SETTING: Alabama State Tuberculosis Control Program, USA. OBJECTIVE: To combine molecular screening data with routine information to assess transmission of Mycobacterium tuberculosis and improve control efforts. DESIGN: Since January 1994, samples from tuberculosis cases statewide have been systematically analyzed by IS6110 restriction fragment length polymorphism (RFLP). All cases during 1994-1995 with a predominate RFLP pattern were evaluated and risk factors assessed. pTBN12 was used to evaluate a large cluster in the Birmingham-Jefferson County (BJC) area. RESULTS: Statewide, a common two-band pattern was found, named JH2 (99/566, 17.5%). The most important risk associated with this pattern was homelessness (odds ratio, 8.9; P < 0.001). In the BJC area, the homeless accounted for 29% (51/175) of new cases diagnosed during the study period. For the BJC homeless, there were 13 unique RFLP patterns, and JH2 was predominant (29/33, 88%) among three clusters. Secondary analysis of the homeless JH2 cluster revealed a large group that included 19 of 24 (79%) isolates analyzed. Compared with the BJC non homeless (n = 124), the homeless were younger (P < 0.001), of male gender (P < 0.001), black race (P = 0.002), and were heavy alcohol (P < 0.001) and non-injection drug (P = 0.001) users. CONCLUSIONS: By screening tuberculosis cases statewide, a common two-band RFLP pattern was identified. Its predominance is explained by an ongoing tuberculosis epidemic among Birmingham's homeless population, highlighting RFLP as a tool for population surveillance. The pattern differences observed by pTBN12 typing clearly demonstrate that the isolates might be related but are not clonal.