SBIR-STTR Award

Recent work at the NCI linking record IDs between the NCI's SEER database and the Medicare data system (Potosky93), has opened the door to epidemiological analysis of many important guest ions in cancer treatment. Unfortunately, working with the Medicare data system is very difficult. The goal of this SBIR program is to make using the Medicare databases easier by building a personal computer (pc) cdrom to teach users about the contents of some of the core Medicare databases, and how to use those databases for health research. No such system exists for the Medicare databases today. In fact, despite the extraordinary content of the Medicare databases, they are under-used by the healthcare and epidemiology research communities owing to 2 main factors: (1) the difficulty of accessing non-public release files, and (2) a lack of practical knowledge about working with Medicare files (how do you do it, how much work does it involve, and what can you learn from the results?). Answers to these questions must take into account the changes (over time) of Medicare data formats and conventions. In this project, we propose to build a cdrom system that would contain (1) a simulation of some non- public Medicare datasets, (2) a sample of records drawn from some public Medicare datasets, (3) text discussion, with examples, of successful methods as well as problems associated with analyzing these data, (4) documentation about the actual data values and layout (codebook) and forms and contact information to help a user to submit an application to make scientific use of the Medicare databases.Proposed commercial application:The Medicare CDROM could be sold through a merchandising partner, as are other cdrom products produced at ISG.National Cancer Institute (NCI)

The purpose of this SBIR grant is to make simulated MEDICARE databases of cancer research information filled with dummy records embedded in the NCHS SETS software. The data values in the dummy records must be distributed by age, race, location, etc. similarly to the distribution of values in the real records so that computations using them will be useful. Such dummy databases could offer significant advantages in cost, ease of use and confidentiality restrictions over using real data, for preliminary/introductory use of MEDICARE datasets -- though of course they could not replace using real data for research studies. In Phase I, dummy datasets were constructed for prostate cancer and female breast cancer from the MEDICARE 1994 Inpatient Standard Analytical File using a Stratified Averaging and Sampling (SAVES) technique. HCFA approved the prostate dummy data- simulation for general release with no confidentiality restrictions. Phase-II will enhance the dummy simulation system to include additional cancers, better representation of racial distributions, enrollment/ population "denominator" data, the SEER-MEDICARE link system (Potosky 93) and other aspects. Perhaps a standardized dummying protocol could be developed, leading to similar simulations for many distribution- sensitive medical datasets PROPOSED COMMERCIAL APPLICATION: The proposed commercial applications are Educational Materials for students and researchers. Project products (CD-ROM and Web) will enable users to become familiar with MEDICARE & SEER data without high costs or privacy liabilities, and to design studies by estimating key properties of data. Tutorials will illustrate methods of cancer research using medicare data.

Thesaurus Terms:
Medicare /Medicaid, computer assisted instruction, computer program /software, computer system design /evaluation, information system, neoplasm /cancer epidemiology breast neoplasm, cancer information system, computer simulation, optical data storage, prostate neoplasm human dataNATIONAL CANCER INSTITUTE