During identical 2-month periods in 1989 and 1991, all applicants for employment at a major teaching hospital participated in preemployment drug screening. In 1989, before establishment of a formal preemployment testing program, screening was conducted without identifying information. Of 593 applicants screened, 64 (10.8%) were confirmed positive for one or more drugs. Marijuana metabolites were detected with the greatest frequency (35 samples, 55% of positive screens), followed by cocaine (36%), then opiates (28%). In 1991, after a formal preemployment testing program was in place, 365 applicants were screened, and 21 (5.8%) were confirmed positive. Opiates were most often detected (48% of positive screens), followed by cocaine (38%), then marijuana metabolites (28%). During both periods, positive urine screens were associated with ethnicity (non-White) and occupational category (blue-collar). Whereas in 1989 positive screens were associated with male gender, in 1991, females were more likely to test positive. The decline in prevalence following implementation of a screening program supports the notion that preemployment testing can serve as a deterrent for drug-using persons in applying for employment.