapid Cycle Evaluation Group, Center for Medicare & Medicaid Innovation, CMS, Baltimore, MD (LZ, DB); National Cancer Institute, Division of Cancer Control and Population Sciences, Surveillance Research Program, Bethesda, MD (DRR); Information Management Services, Calverton, MD (MJB).

The authors declare no conflicts of interest.

The contents of this article are solely the responsibility of the authors and do not necessarily represent the official views of the US Department of Health and Human Services or any of its agencies.

References

Published by Oxford University Press 2020. This work is written by US Government employees and is in the public domain in the US.

Issue Section:

Articles

Email alerts

Citing articles via

More from Oxford Academic