Comparing the Cumulative False-Positive Risk of Screening Mammography Programs Using a Discrete Time Survival Model Allowing for Nonignorable Drop-Out

• Activity Number: 184
• Type: Contributed
• Date/Time: Monday, August 2, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Rebecca Hubbard*+ and Diana Miglioretti
• Companies: Group Health Research Institute and Group Health Research Institute
• Keywords: Discrete survival ; Informative censoring ; Bayesian ; Cancer ; Screening test

Abstract

Mammography is the only screening modality shown to reduce breast cancer mortality among women 50 and older. To evaluate screening mammography programs we must understand both benefits and harms. The most prevalent harm of screening mammography is the risk of a false-positive recall. Estimating the cumulative false-positive risk is complicated by the presence of a non-ignorable censoring mechanism. Drop-out may be related to covariates that also influence test performance or to the event itself. We propose a model for time to the first false-positive result that accounts for non-ignorable drop-out. We demonstrate that in this context the censoring mechanism is identifiable. We develop a Bayesian estimation method for jointly estimating the censoring bias function and false-positive risk and use this model to compare the cumulative false-positive risk of screening mammography programs.