A records analysis was conducted at 6 and 18 months after the introduction of random ward urine testing. This was followed by a true random prospective study. A random subsample of these drug tests was checked in the laboratory. In the first period there were few drug tests and inconsistent record keeping. During the second period, both of these areas improved, but testing was not random. In the third period, there were 217 tests (65 percent of those requested) in 33 days; 116 tests requested were not performed; 42 of these were due to patient refusals. Correlations between "dipstick" and laboratory results as well as research and central recording of tests were high. The rate of drug detection was no higher with true randomization. The hospital's central record was confirmed as a reliable and valid research tool regarding the use of illicit substances. In the year 2000, the central record showed 12 definite cases of illicit substance use in 153,887 patient days, but some for test refusals. The study concluded that random, reliable urine drug testing can be implemented in a secure hospital setting, with generally good cooperation of patients with staff. Contrary to popular belief, little drug use was found among the patients in this setting. Further research might examine whether current extensive preventive measures are cost effective. 2 tables, 10 references, and appended supplementary information on literature search terms and manufacturer's instructions for the drug test