ABSTRACT
To develop a decision rule for predicting urinary culture results in patients suspected of having urinary tract infection, we used discriminant analysis to identify the optimum combination of clinical findings. Thirty variables identified in a pilot study were recorded from 248 patients in a second study. Five findings were independent predictors of positive urinary culture: history of urinary tract infection, back pain, microscopic pyuria, hematuria, and bacteriuria. An additive decision rule that assigned one point for each of the five variables was tested in a third group of 258 patients. These scores stratified patients into subsets with increasing likelihood of positive culture. Higher scores identified patients who can confidently be treated without documentation of bacteriuria. If the rule applies successfully to other populations, cost savings could result from identification of patients who do not require quantitative urinary culture to demonstrate significant bacteriuria.
