The objective of this study was to describe outcomes of tuberculosis (TB) contact investigations, factors correlated with those outcomes, and current successes and ways to improve TB contact investigations. We abstracted clinic records of a representative U.S. urban sample of 1,080 pulmonary, sputum-smear(+) TB patients reported to CDC July 1996 through June 1997 and the cohort of their 6,225 close contacts. We found a median of four close contacts per patient. Fewer contacts were identified for homeless patients. A visit to the patient's residence resulted in two additional (especially child) contacts identified. Eighty-eight percent of eligible contacts received tuberculin skin tests (TSTs). Recording the last exposure date to the infectious patient facilitated follow-up TST provision. Thirty-six percent of contacts were TST(+). Household contacts and contacts to highly smear(+) or cavitary TB patients were most likely to be TST(+). Seventy-four percent of TST(+) contacts started treatment for latent TB infection (LTBI), of whom 56% completed. Sites using public health nurses (PHNs) started more high-risk TST(-) contacts on presumptive treatment for LTBI. Using directly observed treatment (DOT) increased the likelihood of treatment completion. We documented outcomes of contact investigation efforts by urban TB programs. We identified several successful practices, as well as suggestions for improvements, that will help TB programs target policies and procedures to enhance contact investigation effectiveness.