Birth Defects Registered by Double Sampling: A Bayesian Approach Incorporating Covariates and Model Uncertainty