OBJECTIVE: To assess predictors of latent tuberculosis infection (LTBI) completion by using structural equation modeling (SEM) among homeless adults, a group at great risk for LTBI and active tuberculosis (TB). LTBI therapy is effective in stemming the progression to active TB, yet treatment adherence among homeless persons is difficult to attain. DESIGN: By using SEM, the authors assessed predictors of LTBI completion among a sample of 494 homeless adults in Los Angeles, CA, who received either a nurse case-managed program (NCM) or a usual care program. MAIN OUTCOME MEASURES: Latent variables were created with the baseline variables of site type, age, intervention status, dissatisfaction with health care, depression, TB risk assessment, alcohol use, heroin or cocaine use, and TB knowledge. Outcome variables included many of the same baseline variables as well as treatment completion. RESULTS: LTBI treatment completion (100% adherence) was significantly and positively associated with participation in NCM, older age, and less heroin or cocaine use. NCM also predicted greater TB knowledge, greater ease of treatment, and more satisfaction with treatment (NCM completion rate = 64%, control rate = 42%). CONCLUSION: The culturally competent NCM program, combined with active tracking and incentives, was successful in a difficult-to-treat and highly transient population. (Authors)