To the Editor.
—Locke et al1 demonstrate that potential blood donors yield more data about personal risk factors for the human immunodeficiency virus (HIV) in private computer interviews than they do in response to written questionnaires and face-to-face encounters. As these authors acknowledge, it is now well recognized that patients respond favorably to computer-assisted interviews.The development of computer risk-assessment techniques is a relatively recent endeavor, and knowledge about the validation process for these instruments is at an embryonic stage. At this time, discussion should explore the nature of experimental methods for assessing the accuracy of these techniques.The conventional measures for reporting the accuracy of diagnostic instruments are sensitivity, specificity, and positive and negative predictive value. One's first impulse is to attempt to estimate these measures for risk assessment techniques. Locke et al claim that their computer interview is "more sensitive" than the traditional method. They promise that