Related Experiment Video
Updated: May 19, 2026
Detection of Architectural Distortion in Prior Mammograms via Analysis of Oriented Patterns
Published on: August 30, 2013
Distinguishing screening from diagnostic mammograms using Medicare claims data
Joshua J Fenton1, Weiwei Zhu, Steven Balch
1*Departments of Family and Community Medicine, Radiology, and the Center for Healthcare Research and Policy, University of California-Davis, Sacramento, CA †Group Health Research Institute, Seattle, WA ‡Departments of Radiology, Epidemiology and Biostatistics, University of California-San Francisco, San Francisco, CA §Department of Biostatistics, University of Washington, Seattle, WA.
Simple algorithms accurately identify screening mammograms in Medicare claims data. These methods provide high positive predictive value (PPV) for mammography research and quality measurement, even without cancer registry linkage.
Area of Science:
- Health Services Research
- Medical Informatics
- Cancer Screening
Background:
- Medicare claims data offers potential for mammography research and quality assessment.
- Accurately distinguishing screening from diagnostic mammograms in claims data is challenging, especially without cancer registry linkage.
Purpose of the Study:
- To validate algorithms for identifying screening mammograms in Medicare claims data.
- To achieve high positive predictive value (PPV) with and without cancer registry linkage.
Main Methods:
- Claims-derived algorithms were developed using classification and regression tree analyses.
- Data included bilateral mammograms from female fee-for-service Medicare enrollees aged 66+ (1999-2005).
- Validation used Breast Cancer Surveillance Consortium (BCSC) registry data for linkage.
Main Results:
- Without registry linkage, a 3-step algorithm achieved 97.1% sensitivity and 94.9% PPV for screening mammograms.
- With registry linkage, sensitivity increased to 99.7% and PPV to 97.4%.
- Specificity remained similar across both datasets (approx. 63-69%).
Conclusions:
- Simple, claims-derived algorithms can reliably identify screening mammograms.
- High predictive values are achievable using Medicare claims alone or linked with registry data.
- These algorithms support mammography research and quality measurement.
