[Information value of clinical signs and stat tests as indicators of female outpatient urinary tract infection].

This study was designed to evaluate signs and stat tests as an indicator of lower urinary tract infection in female subjects with urogenital complaints at an out-patient clinic. Of various symptoms including hematuria, pollakiuria, dysuria, urinary retention, and micturition pain, pain during micturition was present in 48% of 25 patients with urinary tract infections and in 4% of 27 patients without urinary tract infections, and was the sign with the highest positive predictive value. Comparison and discrimination of the infection and non-infection groups using a single laboratory valuable yielded significant F-statistics for urinary leukocyte esterase (14.5) and leukocyte count in urinary sediment (31.1), and revealed large Mahalanobis' distances for the same variables. Multivariate analysis using a discriminant function of categorical data (Hayashi's Suryoka type 2) revealed that combining occult blood with leukocyte esterase in the urine or combining red cell count with leukocyte count in sediment did not yield substantially smaller misclassification error than did leukocyte esterase alone or leukocyte count alone. It was concluded that neither urinary occult blood nor red cell count in sediment contribute substantially to the prediction of urinary tract infection. For the purpose of detecting urinary tract infection among outpatients, a receiver-operating characteristic analysis demonstrated that the optimal cut-off point in sediment was 3 or more leukocytes per microscopic high power field (x 400). Urinary leukocyte esterase was found to have limitations for use in screening, because its optimal decision level is equivalent to trace esterase reading on the dipstick test.