The diagnostic accuracy of rapid dipstick tests to predict urinary tract infection.

This study summarizes the diagnostic accuracy data for dipstick nitrite and/or leukocyte esterase tests (the index tests) as predictors of bacterial urinary tract infection as defined by quantitative culture (the reference test). On-line search of the literature using the MEDLARS database identified 1,017 citations, 51 of which were relevant and contained sufficient data for further analysis. From each citation, 2 x 2 tables of true-positive, true-negative, false-positive, and false-negative results were extracted. Four categories of index test were assessed: nitrite only, leukocyte esterase only, disjunctive pairing (dipstick positive if nitrite, leukocyte esterase, or both were positive), and conjunctive pairing (dipstick positive only if both nitrite and leukocyte esterase were positive). The true- and false-positive rates were calculated from each 2 x 2 table. Plots of true-positive rates versus false-positive rates demonstrated widely scattered points, indicating heterogeneity. A receiver-operating characteristic curve was fitted to the data using logistic transforms and weighted linear regression. This analysis revealed that the disjunctive pair is the most accurate index test. However, in many clinical settings, the posterior probability of urinary tract infection given a negative dipstick is too high to exclude it. Within most clinically relevant ranges of true- and false-positive rates, a negative urine dipstick test cannot exclude the diagnosis of urinary tract infection in patients with high prior probabilities of contracting this infection. For lower prior probabilities, the clinical efficacy of these rapid tests would best be determined by decision analysis, for which these receiver-operating characteristic functions would serve as valuable analytical tools.