DNA fingerprinting was used to evaluate epidemiologically linked case pairs found during routine tuberculosis (TB) contact investigations in seven sentinel sites from 1996 to 2000. Transmission was confirmed when the DNA fingerprints of source and secondary cases matched. Of 538 case pairs identified, 156 (29%) did not have matching fingerprints. Case pairs from the same household were no more likely to have confirmed transmission than those linked elsewhere. Case pairs with unconfirmed transmission were more likely to include a smear-negative source case (odds ratio [OR] 2.0) or a foreign-born secondary case (OR 3.4) and less likely to include a secondary case <15 years old (OR 0.3). Our study suggests that contact investigations should focus not only on the household but also on all settings frequented by an index case. Foreign-born persons with TB may have been infected previously in high-prevalence countries; screening and preventive measures recommended by the Institute of Medicine could prevent TB reactivation in these cases.