A model for assessing the sensitivity and specificity of tests subject to selection bias