In the spring of 1987, the Athletic Director at Santa Clara University presented a proposal to the university's Athletic Board of Governance to test all student athletes for drug use. Some straightforward techniques of operations research, including a decision tree, were applied to the question of whether to test any single individual for the presence of drugs.
The heart of this analysis was a decision tree which began with the simple decision node having branches "test" and "do not test" and progressed to three outcomes: Drug user identified, false accusation, and unidentified drug user. Tables to determine the probability that a person is a drug user, given that he or she tests positive were developed. Tests having reliabilities between 75% and 99% and possible proportions of drug users in the general population ranging between 5% and 16.6% were included in the tables. The tables of probabilities were then used to determine the reliability that a test would need in order to reduce to an acceptable level the probability of making a false accusation.
Based on this analysis and the ensuing discussion at the Athletic Board of Governance meeting, the Board voted unanimously to recommend to the President of the university not to begin drug testing of student athletes. The Board had simply determined that no available test would reduce the probability of making a false accusation to an acceptable level.
Ultimately, the President of the university adopted the Board's recommendation. The Chairman of the Board later indicated that the analysis of the decision using a decision tree was the prime factor behind the Board's recommendation.