In the absence of predictive values for quantitative tests ordered in particular diseases, we approximated the comparative diagnostic merit of such tests by graphically reviewing hospital experience with those diseases. As a model, we examined the usefulness of the five chemical tests most commonly ordered on admission as part of the "Reye's workup": aspartate and alanine aminotransferases, NH3, prothrombin time, and glucose. We expressed the data as multiples of the upper reference limit (URL) for each of these liver-function tests, setting the URL high enough to overcome method- and age-related differences. Graphs of "first test" data from 141 patients show the great sensitivity of the two aminotransferase assays for Reye's syndrome: only 2% of the results fall below their URL, whereas about 32% of NH3 determinations and 27% of prothrombin times fall below their URL. Thus measuring the aminotransferases apparently is at least 10 times more useful than the other three analytes as "first tests." Current record keeping does not provide a ready means for assessing the specificity of these tests to determine the number of false positives, but does provide useful data for assessing sensitivity and false negatives.