Journal of Medical Systems

, Volume 18, Issue 1, pp 23–32

Analysis of a claims database for the identification of patients with carcinoma of the breast

Authors

  • Lawrence J. Solin
    • From the Department of Radiation Oncologythe University of Pennsylvania School of Medicine
    • U.S. Healthcare
    • Department of MathematicsMillersville University
    • the University of Pennsylvania Cancer Center
  • Antonio Legorreta
    • From the Department of Radiation Oncologythe University of Pennsylvania School of Medicine
    • U.S. Healthcare
    • Department of MathematicsMillersville University
    • the University of Pennsylvania Cancer Center
  • Delray J. Schultz
    • From the Department of Radiation Oncologythe University of Pennsylvania School of Medicine
    • U.S. Healthcare
    • Department of MathematicsMillersville University
    • the University of Pennsylvania Cancer Center
  • Howard A. Levin
    • From the Department of Radiation Oncologythe University of Pennsylvania School of Medicine
    • U.S. Healthcare
    • Department of MathematicsMillersville University
    • the University of Pennsylvania Cancer Center
  • Steven Zatz
    • From the Department of Radiation Oncologythe University of Pennsylvania School of Medicine
    • U.S. Healthcare
    • Department of MathematicsMillersville University
    • the University of Pennsylvania Cancer Center
  • Robert L. Goodman
    • From the Department of Radiation Oncologythe University of Pennsylvania School of Medicine
    • U.S. Healthcare
    • Department of MathematicsMillersville University
    • the University of Pennsylvania Cancer Center
Articles

DOI: 10.1007/BF00999321

Cite this article as:
Solin, L.J., Legorreta, A., Schultz, D.J. et al. J Med Syst (1994) 18: 23. doi:10.1007/BF00999321

Abstract

To develop and optimize algorithms for the identification of newly diagnosed and treated cases of women with carcinoma of the breast, an analysis was performed of cases identified from the claims database of a large health maintenance organization (U.S. Healthcare). An initial algorithm was developed from the patterns of claims which suggested common clinical presentations of carcinoma of the breast, and the positive predictive value was 88% (411/469). To attempt to improve upon the positive predictive value, multiple modifications of the initial algorithm were performed. The best identified modification of the initial algorithm yielded a positive predictive value of 93% (400/432) with a loss of only 3% (11/411) of the true positive cases. These results demonstrate that logic-based algorithms can be used as a valid and efficient method of identifying large numbers of cases from claims data with specific clinical characteristics. The best algorithm identified provides a powerful and accurate tool to perform health care analysis and research on large populations of women with newly diagnosed and treated carcinoma of the breast.

Copyright information

© Plenum Publishing Corporation 1994