Background: Multiple drug-resistance in new tuberculosis (TB) cases accounts for the majority of all multiple drug-resistant TB (MDR-TB) worldwide. Effective control requires determining which new TB patients should be tested for MDR disease, yet the effectiveness of global screening recommendations of high-risk groups is unknown. Methods: Sixty MDR-TB cases with no history of previous TB treatment, 80 drug-sensitive TB and 80 community-based controls were recruited in Lima, Peru between August and December, 2008 to investigate whether recommended screening practices identify individuals presenting with MDR-TB. Odd ratios (OR) and 95% confidence intervals (CI) were calculated using logistic regression to study the association of potential risk factors with case/control variables. Results: MDR-TB cases did not differ from drug-sensitive TB and community controls in rates of human immunodeficiency virus infection, reported hospital or prison visits in the 3 years prior to diagnosis. MDR-TB cases were more likely than drug-sensitive TB controls to have had a recent MDR-TB household contact (OR 4.66, (95% CI 1.56-13.87)); however, only 15 cases (28.3%) reported this exposure. In multivariate modeling, recent TB household contact, but not contact with an MDR-TB case, remained predictive of MDR-TB, OR 7.47, (95% CI 1.91-29.3). Living with a partner rather than parents was associated with a lower risk of MDR-TB, OR 0.15, (95% CI 0.04-0.51). Conclusion: Targeted drug susceptibility testing (DST) linked to reported MDR-TB contact or other high-risk exposures does not identify the majority of new TB cases with MDR disease in Lima where it is endemic. All new TB cases should be screened with DST to identify MDR patients. These findings are likely applicable to other regions with endemic MDR-TB.