Examining accuracy of screening mammography using an event order model

Area of Science:

• Radiology
• Medical Imaging
• Biostatistics

Background:

• Screening mammography is crucial for early breast cancer detection.
• Assessing mammogram performance requires understanding sequential outcomes.
• Factors like patient characteristics, facility, and radiologist can influence assessment accuracy.

Purpose of the Study:

• To propose a performance model for screening mammography based on the order of outcomes.
• To investigate agreement and disagreement rates at various risk factor levels.
• To analyze screening mammography data using a Bayesian framework.

Main Methods:

• Development of a multi-stage performance model for mammography assessment.
• Incorporation of patient, facility, and radiologist variables into the model.
• Application of a Bayesian framework for statistical inference.

Main Results:

• The proposed model can be built at each assessment stage (initial, follow-up, final outcome).
• The model reflects the influence of patient, facility, and radiologist on mammogram outcomes.
• Agreement and disagreement rates can be quantified and analyzed.

Conclusions:

• The developed model provides a structured approach to evaluating screening mammography performance.
• Understanding performance variations across different factors is essential for improving breast cancer detection.
• Bayesian inference offers a robust method for analyzing complex screening data.