(Learn more about this study)
Initial CDAS Request Approval
Feb 7, 2020
A new method to estimate the predictive value in a screening program
The goal of screening tests for a disease such as cancer is early detection and treatment with a consequent reduction in mortality from the disease. An important quantity one would like to know after the screening test is the predictive value (positive or negative predictive value) because screening test might produce false-positive and false-negative results. The available method of estimating this quantity typically assumes that all participants without a verification procedure after a positive screening test are missing-at-random, which is equivalent to the conditional independence of disease status and verification indicator when the testing result is given. In this project we introduce a new model that allows dependence between the disease status and the verification indicator and can easily incorporate with covariate information. The estimator of the predictive value under the new model will be introduced and the asymptotic normality of the estimator will also be studied. We also plan to apply the new model to the data from the prostate and ovarian components of PLCO trial.
1. Introduce a new model which doesn't require the conditional independence of disease status and verification indicator when the testing result is given.
2. Make a comparison between the new method and the traditional method via simulation.
3. Apply the method to data from the prostate and ovarian components of PLCO trial.