Persons named by a patient with tuberculosis (TB) are the focus of traditional TB contact investigations. However, patients who use illicit drugs are often reluctant to name contacts. Between January 2004 and May 2005, 18 isoniazidresistant TB cases with matching Mycobacterium tuberculosis genotypes (spoligotypes) were reported in Miami; most patients frequented crack houses and did not name potentially infected contacts. We reviewed medical records and reinterviewed patients about contacts and locations frequented to describe transmission patterns and make recommendations to control TB in this population. Observed contacts were not named but were encountered at the same crack houses as the patients. Contacts were evaluated for latent TB infection with a tuberculosis skin test (TST). All 18 patients had pulmonary TB. Twelve (67%) reported crack use and 14 (78%) any illicit drug use. Of the 187 contacts evaluated, 91 (49%) were named, 16 (8%) attended a church reported by a patient, 61 (33%) used a dialysis center reported by a patient, and 19 (10%) were observed contacts at local crack houses. Compared to named contacts, observed contacts were eight times as likely to have positive TST results (relative risk=7.8; 95% confidence interval=3.8-16.1). Dialysis center and church contacts had no elevated risk of a positive TST result. Testing observed contacts may provide a higher yield than traditional name-based contact investigations for tuberculosis patients who use illicit drugs or frequent venues characterized by illicit drug use.