Unrecognized TB among hospitalized patients is a well-recognized problem. The issue of including unsuspected cases of TB in the derivation of the prediction rule is a common question. It must be emphasized that the target population for the decision rule are those patients with signs or symptoms suggestive of TB. In any situation, if a clinician did not suspect a diagnosis of TB, he or she would not order the diagnostic test. Therefore, given that TB is not even considered in the differential diagnosis at the time of admission, the rule will not be applied to patients with unrecognized TB in the future. Consequently, the inclusion of these patients in the derivation process would have biased the estimators of the true relationship between the clinical predictors and the presence of TB among patients considered at risk of the disease, rather than improving the sensitivity of the prediction rule.
Wisnivesky JP, McGinn T, Crystal RG. Predicting Tuberculosis at Hospital Admission—Reply. Arch Intern Med. 2002;162(5):611-612. doi: