Determining predictors of true HIV status using an errors-in-variables model with missing data