The first U.S. ELISA test for T. cruzi antibodies was licensed by the Food and Drug Administration (FDA) on December 13, 2006. Blood banks have begun screening in absence of FDA recommendations for best implementation methods. We surveyed 2,029 blood donors at five California sites with three risk-based Chagas risk-screening questions. Semi-Markov models compared the cost-effectiveness of three testing strategies. 30% of donors screened positively. Screening all dominated doing nothing, being less costly, and saving more lives. The choice to "screen and test" compared with "testing all" varied by Chagas prevalence, "screening and testing" being cost-effective for high (0.004) and low (0.00004) prevalences, and "testing all" cost-effective for moderate risk (0.0004). It is cost-effective to screen by ELISA rather than do nothing. The best strategy depends on site-specific risk. Census estimates of Hispanics do not predict donor risk well. We suggest using our screening questions to determine risk level and most cost-effective testing strategy.