Inmates have high rates of latent tuberculosis infection (LTBI), but inmates are often released early and do not complete therapy in the community. This study evaluated the translation of results from a randomized trial to improve therapy completion to usual care in a county jail using Rogers' Diffusion of Innovation theory. Inmates who received a single education in the randomized trial in 1998-1999 (study group) were compared to inmates educated by Jail Discharge Planners in 2002-2003 (usual care group). Outcomes were rates of completion of a visit to the TB clinic and completion of therapy. Subjects in the usual care group were significantly less likely to go to clinic in the 30-day period after release (relative risk 0.84, 95% confidence interval 0.75-0.95). The transfer of an educational protocol did not achieve results seen under study conditions, mostly because of implementation fidelity. The educational session in the usual care period for 81.0% of inmates took 5 min, as compared to 10-15 min during the randomized trial. Differences in personnel administering the protocol, training, high turnover and time available may also account for lower rates seen. Practical clinical trials should focus on the context of care as well as the intervention and should have participation by those who will be implementing results.