This paper compares the use of multiple pools versus a single pool with respect to test security against large-scale item sharing among some examinees in a computer-based test, under the assumption that a randomized item selection method is used. It characterizes the conditions under which employing multiple pools is better than using a single whole pool in terms of minimizing the expected number of compromised items encountered by an examinee. The results obtained in this paper provide guidelines for constructing multiple pools optimally and evaluating the risks associated with some test designs that employ item reuse schemes. Finally, an adjusted-pool approach is proposed to achieve better test security than either the single-pool approach or the multiple-pool approach.