Sir.—I read with interest the editorial by Fulginiti and Ray1 and the article by Evans et al2 in JAMA on which it comments. The findings of Evans et al are personally gratifying to me because they confirm my own belief and practice. Fulginiti and Ray's conclusions of the clinical implications of these findings moreover are appropriate. I must take exception, however, to several statements and omissions in what is otherwise an all too infrequent restatement of our professional imperative to "First of all, do not harm."
The evaluation of the "probability of association" that you ascribe to Katz3 is in reality a calculation of the positive predictive ability (PPA), as follows: PPA = Sn × Pd/(Sn × Pd + [1 − Sp] × [1−Pd]), where Sn indicates sensitivity; Sp, specificity, and Pd, baseline infection rate. The Katz formula3 shown below is true if (and only if) the