Identifying methicillin-resistant Staphylococcus aureus (MRSA) colonization or infection present at admission has become important in reducing subsequent nosocomial transmission, but the most efficient surveillance methods remain to be defined. We performed anterior nares surveillance cultures of all patients upon admission to and discharge from the general internal medicine floor in our community hospital over a 7-week period, and patients completed a questionnaire on MRSA risk factors. Of the 401 patients, 41 (10.2%) had MRSA upon admission. Of the 48 risk measures analyzed, 10 were significantly associated with admission MRSA, and 7 of these were independently associated in stepwise logistic regression analysis. Factor analysis identified eight latent variables that contained most of the predictive information in the 48 risk measures. Repeat logistic regression analysis including the latent variables revealed three independent risk measures for admission MRSA: a nursing home stay (relative risk [RR], 6.18; 95% confidence interval [95% CI], 3.56 to 10.72; P < 0.0001), prior MRSA infection (RR, 3.97; 95% CI, 1.94 to 8.12; P = 0.0002), and the third latent variable (factor 3; RR, 3.14; 95% CI, 1.56 to 6.31; P = 0.0013), representing the combined effects of homelessness, jail stay, promiscuity, intravenous drug use, and other drug use. Multivariable models had greater sensitivity at detecting admission MRSA than any single risk measure and allowed detection of 78% to 90% of admission MRSA from admission surveillance cultures on 46% to 58% of admissions. If confirmed in additional studies, multivariable questionnaire screening at admission might identify a subset of admissions for surveillance cultures that would more efficiently identify most admission MRSA.