OBJECTIVES The cost-effectiveness of tuberculin screening may be substantially reduced by non-compliance of patients and physicians. We have examined the association of these problems with the socio-demographic characteristics of tuberculin reactors. METHODS Community-based tuberculin screening was conducted among students in grades 6 and 10, and in post-secondary health training, as well as young adult workers. A follow-up survey was conducted to determine if tuberculin reactors referred for further evaluation actually reported, if they were prescribed therapy when indicated, and if they took therapy when it was prescribed. Association of reactors' socio-demographic characteristics with these outcomes was analyzed. RESULTS Canadian-born subjects were less likely to report if they were: older (adjusted and standardized odds ratio: 0.7, 95% confidence interval: [0.5, 0.9]), resident in more affluent neighbourhoods (0.7 [0.6, 0.99]), and from single parent households (0.1 [0, 0.9]). Even when indicated, physicians were less likely to prescribe treatment for Canadian-born subjects who reported bacille Calmette-Guérin vaccination, but had not actually received this (0.3 [0.1, 0.7]), or who were from single-parent households (0.1 [0, 0.9]). Physicians were less likely to prescribe treatment for foreign-born who gave a history of BCG vaccination (0.1 [0.1, 0.3]), and were more likely to prescribe treatment for reactors from countries such as Haiti or Vietnam. The only factor significantly associated with compliance was that older Canadian-born subjects were less compliant (0.6 [0.4, 0.97]). CONCLUSIONS Failure to report for further medical evaluation and physician non-compliance were associated with a number of socio-demographic characteristics, and substantially reduced the benefit of a tuberculosis screening program.