Reproducible Research: The Case of Sample Selection in SEER-Medicare Studies

• Activity Number: 75
• Type: Contributed
• Date/Time: Sunday, July 31, 2011 : 4:00 PM to 5:50 PM
• Sponsor: Section on Health Policy Statistics
• Author(s): Hoang Thanh Nguyen*+ and Ying Xu and Linda S. Elting
• Companies: The University of Texas MD Anderson Cancer Center and The University of Texas MD Anderson Cancer Center and The University of Texas MD Anderson Cancer Center
• Keywords: Reproducible Research ; Comparative Effectiveness ; Cancer ; SEER Medicare

Abstract

Recent events at Duke University point to the importance of reproducible research. For our purposes, reproducible research is the ability to replicate the study findings with the same data and methods. We evaluate the reproducibility of published studies that used the SEER-Medicare data. The SEER-Medicare data have been widely used to study all aspects of the cancer control continuum and the economics of cancer care. We review all publications listed under treatment and breast cancer on the SEER Medicare Website of the National Cancer Institute to evaluate the reproducibility of the construction of the sample cohort, which generally is the first step in any study that uses a publicly available large data set. We find that published papers generally provide the selection criteria but rarely the actual programming code or algorithm for sample selection. Even when the selection criteria are presented, they may not be presented with sufficient details for reproducibility. The importance of reproducibility in sample selection for SEER Medicare data is discussed and its implications to health services policy.