© 1998 by Oxford University Press
A database combining cancer tumor registry cases with Medicare claims files is giving researchers a powerful tool for looking at real-world applications of cancer treatment.
Created in 1992, the database was designed to help answer questions about patterns, costs, and outcomes of cancer-related health services. It captures these components and associated clinical characteristics, such as diagnosis and mortality, on approximately 650,000 people age 65 and over. The information is collected by the National Cancer Institute's Surveillance, Epidemiology, and End Results Program (SEER) and the Health Care Financing Administration.
Thirty-three outside investigators have requested and received SEER-Medicare data files, according to Joan Warren, Ph.D., of NCI's Applied Research Branch. Most of the resulting studies were published within the last 2 years.
Researchers are examining such topics as: diffusion of breast-conserving surgery in SEER geographic areas according to hospital characteristics; use and costs of procedures for monitoring patients after initial treatment for breast cancer; patterns of breast cancer screening and how it may influence stage at diagnosis and survival; lifetime costs and treatment patterns for persons with oral cancer; stage at diagnosis for melanoma and insurance status; costs of therapies for metastatic prostate cancer; and effects of early diagnosis of colon cancer on treatment costs and quality of life.
As more people have used the data set, both its promise and practical problems have emerged, Warren said. This past summer, at the first major gathering of SEER-Medicare data users, researchers discussed methodological issues, including nuances in the data caused by changes in Medicare policy and by the ambiguous description of some health care services.
Research Potential
Workshop attendee Beth Virnig, Ph.D., of the University of Minnesota School of Public Health, Minneapolis, said the SEER-Medicare database provided investigators with a way to look at real world applications when comparing cancer treatments. Some research questions, such as whether physicians treat older cancer patients differently from younger patients, do not lend themselves to a randomized study, but are possible because of this database. Analyzing these data are the next logical step for cancer research after randomized trials, she said.
It is difficult to determine from a hospital bill whether someone has a particular type of cancer, but the SEER data provide a specific diagnosis, Virnig noted. The Medicare data provides the subsequent utilization information and using the combined data researchers can get a more complete picture from diagnosis to current health or death.
"There's no limit to the possibilities of what people can do with this data and that's really exciting," Virnig said.
Ann Nattinger, M.D., Medical College of Wisconsin, Milwaukee, who has been working with the data for several years, called it a wonderful resource for studying older women, who generally tend not to participate in clinical trials as much as younger women.
Nattinger and MCW colleague Marilyn Schapira, M.D., are conducting a surveillance follow-up study of breast cancer survivors, looking at the timing and relationship among breast cancer diagnosis, treatment, and subsequent mammography. For this, they needed to understand when the treatment period was over, but the claims data did not provide this level of detail. After conducting an internal validation, they were able to show that women who have had radiotherapy treatment for breast cancer appear more likely to get subsequent mammograms.
"In this combined database, we're trying to exploit both the strength of SEER in ascertaining the date of diagnosis and the strength of Medicare in ascertaining the date of treatment," Nattinger said.
Marshall McBean, M.D., of the University of Minnesota School of Public Health, hailed the data set for potentially providing a more detailed picture of cancer survival.
"As 5- and 10-year survival improves, it is imperative that quality of life at 5 and 10 years also be understood," McBean said.
The SEER-Medicare data release policy has made access easier for researchers, while protecting private medical information, Virnig noted. Nattinger said that data security procedures, such as using the SEER identification numbers and age instead of date of birth, make breaches of confidentiality unlikely.
Variables in the data that might result in the identification of physicians or patients are removed and not routinely released, according to Warren. In addition, NCI, HCFA, and SEER officials review study proposals before releasing the data to determine if they pose a privacy threat. Researchers also must sign a data-release agreement governed by the federal Privacy Act of 1974. Before publication, all manuscripts and presentations must be sent to the original SEER reviewer through NCI for final approval to ensure the confidentiality of patients and providers in SEER areas. NCI grantee status is not required.
"What we've tried to do is implement a balance in protecting patient-provider confidentiality while not imposing overwhelming burdens on the researcher," Warren said.
Next Steps
NCI and HCFA are working to update the linkages and may report on their progress in the summer of 1999. In 1995, the linkage was updated to include all cases reported to the SEER registries through 1993 and available Medicare data through 1994. The SEER program now collects cancer incidence data from 10 standard registries in various geographic areas representing about 14% of the U.S. population.
Officials are considering convening another users workshop, but in the meantime will answer data user questions as they arise.
-- Gwen Moulton
Database Provides Window on Applications of Treatments
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||